ˇ˛a Endocrinology Service, University Hospital Clinic, University of Valencia, Blasco Iba n ez 15, E-46010 Valencia, Spain b Received 25 April 2002; received in revised form 28 October 2002; accepted 4 November 2002 European Journal of Internal Medicine 14 (2003) 101 106 Original article www.elsevier.com / locate / ejim Abdominal obesity, insulin resistance, and metabolic syndrome in a southern European population Juan F. Ascasoa,*, Pedro Romerob, Jose T. Reala, Rosario I. Lorentea, Jose Mart 1nez-Vallsa, a Rafael Carmena Center for Primary Care, Godella, Valencia, Spain Abstract Background: Our objective was to investigate the relationship between abdominal obesity (AO), as measured by waist circumference (WC), insulin resistance (IR), and components of the metabolic syndrome (MS). Methods: A cross-sectional study was carried out with 283 subjects (130 males and 153 females aged 25 65 years) from a primary care outpatient clinic in Valencia (Spain) over a period of 1 year. Body mass index (BMI), waist circumference (WC), blood pressure (BP), total cholesterol, triglycerides, HDL-C, glucose, and insulin were measured by standard methods. IR was defined as HOMA-IR equal to or greater than 3.8. Results: The prevalence of IR was 39.6%. Subjects were divided into groups according to WC. A normal WC was defined as below 88 cm in women and below 102 cm in men; AO was defined as a WC equal to or above 88 cm in women and equal to or above 102 cm in men. The prevalence of IR was 31.7% in the group with normal WC and 54.6% in the AO group (P,0.001). The percentage of subjects with the MS (high BP, dyslipemia or abnormal glucose tolerance) significantly increased (P,0.001) in subjects with AO (48.4 vs. 18.8% in normal WC subjects). AO is an indicator of IR with an odds ratio of 2.59 (95% CI 1.55 4.29). Conclusions: AO, expressed as WC, appears to be a good indicator of risk for IR and the MS, particularly in non-obese subjects (BMI,30). The main independent parameters of risk for IR are WC and TG, whereas those for the MS are IR, WC, and age. Ÿˆ 2003 Elsevier Science B.V. All rights reserved. Keywords: Insulin resistance; Abdominal obesity; Obesity; Metabolic syndrome; HOMA-IR 1. Introduction Obesity is the most prevalent metabolic disease in developed countries [1] and carries an important car- diovascular and global mortality rate, be it directly or through its association with numerous cardiovascular risk factors [2]. For this reason, it constitutes an important cause for concern from the point of view of public health. In a recent national survey [3], the prevalence of obesity (BMI$30 kg / m2 ) in the Spanish population between 25 *Corresponding author. Tel.: 134-96-386-2665; fax: 134-96-386- 4767. E-mail address: ascaso@uv.es (J.F. Ascaso). and 60 years of age is 13.4% (11.5% in men and 15.2% in women). In the Valencia region, the prevalence in the population under 65 years of age is 14.7% in men and 17.8% in women [4]. Abdominal or visceral obesity, clinically defined by the waist / hip ratio ($1 in men or $0.90 in women) or by waist circumference ($88 cm in women and $102 cm in men) [5,6], is related to the metabolic and cardiovascular changes that form part of metabolic syndrome (MS) [7]. The MS is characterized by abdominal obesity, dyslipemia, glucose intolerance or type 2 diabetes, high blood pressure, hyperuricemia or gout, hypercoagulability and fibrinolysis defects, and a high incidence of coronary heart disease (CHD). These characteristics coexist with an increase in 0953-6205/03/$ see front matter Ÿˆ 2003 Elsevier Science B.V. All rights reserved. doi:10.1016 / S0953-6205(03)00022-0 102 J.F. Ascaso et al. / European Journal of Internal Medicine 14 (2003) 101 106 the prevalence of non-alcoholic fatty liver, gallstones, hyperandrogenism, and osteoporosis [8]. Obesity, and particularly abdominal obesity (AO), is an important coronary risk factor, both directly and through its association with other risk factors. The goal of the present study was to assess the relationship between AO, measured by waist circumference (WC), insulin resistance (IR), and the main components of the metabolic syndrome (MS) in a southern European population with a low CHD prevalence. 2. Materials and methods 2.1. Subjects A cross-sectional study was carried out on men and women 25 65 years of age who, for various reasons, visited a primary care outpatient clinic in the metropolitan area of Valencia over a period of 1 year. An opportunistic search method, with sample selection, was carried out using a simple, random sampling method [9]. A total of 283 subjects (130 males and 153 females) were studied. Inclusion criteria were: voluntary participation in the study, normal hepatic and renal function, a complete blood count, assessment of thyroid hormones, and a standard urine analysis. Exclusion criteria were: age outside the given range; subjects following a hypocaloric diet or having ex- perienced either weight gain or loss greater than 10% within the previous 3 months; hypothyroidism, including sub-clinical hypothyroidism (TSH.5); liver, heart, or kidney failure, and neoplasias. Subjects with a previous history of viral hepatitis or cirrhosis were excluded. Subjects with plasma liver enzyme values twice the upper limits of normal were also excluded, as were diabetic subjects treated with insulin. 2.2. Methods Data collected on patients included age, sex, personal medical history, toxic habits, any history of coronary heart disease, ischemic stroke, or peripheral vascular disease, high blood pressure, diabetes, tobacco and alcohol con- sumption. Also noted were degree of physical exercise and a family history of high blood pressure, diabetes, ischemic heart disease, or dyslipemia. A qualitative dietary survey was carried out using a standard protocol described elsewhere [10]. Blood pressure was measured in the supine position and after a rest period of 10 min, with two readings taken 5 min apart. High blood pressure (HBP) was diagnosed if systolic blood pressure (SBP) values were equal to or above 140 mmHg, if diastolic blood pressure (DBP) was equal to or above 90 mmHg, or if the patient was being treated with anti-hypertensive drugs. Weight and height, body mass index calculation (BMI), and waist circumference (WC) were measured using standard methods. Abdominal obesity was defined as WC values equal to or greater than 88 cm in women or equal to or greater than 102 cm in men. These have been recommended as cut-off points due to their association with the metabolic changes related to obesity [11]. Blood samples were taken after a 12-h overnight fast, and plasma was separated immediately by refrigerated centrifugation at 2500 3000 rpm for a period of 10 min. The samples were processed immediately or in the first week following preservation at 220 8C. Total cholesterol (TC) and triglycerides (TG) were determined using en- zymatic methods [12,13] in a Technicon⁄ˆ analyser RAˇ€‹ 1000. HDL-C was measured using separation by precipi- tation with phosphotungstic acid and magnesium chloride [14]. Glucose was measured via an enzymatic method [15], and insulin was determined using an RIA [16]. Subjects with a fasting plasma glucose between 6.1 and 6.9 mmol / l were diagnosed as having abnormal fasting glycemia, while subjects previously diagnosed with dia- betes or having a fasting plasma glucose equal to or greater than 7 mmol / l were diagnosed as diabetics. Insulin resistance (IR) was measured using the HOMA index ( homeostasis model assessment of insulin sensitivi- ty ) with the formula described by Matthews et al. (HOMA-IR5insulin mU / l3glucose mmol / l / 22.5) [17]. In subjects without clinical or biological parameters of IR, the 90th percentile for the HOMA-IR was equal to or greater than 3.8, and this value was considered diagnostic of IR [18]. Metabolic syndrome (MS) was defined as the presence of at least two of the following: hypertension, impaired fasting glycemia (IFG) (fasting plasma glucose$6.1 mmol / l), and dyslipemia (plasma TG after fasting$1.7 mmol/l and/or HDL-C,1 mmol/l. 2.3. Statistical analysis For the descriptive analysis, and after having checked for normality of variables using the Kolmogorov Smirnov test, central measures of normal dispersion were used: average, standard deviation, and a confidence interval of 95% (95% CI) for quantitative variables and percentages in the case of qualitative variables. For the bivariable analysis, when the variables were parametric, the difference of averages test (Student s t- test) was used; in the case of variables with more than two categories, the one-way ANOVA test was carried out. In the case of non-parametric variables, either the Mann Whitney U-test or the Kruskal Wallis H-test was used, depending on whether two or more variables were being compared. The x2-test or Fisher s exact test was used to compare proportions. The correlation between two variables was studied with the Pearson or Spearman test, depending on whether the J.F. Ascaso et al. / European Journal of Internal Medicine 14 (2003) 101 106 103 variables had a normal (parametrical) distribution or not. Logistical regression analysis was used to assess the presence of independent associations between IR and other parameters. All analyses were carried out with the statistical pro- gram SPSS, version 10. 3. Results The study population included 283 subjects (130 men and 153 women) aged 25 65 years. There were no differences in the gender distribution in the study group. Table 1 shows the general characteristics of the subjects studied, divided according to their WC (normal,88 cm in women and ,102 cm in men and abnormal or AO$88 cm in women or $102 cm in men) and gender. Table 2 shows the prevalence of IR (HOMA-IR$3.8) and of the MS, characterized by the presence of at least two of the following changes: hypertension, dyslipemia (TG$1.7 mmol / l or HDL-C,1 mmol / l) and IFG, with men and women grouped together according to the pres- ence of AO. We found a significant increase (P,0.01) in the percentage of changes in relation to the waist perimeter (normal abnormal): IR 31.7% versus 54.6%, MS 18.8% versus 48.4%, as well as the differences in the percentage of subjects with hypertension, dyslipemia, IFG, and obesi- ty (P,0.01). The prevalence of AO, IR, and the MS in the population studied, divided into decades, is shown in Fig. 1. We observed significant differences (P,0.01) between de- cades. It is evident that in the younger population, the prevalence of IR is greater (25%) than AO (13%) and the MS (5%) (P,0.01). On the other hand, in the oldest Table 2 Prevalence of IR, MS, and metabolic components in the group studied, divided according to waist circumference N (M/W) IR MS Hypertension Dyslipemia IFG Normal waist No AO 186 (88/98) 31.7%* 18.8%* 21.5%* 40.3%* 13.9%* Abnormal waist AO 97 (42 / 55) 54.6% 48.4% 54.6% 63.9% 30.9% 60.8% Total group 283 (130 / 153) 39.6% 28.9% 32.8% 48.4% 16.9% 22.2% Obesity 2.1%* BMI$30 Table 1 General characteristics, blood pressure, plasma lipids, glucose, insulin and HOMA index in the subjects studied, divided according to waist circumference WC5waist circumference: normal (,88 cm in women or ,102 cm in men) and abnormal (AO)$88 cm in women or $102 cm in men. M5men, W5women; IR5insulin resistance (HOMA $3.8). MS5 IR metabolic syndrome 2 or more of the described alterations (hypertension, AGT and dyslipemia). IFG5impaired fasting glycemia (fasting plasma glucose$6.11 mmol / l); Dyslipemia5plasma triglycerides$1.7 mmol / l and / or HDL-C,1 mmol / l; Obesity5BMI$30. *P,0.01 normal WC versus abnormal WC. decade studied, no significant differences in the prevalence of these parameters were observed (IR 47%, AO 52%, and MS 49%). WC and BMI were directly related, with a high statisti- cal power, to age, blood pressure, plasma triglycerides, glucose, insulin, and HOMA-IR, and inversely to the concentration of HDL-C. All of these parameters are constituents of the MS. For IR (HOMA-IR), the logistical regression analysis (Table 3) demonstrates a significant correlation (P, 0.0001) with WC and plasma triglycerides. Other parame- ters, such as hypertension and age, were excluded from the model. The odds ratio (OR) for IR was 2.4 (95% CI 1.4 4.0) for WC and 2.7 (95% CI 1.6 4.5) for plasma Waist circumference: men Normal Abnormal 88 42 47.1611.5 48.4612.1 25.162.7* 31.963.7 89.367.6 108.8611.6 24%** 40% 66% 64% Waist circumference: women Total group 283 (M130 / W153) 47.2611.7 26.864.9 89.6613.2 14.5% 40% 128.2622.5 81.5610.9 5.2861.05 1.6560.86 1.2160.32 1.0760.27 5.8761.95 14.566.5 3.962.8 N Age (years) BMI (kg/m2) Waist (cm) Smokers Alcohol,30 g SBP mmHg DBP mmHg TC mmol/l TG mmol/l HDL-C mmol/l Apo B g/l Glucose mmol/l Insulin mU/ml HOMA IR 127.3620.9* 81.869.9** 5.0460.54 1.7260.81 1.0960.29 1.0460.27 5.9062.01** 13.464.7* 3.561.6* 139.5619.7 86.3610.8 5.2761.02 1.9861.01 1.0060.25 1.1160.23 6.6862.24 18.567.9 5.864.1 Normal 98 42.3611.7* 23.662.9* 77.366.4 24%* 23%** 119.1623.5* 76.5610.5* 5.1961.03* 1.3660.69* 1.3760.31** 1.0160.27* 5.3361.07** 12.664.4* 3.061.3* Abnormal 55 54.767.0 31.364.8 97.669.2 2% 11% 137.2617.2 85.9611.1 5.8461.07 1.7960.95 1.2660.25 1.2160.28 6.1762.51 16.568.6 4.864.1 WC: waist circumference: abnormal (AO)$88 cm in women and $102 cm in men. M5men, W5women; SBP5systolic blood pressure; DBP5diastolic blood pressure; TC5total cholesterol; TG5plasma triglycerides; HDL-C5high density lipoprotein cholesterol; Apo B5apolipoprotein B; Glucose5fasting plasma glucose; Insulin5fasting plasma insulin; HOMA 5index of insulin resistance. IR *P,0.01, **P,0.05 normal WC versus abnormal WC. 104 J.F. Ascaso et al. / European Journal of Internal Medicine 14 (2003) 101 106 Fig. 1. Prevalence of abdominal adiposity, IR and the MS according to age (decades). Abbreviations: N5number of subjects (M5males, F5 females); AO5Abdominal obesity; IR5Insulin resistance; MS5 Metabolic syndrome; P,0.01 between decades for AO, IR, and MS prevalence; no significant differences. triglycerides$1.7 mmol/l. In the case of the MS, there was an independent correlation with age (P,0.0001), OR 3.1 (95% CI 0.7 12.6) in the 35 44 year group, OR 5.6 (95% CI 1.5 20.9) in the 45 54 year group, and OR 12.2 (95% CI 3.4 44.2) in the 55 64 year group. The OR for IR (HOMA-IR$3.8) and WC [3.9 (95% CI 2.2 7.1) and 2.6 (95% CI 1.4 4.7), respectively] were also higher signifi- cant. 4. Discussion Obesity, particularly AO [19,20], is related to various cardiovascular risk factors that form part of the MS or IR Table 3 Odds ratios calculated with logistic regression analysis for IR (HOMA) as dependent variable and MS as dependent variable syndrome [21]. The IR syndrome is clinically character- ized by the association of two or more of the following components: changes in glycemia, dyslipemia, hyperten- sion, hyperuricemia, and other metabolic disturbances, together with a high cardiovascular mortality [22]. Recent- ly, other inflammatory components have also been related to obesity and IR [23,24]. However, the clinical definition of the MS remains under discussion and there is con- troversy concerning its standardization [25 28]. The quantification of IR has not been well established. The euglycemic-hyperinsulinemic clamp is the gold stan- dard for its quantification, but its use is difficult in epidemiological studies. For this reason, indirect measure- ments have been advocated, based on the levels of plasma insulin in the basal situation or 2 h after an oral tolerance test with 75 g of glucose or in the HOMA index ( homeo- stasis model assessment of insulin sensitivity ) proposed by Matthews et al. in 1995 [17]. This method has recently been validated with the euglycemic hyperinsulinemic clamp technique by Bonora et al. [29]. Our group [30,31] has shown, in a limited number of subjects, that hyperinsulinism can be defined as basal insulinemia greater than or equal to 16 mU/ml, or greater than or equal to 63 mU/ml 2 h after an oral glucose tolerance test. Our findings coincide with those proposed in the Paris Prospective Study [32], which defined hy- perinsulinism with basal plasma insulin greater than or equal to 16 mU/ml or greater than or equal to 62 mU/ml 2 h after an oral glucose load. These values were associated with a 1.6 increase in cardiovascular risk, independently of plasma glucose levels. In the present study, fasting plasma insulin levels were 14.566.5 mU / ml, with significant differences in the group of subjects with AO compared to the non-obese ones (18.567.9 vs. 13.464.7 mU/ml in men, P,0.01 and 16.568.6 vs. 12.6 mU/ml in women, P,0.01). We have also defined the degree of IR by a HOMA-IR index greater than or equal to 3.8, which corresponds to the 90th percentile of the subgroup of our population without clinical or biological parameters of IR [18]. This value coincides with those found by other authors in the upper fifth of a population of normal weight (HOMA-IR, 95% CI 2.77 36.4) [33]. In our whole group, the mean HOMA-IR value was 3.962.8, quite similar to the 3.8 (0.02 72.7) value ob- tained by Haffner et al. [34] in a Mexican-American population. In our group without AO, the HOMA-IR was 3.561.6 in men and 3.061.3 in women, whereas in the group with AO these values were 5.864.1 and 4.864.1, respectively (P,0.01). IR (HOMA-IR$3.8) was found in 31.7% of subjects with normal WC and in 54.6% of subjects with elevated WC. When studying the main components of the MS, we found a prevalence of 48% in subjects with AO and only 18% in those without AO (P,0.01). The prevalence in other populations varied from 0.8 to 35.3%, depending on the criteria used in establishing the diagnosis of the MS [27]. IR (HOMAIR$3.8)a WC (AO) Plasma triglycerides ($1.7 mmol/l) MS Age 35 44 45 54 55 65 IR (HOMAIR$3.8) WC (AO) OR 2.4 2.7 3.1 5.6 12.2 4.2 3.5 95% CI 1.4 4.1 1.6 4.5 0.7 12.6 1.5 20.9 3.4 44.2 2.2 7.6 1.3 4.5 P-value 0.001 0.000 0.000 0.114 0.000 0.000 0.000 0.003 years years years a Age and hypertension excluded. J.F. Ascaso et al. / European Journal of Internal Medicine 14 (2003) 101 106 105 It is well known that WC and BMI are statistically related to age, blood pressure, plasma triglycerides, glu- cose, insulin values, and HOMA-IR, and inversely related to the plasma concentration of HDL-C. All of these parameters are components of the MS. Logistical regres- sion analysis has shown a correlation between IR (HOMA- IR), WC, and TG (P,0.01), excluding other parameters from the model, such as hypertension and age. These results have also been found by other authors [34]. Plasma insulin and glucose values were not included in the mathematical method, as they were already included in the HOMA-IR formula. The odds ratios (OR) for IR (Table 3) were: WC (AO) 2.41 (1.43 4.05) and TG$1.7 mmol/l of 2.71 (1.61 4.54). On the other hand, the MS was related (P,0.01) to IR, WC, and age, with an OR of: HOMA-IR$ 3.8 OR 3.98 (95% CI 2.21 7.18), WC (AO) OR 2.64 (95% CI 1.46 4.77), and age ($45 years) OR 4.13 (95% CI 2.05 8.33), not including the metabolic components of the MS. According to our results, WC is an excellent predictor of risk for the MS and, probably, for ischemic heart disease, as has been shown recently in different ethnic groups [23]. When subjects with and without AO were divided by BMI values below 25, 25 29.9, and equal to or greater than 30 (Table 4), the prevalence of the MS and IR was sig- nificantly different only in the group with a BMI below 25. In subjects with a BMI of 30 or more (AO), WC measurements did not help to differentiate the prevalence of IR and the MS. Only four subjects with normal WC were identified in this category. On the other hand, in the group with a BMI of 25 29.9, subjects with AO showed a higher prevalence of the MS. The concept of the MS implies an evolutionary, dy- namic element. Accordingly, in our youngest age group (25 34 years), the proportion of subjects with IR (25%) was greater than that of subjects with AO (13%) and the MS (5%). These differences were statistically significant (P,0.001). On the other hand, in the oldest decade studied, there were no differences in the prevalence of these parameters: IR 47%, AO 52%, and the MS 49%. Table 4 Prevalence of IR and the MS in subjects with and without AO, according to the grade of obesity (BMI) These findings support the notion that IR is present at an early stage, preceding the different components of the MS that emerge at a later stage. This is in accordance with findings recently reported by other authors who also found IR without features of the MS [33 36]. We conclude that WC is a good predictor of risk of IR and the MS, especially in non-obese subjects (BMI,30). The main independent parameters of risk for IR are WC and TG, whereas those for the MS are IR, WC, and age. ˇ˛Abstract European Journal of Internal Medicine 18 (2007) 44 47 Original article www.elsevier.com/locate/ejim Abdominal tenderness in ascites patients indicates spontaneous bacterial peritonitis Sven Wallerstedt a,N , Rolf Olsson b, Magnus Simren b, Ulrika Broome c, Staffan Wahlin c, Lars Loof d, Rolf Hultcrantz e, Klas Sjoberg f, Hanna Sandberg Gertzen g, Hanne Prytz h, Sven Almer i a Department of Medicine, Sahlgrenska University Hospital/Ostra, SE-416 85, Goteborg, Sweden b Department of Medicine, Sahlgrenska University Hospital/Sahlgrenska, Goteborg, Sweden c Department of Medicine, Karolinska University Hospital/Huddinge, Huddinge, Sweden d Department of Medicine, Akademiska University Hospital, Uppsala, Sweden e Department of Medicine, Karolinska University Hospital/Solna, Solna, Sweden f Department of Medicine, University Hospital MAS, Malmo, Sweden g Department of Medicine, University Hospital, Orebro, Sweden h Department of Medicine, University Hospital, Lund, Sweden i Department of Medicine, University Hospital, Linkoping, Sweden Received 13 February 2006; received in revised form 1 June 2006; accepted 11 July 2006 Background: Spontaneous bacterial peritonitis (SBP), which has been reported to be present in 10 30% of patients with cirrhotic ascites, may easily be overlooked. An important aim of our study was to determine whether there are any clinical signs which, in clinical practice, may predict or exclude SBP. Methods: We studied 133 patients with cirrhotic ascites from medical units at nine Swedish university hospitals where there had been at least one diagnostic ascites tap with analysis of polymorphonuclear leukocytes in the ascites fluid. The patients had initially been questioned about background factors and physically examined according to a standardized case record form. Samples of blood, urine, and ascites were then drawn for analysis according to a structured schedule. Results: SBP could be excluded in 80% of all the cases and was confirmed in 8% of the 133 patients in the final analysis. Abdominal pain and abdominal tenderness were more common in patients with SBP (pb0.01), but no other physical sign or laboratory test could separate SBP cases from the others. Conclusions: SBP was present in about one-tenth of the hospitalized patients with cirrhotic ascites in this cohort. Performing repeated physical examinations and paying particular attention to abdominal tenderness may be the best way to become aware of the possible development of this complication. © 2006 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. Keywords: Cirrhosis; Diagnosis; Signs; Spontaneous bacterial peritonitis; Symptoms 1. Introduction Although spontaneous bacterial peritonitis (SBP) is an important complication of cirrhotic ascites, having been estimated to be present in 10 30% of cases [1], its prevalence N Corresponding author. Tel.: +46 31 3434000; fax: +46 31 259254. E-mail address: sven.wallerstedt@medic.gu.se (S. Wallerstedt). and clinical picture are somewhat obscure. There are several explanations for this, due to changes in the diagnostic criteria that have been established for SBP in the years since this clinical entity was first described in 1964 [2]. At that time, cultures positive for ascites and blood were obligatory for the diagnosis. Two decades later, Runyon and Hoefs described an ascites culture-negative variant of SBP [3]. One reason for a negative ascites culture could be the low bacterial content in 0953-6205/$ - see front matter © 2006 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.ejim.2006.07.022 SBP ascites, i.e., one to two organisms per mL [4], which, to some extent, can be overcome using a blood culture technique [1]. Nowadays, a positive ascites culture is not mandatory for the diagnosis of SBP, which is defined as neutrocytic ascites with at least 0.250 N 109 polymorphonuclear leukocytes (PMN) per liter in the absence of an abdominal infectious focus [1]. Thus, the diagnosis may be overlooked if abdominal pa- racentesis and analysis of ascites are not performed. In the past, paracentesis was generally restricted to severely ill patients because of an unfounded fear of complications [5]. In these patients, the prevalence of SBP was thought to be high. The modern practice of paracentesis, also in less diseased patients, should result in a lower frequency of SBP. This idea is supported by the finding of SBP in 3.5% of 427 asymptomatic cases among 513 outpatients with cirrhotic ascites who under- went paracentesis [6]. Various predisposing factors, clinical signs, and prognostic markers for SBP development have been studied. Analyses of factors that are associated with SBP have generally shown that the parameters included in the Child Pugh classification [7], albumin, coagulation factors (prothrombin complex), bilirubin, and encephalopathy are the most important. In 1986, Runyon concluded that there was an inverse relation between the protein content in ascites and susceptibility to infections [8]. This is supported, for example, by the finding that a protein concentration in ascites of less than 10 g/L is common in SBP, whereas carcinomatosis-related ascites with a high protein content very seldom becomes infected. One aim of the present study was to evaluate the prevalence of SBP in Swedish hospitalized patients with cirrhotic ascites. Based on our clinical experience, we hypothesized that SBP in modern clinical practice should be infrequent. Another aim was to study to what extent clinical and laboratory data could, in clinical practice, predict or exclude the presence of SBP in these patients. 2. Materials and methods Patients with a diagnosis of cirrhotic ascites and with an analysis of ascites leukocyte count, but without secondary peritonitis, were prospectively included in the study. Most of these patients had been hospitalized because of ascites. Only the first ascites episode during the study period was registered. Initially, 156 patients were included in the study, but two patients were later excluded since the ascites was bloody and no correction [9] had been performed to compensate for added blood leukocytes. Another 21 cases had to be excluded be- cause no differential leukocyte count had been performed, although the ascites leukocyte count was at least 0.250N 109/L. For the final analysis, 133 patients were available, and they were divided into two groups. The first ( non-SBP ) group consisted of patients with a leukocyte or PMN count in ascites below 0.250N 109/L. The second ( SBP ) group consisted of patients with an ascites PMN count of at least 0.250N 109/L. All patients were questioned about background factors and physically examined according to a standardized form (Table 1) before the ascites tap. The ascites volume was determined as the fluid volume received by total paracentesis with complete ascites mobilization as the goal (maximum 21 L in a non-SBP patient). The results of routine analyses of blood, serum, ascites, and urine were also recorded in accordance with a structured schedule (Table 2), and the patients were then classified according to their Child Pugh classification [7]. Body mass index (BMI), calculated after ascites mobilization, was used to reflect malnutrition. The local ethics committees approved the study. For the statistical analysis of possible differences between SBP and non-SBP cases with regard to background factors, symptoms, physical findings, and laboratory results, we used Fisher's non-parametric permutation test [10]. A p-value below 0.05 was regarded as significant. 3. Results SBP could be excluded in 123 of the original cohort of 154 patients (80%). A positive SBP diagnosis was established in 10 of the 133 patients included in the final analysis (7.5%). A positive bacterial ascites culture was found in two of these ten patients (Escherichia coli and alpha streptococci, respective- ly). In two of the eight culture-negative cases, antibiotics had been given before the paracentesis. S. Wallerstedt et al. / European Journal of Internal Medicine 18 (2007) 44 47 45 Table 1 Clinical data on 133 patients with and without SBP Age (years) Sex [male:female] (n) Background history SBP, earlier (n) Alcohol etiology (n) Hepatitis C antibodies (n) Antibiotics due to any infection in last 6 months (n) Cortisone treatment (n) Immunosuppressive treatment (n) Portosystemic encephalopathy, earlier (n) Variceal bleeding (n) Variceal sclerotherapy (n) Present case history Alcohol intake last week (n) Respiratory infection (n) Abdominal pain (n) Intravenous therapy (n) Days at hospital before puncture (median, n) Physical findings Body mass index [(weight in kg)/(length in m)2] Child Pugh-class [A:B:C] (n) Spiders e" 3 (n) Fever (n) e" 38.0 ∞C (n) Abdominal tenderness (n) Portosystemic encephalopathy (n) SBP (n=10) 61 8:2 0 6 4 (9) 3 1 0 2 3 2 2 1 8 6 2.0 (8) 25.4 (8) 0:3:7 6 4 1 7 2 non-SBP (n = 123) 58 86:37 9 (119) 84 25 (114) 40 (121) 7 4 (122) 26 30 (117) 28 (115) 37 7 36 (121) 61 (121) 4.0 (50) 23.5 (85) 0:25:98 59 (115) 29 (119) 7 28 (120) 22 (119) In cases of missing data, the number of patients is given within parentheses. 46 S. Wallerstedt et al. / European Journal of Internal Medicine 18 (2007) 44 47 Table 2 Laboratory tests (mean values with confidence intervals within parentheses) in 133 patients with and without SBP reflect readiness for paracentesis, especially in such cases where SBP is common. Furthermore, our clinical impression is that there is a substantial number of patients with cirrhotic ascites who do not undergo a diagnostic paracentesis. A problem when comparing various prevalence studies is the use of different diagnostic criteria. For example, in a well-performed Danish study from 1987 [11], 4 out of 13 patients with SBP would nowadays have been classified as having bacterial ascites [1,4] rather than SBP, reducing the reported SBP prevalence from 19% to 13%. Other factors, such as geographical differences, may explain the rather high SBP prevalence of 23.7% in a recent Slovakian study, al- though an old disease definition, i.e., a PMN count of at least 0.500N 109/L, was used in that study [12]. This reported pre- valence figure would have been even higher if the modern definition had been used. Several studies have been performed to reveal factors pre- disposing to SBP. Andreu et al. [13] found that a prothrombin complex greater than 1.2 (INR), a serum albumin below 27 g/L, and a serum bilirubin above 42.7 ºmol/L had prognostic sig- nificance. The two last parameters could be used in a formula to calculate the relative risk of a first SBP. The three parameters mentioned are also components of the formula used for the modified Child Pugh classification [7]. Thus, it is not sur- prising that Child Pugh class C has been reported to be a risk factor for SBP [5]. As is apparent from our study, there is a great overlap in these factors between cases with and without SBP. This is also apparent with other reported risk factors, such as protein content in ascites below 10 g/L, a history of variceal bleeding, or an earlier SBP episode [5]. Although SBP may be asymptomatic, this complication must be considered if a patient with cirrhotic ascites develops symptoms or signs of local and/or systemic infection or an unexplained clinical deterioration. Many clinical signs have been studied, but in our study abdominal tenderness, present in 70%, was the only one that was strongly related to SBP. This is consistent with what has been reported by Runyon, who claimed that 39 44% of patients with a diagnosis of SBP showed this physical sign [14]. Abdominal tenderness was, along with jaundice, the most prominent physical sign, 54.5%, in a recent study of symptomatic SBP cases [15]. Also, other symptoms and signs, such as abdominal pain, clinically relevant alterations of gastrointestinal motility, fever, mental deterioration, and renal impairment, may re- flect the development of SBP [1]. Irrespective of the reported SBP risk factors, paracentesis should be performed in every hospitalized patient with cirrhotic ascites. This is also true for patients admitted for reasons other than ascites. It is important to get a PMN count in order to initiate antibiotic treatment and to use an adequate culture technique [1]. According to our results, repeated paracentesis must be con- sidered for patients developing abdominal pain and/or abdom- inal tenderness, even though a minority of these patients have SBP. On the other hand, the absence of abdominal tenderness very strongly contradicts the presence of this complication, not only in outpatients [6] but also in hospitalized patients. B-hemoglobin (g/L) B-leukocyte count (109/L) B-platelet count (109/L) S-AST (ºkat/L) S-ALT (ºkat/L) S-alkaline phosphatases (ºkat/L) S-bilirubin (ºmol/L) Prothrombin complex (INR) S-sodium (mmol/L) S-potassium (mmol/L) S-creatinine (ºmol/L) S-albumin (g/L) S-immunoglobulin G (g/L) Ascites-volume (L) Ascites-protein (g/L) Ascites-albumin (g/L) Ascites-amylase (ºkat/L) U-sodium (mmol/L) U-potassium (mmol/L) SBP (n=10) 116 (102 131) 11.5 (5.8 17.2) 164 (98 230) 4.61 (0 11.06) 1.92 (0 3.95) 13.3 (3.3 23.3) 140 (68 212) 1.4 (1.19 1.70)4 132 (128 136) 4.2 (3.6 4.8) 157 (92 222) 26 (22 31)4 16 (12 21)2 5.5 (2.2 8.9)1 17 (7 26)2 10 (4 16)3 0.6 (0 1.3)1 31 (6 57)1 37 (13 61)1 non SBP (n = 123) 114 (110 117) 10.5 (9.4 11.7)12 183 (164 202) 1.92 (1.56 2.28)12 1.24 (0.43 2.04)12 8.8 (7.8 9.8)12 112 (88 136)12 1.5 (1.46 1.59)12 133 (132 134) 4.0 (3.8 4.1) 124 (106 142)12 24 (23 25)11 17 (15 18)10 6.7 (5.5 7.9)5 13 (11 15)8 8 (5 10)9 0.6 (0.4 0.7)6 65 (54 76)7 33 (29 38)7 1n=5, 2n=6, 3n=8, 4n=9, 5n=59, 6n=65, 7n=71, 8n=72, 9n=90, 10n=96, 11n=115, 12n=122. B, blood; S, serum; U, urine. There were few clinical differences between the two groups (Table 1). Complaints of abdominal pain, as well as the pre- sence of abdominal tenderness, were, however, more common in SBP than in non-SBP patients (80% vs. 30%, p = 0.0050, and 70% vs. 23%, p = 0.0079, respectively). With regard to abdominal tenderness as a possible sign of SBP, the positive predictive value was 20% (7/35) and the negative predictive value 97% (92/95). None of the SBP patients had had an earlier SBP episode, whereas this was reported in the files of 9 of the 119 non-SBP cases. The time period between admission and paracentesis was similar in the two groups. Signs of liver dysfunction, malnutrition, the presence of an alcohol or hepatitis C viral etiology, immune-modulating therapy, suspicion of a suscep- tibility to infections, esophageal varices, and intravenous catheterization were not more common in the SBP group. The results from laboratory tests did not differ between the two groups (Table 2). 4. Discussion We found that SBP, as a complication of cirrhotic ascites, was less frequent than what has generally been published to date [1]. Although some of the local laboratories did not perform a PMN count in cases with an ascites leukocyte count below 0.500N 109/L, SBP could be excluded in more than 80% of the cases since at least the total leukocyte count was below 0.250N 109/L. Thus, the occurrence of SBP in 7.5% of our cases might have been higher if this com- plication had been present in some of the excluded cases. On the other hand, our figure may be too high since the high prevalence of abdominal tenderness in our cohort, 27%, could In conclusion, our study indicates that performing repeated clinical examinations of patients with cirrhotic ascites, paying particular attention to abdominal tenderness, is essential in deciding whether diagnostic paracentesis should be carried out to reveal possible SBP. 5. Learning points " Although infrequent, SBP should be considered in cirrhotic patients with ascites and abdominal tenderness. " The absence of abdominal tenderness, on the other hand, strongly contradicts SBP. Acknowledgements The study was initiated by SILK, the Swedish Internal Medicine Liver Club, which is supported by Meda. We also thank Anders Oden, PhD, for the statistical analyses. Accuracy of preoperative CT T staging of renal cell carcinoma: which features predict advanced stage? * A.J. Bradleya, , , * L. MacDonalda, * S. Whitesideb, * R.J. Johnsona, * V.A.C. Ramanic Show more doi:10.1016/j.crad.2015.03.013 Get rights and content Highlights • Presence or absence of 4 features were correlated with pT stage in 94 cases of RCC. • Sensitivity, specificity, PPV, NPV for predicting pT Stage ≥pT3a were derived for each feature. • Presence of perinephric stranding and tumour necrosis were not reliable signs for pT stage >T3a. • Thickening of Gerota's fascia appeared to be reliable sign of pT Stage ≥pT3a. • Presence of collateral vessels appeared to be a reliable sign pT Stage ≥pT3a. Aims To characterise CT findings in renal cell carcinoma (RCC), and establish which features are associated with higher clinical T stage disease, and to evaluate patterns of discrepancy between radiological and pathological staging of RCC. Materials and methods Preoperative CT studies of 92 patients with 94 pathologically proven RCCs were retrospectively reviewed. CT stage was compared with pathological stage using the American Joint Committee on Cancer (AJCC), 7th edition (2010). The presence or absence of tumour necrosis, perinephric fat standing, thickening of Gerota's fascia, collateral vessels were noted, and correlated with pT stage. The sensitivity, specificity, and positive (PPV) and negative predictive values (NPV) for predicting pT stage ≥pT3a were derived separately for different predictors using cross-tabulations. Results Twenty-four lesions were pathological stage T1a, 21 were T1b, seven were T2a, 25 were T3a, 11 were T3b, four were T3c, and two were T4. There were no stage T2b. Sixty-three (67%) patients had necrosis, 27 (29%) thickening of Gerota's fascia (1 T1a), 25 had collateral vessels (0 T1a), 28 (30%) had fat stranding of <2 mm, 20 (21%) of 2–5mm and one (1%) of >5 mm. For pT stage ≥pT3a, the presence of perinephric fat stranding had a sensitivity, specificity, PPV and NPV of 74%, 65%, 63%, and 76%, respectively. Presence of tumour necrosis had a sensitivity, specificity, PPV, and NPV of 81%, 44%, 54%, and 72%, respectively. Thickening of Gerota's fascia had a sensitivity, specificity, PPV, and NPV of 52%, 90%, 81% and 70%, respectively; and enlarged collateral vessels had a sensitivity, specificity, PPV, and NPV value of 52%, 94%, 88%, and 71% respectively. Conclusion The presence of perinephric stranding and tumour necrosis were not reliable signs for pT stage >T3a. Thickening of Gerota's fascia and the presence of collateral vessels in the peri- or paranephric fat had 90% and 94% specificity, with 82% and 88% PPV, respectively, for the presence of tumour stage for pT stage >T3a. These are considered reliable signs of locally advanced renal cancer. Introduction Kidney cancer (the majority of which is renal cell cancer [RCC]) is the eighth most common cancer in the UK, with 9639 new cases in 2010, the latest year for which statistics are available. Incidence rates have increased since the mid-1970s by 125% in males and 160% in females.1 Increasingly, renal tumours are presenting at smaller size and earlier stage, although this does not always confer survival advantages.2 and 3 Partial nephrectomy (nephron-sparing surgery) is considered the standard of care for T1a tumours, as the long-term incidence and consequences of renal insufficiency is less than after radical nephrectomy.4 Partial nephrectomy can be performed open or with minimally invasive approaches, with equivalent efficacy, although recovery (and hence length of stay in hospital) is shorter with the laparoscopic technique.4, 5, 6 and 7 Thermal ablation with either radiofrequency or cryotherapy is becoming an attractive option for T1a tumours in patients with co-morbidities that prevent surgical removal, with low complication and recurrence rates.8 and 9 These minimally invasive procedures can be performed either with computed tomography (CT) guidance, or laparoscopically with ultrasound guidance. A recent meta-analysis slightly favours cryotherapy as the more efficacious technique, having fewer complications and a lower recurrence rate than radiofrequency ablation.10 Radical nephrectomy (open or laparoscopic) is generally reserved for T3–4 disease, with laparoscopic nephrectomy preferred in some patients with T1b and T2 disease. Supradiaphragmatic (level 3) tumour thrombus extension would involve a combined thoraco-abdominal approach. Accurate radiological staging of RCC is vital to enable appropriate management decisions when cases are discussed at the level 2 multidisciplinary team meeting (MDT). Measurement of tumour size on CT has been shown to correlate well with resected specimens with a tendency to overestimate size slightly in masses of 4–5 cm diameter.11 and 12 CT is known to be less efficacious when attempting to distinguish between tumours confined to the renal parenchyma (T1 and 2) and those that have involved the renal vein, sinus, or extra-capsular fat (T3a). Inflammatory changes surrounding the tumour can mimic tumour extension and the presence of a secondary pseudocapsule can give the impression of an intact capsule, when in fact there is invasion.13 and 14 An audit of the authors' own accuracy at preoperative T staging of RCC in 90 patients, revealed concurrence in 67 cases and disagreement in 23 cases, when compared to final histopathology stage. Many of these were minor changes in size resulting in moving up or down one stage, without management implications, but there were five overcalls and six under-calls of stage pT3a, which might have been managed differently, if this could have been predicted.15The present study was undertaken to evaluate selected imaging features that may be attributed to more advanced radiological stage (for example, presence and thickness of perinephric stranding, thickening of Gerota's fascia, presence of collateral vessels in the perinephric fat, and presence of tumour necrosis) to investigate whether the presence of these features could predict pathological tumour (pT) stage. Materials and methods As this was a service-improvement study and audit, ethical approval was not required, as advised by the local research and development department. Records of the level 2 MDT meeting between 2006–2011 were reviewed and cases of surgical resection of pathologically proven RCC were selected. As a tertiary referral centre, a number of patients had their staging CT performed at outside hospitals, and the studies were incorporated into the picture archiving and communication system (PACS) for the purpose of MDT review. If studies were incomplete, cross-sectional imaging was repeated at the authors' institution. The minimum acceptable dataset for inclusion were a non-enhanced abdominal CT examination, arterial-phase CT thorax and parenchymal phase of the abdomen, with a maximum section thickness of 5 mm. The standard CT protocol for staging RCC is given in Table 1. Table 1. CT protocol for staging renal cancer. Parameters Routine Scan Reconstruction interval Topogram Thorax and abdomen 40 Ma 120 kW 1 mm 512 length Pre-contrast 100 mAs 120 kW 1.5 detector Diaphragms to Iliac crest 5 mm 1 mm Soft-tissue windows Arterial thorax and abdomen 140 mAs 120 kW 1.5 detector 20 sec delay Thoracic inlet through kidneys 5 mm 1 mm Lung and soft-tissue windows Parenchymal abdomen 180 mAs 120 kW 1.5 detector 110 sec delay Diaphragms to iliac crest 5 mm 1 mm Soft-tissue windows Contrast medium 100 ml of 300 mg iodine/ml strength iodinated non-ionic contrast medium Table options All imaging studies were reviewed by one senior radiology trainee, together with one of three consultant uroradiologists, with 20, 15, and 6 years of experience at consultant level. Reviewers were blinded to the original CT report, the MDT review, and the histopathological information. The images were reviewed using a GE PACS with RA1000 PACS workstations (GE Healthcare, Barrington, IL, USA). Three-dimensional (3D) soft-copy reconstructions in the coronal and sagittal planes were routinely reviewed alongside the axial images. Maximum dimensions in two of the available three planes were recorded for T staging. In addition to attributing a T stage for each tumour, images were analysed for the following: presence and thickness of perinephric stranding <2, 2–5, and >5 mm; necrosis within the tumour thickening; transgression of Gerota's fascia; and enlarged perinephric collateral vessels. Due to the timing of when the data were collected, both the sixth and seventh versions of American Joint Committee on Cancer (AJCC) were used in the initial reports. For the purpose of the present study, imaging and surgical pathological staging were recorded according to the seventh edition16 (Table 2) reclassifying from the sixth edition as required. Note was made of whether the assessed clinical T stage concurred with the final histopathological report, allowing for changes due to the seventh edition. Discrepancies were recorded, and whether this resulted in an upstaging or downstaging of the tumour. All patients were operated upon and had their pathology reported at University Hospital of South Manchester. The operation performed and histological subtype of RCC were recorded. Table 2. The American Joint Committee on Cancer 7th edition TNM staging for renal cell cancer (T stage). Tx Primary tumour cannot be assessed T0 No evidence of primary tumour T1 Tumour ≤ 7 cm in greatest dimension, limited to kidney T1a Tumour ≤ 4 cm in greatest dimension, limited to kidney T1b Tumour > 4 cm but ≤ 7 cm in greatest dimension, limited to kidney T2 Tumour > 7 cm in greatest dimension, limited to kidney T2a Tumour > 7 cm but ≤ 10 cm in greatest dimension, limited to kidney T2b Tumour > 10 cm in greatest dimension, limited to kidney T3 Tumour extends into major veins or perinephric tissues but not into the ipsilateral adrenal gland and not beyond Gerota's fascia T3a Tumour grossly extends into the renal vein or its segmental branches, or tumour invades perirenal &/or renal sinus fat but not beyond Gerota's fascia T3b Tumour grossly extends into the vena cava below the diaphragm T3c Tumour grossly extends into the vena cava above the diaphragm or invades the wall of the vena cava T4 Tumour invades beyond Gerota's fascia (including contiguous extension into the ipsilateral adrenal gland) Table options Statistical analysis Categorical parameters (pT stage) were summarised using frequencies, percentages and cross-tabulations. Additionally the pT stage coding (0 to 7) was considered an ordinal score, and analysed using Kruskal–Wallis and Mann–Whitney U-tests. The overall analyses used the conventional two-sided 5% significance level. Adjustments for multiple pairwise comparisons used a quasi-Bonferroni correction. The 95% confidence intervals (CI) for percentages were calculated using the Wilson method in CIA version 2.0, all other summaries and analyses were produced using IBM SPSS version 20. In addition, pairwise comparisons of pT stage between fat stranding categories were carried out using Mann–Whitney U-tests, adjusting the critical value to 0.017 (0.05/3) due to multiple testing. To assess predictions of pT staging using sensitivity and specificity measures, the pT stages were dichotomised into stages 5 mm was included in the group of 2–5 mm. The presence of perinephric fat standing according to pathological T stage is given inTable 4. Imaging features according to pathological T stage are given in Table 5. Table 6presents the predictors with dichotomised pathological T stages. Table 4. Presence of perinephric fat stranding according to pathological T stage (percentages calculated within each pT stage). Pathological stage Perinephric fat stranding None (n = 45) <2 mm (n = 28) ≥2 mm (n = 21) T1a 17 (70.8%) 6 (25%) 1 (4.2%) T1b 14 (66.7%) 5 (23.8%) 2 (9.5%) T2a 3 (42.9%) 2 (28.6%) 2 (28.6%) T2b 0 0 0 T3a 6 (24%) 8 (32%) 11 (44%) T3b 2 (18.2%) 6 (54.5%) 3 (27.3%) T3c 2 (50%) 1 (25%) 1 (25%) T4 1 (50%) 0 1 (50%) Table options Table 5. Imaging features according to pathological T stage (percentages in brackets). Pathological stage Necrosis Thickening of Gerota's fascia Enlarged collateral vessels No (n = 31) Yes (n = 63) No (n = 67) Yes (n = 27) No (n = 69) Yes (n = 25) T1a 16 (67) 8 (33) 23 (96) 1 (4) 24 (100) 0 T1b 6 (28) 15 (72) 18 (86) 3 (14) 21 (100) 0 T2a 1 (14) 6 (86) 6 (14) 1 (86) 4 (57) 3 (43) T2b 0 0 0 0 0 0 T3a 5 (20) 20 (80) 12 (48) 13 (52) 16 (64) 9 (36) T3b 1 (9) 10 (91) 4 (36) 7 (64) 3 (27) 8 (73) T3c 2 (50) 2 (50) 3 (75) 1 (25) 1 (25) 3 (75) T4 0 2 (100) 1 (50) 1 (50) 0 2 (100) Table options Table 6. Summaries and analyses for pT stage cut-off. Pathological stage cut-off p-Value pT3a) and perinephric standing, thickening of Gerota's fascia, enlarged collateral vessels (p < 0.001, Pearson's Chi squared tests, Table 6), and necrosis (p < 0.010). Patients with stages None 73.8% (58.9%, 84.7%) 65.4% (51.8%, 76.8%) 63.3% (49.3%, 75.3%) 75.6% (61.3%, 85.8%) Necrosis Yes 81% (66.7%, 90%) 44.2% (31.6%, 57.7%) 54% (41.8%, 65.7%) 74.2% (56.8%, 86.3%) Thickening of Gerota's fascia Yes 52.4% (37.7%, 66.6%) 90.4% (79.4%, 95.8%) 81.5% (63.3%, 91.8%) 70.1% (58.3%, 79.8%) Enlarged collateral vessels Yes 52.4% (37.7%, 66.6%) 94.2% (84.4%, 98%) 88% (70%, 95.8%) 71% (59.4%, 80.4%) Table options Thickening of Gerota's fascia had a sensitivity, specificity, PPV, and NPV of 52%, 90%, 81%, and 70%, respectively, for pT stage ≥pT3a (Table 7). Enlarged collateral vessels had a sensitivity, specificity, PPV, and NPV value of 52%, 94%, 88%, and 71%, respectively, for pT stage ≥pT3a (Table 7). Discrepancy review Twenty-seven (29%) of the 94 tumours were incorrectly staged using cross-sectional imaging. The type of discrepancy when compared with pathological stage is given inTable 8. There were five overcalls of size (Fig 4), and one undercall, and four overcalls of renal sinus invasion with one undercall. For perinephric extension, there were three overcalls and one undercall, one overcall and six undercalls of segmental renal vein invasion, and one undercall of right atrial venous thrombus. Regarding adjacent organ involvement, there were two over- and two undercalls (Fig 5). Table 8. Discrepancy numbers and stages by pT stage. pT stage Discrepancy type Number Totals T1a Overcall size 4 Overcall perinephric extension 2 6 T1b Overcall renal sinus infiltration 2 2 T2a Overcall size 1 Overcall renal sinus invasion 1 Overcall perinephric extension 1 Undercall size 1 4 T2b – – T3a renal vein Overcall segmental RV invasion 1 Undercall segmental RV invasion 6 T3a renal sinus Overcall invasion 1 Undercall invasion 1 T3a perinephric extension Overcall adjacent organ invasion 2 Undercall extension 1 11 T3b – – T3c Undercall right atrial thrombus 1 1 T4 Undercall adrenal invasion 2 2 27 Table options Figure 4. Large right renal tumour confined to the kidney. There is no perinephric fat standing. This was staged as T2b (10.8 cm) but pathologically was under 10 cm and thus T2a. CT tends to overestimate the size of tumours slightly. Figure options Figure 5. (a–b) Axial CT images of a large left-sided tumour. Radiologically staged as T4 with small bowel involvement (white arrow), pathologically T3a (extracapsular extension and proximal renal vein involvement). At surgery, Gerota's fascia was stretched over the nodule, thought to have transgressed it to involve the adjacent small bowel, accounting for the discrepancy. In addition to fat stranding and collateral vessels, note the thickening of Gerota's fascia (black arrow). Figure options Discussion Eight-five percent of tumours were of conventional clear cell type, with small numbers of other subtypes, including papillary and chromophobe, as has typically been found in other series.17 and 18 Five of the seven papillary tumours were stage T1a. Three tumours were clear cell with sarcomatoid features (one T3a and two T3b), and one stage T4 tumour was pleomorphic with sarcomatoid, clear cell and papillary features. Stages T1a, T1b, and T3a accounted for approximately 75% of cases. There were no cases of T2b (tumours >10 cm confined to kidney), which is in keeping with the increasing likelihood of extracapsular extension with increasing size of tumour. Zhang et al.19 documented a 91.9% incidence of extracapsular extension in tumours >10 cm. Known prognostic variables for RCC include tumour size, stage, and histological subtype. The AJCC staging takes size into account for pTa–pT2b. Tumours can be stage pT3a due to renal vein invasion, renal sinus invasion or extracapsular extension, which may be microscopic and thus not readily appreciated on CT. For pT, stage >T3a, size does not always correlate with stage of tumour, hence the choice of features that may help predict more advanced stage. Minor stranding in the perinephric fat around the whole kidney is a recognised feature of ureteric obstruction, particularly acute obstruction secondary to calculus disease,20 and 21or infections of the renal parenchyma. It is postulated to be due to dilated lymphatics, and can persist after the obstruction resolved. For this study, perinephric stranding was only included if it was local to the tumour and not generalised. It is acknowledged that there may be some inaccuracies as some stranding attributed to a tumour may be benign and incidental.22 One of the two patients with stage T4 had no stranding, and one of the 24 stage T1a patients had stranding of 2–5 mm. The presence of fat stranding had a 65% specificity and 63% PPV for pT stage >T3a and is not a reliable radiological sign for locally advanced renal cancer. Tumour necrosis visible at CT was seen in 63 patients. This is typically seen in larger tumours, and is known to be a poor prognostic factor in staging systems.22, 23 and 24Although tumour necrosis is reported to be unusual in small tumours, eight of the 24 T1a tumours did show this (33%), compared with 34 of the 42 T3 and T4 tumours (81%). The presence of tumour necrosis had a low specificity and PPV of 44% and 54%, respectively, and is not a reliable sign for pT stage >T3a. Thickening of Gerota's fascia (27/94 patients), and the presence of collateral vessels in the peri- or paranephric fat (25/94 patients) had a 90% and 94% specificity, and 81% and 88% PPV, respectively, for the presence of pT stage >T3a. These are considered reliable signs of locally advanced renal cancer. To the authors' knowledge, no other study has evaluated presence of these features as a predictor of pT stage >T3a. A study by Nazim et al. assessed the accuracy of CT for capsular invasion in 98 tumours, with a sensitivity, specificity, PPV, and NPV of 68%, 85%, 76%, and 81% respectively. When they looked specifically at renal vein involvement, the sensitivity, specificity, PPV, and NPV were higher at 84%, 98%, 89%, and 96%.25 Nodal and metastatic disease were not included, partly due to the low numbers of positive cases in the present series, and partly due to well-established size criteria for considering nodal disease to be significant on CT or MRI.26 In practice, many nodes seen at or exceeding the 9–11 mm limit in the upper and lower para-aortic and porto-caval regions will be reactive, but the presence of such nodes may preclude nephron-sparing approaches and in some instances warrant open radical nephrectomy with lymphadenectomy for staging purposes. A large EORTC study showed no benefit for routine staging lymphadenectomy in patients with renal carcinoma.27 A limitation of the present study is that the patient numbers in certain groups are relatively small. Histology and operative notes were available dating back to 1999, but radiological images were only available as hard copy, without MPR capability, thus cases were limited to soft copy, which could be viewed on a PACS workstation. Another limitation is that the cases were reviewed by three different uroradiologists (together with one senior radiology trainee), which may have introduced some bias. All three are, however, active members of the level 2 specialist MDT with extensive experience in the CT staging of RCC. As several imaging studies were performed at outside institutions and interpreted at the authors' tertiary centre, several different protocols were followed. As long as the minimum specified dataset (arterial-phase CT thorax, pre and parenchymal phase CT abdomen; Table 1) was available, the case was included in the study. It is acknowledged that the present CT technique is radiation intensive. Several studies have demonstrated that the optimum combination of phases for detecting and characterising small renal masses and complex renal cysts is the unenhanced and parenchymal phase.28 and 29 However, for the staging of renal cancer, the addition of an arterial phase is valuable in assessing the renal veins and inferior vena cava (IVC) for the presence of tumour thrombus, and demonstrates the renal arterial anatomy, which is particularly important for laparoscopic surgical approaches.30, 31, 32 and 33 The accuracy of the present study at predicting exact pathological stage using cross-sectional imaging was 71% (27 discrepancies out of 94). CT tends to oversize small tumours, which needs to be taken into account when making treatment decisions, as tumours measuring just above 4 cm are quite likely to be T1a pathologically, and should not be denied consideration of local therapy.11, 12 and 33 Microscopic extracapsular extension or renal sinus invasion is also impossible to detect, along with segmental renal vein invasion of small vessels. For larger tumours abutting adjacent organs, it is difficult to evaluate radiologically whether invasion is present or not. Discussion at the specialist MDT will recognise any limitation of CT staging and ensure the correct management choice. In conclusion, radiological staging of RCC cannot detect microscopic T3a disease, and can overestimate the size of the tumour, which is most important in T1a tumours. The presence of perinephric stranding and tumour necrosis were not reliable signs for pT stage >T3a. Thickening of Gerota's fascia and the presence of collateral vessels in the peri- or paranephric fat had 90% and 94% specificity, with 82% and 88% PPV, respectively, for the presence of tumour stage for pT stage >T3a. These are considered reliable signs of locally advanced renal cancer. CLINICAL RESEARCH STUDY A Combined Cardiorenal Assessment for the Prediction of Acute Kidney Injury in Lower Respiratory Tract Infections Tobias Breidthardt, MD,a,b,c Mirjam Christ-Crain, MD, Prof,d Daiana Stolz, MD, Prof,e Roland Bingisser, MD, Prof,f Beatrice Drexler, MD,a Theresia Klima, MD,a,b Catharina Balmelli, MD,a Philipp Schuetz, MD,d Philip Haaf, MD,a Michael Sch√§rer, MsC,a Michael Tamm, MD, Prof,e Beat M√ºller, MD, Prof,g Christian M√ºller, MD, Profa a Department of Internal Medicine and bDivision of Nephrology, University Hospital, Basel, Switzerland; cDepartment of Renal Medicine, Royal Derby Hospital, Derby, United Kingdom; dDivision of Endocrinology, eDivision of Pneumology, and fEmergency Department, University Hospital, Basel, Switzerland; gMedical University Clinic, Kantonsspital, Aarau, Switzerland. ABSTRACT BACKGROUND: The accurate prediction of acute kidney injury (AKI) is an unmet clinical need. A combined assessment of cardiac stress and renal tubular damage might improve early AKI detection. METHODS: A total of 372 consecutive patients presenting to the Emergency Department with lower respiratory tract infections were enrolled. Plasma B-type natriuretic peptide (BNP) and neutrophil gelatinase-associated lipocalin (NGAL) levels were measured in a blinded fashion at presentation. The potential of these biomarkers to predict AKI was assessed as the primary endpoint. AKI was defined according to the AKI Network classification. RESULTS: Overall, 16 patients (4%) experienced early AKI. These patients were more likely to suffer from preexisting chronic cardiac disease or diabetes mellitus. At presentation, BNP (334 pg/mL [130-1119] vs 113 pg/mL [52-328], P œΩ.01) and NGAL (269 ng/mL [119-398] vs 96 ng/mL [60-199], P œΩ.01) levels were significantly higher in AKI patients. The predictive accuracy of presentation BNP and NGAL levels was comparable (BNP 0.74; 95% confidence interval [CI], 0.64-0.84 vs NGAL 0.74; 95% CI, 0.61-0.87). In a combined logistic model, a joint BNP/NGAL approach improved the predictive accuracy for early AKI over either biomarker alone (area under the receiver operating characteristic curve: 0.82; 95% CI, 0.74-0.89). The combined categorical cut point defined by BNP œæ267 pg/mL or NGAL œæ231 ng/mL correctly identified 15 of 16 early AKI patients (sensitivity 94%, specificity 61%). During multivariable regression analysis, the combined BNP/NGAL cutoff remained the independent predictor of early AKI (hazard ratio 10.82; 95% CI, 1.22-96.23; P œ≠ .03). CONCLUSION: A model combining the markers BNP and NGAL is a powerful predictor of early AKI in patients with lower respiratory tract infection. ¬© 2012 Elsevier Inc. All rights reserved. ‚Ä¢ The American Journal of Medicine (2012) 125, 168-175 KEYWORDS: Acute kidney injury; B-type natriuretic peptide; Lower respiratory tract infection; Plasma neutrophil gelatinase-associated lipocalin; Prediction Acute kidney injury (AKI) is an increasingly common complication in hospitalized patients. Over the last decade the incidence of AKI has steadily increased from approximately 60 to 500 events per 100,000 population.1,2 Current estimates Funding: None. Conflict of Interest: None. Authorship: All authors had access to the data and played a role in writing this manuscript. Requests for reprints should be addressed to Tobias Breidthardt, MD, Department of Internal Medicine, University Hospital, Petersgraben 4, Basel, Switzerland CH - 4031. E-mail address: breidthardtt@uhbs.ch 0002-9343/$ -see front matter ¬© 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2011.07.010 assume that annually, about 17 million hospital admissions in the US are complicated by AKI, resulting in additional health care costs of 10 billion dollars.3 In high-risk patients, such as those hospitalized with acute myocardial infarction, acute heart failure, or sepsis, the incidence of AKI can reach up to 50%. Even in nonsevere community-acquired pneumonia, the incidence of AKI is over 30%.4 Importantly, AKI is strongly associated with increased morbidity, mortality, and a progression to end-stage renal disease.5 Unfortunately, the accurate prediction of AKI is an unmet clinical need. Recently, cardiac stress and the inflammatory cytokine response have been identified as important triggers for the Breidthardt et al Acute Kidney Injury in Lower Respiratory Tract Infections 169 occurrence of sepsis-induced AKI.6 B-type natriuretic pepClinical Evaluation tide (BNP), a 32-amino-acid polypeptide, has been shown Patients were evaluated in the ED by at least 2 physicians, consistently to mirror the extent of cardiac stress in various a resident in internal medicine and an internal medicine disease states.7 The main stimulus for the secretion of BNP specialist. All patients were subject to an initial clinical is cardiac stress reflected by myocardial stretch and pressure assessment including medical history, physical examinaor volume overload.7 Additiontion, electrocardiography, pulse ally, pro-inflammatory cytokines oximetry, blood tests including arand the activation of the sympaterial blood gas analysis, and chest CLINICAL SIGNIFICANCE thetic nervous system have reradiograph. Serum creatinine levcently been identified as triggers els were measured at presentation ‚óè In patients with lower respiratory tract of BNP secretion.8 to the ED and daily for the next 2 infections, plasma B-type natriuretic At the same time, tubularly sedays at the central laboratory of peptide (BNP) and neutrophil gelaticreted neutrophil gelatinase-associthe University Hospital Basel. nase-associated lipocalin (NGAL) levels ated lipocalin (NGAL; also known Following routine clinical and labat presentation are significantly eleas human neutrophil lipocalin, lioratory practice, serum creatinine vated in patients experiencing early pocalin-2, siderocalin, or 24p3) has levels were measured within 2 acute kidney injury (AKI). been identified as a novel and spehours after the respective blood cific biomarker for the early detecdraw. The study investigators ‚óè In this setting, BNP and NGAL levels tion of AKI. Its predictive potential were neither directly involved in powerfully predict early AKI. has been shown in several settings patient care in the ED nor did they ‚óè A combined cardiorenal assessment usincluding cardiac surgery,9,10 conhave any influence on the decision ing BNP and NGAL is a powerful predictrast agent administration,11,12 unseto discharge patients from the tor of early AKI, surpassing the prediclected patients in the emergency deward. partment (ED),13 and in critically ill tive potential of either marker alone. patients.14,15 Endpoints Hence, we hypothesized that We defined and graded AKI acthe combination of a marker of cording to the Acute Kidney Injury Network classification cardiac stress such as BNP with a pathophysiologically as an in-hospital increase of serum creatinine over 26.4 different biomarker reflecting acute renal tubular damage ‚êÆmol/L (0.3 mg/dL) occurring during the first 48 hours, such as NGAL might allow the accurate prediction of AKI which persisted after adequate volume resuscitation.18 in patients presenting to the ED with lower respiratory tract Acute increases in serum creatinine that responded to volinfection (LRTI). ume treatment were considered prerenal aggravations of kidney function. The estimated glomerular filtration rate was calculated using the abbreviated Modification of Diet in METHODS Renal Disease Study equation.19 The systemic inflammatory response syndrome (SIRS) and its criteria were defined Setting and Study Population according to current guidelines.20 The potential of presenThis study specifically investigated the potential of plasma tation BNP and NGAL levels to predict early AKI was BNP and NGAL levels to predict early AKI in hospitalized assessed as the primary endpoint. The study was carried out patients with LRTI. We screened 510 consecutive patients according to the principles of the Declaration of Helsinki presenting to the ED with signs of LRTI from November and approved by the local ethics committee. Written in2003 to March 2005 following previously described meth16,17 formed consent was obtained from all participating patients. To be eligible for the enrollment, patients had odology. Two-year follow-up was complete in all patients. to be over 18 years old and present with one or several of the following recently acquired respiratory signs or sympMeasurement of B-type Natriuretic Peptide toms: cough, sputum production, dyspnea, core body temperature œæ38.0¬∞C, auscultatory findings of abnormal breath and Neutrophil Gelatinase-associated sounds, or a white blood cell count œæ10 or œΩ4 œ´ 109 Lipocalin Levels cells/L. Patients with cystic fibrosis, active pulmonary tuAt presentation to the ED, blood samples were collected berculosis, or hospital-acquired pneumonia were excluded into bottles containing potassium ethylenediaminetetfrom the evaluation. Severely immunocompromised paraacetic acid (EDTA). After centrifugation, samples were tients or patients undergoing chronic dialysis also were immediately frozen at œ™80¬∞C until later batch analysis. excluded. BNP was detected in EDTA plasma from all patients with a The 372 patients hospitalized for over 48 hours, receivfluorescence immunoassay (Triage BNP; Alere San Diego ing serial serum creatinine measurements, and in whom Incorporated; San Diego, Calif). In brief, the coefficient of values for BNP and NGAL were available, constitute the variation within a given assay has been reported to be 9.5%, 12.0%, and 13.9% for levels of 28.8, 584.0, and 1180.0 study population for this analysis. 170 The American Journal of Medicine, Vol 125, No 2, February 2012 pg/mL, respectively, while the coefficient of variation among assays is 10.0%, 12.4%, and 14.8%, respectively. NGAL was detected in EDTA plasma in blinded fashion from all patients with a fluorescence-immunoassay at Alere‚Äôs central laboratory in San Diego (Triage NGAL; Alere San Diego Incorporated). According to Alere, the lower detection limit of the assay is 60 ng/mL; the upper detection limit is 1300 ng/mL. The product package insert describes the lot-to-lot coefficients of variance as 6.4%, 5.8%, and 3.9% for low, medium, and high NGAL values, respectively. The 95th percentile in 120 healthy individuals aged between 18 and 83 years was 159 ng/mL. All investigated cardiovascular and antibiotic drugs tested at concentrations representing at least 2 times the maximal therapeutic dose did not show cross-reactivity or interference with the assay. Statistical Analysis The statistical analyses were performed using the SPSS/PC (version 19.0, SPSS Inc., Chicago, Ill) software package. A statistical significance level of .05 was used. Discrete variables are expressed as counts (percentage) and continuous variables as means œÆ SD or median and interquartile range (IQR), unless stated otherwise. Frequency comparisons were made using t-test, Kruskal-Wallis test, Mann-Whitney U test, and chi-squared test as appropriate. The relationship between the biomarker levels and the degree of acute kidney dysfunction was assessed using the Jonckheere-Terpstra test. All hypothesis testing was 2-tailed. Areas under receiver operating curves were compared using MedCalc software (MedCalc Software, Mariakerke, Belgium). To estimate the potential clinical relevance of BNP measurements, we used likelihood-ratio tests to determine whether logistic regression models that included measurements of BNP and NGAL provided a significantly better fit than did logistic regression models limited to either biomarker alone. Logistic regression analysis was applied to identify predictors of early AKI in univariate and multivariable analysis. Multivariable analysis included all candidate variables (P œΩ.1) identified in univariate analysis. The Spearman rank correlation was used to perform correlation analyses. RESULTS Baseline Characteristics Detailed baseline characteristics of the study population are summarized in Table 1. The age of the 372 patients ranged from 20 to 95 years and the incidence of cardiovascular comorbidities was moderate. Overall, 98 (26%) patients fulfilled 0/1 SIRS criteria at presentation, while 274 (74%) patients fulfilled at least 2 SIRS criteria. The majority of patients (205 patients, 55%) fulfilled 2 or 3 criteria. The patient population was well matched for sex. Overall, 11 (3%) patients experienced a transient prerenal aggravation of kidney function, while 16 (4%) patients suffered early AKI. Of these, 12 cases were class 1, while only 4 reached class 2. AKI patients were more likely to suffer from a previously diagnosed chronic kidney disease or diabetes mellitus. Outpatient pretreatment with antibiotics was not associated with an increased risk of AKI. Interestingly, laboratory markers of inflammation (leukocyte count, C-reactive protein) and disease severity (number of SIRS criteria fulfilled) were similar between AKI and non-AKI patients. BNP and Plasma NGAL Levels in Patients Experiencing Acute Kidney Injury BNP values at presentation were significantly correlated to age (r œ≠ 0.50, P œΩ.001), admission creatinine values (r œ≠ 0.42, P œΩ.001), CRP values (r œ≠ 0.21, P œΩ.001), and NGAL values (r œ≠ 0.32, P œΩ.001). NGAL values showed similar correlations with age (r œ≠ 0.18, P œΩ.001), presentation creatinine values (r œ≠ 0.50, P œΩ.001), and CRP values (r œ≠ 0.30; P œΩ.001). Spot measurements of BNP and plasma NGAL levels at presentation showed a significant trend association with the severity of renal dysfunction observed during the first 48 hours (Figure 1A, B). Importantly, BNP levels were especially increased in patients with early stages of renal dysfunction, while plasma NGAL values were primarily elevated in the advanced stages of acute renal impairment. However, levels of both biomarkers at presentation were significantly higher in patients experiencing early AKI compared with patients without AKI (BNP: 334 pg/mL, IQR 130-1119 vs 113 pg/mL, IQR 52-328, P œΩ.01; NGAL: 269 ng/mL, IQR 119-398 vs 96 ng/mL, IQR 60-199, P œΩ.01). BNP and Plasma NGAL as Predictors of Acute Kidney Injury To evaluate the potential spot measurements of BNP and plasma NGAL at presentation to predict the occurrence of AKI, receiver operating characteristic (ROC) analyses were performed. Figure 2 shows the areas under the ROC curve for both markers to detect different stages of renal impairment. The area under the ROC curve (AUC) for the prediction of AKI was comparable for the 2 markers (BNP 0.74; 95% confidence interval [CI], 0.64-0.84 vs NGAL 0.74; 95% CI, 0.61-0.87). The calculated BNP cut point of 267 pg/mL, which maximized the sum of sensitivity and specificity, achieved a sensitivity of 0.69 and a specificity of 0.73. Similarly, the best calculated NGAL cut point of 231 ng/mL obtained a sensitivity of 0.63 and a specificity of 0.80. By applying these cutoff values, BNP correctly identified 11 of the 16 patients, while NGAL identified 10 of the 16 patients experiencing early AKI. Of note, categorical NGAL values identified all patients experiencing AKI stage 2. In patients presenting with a history of congestive heart failure, the predictive potential of both biomarkers remained comparable (BNP: 0.78, 95% CI, 0.67-0.89; NGAL 0.70, 95% CI, 0.53-0.86). In a combined logistic model, an integrated biomarker approach improved the predictive accuracy for the occurrence of early AKI over either biomarker alone (AUC: 0.82; 95% CI, 0.74-0.89) (Figure 2). This improvement did not Breidthardt et al Table 1 Acute Kidney Injury in Lower Respiratory Tract Infections 171 Demographic and Clinical Characteristics of 372 Lower Respiratory Tract Infection Total number of patients Age (years) Male Medical history Arterial hypertension Heart failure Diabetes mellitus Chronic kidney disease Neoplastic disease Antibiotic pretreatment Symptoms Cough Sputum Dyspnea Fever Signs and vital status Rales Temperature (¬∞C) Heart rate (per minute) Respiratory rate (per minute) Systolic blood pressure (mm Hg) Laboratory values Leukocytes (œ´109/L) C-reactive protein (mg/L) Creatinine (‚êÆmol/L) NGAL (ng/mL) BNP (pg/mL) SIRS criteria Outcome Intensive care unit admission Index-hospitalization (days) 2-Year mortality All Patients No AKI 372 73 [64-80] 215 (58) 356 (96) 74 [65-81] 206 (58) 16 (4) 75 [68-85] 9 (56) 103 78 69 79 61 79 (28) (21) (19) (21) (16) (21) 100 73 65 67 58 74 (28) (21) (18) (19) (16) (21) 3 5 4 12 7 5 (19) (31) (25) (75) (19) (32) .47 .34 .51 œΩ.01 .70 .24 322 265 305 253 (87) (71) (82) (68) 308 255 292 242 (87) (72) (82) (68) 14 10 13 11 (88) (62) (81) (69) 1.0 .41 .87 1.0 215 38.0 98 23 134 (58) [37.3-38.8] [84-111] [20-28] [118-148] 207 38.1 98 23 133 (58) [37.3-39.0] [84-112] [20-28] [119-148] 11.2 [8.6-15.3] 76 [28-183] 97 [77-133] 125 [60-273] 146 [56-428] 2.4 œÆ 1.2 11.2 [8.6-14.7] 75 [24-183] 96 [76-129] 96 [60-199] 113 [52-328] 2.4 œÆ 1.2 31 (8) 10 [5-16] 153 (41) 27 (8) 11 [6-16] 143 (40) AKI 8 37.9 90 21 110 P-Values .18 .95 (50) [37.3-38.7] [78-108] [20-25] [103-135] .31 .78 .29 .71 .01 12.8 [7.0-19.1] 79 [62-184] 182 [147-302] 269 [119-398] 334 [130-1119] 2.3 œÆ 1.3 .58 .36 œΩ.01 œΩ.01 œΩ.01 .69 2 (13) 18 [14-22] 10 (63) .35 œΩ.01 .07 Abbreviations: AKI œ≠ acute kidney injury; BNP œ≠ B-type natriuretic peptide; eGFR œ≠ estimated glomerular filtration rate; NGAL œ≠ neutrophil gelatinase-associated lipocalin; SIRS œ≠ systemic inflammatory response syndrome. Data are presented as median [interquartile range], number of patients (%) as appropriate. reach statistical significance (P vs BNP œ≠ 0.14; P vs NGAL œ≠ 0.15). However, a combined categorical cut point defined by BNP œæ267 pg/mL or NGAL œæ231 ng/mL correctly identified 15 of 16 patients experiencing early AKI (sensitivity 94%, specificity 61%) (Table 2). When entering these biomarkers into a univariate binary regression analysis, continuous and categorical BNP and NGAL levels significantly predicted the occurrence of early AKI (Table 3). The predictive potential of the combined BNP/NGAL cutoff remained even after adjustment for admission creatinine levels (hazard ratio 10.82; 95% CI, 1.2296.23; P œ≠ .03). DISCUSSION In this investigation we examined the potential of BNP and plasma NGAL levels at presentation to predict the occurrence of early AKI in 372 patients with LRTI. There are several key findings in this study. First, early AKI is associated with a trend towards heightened long-term mortality. Second, BNP and NGAL levels at presentation are significantly higher in patients experiencing early AKI compared with non-AKI patients. Third, BNP and NGAL levels at presentation are powerful predictors of early AKI. Fourth, the predictive potential of BNP and NGAL levels is independent of creatinine values at presentation and persists after the exclusion of patients with a history of congestive heart failure. Fifth, a combined cardiorenal assessment using BNP and NGAL is a powerful predictor of early AKI, surpassing the predictive potential of either marker alone. To our knowledge, our results represent the first study investigating the potential of a combined cardiorenal assessment in the prediction of early AKI. Infection-induced AKI has recently been proposed to be triggered by pathophysiological patterns distinctly different from noninfectious AKI. A recent study enrolling over 120,000 critically ill patients found patients suffering from sepsis-induced AKI to display more pronounced signs of cardiac stress, inflammation, renal and pulmonary impairment, and disease acu- 172 The American Journal of Medicine, Vol 125, No 2, February 2012 Figure 1 Box plots showing B-type natriuretic peptide (BNP) (A) and plasma neutrophil gelatinase-associated lipocalin (NGAL) (B) levels at presentation according to maximum renal impairment during the observational period. AKI œ≠ acute kidney injury. ity6 compared with patients suffering from AKI of other causes. BNP, a 32-amino-acid polypeptide, has been shown consistently to mirror the extent of cardiac stress; its levels integrate systolic and diastolic left ventricular function, heart rate, right ventricular, and valvular function.7 Consequently, the assessment of BNP has become a cornerstone in the management of heart failure patients7,21 and is endorsed by heart failure guidelines.22 While LRTI patients do not display the overt clinical picture of heart failure, LRTI is a significant stressor to the cardiovascular system through low peripheral vascular resistance, increased cardiac output, and the occurrence of arteriovenous shunts in inflamed areas.23 Additionally, cardiodepressing effects have been described for pro-inflammatory mediators such as lipopolysaccharides,24,25 tumor necrosis factor (TNF)-‚ê£, and interleukin (IL)-1‚ê§.26,27 Hence, elevated BNP levels in the setting of LRTI appear to identify a patient subgroup with limited cardiac reserves. Increased cardiac stress propels these patients over their threshold at which inadequate renal oxygen supply and AKI occur. By reflecting cardiac stress, BNP levels mirror derangements in an important pathophysiological pathway to inflammatory AKI. A further reason for elevated BNP levels in LRTI is their response to inflammatory stimuli. In agreement with previous observations, we found BNP levels to correlate positively with CRP values.28 Moreover, pro-inflammatory cytokines such as IL-1b, IL-6, and TNF-‚ê£ have been shown to induce BNP secretion from cultured myocytes in vitro.8,29 Concordantly, bacterial endotoxin was found to directly increase the expression of BNP mRNA in rat myocytes.30 Thus, BNP levels appear to reflect the extent of systemic inflammation in LRTI, an additional contributor to the pathogenesis of sepsis-induced AKI. NGAL is a small molecule that belongs to the welldefined superfamily of proteins called lipocalins.31 Human NGAL was originally identified as a 25-kDa protein associated with purified gelatinase obtained from the supernatant of activated neutrophils, which normally are the main cellular source of circulating NGAL.31 Subsequent genomic, transcriptomic, and proteomic studies have identified tubularly secreted NGAL as an early marker of AKI.32,33 Importantly, our results are in agreement with the findings of an earlier study enrolling 661 patients presenting to the ED with suspected sepsis.34 In that study, Shapiro et al34 also found spot measurements of serum NGAL to be significantly higher in patients experiencing early AKI. However, the predictive potential of admission NGAL to predict early AKI described in their study slightly surpassed our findings (AUC 0.82 vs 0.72, respectively). This difference is probably caused by the different enrollment criteria of both studies, with the Shapiro study34 restricting enrollment to patients fulfilling at least 2 SIRS criteria. Importantly, septic AKI has recently been described to be associated with significantly higher NGAL values compared with nonseptic AKI.35 Hence, the higher percentage of septic patients (24% of patients in this study fulfilled œΩ2 SIRS criteria) might explain the better predictive potential observed in their study. Additionally, the Shapiro study defined early AKI as an increase of serum creatinine over 0.5 mg/dL compared with the 0.3 mg/dL advised by the Acute Kidney Injury Network guidelines. This stricter AKI definition also might have contributed to the slightly better predictive potential observed by Shapiro,34 as we found plasma NGAL levels to be especially elevated in patients experiencing more advanced stages of early AKI. Despite these minor differences, both studies advocate plasma NGAL as a powerful predictor of early AKI in patients presenting to the ED with signs of infection. Breidthardt et al Acute Kidney Injury in Lower Respiratory Tract Infections 173 Admission NGAL AUC: 0.74 [0.61-0.87] Sensitivity Admission BNP AUC: 0.74 [0.64-0.84] Combination NGAL/BNP AUC: 0.82 [0.74-0.89] Admission Creatinine AUC: 0.89 [0.83-0.95] 1- Specificity Figure 2 Receiver operating characteristic curves displaying the ability of presentation B-type natriuretic peptide (BNP), neutrophil gelatinase-associated lipocalin (NGAL), and the combination of both to predict the occurrence of early acute kidney injury (AKI). Most notably, we found a combined cardiorenal assessment using BNP and NGAL levels to significantly improve the prediction of early AKI in LRTI patients over either marker alone. We believe that this additive effect reflects their separate pathophysiological pathways to overexpression in AKI: BNP primarily reflects cardiac stress, while NGAL reflects renal tubular damage. A combined assessment will therefore detect patients with limited cardiac reserves and a propensity to inadequate renal perfusion in the setting of increased cardiac demands, as well as patients experiencing tubular damage. As shown by the 6 patients exceeding both cut points, these pathways to AKI are not mutually exclusive. Indeed, ischemic tubular damage occurs in about 50% of critically ill patients experiencing in-hospital AKI.36 On the other hand, nephrotoxic tubular injury (eg, following aminoglycoside antibiotic therapy) or acutely increased cardiac demands appear to be able to induce AKI in patients exceeding only one cut point. It can be speculated that all patients exceeding only the BNP cut point will eventually also display increased NGAL levels as renal injury persists and progresses to tubular damage. Table 2 Combined BNP/NGAL Approach Class No AKI (356) AKI (16) Above no cut point Above either cut point Above both cut points 218 (61) 110 (31) 28 (8) 1 (1) 9 (56) 6 (37) -AKI œ≠ acute kidney injury; BNP œ≠ B-type natriuretic peptide; NGAL œ≠ neutrophil gelatinase-associated lipocalin. BNP cut point of 267 pg/mL, NGAL cut point of 231 ng/mL. Importantly, 2 recent studies comparing an isolated NGAL approach in critically ill patients to admission creatinine14 or clinical prediction models,15 found classical predictors of AKI to at least equal if not surpass the poten- Table 3 Prediction of the Occurrence of Acute Kidney Injury in Univariate and Multivariable Regression Analysis Predictor Hazard Ratio (95% CI) Univariate analysis Systolic blood pressure (mm 0.98 (0.96-1.00) Hg) Creatinine (‚êÆmol/L) 1.01 (1.00-1.01) NGAL (ng/mL) (for an increase 1.37 (1.14-1.63) of 100 ng/mL) NGAL above 231 ng/mL 6.80 (2.39-19.36) BNP (pg/mL) (for an increase 1.10 (1.04-1.17) of 100 pg/mL) BNP above 267 pg/mL 5.96 (2.02-17.59) Cut point combination BNP/ 23.70 (3.10-181.41) NGAL Multivariable regression analysis Creatinine (‚êÆmol/L) 1.00 (1.00-1.01) NGAL (ng/mL) (for an increase 0.96 (0.72-1.27) of 100 ng/mL) BNP (pg/mL) (for an increase 1.05 (0.97-1.13) of 100 pg/mL) Cut point combination BNP/ 10.82 (1.22-96.23) NGAL P-Value .03 œΩ.01 œΩ.01 œΩ.01 œΩ.01 œΩ.01 œΩ.01 .02 .14 .26 .03 BNP œ≠ B-type natriuretic peptide; NGAL œ≠ neutrophil gelatinaseassociated lipocalin. 174 The American Journal of Medicine, Vol 125, No 2, February 2012 tial of NGAL. Similarly, our group recently found isolated BNP levels to fail to predict in-hospital AKI in patients with acute heart failure.37 This suggests that, in settings marked by multifactorial AKI pathogeneses, a combined cardiorenal assessment might be able to improve early AKI prediction. Further studies will be needed to verify this hypothesis. Several limitations merit consideration. First, this was primarily a single-center study. However, as baseline characteristics, mortality rates, and the frequency of AKI were similar to those of previous studies,34 we consider our results representative. Second, despite enrolling patients of all age groups, most our patients were elderly and suffered from renal impairment at presentation. While this represents the majority of patients requiring hospitalization due to LRTI, we cannot selectively comment on the diagnostic potential of a combined cardiorenal assessment in a younger, less comorbid population. Third, treating physicians were blinded to the BNP and NGAL test results. We can therefore not assess the impact of a biomarker-guided treatment strategy on AKI rate and long-term outcome. To ascertain blinded measurements of BNP and NGAL levels, all samples had to be stored until after the conduction to the study. This might have affected the predictive potential of BNP. However, we feel that blinded biomarker measurements were necessary to rule out any treatment bias. Fourth, we cannot comment on the performance of BNP and NGAL among patients with terminal kidney failure requiring dialysis, because such patients were excluded from our study. In conclusion, BNP and NGAL levels at presentation are powerful predictors of early AKI in patients with LRTI. A combined BNP/NGAL approach surpasses the predictive potential of either marker alone. At-risk patients identified by this approach are at a 10-fold increased risk of early AKI. References 1. Hsu CY, McCulloch CE, Fan D, Ordonez JD, Chertow GM, Go AS. Community-based incidence of acute renal failure. Kidney Int. 2007; 72(2):208-212. 2. Waikar SS, Curhan GC, Wald R, McCarthy EP, Chertow GM. Declining mortality in patients with acute renal failure, 1988 to 2002. J Am Soc Nephrol. 2006;17(4):1143-1150. 3. Chertow GM, Burdick E, Honour M, Bonventre JV, Bates DW. Acute kidney injury, mortality, length of stay, and costs in hospitalized patients. J Am Soc Nephrol. 2005;16(11):3365-3370. 4. Murugan R, Karajala-Subramanyam V, Lee M, et al. Acute kidney injury in non-severe pneumonia is associated with an increased immune response and lower survival. Kidney Int. 2010;77(6):527-535. 5. Coca SG, Yusuf B, Shlipak MG, Garg AX, Parikh CR. Long-term risk of mortality and other adverse outcomes after acute kidney injury: a systematic review and meta-analysis. Am J Kidney Dis. 2009;53(6): 961-973. 6. Bagshaw SM, George C, Bellomo R. Early acute kidney injury and sepsis: a multicentre evaluation. Crit Care. 2008;12(2):R47. 7. Mueller C, Breidthardt T, Laule-Kilian K, Christ M, Perruchoud AP. The integration of BNP and NT-proBNP into clinical medicine. Swiss Med Wkly. 2007;137(1-2):4-12. 8. He Q, LaPointe MC. Interleukin-1beta regulation of the human brain natriuretic peptide promoter involves Ras-, Rac-, and p38 kinasedependent pathways in cardiac myocytes. Hypertension. 1999;33(1 Pt 2):283-289. 9. Wagener G, Jan M, Kim M, et al. Association between increases in urinary neutrophil gelatinase-associated lipocalin and acute renal dysfunction after adult cardiac surgery. Anesthesiology. 2006;105(3):485491. 10. Mishra J, Dent C, Tarabishi R, et al. Neutrophil gelatinase-associated lipocalin (NGAL) as a biomarker for acute renal injury after cardiac surgery. Lancet. 2005;365(9466):1231-1238. 11. Bachorzewska-Gajewska H, Malyszko J, Sitniewska E, Malyszko JS, Dobrzycki S. Neutrophil-gelatinase-associated lipocalin and renal function after percutaneous coronary interventions. Am J Nephrol. 2006;26(3):287-292. 12. Hirsch R, Dent C, Pfriem H, et al. NGAL is an early predictive biomarker of contrast-induced nephropathy in children. Pediatr Nephrol. 2007;22(12):2089-2095. 13. Nickolas TL, O‚ÄôRourke MJ, Yang J, et al. Sensitivity and specificity of a single emergency department measurement of urinary neutrophil gelatinase-associated lipocalin for diagnosing acute kidney injury. Ann Intern Med. 2008;148(11):810-819. 14. de Geus HR, Bakker J, Lesaffre EM, le Noble JL. Neutrophil gelatinase-associated lipocalin at ICU admission predicts for acute kidney injury in adult patients. Am J Respir Crit Care Med. 2011;183(7):907914. 15. Siew ED, Ware LB, Gebretsadik T, et al. Urine neutrophil gelatinaseassociated lipocalin moderately predicts acute kidney injury in critically ill adults. J Am Soc Nephrol. 2009;20(8):1823-1832. 16. Christ-Crain M, Stolz D, Bingisser R, et al. Procalcitonin guidance of antibiotic therapy in community-acquired pneumonia: a randomized trial. Am J Respir Crit Care Med. 2006;174(1):84-93. 17. Stolz D, Christ-Crain M, Bingisser R, et al. Antibiotic treatment of exacerbations of COPD: a randomized, controlled trial comparing procalcitonin-guidance with standard therapy. Chest. 2007;131(1): 9-19. 18. Mehta RL, Kellum JA, Shah SV, et al. Acute Kidney Injury Network: report of an initiative to improve outcomes in acute kidney injury. Crit Care. 2007;11(2):R31. 19. Levey AS, Bosch JP, Lewis JB, Greene T, Rogers N, Roth D. A more accurate method to estimate glomerular filtration rate from serum creatinine: a new prediction equation. Modification of Diet in Renal Disease Study Group. Ann Intern Med. 1999;130(6):461-470. 20. Bone RC, Balk RA, Cerra FB, et al. Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine. Chest. 1992; 101(6):1644-1655. 21. Porapakkham P, Zimmet H, Billah B, Krum H. B-type natriuretic peptide-guided heart failure therapy: a meta-analysis. Arch Intern Med. 2010;170(6):507-514. 22. Pfister R, Schneider CA. ESC guidelines for the diagnosis and treatment of acute and chronic heart failure 2008: application of natriuretic peptides. Eur Heart J. 2009;30(3):382-383; author reply 383. 23. Ince C. The microcirculation is the motor of sepsis. Crit Care. 2005; 9(Suppl 4):S13-S19. 24. Kumar A, Haery C, Parrillo JE. Myocardial dysfunction in septic shock. Crit Care Clin. 2000;16(2):251-287. 25. Jianhui L, Rosenblatt-Velin N, Loukili N, et al. Endotoxin impairs cardiac hemodynamics by affecting loading conditions but not by reducing cardiac inotropism. Am J Physiol Heart Circ Physiol. 2010; 299(2):H492-H501. 26. Finkel MS, Oddis CV, Jacob TD, Watkins SC, Hattler BG, Simmons RL. Negative inotropic effects of cytokines on the heart mediated by nitric oxide. Science. 1992;257(5068):387-389. 27. Walley KR, Hebert PC, Wakai Y, Wilcox PG, Road JD, Cooper DJ. Decrease in left ventricular contractility after tumor necrosis factoralpha infusion in dogs. J Appl Physiol. 1994;76(3):1060-1067. 28. Shor R, Rozenman Y, Bolshinsky A, et al. BNP in septic patients without systolic myocardial dysfunction. Eur J Intern Med. 2006;17(8):536-540. Breidthardt et al Acute Kidney Injury in Lower Respiratory Tract Infections 29. Ma KK, Ogawa T, de Bold AJ. Selective upregulation of cardiac brain natriuretic peptide at the transcriptional and translational levels by pro-inflammatory cytokines and by conditioned medium derived from mixed lymphocyte reactions via p38 MAP kinase. J Mol Cell Cardiol. 2004;36(4):505-513. 30. Tomaru Ki K, Arai M, Yokoyama T, et al. Transcriptional activation of the BNP gene by lipopolysaccharide is mediated through GATA elements in neonatal rat cardiac myocytes. J Mol Cell Cardiol. 2002;34(6):649-659. 31. Bolignano D, Donato V, Coppolino G, et al. Neutrophil gelatinaseassociated lipocalin (NGAL) as a marker of kidney damage. Am J Kidney Dis. 2008;52(3):595-605. 32. Mishra J, Ma Q, Prada A, et al. Identification of neutrophil gelatinaseassociated lipocalin as a novel early urinary biomarker for ischemic renal injury. J Am Soc Nephrol. 2003;14(10):2534-2543. 33. Supavekin S, Zhang W, Kucherlapati R, Kaskel FJ, Moore LC, Devarajan P. Differential gene expression following early renal ischemia/ reperfusion. Kidney Int. 2003;63(5):1714-1724. 175 34. Shapiro NI, Trzeciak S, Hollander JE, et al. The diagnostic accuracy of plasma neutrophil gelatinase-associated lipocalin in the prediction of acute kidney injury in emergency department patients with suspected sepsis. Ann Emerg Med. 2010;56(1):52.e159.e1. 35. Bagshaw SM, Bennett M, Haase M, et al. Plasma and urine neutrophil gelatinase-associated lipocalin in septic versus non-septic acute kidney injury in critical illness. Intensive Care Med. 2010; 36(3):452-461. 36. Santos WJ, Zanetta DM, Pires AC, Lobo SM, Lima EQ, Burdmann EA. Patients with ischaemic, mixed and nephrotoxic acute tubular necrosis in the intensive care unit‚Äîa homogeneous population? Crit Care. 2006;10(2):R68. 37. Breidthardt T, Socrates T, Noveanu M, et al. Effect and clinical prediction of worsening renal function in acute decompensated heart failure. Am J Cardiol. 2011;107(5):730-735. A Comparative Assessment of Survival Between Propensity Score-Matched Patients With Peritoneal Dialysis and Hemodialysis in Taiwan Yu-Kang Chang, MS, Chih-Cheng Hsu, MD, DrPH, Shang-Jyh Hwang, MD, Pei-Chun Chen, PhD, MSPH, Chiu-Chin Huang, MD, Tsai-Chung Li, PhD, MS, and Fung-Chang Sung, PhD, MPH Abstract: Studies comparing mortality for Asian populations with endstage renal disease (ESRD) on hemodialysis (HD) and peritoneal dialysis (PD) are limited. We compared mortality between patients treated with PD and HD in Taiwan, the population with the highest incidence of ESRD worldwide. Using the population-based insurance claims data of Taiwan from 1997 to 2006, we identified 4721 patients treated with PD and randomly selected 4721 patients treated with HD who were frequency-matched to the PD patients based on their propensity scores. In follow-up analyses we measured mortalities and hazard ratios associated with comorbidities in 2 different 5-year cohorts (1997Y2001 and 2002Y2006). In the 10-year period from 1997 to 2006, the overall mortality rates were similar in patients treated with PD and in patients treated with HD (12.0 vs. 11.7 per 100 person-years, respectively), with a PD-to-HD hazard ratio of 1.02 (95% confidence interval ECI^, 0.96Y1.08). In the first 5-year period (1997Y2001), the hazard ratio for mortality was higher for PD (1.33; 95% CI, 1.21Y1.46), but there was no difference between PD and HD in the 2002Y2006 cohort. Of note, younger patients who received PD had better survival than younger patients who received HD; this was especially true for patients aged younger than 40 years. In summary, in this Asian population, no significant survival differences were noted between propensity score-matched PD and HD patients. The selection of a dialysis modality must be tailored to the individual patient. Studies in which patients who are appropriate for either modality are randomly assigned to HD or PD may provide helpful information to clinicians and patients. (Medicine 2012;91: 144Y151) Abbreviations: CI = confidence interval, CCI = Charlson Comorbidity Index score, DM = diabetes mellitus, ESRD = endstage renal disease, HD = hemodialysis, NHI = National Health Insurance, NHRI = National Health Research Institute, PD = peritoneal dialysis. From the Institute of Population Health Sciences (YKC, CC Hsu), National Health Research Institutes, Zhunan, Taiwan; Department of Public Health (YKC, PCC, FCS), Division of Nephrology (CC Huang), and Institute of Biostatistics (TCL), China Medical University and Hospital, Taichung, Taiwan; and Division of Nephrology (SJH), Kaohsiung Medical University Hospital, Kaohsiung, Taiwan. Presented in part in abstract form at the Annual Meeting of the American Society of Nephrology, San Diego, CA, October 27-November 1, 2009. Funding and conflicts of interest: This study was supported by grants NSC 98-2621-M-039-001, CMUH 1MS1, DOH100-TD-B-111-004, and DOH100-TD-C-111-005 of Taiwan. The authors have no other funding or conflicts of interest to disclose. Reprints: Professor Fung-Chang Sung, PhD, MPH, China Medical University College of Public Health, 91 Hsueh-Shih Road, Taichung 404, Taiwan (e-mail: fcsung@mail.cmu.edu.tw). Copyright * 2012 by Lippincott Williams & Wilkins ISSN: 0025-7974 DOI: 10.1097/MD.0b013e318256538e 144 www.md-journal.com INTRODUCTION E nd-stage renal disease (ESRD) has emerged as a disease with a high medical burden worldwide. The disease is most prevalent in Taiwan, the United States, Germany, and Japan.6,13,26,31 Since 2002, Taiwan has had the highest incidence and prevalence of the disease in the world.6 In the United States, approximately 112,000 people are diagnosed with ESRD annually. Most patients with ESRD undergo chronic dialysis, either hemodialysis (HD) or peritoneal dialysis (PD).4 The preference for using HD or PD varies considerably among patients.5Y7,12,21 In 2006, HD was used for 92.4% of dialysis patients in Taiwan,7 91.7% in the United States,5 and 18.7% in Hong Kong. There has been considerable interest in evaluating the treatment effectiveness of PD and HD in the past 2 decades, and many observational studies have been conducted.4,14,17,18,21,23,25,28,35 Several studies have reported that PD appears to be associated with equal or better survival during the first year17,19,22Y25,31,34 or the first 2 years of dialysis compared with HD.4,14,18,28 Studies comparing the length of survival of patients on HD and PD have yielded conflicting results,1,4,11,14,17,18,21,25,27,28,32,33 particularly because of the absence of similar comorbidity burdens in the 2 groups. McDonald et al22 analyzed the population-based registries of Australia and New Zealand and found that treatment with PD is more effective than HD in year 1 during a 5-year follow-up period. Several studies have compared PD patients with propensity score-matched HD patients for survival probability.8,27,35 Weinhandl et al35 used the propensity score-matching technique to establish retrospective study cohorts and performed PD-to-HD survival comparison. Their population-based study suggests better survival for patients treated with PD than those treated with HD. Comparative studies on HD and PD survival for Asian populations using population data for a longer period are limited. To the best of our knowledge, comparisons of secular trends in mortality between PD and HD using a propensity scorematched design have not been reported for Asian patients. Therefore, we used a propensity score-matched design to identify PD and HD patients from the universal insurance claims data of Taiwan to investigate the temporal change in mortality in patients in 2 sets of 5-year cohorts (1997Y2001 and 2002Y2006). We measured the PD-to-HD hazard ratios for mortality in the follow-up years to compare the trend between these 2 cohorts. We also investigated the role of comorbidities, particularly diabetes mellitus (DM), in association with the survival differences between PD and HD treatments over time.23 METHODS Study Design and Patient Selection We conducted a historical retrospective cohort study using claims data from 1997 to 2006 obtained from the Taiwan National Medicine & Volume 91, Number 3, May 2012 Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine & Volume 91, Number 3, May 2012 Health Insurance. This is a reformed universal insurance program that unified 13 insurance systems in March 1995 and covered more than 96% of all Taiwan residents in 1996. The study population comprised patients with newly diagnosed ESRD who received dialysis treatment between January 1, 1997, and December 31, 2006. Mortality was through December 31, 2006. PD and HD patients were identified by special treatment codes registered with and defined by the Taiwan National Health Insurance: treatment codes for HD were 58001C, 58014C, 58019C-58025C, and 58027C; and treatment codes for PD were 58002C, 58009A, 58009B, 58010A, 58010B, 58011A, 58011B, 58011AB, 58011C, 58012A, 58012B, 58017B, 58017C, and 58026C. We identified 60,369 dialysis patients, and excluded patients who had received kidney transplantation (n = 917). Of the remaining 59,452 patients, 3443 (5.8%) were on PD, 54,370 (91.5%) were on HD, 1278 (2.1%) switched from PD to HD, and 361 (0.6%) switched from HD to PD. All HD patients were incenter patients who received the dialysis treatment in hospitals or clinics. Propensity Score Matching Only 7.9% (4721/59,452) of the patients with ESRD had been treated with PD. Patients treated with PD were different Survival by Dialysis Type from patients treated with HD in terms of health status, particularly comorbidities, which are important covariates associated with survival. Thus, we established study cohorts that were frequency-matched with propensity scores. Based on patients treated only or initially with PD (n = 4721), we randomly selected 4721 patients treated with HD who were frequencymatched to PD patients by propensity score. The propensity score was estimated using a logistic regression model including covariates generally considered to be important factors associated with survival.8,14,22,27,35 These covariates included the baseline information on age, sex, DM, hypertension, coronary artery disease, congestive heart failure, ischemic heart disease, cerebrovascular disease, malignancy, and anemia. We identified all illnesses using the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM). All baseline covariates were extracted from the same year the subjects were identified. Predictive performance of the model was assessed using the c-statistic.3 Nearest-neighbor algorithm was applied to construct matched pairs, assuming that the proportion of 0.95 to 1.0 is perfect.2 After the frequency-matching, an HD cohort was formed. The 1639 patients switching between PD and HD were combined with the initial intent-to-treat group for data analysis before the HD cohort was selected. FIGURE 1. Peritoneal dialysis-to-hemodialysis hazard ratio (HR) for mortality by diabetes mellitus (DM) status in the periods 1997Y2001 and 2002Y2006. * 2012 Lippincott Williams & Wilkins www.md-journal.com Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 145 Medicine Chang et al Charlson Comorbidity Index Score We estimated the Charlson Comorbidity Index score10 (CCI) for each patient using 4 weighted scores. Diseases with 1 point each were myocardial infarction, congestive heart failure, peripheral vascular disease, dementia, cerebrovascular disease, chronic pulmonary disease, connective tissue disease, peptic ulcer disease, and mild liver disease. Diseases with 2 points were hemiplegia, moderate or severe renal disease (which all patients had by definition), DM, any kind of tumor, leukemia, and lymphoma. Moderate or severe liver disease scored 3 points, and metastatic solid tumor scored 4 points. Statistical Analysis Recent studies have suggested that age, time (duration of dialysis), and presence of comorbidities, particularly DM, are the factors affecting dialysis modality and mortality.22,23,35 The difference between the effectiveness of PD and HD treatments may decrease over time.23 Based on the dates the patients were identified, we stratified the study subjects into 2 cohorts for the 5-year periods of 1997Y2001 and 2002Y2006 to observe the changes over time.23 We also compared the mortality differences associated with demographic characteristics, CCI, and comorbidities between PD and HD patients based on intent-to-treat analysis. The sample distributions used the chi-square test to verify the matching efficiency. The group of PD patients combined & Volume 91, Number 3, May 2012 with those who switched from PD to HD was compared with the group of HD patients combined with those who switched from HD to PD. Entry age was defined as the age at the beginning of dialysis. Exit age was defined as the date of the last claim, death, loss to follow-up, or end of follow-up. Person-years of follow-up were determined by calculating the time between the date of the start of dialysis and the date of death, censor, or end of follow-up. Mortality rates were calculated and compared between PD and HD patients by demographic status, CCI, and baseline comorbidity for the 1997Y2001 cohort and the 2002Y2006 cohort. We used the Cox proportional hazard method to estimate the PD-to-HD hazard ratio for mortality and the 95% confidence interval (CI), and examined the assumption using the Schoenfeld residuals test.15 Further data analysis was performed to examine the change in hazard ratios over time. The Cox proportional hazard method measured the age-specific (aged G40, 40Y65, and Q65 yr) and CCI-specific PD-to-HD hazard ratios for mortality. The analyses accounted for time dependency, depicting the annual PD-to-HD mortality hazard ratios, from the first year to the fifth year of follow-up, by DM status controlling for sex and age (Figure 1). The Kaplan-Meier model was used to compare survival between patients on PD and those on HD by DM status (Figure 2). Additional effects of hypertension and cardiovascular disease, controlling for sex and age, were also measured FIGURE 2. Kaplan-Meier survival curves for peritoneal dialysis (PD) and hemodialysis (HD) patients by diabetes mellitus (DM) status in the periods 1997Y2001 and 2002Y2006. 146 www.md-journal.com * 2012 Lippincott Williams & Wilkins Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine & Volume 91, Number 3, May 2012 (Figure 3). The nearest-neighbor algorithm procedure was conducted using the STATA v. 8.1 software package (StataCorp, College Station, TX). Other analyses were conducted using the SAS v. 9.13 software package (SAS Institute Inc., Cary, NC). Ethical Considerations The insurance claims data were linked using scrambled patient identification to ensure confidentiality and to prevent ethical violations. The current study was exempted from full review by the institutional review board of the National Health Research Institute, Taiwan. RESULTS BASELINE CHARACTERISTICS OF THE STUDY POPULATION There were no significant differences between PD and HD patients in both the 1997Y2001 and the 2002Y2006 cohorts for Survival by Dialysis Type distributions of sex, follow-up years, and income (Table 1). The mean age of HD patients was slightly, but significantly, higher than that of PD patients in the 2002Y2006 cohort. PD and HD patients were also similar in distributions of CCIs, and in most comorbidities, such as DM, hypertension, ischemic heart disease, congestive heart disease, and cancer (Table 2). However, cerebrovascular disease was more prevalent in HD patients than in PD patients in the 2002Y2006 cohort, whereas congestive heart failure was more prevalent in PD patients than in HD patients in the 1997Y2001 cohort. Hazard Ratios of PD Cohort Compared to HD Cohort The overall mortality rate was slightly, but not significantly, lower in the HD cohort than in the PD cohort (11.7 vs. 12.0 per 100 person-years, respectively). The hazard ratio of PD vs. HD was 1.02; 95% CI, 0.96Y1.08 (Table 3). The multivariable Cox model-measured PD-to-HD hazard ratio declined from 1.33 (95% CI, 1.21Y1.46) in the 1997Y2001 cohort to 0.99 (95% CI, 0.87Y1.14) in the 2002Y2006 cohort. The stratified age-specific analysis shows that the PD-to-HD hazard ratio decreased with age in the 1997Y2001 cohort, whereas it increased with age in the 2002Y2006 cohort. The hazard ratio of PD patients in both cohorts vs. the HD patients in both cohorts increased with age from 0.85 (95% CI, 0.70Y0.97) among those aged less than 40 years to 1.20 (95% CI, 1.05Y1.31) among those aged 65 years and older. There was little change in the death rate in PD patients aged less than 40 years in the different 5-year cohorts (3.7 per 100 person-years in 1997Y2001 vs. 4.0 per 100 person-years in 2002Y2006), whereas there was a large increase in the death rate of younger HD patients over time (3.2 per 100 person-years in 1997Y2001 vs. 10.2 per 100 person-years in 2002Y2006). By contrast, there was little change in the death rate in older HD patients (aged Q65 yr) in the different 5-year cohorts (26.7 per 100 person-years in 1997Y2001 vs. 27.7 per 100 personyears in 2002Y2006), whereas there was a large increase in the death rate of older PD patients over time (21.8 per 100 person-years in 1997Y2001 vs. 32.1 per 100 person-years in 2002Y2006). The PD-to-HD hazard ratios by CCI in the 2002Y2006 cohort were generally lower than those in the 1997Y2001 cohort (Table 4). For elderly patients, PD patients had a higher risk of death, except those with the CCI of 2 or less in the 2002Y2006 cohort. PD-to-HD Hazard Ratios by Cohort Period and Comorbidity FIGURE 3. Peritoneal dialysis-to-hemodialysis (PD-to-HD) hazard ratio for mortality by comorbidity of diabetes mellitus (DM), hypertension (HP), and cardiovascular disease (CVD) for patients in the 1997Y2001 cohort and 2002Y2006 cohort (*p for trend G0.05). * 2012 Lippincott Williams & Wilkins In Figure 1 we compare the PD-to-HD hazard ratios for mortality of patients whose disease progressed from the first to fifth year based on the cohort period and DM status. The hazard ratios in the 1997Y2001 cohort were significantly higher for patients without DM (‚Äò‚Äònon-DM‚Äô‚Äô) in all 5 years or for DM patients in the first 3 years. The PD patients were generally no longer at higher risk of mortality compared with the HD patients in the 2002Y2006 cohort. The Kaplan-Meier analysis revealed that HD patients had better survival than PD patients in the 1997Y2001 period (Figure 2). The survival probabilities of PD and HD patients were similar in the 2002Y2006 period. The hazard ratio for PD versus HD patients increased annually with comorbidity (Figure 3). The effect was stronger for patients simultaneously affected with DM, hypertension, and cardiovascular disease. For patients without these comorbidities, those on PD had better survival than those on HD in the 2002Y2006 www.md-journal.com Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 147 Medicine Chang et al & Volume 91, Number 3, May 2012 TABLE 1. Sociodemographic Characteristics of PD and HD Patients in 2 Different 5-Year Cohorts 1997Y2001 (%) Variable Sex, male Age, yr G40 40Y64 65+ Mean T SD, yr Follow-up, yr (mean T SD) Income, NTD‚Ä† 928,800 19,200Y28,800 G19,200 2002Y2006 (%) Combined (%) PD HD PD HD PD HD (n = 2134) (n = 2107) P* (n = 2587) (n = 2614) P* (n = 4721) (n = 4721) P* 39.7 38.9 0.46 41.2 39.8 0.10 40.5 39.4 0.48 25.8 52.8 21.4 50.9 T 16.4 4.2 T 2.3 23.5 54.5 22.0 51.6 T 15.6 4.2 T 2.3 0.23 20.1 55.6 24.3 53.0 T 16.0 3.0 T 1.4 17.6 56.4 26.0 54.0 T 15.2 3.0 T 1.4 0.08 22.5 54.3 22.9 52.3 T 16.1 3.5 T 2.3 21.8 54.7 23.5 52.7 T 15.6 3.5 T 2.4 0.21 33.9 40.5 25.6 32.9 41.2 25.9 0.11 32.9 42.1 25.0 31.7 43.5 24.8 33.5 41.2 25.3 32.1 42.3 25.6 0.43 0.18 0.14 0.03 0.20 0.23 0.05 0.31 *Chi-square test for PD and HD patients. ‚Ä†NTD = New Taiwan Dollars; 1 US dollar is about 30Y34 NTD. cohort; the beneficial effect was significant in the first year (hazard ratio, 0.62; 95% CI, 0.44Y0.80). DISCUSSION In the current study, fewer than 10% (4721/59,452; 7.9%) of ESRD patients had been treated with PD, and approximately one-third of these patients switched to HD treatment. This study was conducted based on the number of available patients treated with PD. Thus, our study cohorts involved propensity score-matched patients treated with PD and HD and all available patients who switched treatment, with 78.0% (1278/1639) switching from PD to HD. Although there was a higher hazard ratio for mortality for patients treated with PD compared to those treated with HD in the 1997Y2001 cohort, this was not observed in the 2002Y2006 cohort. In the combined group for the 10 years between 1997 and 2006, there was no difference in survival between patients treated with PD or HD. However, younger patients who received PD experienced better survival than younger patients who received HD; this was especially true among those aged younger than 40 years. These age-specific results are consistent with results of a Dutch study20 that found that PD is associated with greater survival of patients aged in their 40s and 50s, particularly those TABLE 2. Comorbidity of PD and HD Patients in 2 Different Cohorts 1997Y2001 (%) Variable CCI‚Ä† e2 3Y5 6Y9 10+ Diabetes Hypertension Cerebrovascular disease Ischemic heart disease Congestive heart failure Anemia Cancer Stomach malignancy Kidney malignancy Bladder malignancy 2002Y2006 (%) Combined (%) PD HD PD HD PD HD (n = 2134) (n = 2107) P* (n = 2587) (n = 2614) P* (n = 4721) (n = 4721) P* 26.2 32.4 27.1 14.3 31.7 72.8 7.2 13.8 11.4 36.6 4.6 0.1 1.1 0.5 26.3 32.4 26.4 14.9 33.4 72.9 7.5 12.7 9.2 35.0 4.5 0.1 1.1 0.7 0.66 15.8 32.9 31.7 19.6 35.0 74.4 14.7 23.8 18.9 38.1 7.3 0.2 1.2 0.7 15.5 29.9 34.9 19.8 37.1 72.2 17.2 24.7 19.5 37.3 8.1 0.2 1.5 1.0 0.07 20.5 32.7 29.6 17.2 33.5 74.1 11.3 19.3 15.5 37.5 6.1 0.1 1.1 0.6 20.3 31.0 31.1 17.6 34.7 72.5 12.9 19.2 14.9 36.3 6.5 0.2 61.3 30.8 0.11 0.08 0.44 0.50 0.14 0.03 0.14 0.69 0.97 0.86 0.54 0.05 0.08 0.01 0.22 0.26 0.23 0.13 0.83 0.26 0.13 0.10 0.14 0.14 0.86 0.34 0.45 0.53 0.83 0.37 0.25 Abbreviations: CCI = Charlson Comorbidity Index score. *Chi-square test for PD and HD patients. ‚Ä†Score for dialysis excluded. 148 www.md-journal.com * 2012 Lippincott Williams & Wilkins Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine & Volume 91, Number 3, May 2012 Survival by Dialysis Type TABLE 3. Mortality Rates of PD and HD Patients and Multivariable Cox Model-Measured PD-to-HD Hazard Ratios in 2 Different Cohorts 1997Y2001 PD Sex Female Male All Age, yr G40 40Y64 65+ Income, NTD G19,200 19,200Y28,800 928,800 CCI e2 3Y5 6Y9 Q10 2002Y2006 HD PD vs. HD Death Death Rate Rate* Rate* Ratio‚Ä† HR (95% CI)‚Ä° PD P¬ß Combined HD PD vs. HD Death Death Rate Rate* Rate* Ratio‚Ä† HR (95% CI)‚Ä° PD vs. HD HR (95% CI)‚Ä° P¬ß 11.0 12.9 11.9 12.1 10.9 10.3 0.91 1.18 1.16 1.32 (1.17Y1.49) 0.67 1.35 (1.17Y1.55) 1.33 (1.21Y1.46) 11.4 12.9 12.5 14.4 14.6 14.5 0.94 0.88 0.86 1.03 (0.88Y1.21) 0.68 0.96 (0.90 to 1.03) 0.96 (0.90Y1.03) 1.06 (0.95 to 1.16) 0.99 (0.87Y1.14) 1.02 (0.96 to 1.08) 3.7 10.6 21.8 3.2 10.1 26.7 1.16 1.05 0.82 1.21 (1.01Y1.43) 0.01 1.18 (1.02Y1.37) 0.89 (0.99Y1.20) 4.0 9.6 32.1 10.2 11.2 27.7 0.39 0.86 1.16 0.89 (0.54Y1.09) 0.01 0.85 (0.70 to 0.97) 0.85 (0.72Y0.99) 0.94 (0.88 to 0.99) 1.22 (1.08Y1.38) 1.20 (1.05 to 1.41) 12.9 10.8 12.5 12.2 10.0 9.5 1.06 1.08 1.32 1.08 (0.82Y1.38) 0.57 1.10 (0.84Y1.31) 1.15 (0.77Y1.53) 13.2 11.0 12.6 12.5 13.3 24.0 1.06 0.83 0.53 1.02 (0.88Y1.19) 0.21 1.09 (0.82 to 1.25) 1.02 (0.77Y1.35) 0.98 (0.80 to 1.17) 0.88 (0.53Y1.22) 0.97 (0.79 to 1.10) 6.3 9.2 16.4 27.2 4.7 7.3 15.0 23.0 1.34 1.26 1.09 1.18 1.20 (1.02Y1.40) 0.36 1.17 (1.01Y1.35) 1.22 (1.03Y1.44) 1.25 (1.00Y1.61) 3.2 8.1 14.5 26.9 7.8 10.9 16.7 34.1 0.41 0.74 0.87 0.79 0.57 (0.35Y0.94) 0.09 1.02 (0.79 to 1.29) 0.96 (0.73Y1.26) 1.04 (0.88 to 1.21) 1.00 (0.83Y1.20) 1.05 (0.91 to 1.22) 1.19 (0.97Y1.47) 1.21 (1.05 to 1.39) Abbreviations: See previous tables. HR = hazard ratio. *Per 100 person-years. ‚Ä†PD vs. HD. ‚Ä°Level-specific hazard ratio. ¬ßP for interaction. in their 40s in the first 15 months of treatment. This beneficial effect begins to disappear for diabetic patients in their 60s. The results of the beneficial effect of PD in relation to the length of treatment are also consistent with the Dutch study,19 but only for nondiabetic young patients. The mortality advantage of PD treatment also appears in young patients during the 90Y365 day treatment period in the Australia and New Zealand Dialysis and Transplant Registry.22 In a United States propensity-matched comparison, Weinhandl et al35 found that PD improved the survival rate of patients aged G65 years, particularly those without comorbidities. The Dutch study did not use a propensity-matched design,19 whereas the studies from the United States and from Australia and New Zealand did.22,35 Consistent with studies in Western populations, we found that PD is associated with lower mortality for younger patients and those with a lower number of comorbid conditions16,34 in the first year17,19,22,25,31,34 or the first 2 years4,14,18,28 after the start of therapy. Diabetes is a well-known comorbid condition in patients with ESRD.1,4,16,19,21,25,29,32,34 Cross-classifying DM and age or sex reveals that mortality risk was significantly lower in PD patients than in HD patients in younger groups without DM. The opposite effect was observed in elderly DM TABLE 4. Multivariable Cox Model-Measured PD-to-HD Hazard Ratios by CCI and Age in 2 Different Cohorts 1997Y2001 G40 yr 40Y65 yr 2002Y2006 Q65 yr G40 yr Q65 yr 40Y65 yr CCI HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) HR (95% CI) 2 3Y5 6Y9 10+ 0.93 (0.57Y1.51) 0.92 (0.56Y1.51) 0.79 (0.39Y1.60) 0.89 (0.31Y2.56) 1.95 (1.41Y2.71)* 1.45 (1.13Y1.85)* 1.18 (0.95Y1.48) 1.22 (0.92Y1.62) 1.21 (0.76Y1.91) 1.36 (1.01Y1.83)* 1.45 (1.10Y1.90)* 1.47 (1.15Y1.88)* 0.96 (0.41Y2.23) 0.81 (0.57Y1.85) 0.71 (0.49Y1.74) 1.30 (0.37Y2.51) 0.50 (0.24Y1.03) 0.82 (0.56Y1.20) 0.74 (0.55Y0.99)* 0.94 (0.65Y1.35) 0.48 (0.12Y1.33) 1.56 (0.98Y2.49) 1.34 (1.03Y1.74)* 1.36 (1.05Y1.75)* Abbreviations: See previous tables. Note: All groups are adjusted for sex, age, year, and income. *P G 0.05. * 2012 Lippincott Williams & Wilkins www.md-journal.com Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 149 Medicine Chang et al patients. However, previous registry studies in the Netherlands and Canada reported no significant difference in survival rate between patients undergoing PD and HD among older diabetic patients.14,28 The Canadian study comprised 25% DM patients with 10 years of follow-up, whereas the Netherlands study had 19% DM patients also with 10 years of follow-up. In the current study we followed patients for 1Y10 years, and nearly 35% of dialysis patients were diabetic at baseline. This step allowed for more reliable statistical power to compare the mortality risk between PD and HD treatment in elderly patients with DM. Diabetes was also prevalent in dialysis patients in the United States registry studies (36%Y39%).32,34 The PD patients in the United States studies had higher mortality than HD patients in the older group, similar to our observations. We further compared the results of the Cox model propensity score-matched and -unmatched analyses and found no significant difference in the overall PD-to-HD mortality (hazard ratio, 1.02 and 0.98, respectively) (data not shown for unmatched design). In the matched design, the treatment effectiveness differed among patient subgroups stratified by age, CCI, and DM status. However, the unmatched model yielded a higher PD-to-HD mortality risk than the matched model among older patient subgroups, and among those with DM and other comorbidities. The possible reason for this finding is the disparity in inclination toward treatment modality among patient subgroups. Patients with older age or more comorbidities tend to choose HD as their initial dialysis modality, resulting in an associated greater mortality risk. This finding is consistent with that of other studies.28 In the current study, the discrepancy between PD and HD treatments progressively decreased over time in the second 5 ˇ˛Abstract European Journal of Internal Medicine 19 (2008) 15 21 Original article www.elsevier.com/locate/ejim A comparative study of bacteremic and non-bacteremic pneumococcal pneumonia Francisco Jover a,N , Jose-Maria Cuadrado a, Lucio Andreu a, Silvia Martinez a, Ruth Canizares a, Victoria Ortiz de la Tabla b , Coral Martin b , Pablo Roig a , Jaime Merino a a Infectious Diseases Division, Internal Medicine Department, Hospital of San Juan, Alicante, Spain b Microbiology Division, Internal Medicine Department, Hospital of San Juan, Alicante, Spain Received 25 October 2006; received in revised form 7 March 2007; accepted 15 March 2007 Available online 19 September 2007 Background: Few attempts have been made to compare bacteremic and non-bacteremic pneumococcal pneumonia, mainly because it is difficult to gain agreement on which cases represent non-bacteremic pneumococcal pneumonia. Recently, an immunochromatographic assay for the detection of Streptococcus pneumoniae urinary antigen has been successfully evaluated for the diagnosis of pneumococcal pneumonia. The aim of our study was to examine and compare clinical and radiological features, risk factors, and outcome associated with bacteremic and non-bacteremic groups. Methods: A retrospective study (1995 2003) analyzing the clinical records of patients diagnosed with pneumococcal pneumonia in our institution was performed. S. pneumoniae were identified by blood cultures (bacteremic group) and detection of urinary antigen (non- bacteremic group). Results: There were 82 patients (57 bacteremic and 25 non-bacteremic). In seven non-bacteremic cases, another etiology was detected, i.e., Legionella (n=1) and Chlamydia pneumoniae (n=6). Bacteremic patients were significantly younger (p=b0.001), more likely to have liver disease (p=0.028), current smokers (p=0.024), alcohol and intravenous drug abusers (p=0.014 and pb0.001, respectively), and infected with HIV (pb0.001). Non-bacteremic patients were more likely to have congestive heart failure (p=0.004), chronic obstructive pulmonary disease (p=0.033) and to be former smokers (p=0.004). Bacteremic cases needed more prolonged intravenous antibiotic treatment (6 days vs. 4.5 days; p=0.006) than non-bacteremic cases and their length of stay was also longer. Conclusion: In our study, smoking was the leading risk factor for pneumococcal pneumonia. However, current smokers have an increased risk of bacteremic forms and former smokers and patients with COPD developed non-bacteremic forms more frequently. Bacteremic patients need more prolonged intravenous antibiotic treatment than non-bacteremic patients. © 2007 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. Keywords: Community-acquired pneumonia; Streptococcus pneumoniae; Bacteremia; Antibiotic treatment 1. Introduction Community-acquired pneumonia (CAP) is a common disease, representing the most frequent cause of hospital N Corresponding author. C/ Madre Teresa de Calcuta N∞ 4, Bloque 4, Esc 1, 2∞ H, 03016. Alicante, Spain. Tel.: +34 965250654/965656843; fax: +34 956938652. E-mail address: fjoverdiaz@coma.es (F. Jover). admission and mortality of infectious origin. Streptococcus pneumoniae is the leading cause of CAP and is responsible for 30 40% of CAP where the etiology can be established following a routine diagnostic work-up. However, S. pneumoniae may be underdiagnosed, and it is thought that it could be responsible for at least one of three episodes of CAP without an etiological diagnosis [1]. The reason for this underdiagnosis could be the limitations of conventional diagnostic tests. The diagnosis of pneumococcal pneumonia 0953-6205/$ - see front matter © 2007 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.ejim.2007.03.015 16 F. Jover et al. / European Journal of Internal Medicine 19 (2008) 15 21 is challenging and complicated by the lack of a highly sensitive and specific diagnostic gold standard method. The isolation of S. pneumoniae from blood or pleural fluid is highly specific, but lacks sensitivity, being positive in only about one-fourth of cases. Prior antibiotic therapy signifi- cantly reduces its sensitivity. Diagnosis based only on sputum culture is controversial due to both nasopharyngeal carriage of pneumococci in healthy individuals and inadequate sputum specimen collection. Inva- sive methods, such as bronchoalveolar lavage or transthoracic needle aspiration, are generally considered to be the most reli- able, but they require specialist training and may have side effects. Based on data accumulated in the pre-antibiotic era, it is believed that non-bacteremic cases account for 70 80% of pneumonia caused by S pneumoniae [2]. Thus, new diagnostic tests, such as polymerase chain reaction (PCR) and urine antigen assays, are promising in detecting these cases. Detection of S. pneumoniae antigens in the urine of patients with pneumonia was first described in 1917 [3]. Recently, a rapid immunochromatographic membrane test, the NOW Streptococcus pneumoniae urinary antigen test (Binax, USA), has become available and can detect S. pneumoniae antigen in urine samples [4]. The test is simple to perform and detects the S. pneumoniae C-polysaccharide, which is found in the cell wall and is common to all sero- types, providing results within 15 min. In previous studies, the overall sensitivity and specificity of the test ranged from 70% to 80% and from 70% to 100%, respectively [1,2,4 8]. Few studies have been designed to compare bacteremic and non-bacteremic pneumococcal pneumonia, perhaps because it has been difficult to identify non-bacteremic cases. The aim of this study was to compare clinical, radio- logical, microbiological, and epidemiological features of bac- teremic and non-bacteremic pneumonia due to S. pneumoniae. 2. Methods 2.1. Design We undertook a retrospective study from January 1995 through February 2003 of patients diagnosed with pneumo- coccal pneumonia in Hospital Clinico de San Juan (Alicante, Spain), a 350-bed, university-affiliated teaching hospital. 2.1.1. Patients Patients were included in this study if they met the following criteria: n Patients with community-acquired pneumonia (CAP): A case of CAP was defined as an illness occurring in a patient older than 18 years presenting with symptoms of a lower respiratory tract infection, an infiltrate on chest radiography, and non-hospitalization of the patient in the preceding 72 h. n Bacteremic pneumococcal pneumonia: Patients with CAP and one or more blood cultures yielding S. pneumoniae. n Non-bacteremic pneumococcal pneumonia: Criteria for inclusion in this group were stringent in order to identify a group of patients in whom the diagnosis of non-bacteremic pneumococcal pneumonia could be regarded as unques- tionable. This group included patients with CAP and detection of S. pneumoniae antigen in urine using a commercially available immunochromatographic assay (Binax NOW Streptococcus pneumoniae urinary antigen test; Binax) and at least three negative blood cultures, with the blood having been obtained before antibiotic therapy was begun. Because the Binax NOW Streptococcus pneumoniae urinary antigen test was available only after September 2002, all non-bacteremic cases were collected from this date through February 2003. 2.1.2. Baseline data Patients' sociodemographic and clinical characteristics and laboratory findings were collected by medical record review and included: 1) Predisposing factors such as current smoker (currently or up until 6 months before admission), former smoker (N 6 months), severe alcohol or drug abuse, homelessness, nursing home residents, diabetes mellitus, renal or hepatic chronic disease, human immunodeficiency virus (HIV) infection, chronic obstructive pulmonary disease (COPD), solid or hematological malignancy, coronary artery disease, number of hospital admissions in the previous year, congestive heart failure, neurological disease (stroke, dementia), prolonged use of glucocorti- costeroids or other immunosuppressive drugs, asthma, neutropenia, and functional or anatomic asplenia. 2) Hematological and chemical tests, including renal and hepatic function, albumin, arterial oxygen saturation, pH, and bilirubin, were performed. 3) Chest radiographs from the time of admission were reviewed and radiological findings compatible with pneumonia were also collected. The type of infiltrate was classified as segmental (d"1 segment within 1 lobe), lobar (e"2 segments within 1 lobe), or multilobar. The presence or absence of pleural effusion was also determined. 4) Microbiological investigations included a total of three blood samples for aerobic and anaerobic culture, a urine sample for detection of S. pneumoniae urinary antigen and Legionella pneumophila serotype 1 urinary antigen (when available), and serum samples (obtained during the acute stage of illness and 2 4 weeks later) for serol- ogical testing. A complement fixation test was performed to detect antibodies against Mycoplasma pneumoniae, C. pneumoniae, Chlamydia psittaci, and Coxiella burnetii. An indirect immunofluorescence test was used to detect antibodies against L. pneumophila. The presence of pneumococcal C-polysaccharide in urine was determined using a commercially available immunochro- matographic assay (Binax NOW Streptococcus pneumoniae urinary antigen test). Only concentrated urine samples were used. Urine samples were boiled for 5 min and centrifuged for 15 min and concentrated 25-fold. The Binax NOW Strepto- coccus pneumoniae urinary antigen test was performed in accordance with the instructions of the manufacturer. The test result was interpreted by the presence or absence of visually detectable pink-to-purple-colored lines, and only medium- positive or very intensely positive readings, as compared with the positive control line, were considered to be positive. A weakly positive reading was considered to be a negative result. 2.1.3. Antimicrobial susceptibility testing Antibiotic susceptibility data from all bacteremic cases were recorded. Minimum inhibitory concentrations (MIC) were determined by broth microdilution in cation-adjusted Mueller Hinton broth supplemented with 5% lysed horse blood according to the 2002 National Committee for Clinical Laboratory Standard (NCCLS) breakpoints. Intermediate resistance to penicillin was defined as a MIC from 0.12 to 1 ºg/mL and high level resistance as a MIC of 2.0 ºg/ml or more. Resistance to erythromycin was defined as a MIC of 1 ºg/ml or more. 2.1.4. Statistical analysis Univariate analyses were performed using SPSS, version 10.1. For comparisons between case categories, the chi- square/Fisher's exact test was used, as necessary, for discrete variables and Student's t-test was used for continuous variables. A two-tailed p value below 0.05 was considered statistically significant for all analyses. 3. Results There were 82 patients in our study, 53 male and 29 female. Fifty-seven were bacteremic and 25 non-bacteremic. An increased incidence of pneumococcal pneumonia was detected during the study period (Fig. 1), although the availability of Binax NOW Streptococcus pneumoniae Fig. 1. Incidence case distribution in the study period 1995 2003. Table 1 Predisposing factors and difference between bacteremic and non-bacteremic patients F. Jover et al. / European Journal of Internal Medicine 19 (2008) 15 21 17 Sex (male) Age (years) Chronic liver disease Congestive heart failure Current smokers Former smokers COPD Intravenous drug use Alcohol abuse HIV infection Overall Bacteremic Non-bacteremic (n=95) (n=70) (n=25) 56% 71% 64% 63 39.8 70.8 11.6% 16% 0% 13.7% 7% 32% 33.7% 40% 16% 25.3% 17% 48% 24.2% 18% 40% 13.7% 19% 0% 24.2% 30% 8% 16.8% 23% 0% p value NS b 0.001 0.028 0.004 0.024 0.024 0.033 0.014 0.021 b 0.001 urinary antigen test allowed the diagnosis of non-bacteremic cases in the last 2 years. As previously described, 80% of the cases were diagnosed in cold seasons. The overall mean age was 63 years (range 27 93 years), with a significant difference between bacteremic and non-bacteremic patients (57.5 vs. 77.2 years, p b 0.001). In seven non-bacteremic cases, another etiology was found: one case of Legionella pneumonia was confirmed by urinary antigen test and indirect immunofluorescence test. Six cases of seroconver- sion to C. pneumonia were diagnosed by serology. 3.1. Predisposing factors Sixty-nine patients (84%) had at least one predisposing factor for pneumococcal pneumonia. Sixty-four patients (78%) had two or more predisposing conditions, with a higher number among bacteremic than non-bacteremic ones (36% vs. 24%, respectively, p = 0.207) (Table 1). Overall, cigarette smoking was the most prevalent risk factor, present in 59% of patients (33.7% of current smokers and 25.3% of former smokers). Bacteremic cases signifi- cantly tended to be current smokers (49.1% vs. 16% for non- bacteremic cases; p=0.004), while non-bacteremic cases included significantly more former smokers (48% vs. 21.1% for bacteremic cases; p = 0.033). Chronic obstructive pulmo- nary disease (COPD) was diagnosed in 24% of patients, with a significant difference between non-bacteremic and bacter- emic groups (40% vs. 22.8%;p=0.033). Six patients had a malignancy process (3 hematological and 3 solid neoplasms). Alcohol abuse was present in 23 patients and it was significantly more common among bacteremic patients [21 of 57 (36.8%) compared with 2 of 25 (8%), p=0.021]. Fifteen patients from the bacteremic group had documented chronic liver disease, but none of the non- bacteremic group had it (p=0.028). Thirteen patients from bacteremic group had a history of intravenous drug use, but none of the non-bacteremic group did (p = 0. 014). A history of congestive heart failure was present in 8.8% (5 cases) of the bacteremic group and in 32% (8 cases) of the non- bacteremic group (p = 0.004). Sixteen patients (bacteremic 18 F. Jover et al. / European Journal of Internal Medicine 19 (2008) 15 21 group) had HIV infection and six of the them fulfilled AIDS criteria. Twenty-three patients had been discharged from the hospital within the previous year because of chronic illness. 3.2. Clinical features Pneumonia features were present in 86% of cases. Delirium was present in 21% of cases. The number of patients whose temperature exceeded 38.5 ∞C on admission was significantly higher among bacteremic than non- bacteremic patients (41% vs. 20%, p=0.044). Heart rate on admission was significantly higher among bacteremic than non-bacteremic patients (113±16 beats vs. 100±26 beats; p b 0.001). Admission to an intensive care unit was necessary in three cases. In four cases, pneumococcal pneumonia was recurrent. 3.3. Laboratory studies at admission Anemia, defined as a hematocrit level lower than 30%, was present in 13% of cases. Bacteremic cases had significantly lower hematocrit levels than non-bacteremic cases (36.6 vs. 40.1%; p b 0.05). WBC counts on admission ranged from 780 to 52,800 per mm3. Platelet counts were significantly lower among bacteremic cases (200.133 vs. 246.826; p=0.049). Serum urea and creatinine were higher than 60 mg/dL and 1.5 mg/dL, respectively, in 23% and 32% of patients. 3.4. Radiographic findings Pulmonary infiltration involved more than one lobe in 28 cases and 23 were bacteremic (23/52 vs. 5/25; p = 0.122). Eleven patients had pleural effusion on plain chest X-ray, with a higher proportion among bacteremic cases, although this difference was not statistically significant (14% vs. 4%; p = 0.155). No other differences were detected in radiological patterns between the bacteremic and non-bacteremic groups. 3.5. Empirical antibiotic treatment and susceptibility On admission, most patients (93%) received intravenous antibiotic treatment. Antibiotics used as empirical treatment were third-generation cephalosporins in 52 cases, macrolides in 43 cases, amoxicillin/clavulanate in 26 cases, and fluoroquino- lones in 9 cases. In one-half of the cases, dual therapy was prescribed, mainly a combination of third-generation cepha- losporins and macrolides. Combination therapy was prescribed more often in older patients (59 vs. 50 years; p = 0.037) and in patients with more than two predisposing factors (59% vs. 41%; p = 0.006), but no differences were noted between bacteremic and non-bacteremic patients. In bacteremic cases, intravenous administration in order to become afebrile was significantly longer than in non-bacteremic cases (6 vs. 4.5 days; p b 0.006). At discharge, a switch to oral therapy was prescribed to 84% of patients, 73% of whom were on monotherapy Table 2 Variables associated with pneumonia-related mortality Clinical pneumonia features (atypical/typical) Delirium (yes/no) Length of fever N 4 days Previous inbed stay Mortality 33% vs. 5% 33% vs. 5% 23% vs. 6% 33% vs. 7% 44% vs. 6% 60% vs. 6% 66% vs. 7% 21% vs. 5% 14% vs. 2% 27% vs. 7% p 0.002 0.018 0.05 0.023 0.004 0.005 0.023 0.022 0.05 0.05 Neurologic disease Cancer Nosocomia pneumonia Urea N 60 mg/dl Heart rate on admission Chronic liver disease N 120 bpm (amoxicillin/clavulanate in 40% and cefuroxime in 32%). Fifteen patients received two antibiotics (≤ lactamic and macrolide combination). Significantly more non-bacteremic patients than bacteremic patients received dual therapy (43% vs. 9%; p b 0.001). Fifty-seven pneumococcal isolates (bacteriemic group) were tested for penicillin susceptibility. Some 74% of them were susceptible [minimum inhibitory concentration (MIC) d"0.06 ºg/mL], 19.7% were interme- diately resistant, and 6.1% were resistant. 3.6. Clinical outcome and length of hospital stay Pneumonia-related mortality during hospitalization was 11% (9 cases). It was higher in the bacteremic group (7 cases, 13.4%) than in the non-bacteremic group (2 cases, 8%), although this difference was not significant. The mean hospital stay was 7.59±3.45 days, but it was significantly longer among bacteremic cases than non-bacteremic cases (7.9 ± 3.7 days vs. 6.8 ± 2.6 days; p = 0.017) (Table 2). 4. Discussion In this comparative study of bacteremic and non- bacteremic patients, we used the Binax NOW Streptococ- cus pneumoniae urinary antigen test as a diagnostic tool for non-bacteremic cases. The manufacturer's own study confirmed that the test was able to detect 44 different strains of S. pneumoniae, representing the 23 serotypes responsible for at least 90% of pneumococcal infections. Some studies have reported excellent sensitivity and specificity. Dom- inguez et al. [2] found an excellent overall specificity (97.2%) and sensitivity (80.4%). Sensitivity was higher among bacteremic (82%) than non-bacteremic cases (78%). Murdoch et al. [6] reported a high sensitivity of the test (80%) when blood cultures were positive. The absence of S. pneumoniae antigen in the urine of controls provided the highest specificity (100%). Recently, Marcos et al. [1] demonstrated that overall specificity of the test to diagnose pneumococcal CAP in concentrated urine using culture results as a reference was 82%. For bacteremic cases, sensitivity reached 100%, in contrast to 70% for non- bacteremic cases. Gutierrez et al. [8], using conventional microbiological criteria (blood and sputum cultures) as the gold standard, found a sensitivity of 70.4% and a specificity of 89.7%. The positive and negative predictive values were 54.3% and 94.6%, respectively. The high sensitivity and specificity of the NOW Strepto- coccus Pneumoniae urinary antigen test may be due, in part, to the detection of the cell wall C polysaccharide common to all S. pneumoniae strains, rather than to type-specific capsular polysaccharides. Some studies have used concentration of urine, a simple, rapid, and easily performed procedure that can improve the sensitivity of the S. pneumoniae urinary antigen test. Marcos et al. [1] noted a 1.4-fold increase in yield following concentration of urine (from 38% to 53%). The availability of the NOW Streptococcus pneumoniae urinary antigen test increased by twofold the incidence of pneumococcal pneumonia, although the majority of cases were non-bacteremic. In seven non-bacteremic cases, another etiology was found. One patient had L. pneumoniae pneumonia confirmed by direct immunofluorescence and urinary antigen test, and six patients presented seroconver- sion to C. pneumoniae. Because the NOW Streptococcus pneumoniae urinary antigen test specifically detects the cell wall C-polysaccharide common to all S. pneumoniae strains and only other streptococci (Streptococcus oralis and Streptococcus mitis), we think that mixed infection can be a reliable explanation. In fact, some studies have proposed C. pneumoniae as a facilitating pathogen for S. pneumoniae infection. In our series, as much as 84% of the patients had at least one predisposing factor and 67.4% had two or more, coinciding with previous reports [9]. We found that bacteremic and non-bacteremic pneumo- coccal pneumonia patients differed in some baseline characteristics. As previously reported [10,11], bacteremic patients were significantly younger, although the higher proportion of HIV-infected and intravenous drug users among this group could influence this. As described in other series [9,12], the most frequent predisposing factor was cigarette smoking (59%), although current smokers (33.7%) were more prevalent than former smokers (25.3%). We found that bacteremic cases were more likely than non-bacteremic cases to be current smokers. In contrast, non-bacteremic cases tended more significantly to be former smokers. Although the specific biological mechanisms by which exposure to tobacco smoke increases the risk of pneumococcal disease are not definitively known, it is thought that cigarette smoke impairs mucociliary clearance, enhances bacterial adherence, and disrupts the respiratory epithelium. In addition, smokers have higher rates of pneumococcal carriage than non- smokers, and smokers may be more susceptible to viral infections of the respiratory tract, such as influenza. These factors can explain the increased risk of invasive pneumococ- cal disease in smokers. COPD was present in one-half of the patients and it was significantly related to non-bacteremic patients. Smoking is the most common cause of COPD and the rate of pneumococcal disease is high among these patients. A causative role of smoking in the pathogenesis of pneumococcal infections seems highly plausible. Indeed, a case-control study [13] found an increased risk of pneumo- coccal infection among current and former smokers. In another case-control study [14], sponsored by the Centers for Disease Control and Prevention and conducted in three North American metropolitan areas, current smoking was associated with a quadrupling of the odds of invasive infection, and non-smokers who were exposed to second- hand smoke had 2.5 times the odds of disease of non- smokers without such exposure. Former smokers continued to have an excess risk for at least 10 years. There was a clear dose response relation with respect to the amount of smoking, the duration of passive exposure, and the length of time since the cessation of smoking. These results provided the evidence that smoking promotes pneumococcal disease and established smoking as the most important risk factor for invasive disease. As previously described [9,14], alcoholism and chronic liver disease were significantly higher in the bacteremic group. Leukopenia and neutrophil dysfunction in these subgroups of patients can explain the predominance of bacteremia. Previous admission to the hospital in almost one- quarter of patients reflects the high prevalence of chronic conditions in this population (COPD, congestive heart failure, and HIV infection). The proportion of bacteremic patients with a temperature higher than 38.5 ∞C was significantly higher than among non-bacteremic patients. This fact must be stressed because, in both groups, blood cultures were collected at admission. Moreover, bacteremic patients had higher heart rate levels on admission than non-bacteremic cases. We think that the degree of fever could be a distinctive feature between bacteremic and non-bacteremic patients. In our study, bacteremic patients were more likely to be anemic than non-bacteremic patients and to have lower platelet levels. No other differences were noted in laboratory results. In their series, Musher [9] and Brandeburg [15] found that bacteremic patients were more likely to have anemia, lower albumin, and elevated blood urea and serum creatinine levels. Although not statistically significant, pleural effusion was more frequent in bacteremic patients. Musher [9] detected a significantly higher proportion of bacteremic patients with pleural effusion and air bronchogram, although neither was confirmed later [13,14]. At admission, most patients received intravenous antibi- otic treatment (mainly the combination of third-generation cephalosporins and macrolides). Although no differences were noted in the two groups, as previously reported [14], bacteremic cases needed longer intravenous administration in order to become clinically stable than non-bacteremic cases. At discharge, most patients were switched to oral therapy, mainly monotherapy with ≤ lactamic (amoxicillin/clavula- nate or cefuroxime). However, patients who received combination therapy (≤ lactamic and macrolide combination) F. Jover et al. / European Journal of Internal Medicine 19 (2008) 15 21 19 20 F. Jover et al. / European Journal of Internal Medicine 19 (2008) 15 21 were more likely to be non-bacteremic. In our study, 26% of pneumococcal isolates had decreased susceptibility to penicillin (19.7% intermediately resistant and 6.1% resis- tant). For this reason, we think that combination therapy is not justified if amoxicillin/clavulanate or enhanced-spectrum fluorquinolones are used in monotherapy, given also that the macrolide resistance percentage was 18%. Overall pneumonia-related mortality during hospitalization was 9.5%; it was higher in the bacteremic group (10%), although this difference was not significant. Musher [9] found a 7-day mortality rate of 12%, similar to our results, and an overall 90-day mortality of 21%. Significant differences were noted between bacteremic and non-bacteremic groups in both periods. Watari et al. [16], studying 37 bacteremic and non- bacteremic pneumococcal pneumonia patients, detected an overall mortality of 14%, although it was significantly higher among bacteremic patients (27.3% vs. 7.7%). In contrast, although Brandeburg et al. [15] detected higher mortality in bacteremic patients, this difference was not statistically significant. In our results, atypical presentation features, delirium, prolonged fever, previous inbed stay, cancer, acute renal failure, and neurological or chronic liver disease were significantly related to a higher mortality rate. The mean hospital stay was almost 8 days, although it was significantly longer among bacteremic cases, reflecting the need for more prolonged intravenous treatment and more complications. There are some limitations that need to be acknowledged and addressed regarding the present study. The first concerns the historical and retrospective nature of the study. Another limitation could be the lower number of non-bacteremic cases compared to bacteremic cases. Nevertheless, because we wanted to be as strict as possible, we refused to include some non-bacteremic cases in the last 2 years of the study. Explicit inclusion criteria for selection of non-bacteremic cases made it dififcult to collect more cases. Finally, data dealing with treatment and outcome should be interpreted with caution since antibiotic therapy was not randomized. In conclusion, among patients with pneumococcal pneumonia, cigarette smoking is the most prevalent risk factor, although current smokers are more likely to be bacteremic and former smokers and COPD patients non- bacteremic. We think that these differences may be related to a dose response relation with respect to the amount of smoking, the duration of passive exposure, and the length of time since the cessation of smoking, as previously described for invasive pneumococcal disease [15]. Bacteremic pneu- monia patients need longer intravenous antibiotic adminis- tration to become clinically stable than non-bacteremic patients, resulting in a longer hospital stay. 5. Learning points " S. pneumoniae is the leading cause of CAP (30 40%). However, underdiagnosis is frequent and is thought to be responsible for at least one of three episodes of CAP without an etiological diagnosis. " Detection of S. pneumoniae antigens in the urine is good in terms of sensitivity and specificity and can help in diagnosing additional cases. " Among patients with pneumococcal pneumonia , cigarette smoking is the most prevalent risk factor. However. current smokers are more likely to be bacteremic and former smokers and COPD patients non-bacteremic. These differences could be related to a dose response relation to smoking. " As previously reported, bacteremic patients were sig- nificantly younger, although the higher proportion of HIV-infected and intravenous drug users in this group could influence this. " The main limitation to the present study is its historical and retrospective nature. Consequently, data dealing with treatment and outcome should be interpreted with caution since antibiotic therapy was not randomized. ˇ˛Acquired disorders of coagulation Vickie McDonald Abstract Normal coagulation is a balance between pro- and anti-thrombotic mech- anisms. Haemorrhage occurs when factors that promote thrombus forma- tion are dysfunctional/absent and may be due to inherited or acquired factors. The most common acquired abnormalities seen in the clinical setting are covered in this article including vitamin K deficiency and warfarin therapy, liver disease, disseminated intravascular coagulation, platelet disorders and vascular disorders. Patients bleeding on warfarin therapy need urgent INR testing and reversal with vitamin K and/or prothrombin complex concentrate. Patients with liver disease have complex haemostatic changes and the management of bleeding depends on the site and severity of bleeding. Disseminated intravascular coagula- tion may complicate many clinical situations and needs prompt action when patients are bleeding. Acquired dysfunction of platelets is commonly encountered in clinical practice, often in association with drug therapy such as aspirin. Keywords antiplatelet drugs; bleeding; disseminated intravascular coagulation; liver disease; vascular abnormalities; vitamin K deficiency; warfarin Introduction The normal coagulation process is a balance between pro- and anti- thrombotic mechanisms. Haemorrhagic disorders occur when there is deficiency or dysfunction of part or parts of the haemostatic system (platelets, coagulation factors or the blood vessel wall). Acquired disorders of coagulation are much more likely to be encountered in clinical practice than inherited disorders, and clin- ical features vary from simple bruising to life-threatening bleeding. History Establish the type of bleeding, precipitating factors and the severity and frequency of bleeding. The patient s medical history may give a clue, if it affords evidence of liver or kidney disease, and a careful drug history is critical. Examination Look for evidence of associated systemic disease, such as signs of liver disease or lymphadenopathy. Examine the mucosal surfaces such as the skin, mouth and joints for bruises, petechiae and signs of bleeding. Bruises suggestive of an abnormal bleeding Vickie McDonald MA PhD MRCP FRCPath is a Consultant Haematologist at Guy s and St Thomas NHS Foundation Trust, London, UK. Competing interests: none declared. tendency are usually large, and occur with minimal trauma and in atypical sites, such as the trunk. Investigations Initial investigations should include: full blood count and blood film examination coagulation screen prothrombin time (PT) activated partial thromboplastin time (APTT) fibrinogen If PT or APTT are prolonged, a 50/50 mix of the patient s plasma with normal plasma will distinguish between an inhibitor or a clotting factor deficiency Renal and liver function. Additional investigations, such as factor assays or platelet studies, will depend on the initial history, examination and baseline investigations. Vitamin K deficiency and warfarin therapy Vitamin K is essential for the function of coagulation factors II, VII, IX and X, and anticoagulants proteins C and S. Deficiency occurs as a result of malabsorption or (rarely) dietary deficiency and leads to a prolonged PT. Warfarin inhibits the effect of vitamin K on clotting factors and the international normalized ratio (INR) is used to monitor its effect. The risk of major haemorrhage in patients taking warfarin is about 2% per year and management depends on the INR and severity of bleeding.1 If the INR is high with no bleeding, warfarin should be stopped and recommenced when the INR falls back into range. If the INR is greater than 5.0, small doses (1 or 2 mg orally) of vitamin K may be needed with INR repeated 24 hours later. If the patient has major bleeding, vitamin K and dried prothrombin complex (prothrombin complex concentrate (PCC), which contains factors II, VII, IX and X) should be given to reverse the effect rapidly.2 PCC acts quickly and a repeat INR can be performed 10 minutes after administration to see its effect, but its response lasts only a few hours and vitamin K should be given alongside to establish a more durable reduction in the INR.1 Fresh frozen plasma (FFP) is no longer recommended for the reversal of warfarin.1 PCC is associated with potential for thrombosis and so is BLEEDING DISORDERS What s new? C Prothrombin complex concentrate should be used in preference to fresh frozen plasma for reversal of warfarin because of its more rapid onset of action, better efficacy, and because it has better viral inactivation C Increasing repertoire of anti-platelet agents for cardiac disease means acquired disorders of platelet function are increasingly common C Liver disease is becoming increasingly common and the coag- ulation abnormalities associated with this are likely to become more of a clinical burden MEDICINE 41:4 228 ! 2013 Elsevier Ltd. All rights reserved. used with caution in some groups. Neonates are born with a low vitamin K concentration and are at risk of haemorrhage if this is not supplemented.3 Liver disease Abnormal haemostasis in liver disease is usually multifactorial (Table 1) and not all patients have associated bleeding.4 The management of bleeding in patients with liver disease depends on its site and severity. Vitamin K 10 mg orally (or intravenously) for 3 days may help patients who have associated vitamin K deficiency. In life-threatening or major bleeding, any potential source of bleeding, such as varices, should be sought and treated. In addition, platelets should be given if the platelet count is less than 50e100 109/litre but recovery may be poor if spleno- megaly is present. FFP can be used to replace clotting factors in major haemorrhage when the PT (`APTT) is prolonged but large volumes are often required. PCC has been used off-licence in patients with liver disease who are bleeding but should be dis- cussed with a haematologist or expert in this area. In addition, fibrinogen replacement with cryoprecipitate (or off-licence use of fibrinogen concentrate) may be required in those patients with low fibrinogen concentration. Disseminated intravascular coagulation Disseminated intravascular coagulation (DIC) is due to inappro- priate and excessive activation of the haemostatic system.5,6 It may lead to thrombosis, haemorrhage or both, and purpura fulminans is the syndrome of DIC with skin necrosis. There are many triggers, including trauma, infection, malignancy or placental abruption.5,7 Activation of coagulation leads to micro- thrombus formation with subsequent consumption of clotting factors and increased breakdown of fibrin. The clotting factors become depleted, platelets are consumed and fibrin deposition in the circulation reduces tissue perfusion and causes mechanical haemolysis. The patient may bleed from any site but this is often mucocutaneous. The clotting screen shows a prolonged APTT, with or without a prolonged PT, and low fibrinogen. In addition, investigations show thrombocytopaenia, anaemia due to red cell breakdown (microangiopathic haemolytic anaemia) and raised D-dimer. Emergency management is discussed in MEDICINE 2013; 41(5). Massive blood loss Causes of abnormal clotting after massive blood loss include dilu- tion, hypocalcaemia and acidosis and are discussed in the article on Blood transfusion on pages 242e247 of this issue and also in the article on Emergency management in MEDICINE 2013; 41(5). Acquired disorders of platelet function Acquired platelet disorders (listed below) are much more common than congenital ones (see Inherited Bleeding Disorders on pages 231e233 of this issue). Antiplatelet drugs (e.g. aspirin) irreversibly acetylate platelet cyclo-oxygenase, impairing thromboxane A2 production and platelet aggregation. The effects of aspirin last approximately 7e10 days. Low-dose aspirin can prevent thrombosis, but may lead to haemorrhage, espe- cially from the gastrointestinal tract. Other antiplatelet drugs include dipyridamole, which inhibits phosphodies- terase, clopidogrel and GPIIb/IIIa inhibitors. Uraemia can result in a functional defect in platelets which may improve with dialysis or DDAVP (desmopressin). Cardiopulmunary bypass e the bypass circuit may result in platelet trauma, leading to thrombocytopenia, platelet fragmentation and platelet function abnormalities.8 With excessive bleeding, platelet transfusions may be required. Myeloproliferative disorders e may be associated with abnormal platelet function, which can result in thrombosis or haemorrhage.9 Haemostasis usually improves with correction of the platelet count. Vascular disorders Intact vascular endothelial function and the ability of vessels to contract are essential in controlling blood loss. Abnormal vessel walls can result in excessive bleeding and routine clotting tests are normal. Hereditary disorders include hereditary haemorrhagic telangiectasia (OslereRendueWeber syndrome), characterized by fragile telangiectasia and arteriovenous mal- formations, and collagen vascular disorders (e.g. EhlerseDanlos syndrome), characterized by hyperextensible joints and elastic skin. Acquired disorders include corticosteroid-induced purpura, HenocheScho¨ nlein purpura and scurvy, which leads to an acquired collagen abnormality. A BLEEDING DISORDERS Causes of abnormal coagulation in liver disease Causes of abnormal coagulation in liver disease C Vitamin K deficiency from C malabsorption C Defective synthesis of clotting factors C Thrombocytopenia due to C portal hypertension, hypersplenism and reduced thrombopoietin production C C Abnormal fibrinolysis and/ C or intravascular coagulation Coagulation abnormalities seen in liver disease Prolonged prothrombin time (PT): due to vitamin K deficiency leading to reduced synthesis of factors II, VII, IX and X Prolonged activated partial thromboplastin time and PT: due to reduced synthesis of all clotting factors Low platelet count Reduced fibrinogen due to disseminated intravascular coagulation C Prolonged thrombin time due to dysfibrinogenaemia C Reduced protein C and S levels Table 1 MEDICINE 41:4 229 ! 2013 Elsevier Ltd. All rights reserved. MEDICINE 41:4 230 ! 2013 Elsevier Ltd. All rights reserved. BLEEDING DISORDERS ˇ˛Acromegaly Belayet Hossain William M Drake Abstract Acromegaly is a rare, chronic, debilitating condition which, if untreated, not only causes significant morbidity but reduces life expectancy by about 10 years. Because the disease process is insidious and early pre- senting features can be atypical, physicians, dentists and surgeons should consider this diagnosis if any of the typical features of acromegaly are present. This is important because surgery remains the only hope of cure and surgical outcome varies widely with the size of the adenoma, which is related to the duration of the disease (90% versus 40e45% in microadenoma versus macroadenoma). The diagnosis of acromegaly is based on three key findings: clinical features, elevated insulin-like growth factor I and inability to suppress serum growth hormone to less than 1 mU/litre following a 75 g oral glucose challenge. Following biochemical confirmation of the disorder, magnetic resonance imaging of the pituitary is performed to assess the size and regional anatomy in anticipation of future surgery. Medical control of acromegaly has improved significantly during the last few years with the introduction of long-acting somato- statin analogues and the relatively newer agent, pegvisomant. Radiation therapy is a potential adjuvant therapy for patients with residual disease, but can take 5e10 years to have its full effect. Keywords acromegaly; pituitary Acromegaly is a rare, chronic, debilitating condition in adults associated with significantly increased morbidity and mortality. It is almost invariably caused by a growth hormone (GH)- secreting pituitary tumour. The same pathological process occurring before epiphysial closure results in pituitary gigantism. Life expectancy in untreated individuals with or without associ- ated co-morbidities is reduced by several years, as shown in Figures 1 and 2. Epidemiology Acromegaly typically becomes apparent at the age of 25e40 years. In Europe, the incidence is 3e4/million/year and the prevalence 40e60/million. Males and females are affected equally. Aetiology In 99% of cases, acromegaly is caused by a GH-secreting pitui- tary adenoma. In most patients, the tumour secretes GH alone, Belayet Hossain MRCP is Specialist Registrar in Diabetes and Endocri- nology in the North East Thames Region, London, UK. Competing interests: none declared. William M Drake DM FRCP is Consultant Physician in the Department of Endocrinology at St Bartholomew s Hospital, London, UK. Competing interests: none declared. Figure 1 but in some cases co-secretion of prolactin or thyroid-stimulating hormone (TSH) may occur. Acromegaly caused by ectopic secretion of GH-releasing hormone is extremely rare. Occasionally, acromegaly may occur in the context of a rare genetic syndrome (Table 1). Clinical features The clinical features of acromegaly are listed in Table 2. The condition develops insidiously, such that the average time from initial onset of symptoms to diagnosis is typically 5e10 years, and often longer. In view of the non-specific nature of many of the common symptoms (e.g. fatigue, joint pain), it is recom- mended that acromegaly is considered by physicians, surgeons and dental practitioners attending patients with the following symptoms and/or signs: ˇ€‹ increased sweating ˇ€‹ carpal tunnel syndrome ˇ€‹ dental malocclusion Figure 2 Acromegaly co-morbidities. PITUITARY Acromegaly significantly impacts survival 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0 5 10 15 20 25 Length of survival (years) General population Acromegaly and diabetes All acromegaly Acromegaly and cardiac disease Hypertension and heart disease Sleep apnoea Acromegaly co-morbidities Insulin-resistant diabetes Cerebrovascular events and headache Arthritis MEDICINE 37:8 407 O 2009 Published by Elsevier Ltd. Probability of survival Table 1 ˇ€‹ multiple skin tags ˇ€‹ sleep apnoea ˇ€‹ recurrent colonic polyps ˇ€‹ oligomenorrhoea/amenorrhoea ˇ€‹ type 2 diabetes mellitus (insulin resistant) in the absence of a family history and/or appropriate phenotype. Clinical assessment of patients with suspected acromegaly should focus particularly on the symptoms and syndromes listed in Table 2, but should also include detailed neuro-ophthalmic evaluation and address the symptoms and signs that may suggest any of the common complications of acromegaly (hypertension, cardiac failure, large joint osteoarthritis, type 2 diabetes). Investigations and diagnosis The major challenge in acromegaly is early diagnosis of a disease that develops insidiously. GH secretion is pulsatile, with six to ten pulses of release separated by long periods during which GH concentrations are virtually undetectable. This means that measurement of a random GH level has little value in the diagnosis of acromegaly. Although insulin-like growth factor I (IGF-I) mediates almost all of the actions of GH, the assays for this peptide are less robust than those for GH, and circulating IGF-I levels vary considerably with factors such as nutrition, insulin resistance, liver/renal disease and exogenous oestrogen administration. For this reason, diagnostic tests for acromegaly include both serum IGF-I concentration and the 75 g oral glucose tolerance test. Variations in assay methodology dictate that no overall consensus exists for the nadir GH during an oral glucose tolerance test, but most endocrine physicians regard a value of 0.3 mg/litre (1 mU/ litre) as an appropriate threshold for the diagnosis. Further assessment Once the diagnosis is established, further investigations (Table 3) are directed at: ˇ€‹ defining the severity of the disease (usually achieved by calcu- lating the average of several serum GH measurements taken throughout the day, which complements IGF-I measurement) ˇ€‹ assessing anterior pituitary function ˇ€‹ documenting the pituitary and peripituitary anatomy, with a view to possible surgery ˇ€‹ assessing the presence or absence of complications. Management Acromegaly is best managed in a centre with an experienced endocrinologist, a pituitary surgeon and a radiotherapist. The aims of treatment are: ˇ€‹ control of the tumour and its mechanical effects ˇ€‹ relief of symptoms ˇ€‹ reducing serum GH/IGF-I to safe levels associated with improved prognosis ˇ€‹ preservation of normal pituitary function. Surgery Surgical adenomectomy remains the treatment of choice in most patients. With an experienced surgeon, the cure rate (defined by normal age-adjusted and sex-adjusted IGF-I and post-glucose GH nadir of <1 mg/litre) is about 90% in patients with a micro- adenoma and 40e50% in those with a macroadenoma. Drug therapy Various medical therapies are available for patients who are not cured by surgery or who are unwilling to have or unfit for general anaesthesia (Table 4). Dopamine agonists achieve satisfactory biochemical control in a minority patients, but their relative inexpense and oral administration dictate that a therapeutic trial is reasonable in most cases. Somatostatin analogues (e.g. octreotide, lanreotide) are generally regarded as the gold-standard medical therapy, although many patients (up to 40%) do not experience satisfactory control with this class of drug. For these PITUITARY Acromegaly: familial syndromes Isolated familial somatotropinomas C 11q13.1e11q13.3, potential second locus at 2p16e12 MEN-1 C Due to mutation of tumour suppressor gene MENIM on chromosome 11q13 C Pituitary tumours, hyperparathyroidism and pancreatic tumours Carney s syndrome C Due to germ line mutation of the protein kinase a regulatory subunit (PRKARI1A) at 17q23e24. A second candidate is 2p15e16 C Spotty skin pigmentation, cardiac myxomas C Acromegaly, thyroid and testicular tumours, and (ACTH)-indepen- dent Cushing s due to primary pigmented nodular adrenocortical disease McCuneeAlbright C Due to activating mutations in the complex GNAS locus at 20q13.3 C Fibrous dysplasia, cafe au lait marks and precocious puberty C May include acromegaly or Cushing s Clinical features of acromegaly Symptoms C Increased sweating C Change in shoe/ring size C Headache C Tiredness/lethargy C Joint pain C Symptoms of other anterior pituitary hormone deficiency Signs C Characteristic facial appearance e frontal bossing, enlarged nose, deep nasolabial furrows, prognathism, increased interdental separation C Oily skin C Deep voice C Enlargement of hands, feet, tongue, lips C Signs resulting from compression of optic apparatus by pituitary tumour Table 2 MEDICINE 37:8 408 O 2009 Published by Elsevier Ltd. PITUITARY Investigations in acromegaly Establish diagnosis C Serum IGF-I C Non-suppression of GH following 75 g oral glucose tolerance test Establish severity of disease C Mean of several serum GH samples C Serum IGF-I Pituitary anatomy C MRI C Visual field perimetry Pituitary function C Serum prolactin C Serum thyroid-stimulating hormone plus free thyroxine C Serum cortisol at 9:00 a.m. C Serum gonadotrophins and sex hormones Metabolic consequences of high GH C Fasting glucose ˇ€‹ oral glucose tolerance test C HbA1C Identify coexistent disease C Serum calcium (multiple endocrine neoplasia type 1) Identify any complications C ECG ˇ€‹ echocardiography C Chest radiography C Radiography of large joints C Sleep study (if sleep apnoea suspected) IGF-I, insulin-like growth factor I; GH, growth hormone. Table 3 patients, the new therapy pegvisomant, which functions as a GH receptor antagonist, is now available. Radiotherapy Pituitary irradiation is a useful adjunctive treatment in some patients with acromegaly who are not cured by surgery. Two techniques are used: standard external beam irradiation (typi- cally given in 25 fractions over 5 weeks) and highly focused stereotactic radiotherapy given as a single dose. Both work relatively slowly and both may cause hypopituitarism. The number of patients receiving pituitary irradiation has declined as more satisfactory medical therapies have become available, but it Drug treatment of acromegaly Class, name Dopamine agonist C Cabergoline Somatostatin analogues C Octreotide C Lanreotide GH receptor antagonist C Pegvisomant Mechanism of action Binds to D2 receptor on somatotrophs Bind to SMS receptors 2 and 5 on somatotrophs No effect on GH secretion; binds to GH receptors to prevent functional activation and IGF-I generation Efficacy Common side effects (% normal IGF-I) 20e40 Nausea, constipation (less commonly, mood changes) 60 Nausea, diarrhoea, gallstones 97 Hepatitis (about 1%) Parameter monitored GH and IGF-I GH and IGF-I IGF-I only (serum GH increases with pegvisomant therapy) IGF-I, insulin-like growth factor I; GH, growth hormone. remains an important treatment option, particularly in patients with large, aggressive, invasive tumours. ÔªøEuropean Journal of Internal Medicine 16 (2005) 359 ‚Äì 360 www.elsevier.com/locate/ejim Brief report ACTH-secreting pituitary adenoma within an ovarian teratoma Renato Candrina a, Intissar Sleiman a,*, Fausto Zorzi b aSezione di Endocrinologia e Malattie del Metabolismo, Brescia, Italy bServizio di Anatomia ed Istologia Patologica, Casa di Cura Poliambulanza, Brescia, Italy Received 27 September 2004; received in revised form 17 December 2004; accepted 17 December 2004 Abstract The differential diagnosis of Cushing‚Äôs syndrome is one of the most difficult tasks in medicine, and it is especially problematic in cases with ‚Äò‚Äòoccult‚Äô‚Äô ectopic ACTH syndrome. We describe the case of a 26-year-old woman who was found to suffer from ectopic ACTH syndrome due to pituitary microadenoma, localized within a mature ovarian teratoma. Cushing‚Äôs syndrome caused by ovarian neoplasia is unusual, but when it occurs, it is most often due to excessive cortisol production by the ovary. Only rarely has ectopic ACTH syndrome in association with an ovarian tumor been described. D 2005 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. Keywords: Ectopic ACTH syndrome; Ovarian teratoma; Hypercorticosurrenalism 1. Introduction The differential diagnosis of chronic glucocorticoid excess (Cushing‚Äôs syndrome) is one of the most difficult tasks in medicine. It is especially problematic in cases with ‚Äò‚Äòoccult‚Äô‚Äô ectopic ACTH syndrome. A benign ACTH- secreting carcinoid may follow a slowly progressive course that is clinically indistinguishable from that of patients with classic Cushing‚Äôs disease [1]. We present here a case of ectopic ACTH syndrome due to an ACTH-secreting pituitary microadenoma that was localized within a mature ovarian teratoma. 2. Case report A 26-year-old woman presented with oligo-amenorrhea, acne, hirsutism, and weight gain after a 2-year course of oral contraceptives. A tentative diagnosis of congenital adrenal hyperplasia was made at another hospital in March 1996, and she was treated for a few months with dexamethasone. * Corresponding author. Casa di Cura Poliambulanza, Via Bissolati 57, I25124 Brescia, Italy. Fax: +39 0303515351. E-mail address: sleimanintissar@libero.it (I. Sleiman). A standard, low-dose dexamethasone suppression test revealed failure of cortisol suppression. No further testing was deemed necessary, and thus, no high-dose dexamethasone suppression test was performed. MRI examination of the sellar region showed a focal intensity alteration in the left pituitary lobe and the patient underwent transnasal left hemihypophysectomy for presumed Cushing‚Äôs disease in January 1997. The pathology report described diffuse hyperplasia of corticotrophic cells. After pituitary surgery, polyuria developed and the patient was treated with desmopressin nasal spray for 3 months. Supplemental 50-Ag thyroxine therapy was also prescribed. Regular menses resumed, but acne and hirsutism improved only slightly. Serum cortisol was not suppressed after overnight low-dose dexamethasone, and cortisol urinary excretion increased (three times normal values). The patient was referred to our hospital in March 1999 for recurrent amenorrhea and hypercorticosurrenalism. On examination, she had cushingoid features with hirsutism. Our review of the surgical specimen of the pituitary gland demonstrated normal pituitary tissue. Initial investigations showed normal morning cortisol and ACTH levels, but high midnight serum cortisol level (175 ng/ml, normal <50 ng/ ml) and an unequivocally elevated 24-h urinary cortisol of 486 Ag/24 h (normal range: 28‚Äì213 Ag/24 h). Serum 0953-6205/$ -see front matter D 2005 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.ejim.2004.12.004 360 R. Candrina et al. / European Journal of Internal Medicine 16 (2005) 359 ‚Äì 360 Fig.1.Tumorcellsshowintenseanddiffuseimmunostainingforanti- ACTHantibodies(40). cortisol was not suppressed after an overnight high-dose dexamethasone suppression test [2] (from 281 to 197 ng/ml, suppression: <50 ng/ml), and no increase in either ACTH or cortisol could be documented during a corticotropin-releasing hormone (CRH) test [3]. A thoraco-abdominal CT scan revealed a moderate, asymmetric adrenal hyperplasia. Furthermore, left ovarian enlargement was noted. Fine needle aspiration under ultrasonic guidance of the lesion was performed and showed abundant aggregates of squamous cells with finely vacuolated cytoplasm and occasional nuclear atypia, consistent with benign cystic teratoma (dermoid cyst). A total body, radio-labelled octreotide scan was negative, and a radioiodine cholesterol adrenal scan showed bilateral activity. Left ovarian vein catheterization was attempted in order to document an ACTH secretory activity but failed due to the presence of an anatomical variant of the renal vein (the presence of a retroaortic left renal vein). In December 1999, the patient underwent laparoscopic resection of the dermoid cyst. The lesion contained a microcystic area in which different tissues could be detected. Among them, neural tissue of the glial type harbored a 5-mm pituitary adenoma with positive ACTH immunoreactivity (Fig. 1). Thyroid tissue (struma ovarii) with a tiny papillary carcinoma (0.6 mm) was also present. Shortly after surgery, the patient resumed her menses, lost weight (from 85 to 65 kg), and noticed hair and skin changes. Under adrenal replacement therapy with cortisone acetate (initially 37.5 mg/day), serum and urinary cortisol normalized, while ACTH levels transiently rose to five times the upper normal level. 3. Discussion Although the ectopic ACTH syndrome is a relatively infrequent cause of hypercorticosurrenalism, overlooking this diagnosis may expose our patients to unnecessary surgery and delay the correct therapy, as shown in this case report. It is, therefore, mandatory to explore this possibility in every case of ACTH-dependent Cushing‚Äôs syndrome. In our case, two simple tests, i.e., the overnight high-dose dexamethasone suppression test [2] and the corticotropinreleasing hormone test [3], pointed to the possibility of an extrapituitary ACTH-secreting neoplasia. This, coupled with the lack of cure after pituitary exploration and hemihypophysectomy, indicated the presence of an ‚Äò‚Äòoccult‚Äô‚Äô ectopic ACTH syndrome, even when the ACTH level was normal or only slightly above normal. After the CT scan indicated an enlargement of the left ovary, this organ appeared the natural candidate for harboring an ACTH-secreting neoplasia. Cushing‚Äôs syndrome caused by ovarian neoplasia is very rare, but when it occurs, it is most often due to excessive cortisol production by the ovary. Only rarely has the ectopic ACTH syndrome been described in association with an ovarian tumor of various types, Sertoli-cell carcinoma, androblastoma, carcinoid, or steroid-cell tumor [4]. We failed to document ACTH secretion, either directly through selective blood sampling or indirectly through radio-labelled octreotide or via immunocytochemical studies on fine needle aspiration. This last study indicated the presence of a dermoid cyst. The occurrence of a functioning ACTH-secreting pituitary adenoma arising within an ovarian benign cystic teratoma is exceptionally unusual. Only a few cases of this type of tumor contain anterior pituitary tissue [5] and, to our knowledge, only three cases of Cushing‚Äôs syndrome due to ACTH hypersecretion by an ovarian teratoma have been described [6]. Our case is also unusual due to the presence of thyroid tissue harboring a tiny papillary carcinoma. In sum, our case epitomizes the many difficulties in the differential diagnosis of adrenocortical hyperfunction and of the ectopic ACTH syndrome. CLINICAL RESEARCH STUDY Acute Aortic Dissection in Blacks: Insights from the International Registry of Acute Aortic Dissection Eduardo Bossone, MD, PhD,a Reed E. Pyeritz, MD, PhD,b Patrick O‚ÄôGara, MD,c Kevin M. Harris, MD,d Alan C. Braverman, MD,e Linda Pape, MD,f Mark J. Russo, MD, MS,g G. Chad Hughes, MD,h Thomas T. Tsai, MD, MSc,i Daniel G. Montgomery, BS,j Christoph A. Nienaber, MD,k Eric M. Isselbacher, MD,l Kim A. Eagle, MDj; International Registry of Acute Aortic Dissection (IRAD) Investigators a Heart Department, University Hospital, Salerno, Italy; bUniversity of Pennsylvania, Philadelphia; cBrigham & Women‚Äôs Hospital, Boston, Mass; dMinneapolis Heart Institute, Minneapolis, Minn; eWashington University, St. Louis, Mo; fUniversity of Massachusetts, Worcester; g University of Chicago, Chicago, Ill; hDuke University, Durham, NC; iUniversity of Colorado, Aurora; jUniversity of Michigan, Ann Arbor; k University Hospital Eppendorf-Rostock, Rostock, Germany; lMassachusetts General Hospital, Boston. ABSTRACT BACKGROUND: Few data exist on race-related differences in acute aortic dissection patients. METHODS: We evaluated black (n ¬º 189, 14%) or white (n ¬º 1165, 86%) patients (mean age 62.8 √Ü 15.3 years; 36.4% women) enrolled in 13 US centers participating in the International Registry of Acute Aortic Dissection. We excluded patients of other racial descent. RESULTS: Type B acute aortic dissection was more frequent in the black cohort (52.4% vs 39.3%, P ¬º .001). Black patients were younger (mean age 54.6 √Ü 12.8 years vs 64.2 √Ü 15.2 years, P <.001) and more likely to have a history of cocaine abuse (12% vs 1.6%, P <.001), hypertension (89.7% vs 73.9%, P <.001), and diabetes (13.2% vs 6.4%, P ¬º .001). Conversely, they were less likely to have bicuspid aortic valve (1.8% vs 5.8%, P ¬º .029), iatrogenic dissection (0.5% vs 4.5%, P ¬º .010), and prior aortic dissection repair (7.7% vs 12.8%, P ¬º .047). Presenting features were similar except for more abdominal pain (44.6% vs 30.6%, P <.001) and left ventricular hypertrophy on echocardiogram (44.2% vs 20.1%, P <.001) in blacks. Management was similar. Hypotension/shock/tamponade was less common (7.6% vs 20.1%, P <.001), whereas acute kidney failure was more common (41.0% vs 21.7%, P <.001) in blacks. Mortality was similar in-hospital (14.3% vs 19.1%, P ¬º .110, odds ratio 0.704, 95% conÔ¨Ådence interval 0.457-1.085) and at 3 years postdischarge, as evaluated by Kaplan-Meier survival analysis (22.0% vs 14.3%, P ¬º .224, SE ¬º 0.062 and 0.018). CONCLUSIONS: Our study shows differences in type, etiology, and presentation of blacks and whites with acute aortic dissection, yet similar mortality for these cohorts. √ì 2013 Elsevier Inc. All rights reserved. The American Journal of Medicine (2013) 126, 909-915 KEYWORDS: Aorta; Epidemiology; Mortality Prior studies have provided important insights into the clinical characteristics, treatments, and outcomes of blacks compared with whites with stable and unstable coronary Funding: W.L. Gore & Associates, Inc., Varbedian Aortic Research Fund, Hewlett Foundation, Mardigian Foundation, UM Faculty Group Practice, Terumo. ConÔ¨Çict of Interest: None. Authorship: All authors had access to the analyzed data and a role in writing the manuscript. Requests for reprints should be addressed to Eduardo Bossone, MD, PhD, Cardiology Division, ‚ÄúCava de‚ÄôTirreni and AmalÔ¨Å Coast‚Äù, University of Salerno, Via Pr. Amedeo, 36, Lauro (AV) 83023, Italy. E-mail address: ebossone@hotmail.com 0002-9343/$ -see front matter √ì 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.amjmed.2013.04.020 syndromes and heart failure.1-5 These studies have suggested that blacks with these syndromes have higher morbidity and mortality despite their younger age.1-5 Importantly, these studies have suggested that the excess adverse clinical events, including higher mortality, among blacks with these disease processes were associated with greater prevalence of comorbid conditions, signiÔ¨Åcant delays in seeking care, lower quality of care (including in the use of evidence-based therapies and invasive cardiac procedures), and socioeconomic factors including poverty, lower education, and inadequate access to long-term health care after the acute event.1-5 However, few data exist on 910 The American Journal of Medicine, Vol 126, No 10, October 2013 differences in clinical presentation, diagnostic Ô¨Åndings, hospital mortality. Investigators were required to check a management, and outcomes in acute aortic dissection in single box on the case report form indicating whether a blacks. This is due to the rarity of this syndrome in general, patient was white, black, Hispanic, American Indian, Asian, resulting in few patients of African American descent with or other, based on their perception of patient‚Äôs race. acute aortic dissection being admitted even to referral cenCompleted data entry forms were forwarded to the IRAD ters specializing in aortic disease. coordinating center at the University of Michigan. Complete The International Registry of data forms were entered into an Acute Aortic Dissection (IRAD) online database. CLINICAL SIGNIFICANCE includes data on patients with acute aortic dissection from cen Little is known about race-related difStatistical Analysis ters of aortic excellence around ferences in the presentation, treatment, We categorized patients into 2 6-8 the globe, including 13 North and outcomes of patients with acute racial groups: black and white. American centers allowing evaluaortic dissection. Summary statistics are presented ation of acute aortic dissection as frequency and percentage, or as Race-related differences in other caramong a modestly large number of median and interquartile range. black patients. The purpose of this diovascular diseases are often associChi-squared analysis was utilized study was to examine the racial ated with modiÔ¨Åable factors that can be for comparing categorical variheterogeneity in etiology, clinical addressed through modiÔ¨Åcations in ables or Fisher‚Äôs exact test where features, diagnostic Ô¨Åndings, treatcare. appropriate. Continuous variables ments, and outcomes (including were analyzed using Student‚Äôs IdentiÔ¨Åcation of dissection-related dislong-term mortality) of blacks and t test for normally distributed data crepancies between races will assist whites with acute aortic dissection. while a nonparametric test of meWe hypothesized that among pawith risk factor modiÔ¨Åcation in black dians was used to compare tients with acute aortic dissection, patients and overall. continuous variables with skewed the etiology, clinical and diagnostic distributions. Missing values were features, and treatments of blacks not defaulted to negative and denominators reÔ¨Çect cases rewould differ considerably from whites, and that these difported. To analyze in-hospital mortality, a multiple logistic ferences in baseline features and treatments would potentially regression was performed utilizing variables from previously account for the some of the variation in outcomes between published multivariable models from the IRAD data for the 2 cohorts. acute aortic dissection.7,9-11 For in-hospital mortality, adjusted odds ratios and 95% conÔ¨Ådence intervals were constructed. For long-term death, a Kaplan-Meier survival METHODS analysis was performed to compute survival estimates, hazPatient Selection ard ratios, and 95% conÔ¨Ådence intervals. A P value of <.05 We analyzed all patients with type A (involving the was considered to indicate signiÔ¨Åcant difference. All anaascending aorta) and type B acute aortic dissection (occurring lyses were performed using SPSS version 20.0 (IBM, Inc., distal to the left subclavian artery) as deÔ¨Åned by the Stanford Armonk, NY). ClassiÔ¨Åcation and enrolled between January 3, 1996 and February 12, 2011 at 13 US centers participating in IRAD. RESULTS The institution and structure of IRAD have been described previously.6 Patients were identiÔ¨Åed prospectively Demographics and Etiology of Aortic Dissection at presentation, or retrospectively by searching hospital (Table 1) discharge diagnosis records and surgical or imaging dataOf 1354 patients with acute aortic dissection enrolled in the bases. Diagnosis was based on symptom onset, imaging 13 US centers participating in IRAD, 189 were African study, visualization at surgery, or postmortem examination. American (14%). Black patients more frequently suffered Because almost all black patients in IRAD were enrolled in from Type B dissections compared with white patients the US, patients enrolled in other countries were excluded. (52.4% vs 39.3%, P ¬º .001); Type A was correspondingly Patients of ethnicity other than black or white also were less common in the black population (47.6% vs 60.7%, excluded due to very low frequencies. P ¬º .001) compared with whites. All Ô¨Åndings noted in the manuscript refer to the overall patient cohort unless otherData Collection wise noted. Black patients were signiÔ¨Åcantly younger, with 10.8% (vs 6.4% whites) under 40 years of age and only 11% Data were collected on a standard questionnaire developed (vs 42.3% whites) >70 years. Weight was signiÔ¨Åcantly by the IRAD investigators.6 Data collected included patient different between groups (88.0 √Ü 22.5 kg in white patients demographics, history, clinical presentation, physical Ô¨Åndvs 96.5 √Ü 25.0 kg in black patients, P ¬º .004) despite seeing ings, imaging study results, medical and interventional no difference in patient height. Among black patients, management, in-hospital clinical events, length of stay, and Bossone et al Table 1 Racial Differences in Acute Aortic Dissection 911 Demographics and History for Overall Study Population Variable Type A Type B Demographics Age, mean √Ü SD, years Age <40 years Age !70 years Male sex Male !70 years Female !70 years Height (cm) Weight (kg) Referred from primary site to IRAD center Etiology and patient history Hypertension Diabetes Atherosclerosis Cocaine abuse* Marfan‚Äôs Bicuspid aortic valve Known aortic aneurysm Iatrogenic‚Ä† Prior cardiovascular surgery Aortic aneurysm or dissection repair CABG PCI Overall, n (n ¬º 1354, 100%) White (n ¬º 1165, 86%) Black (n ¬º 189, 14.0%) P Value 797 (58.9%) 557 (41.1%) 707 (60.7%) 458 (39.3%) 90 (47.6%) 99 (52.4%) 62.8 √Ü 15.3 94 (7.0%) 501 (38.0%) 861 (63.6%) 251 (29.9%) 250 (52.2%) 174.7 √Ü 11.4 89.5 √Ü 23.2 1102 (81.4%) 64.2 √Ü15.2 74 (6.4%) 481 (42.3%) 740 (63.5%) 245 (33.8%) 236 (57.0%) 174.5 √Ü 11.3 88.0 √Ü 22.5 947 (81.3%) 54.6 √Ü 12.8 20 (10.8%) 20 (11.0%) 121 (64.0%) 6 (5.2%) 14 (21.5%) 175.4 √Ü 12.2 96.5 √Ü 25.0 155 (82.0%) <.001 .031 <.001 .894 <.001 <.001 .603 .004 .813 1008 95 366 40 50 60 258 52 842 71 335 18 47 57 232 51 166 24 31 22 3 3 26 1 <.001 .001 <.001 <.001 .099 .029 .052 .010 (76.1%) (7.3%) (28.5%) (3.1%) (3.8%) (5.2%) (19.8%) (4.0%) 158 (12.1%) 99 (7.6%) 42 (3.7%) (73.9%) (6.4%) (30.3%) (1.6%) (4.2%) (5.8%) (20.7%) (4.5%) 144 (12.8%) 88 (7.8%) 40 (4.1%) (89.7%) (13.2%) (17.1%) (12.0%) (1.6%) (1.8%) (14.4%) (0.5%) 14 (7.7%) 11 (6.0%) 2 (1.2%) .001 .001 .047 .395 .060 CABG ¬º coronary artery bypass grafting; IRAD ¬º International Registry of Acute Aortic Dissection; PCI ¬º percutaneous coronary intervention. *Current and past use, including any patient who has taken cocaine to the detriment of his or her health and social functioning. ‚Ä†It was a result of any form of medical treatment and speciÔ¨Åcally resulted from cardiac catheterization, percutaneous transluminal coronary angiography, cardiac surgery, or other treatment during hospitalization. cocaine abuse was more frequent (12% vs 1.6%, P <.001), as was baseline hypertension and diabetes. In contrast, a history of atherosclerosis, bicuspid aortic valve, existing aortic aneurysm, iatrogenic origin, or prior surgical repair of dissection or aneurysm was more common in the white patient cohort. It should be noted that cocaine abuse was more often present in blacks than in whites affected by both Type A (11.4%, vs 1.2%; P <.001) and Type B dissection (12.6% vs 2.3%; P <.001). Clinical Presentations and Diagnostic Imaging Findings (Tables 2, 3) Almost all black patients (97.3% vs 93.2% in whites, P ¬º .033) complained of pain at presentation. In particular, back pain and abdominal pain were more frequent in blacks compared with whites. In contrast, syncope, a murmur of aortic insufÔ¨Åciency on auscultation, and signs of hemodynamic instability (hypotension-shock-tamponade) were less frequent in blacks. SpeciÔ¨Åcally, blacks affected by Type A dissection present rarely with shock, as compared with whites (1.2% vs 8.4%, P ¬º .022). An abnormal aortic contour on chest radiograph and signs of left ventricular hypertrophy on electrocardiogram were more common in blacks with acute aortic dissection. Electrocardiographic evidence of prior myocardial infarction and low voltage were less frequent in blacks. However, blacks affected by either Type A or Type B dissection and overall had an abnormal electrocardiogram (ECG) at presentation with higher frequency. Diagnostic imaging studies were used similarly, with no race-related differences in the number, type, or order of imaging modalities used. The widest diameter of the ascending aorta when comparing the 2 groups of patients was smaller in blacks overall compared with whites (median ¬º 4.0; Q1Q3 ¬º 3.6-4.7 vs median ¬º 4.5; Q1-Q3 ¬º 3.8-5.4; P <.001). When examined separately by Type A and Type B acute aortic dissection, however, there were no signiÔ¨Åcant statistical differences in widest diameter of ascending aorta. In-hospital Treatment and Short- and Long-term Outcomes (Table 4) Acute kidney failure in-hospital was more common in blacks overall (41.0% vs 21.7%, P <.001) and in Type A (46.5% vs 23.2%, P <.001) or Type B dissection (35.9% 912 Table 2 The American Journal of Medicine, Vol 126, No 10, October 2013 Presenting Symptoms and Physical Examination Variable Presenting symptoms Any pain reported Abrupt onset Chest pain Back pain Abdominal Radiating Migrating Syncope Physical examination Ô¨Åndings Any pulse deÔ¨Åcit Auscultated murmur of aortic insufÔ¨Åciency All neurologic deÔ¨Åcits Hypotension Shock Tamponade Hypotension/shock/tamponade Time from presentation to diagnosis, hours Overall White 1223 1042 935 751 406 544 85 130 1043 895 796 629 327 459 70 122 257 230 165 146 58 19 223 3.0 (93.8%) (83.0%) (73.0%) (59.6%) (32.6%) (44.1%) (7.0%) (10.3%) (23.9%) (19.8%) (12.7%) (12.0%) (4.8%) (1.6%) (18.4%) (1.0-13.0) vs 19.5%, P <.001). Overall, hypotension (25.1% vs 15.3%, P ¬º .004) and cardiac tamponade (10.3% vs 5.6%, P ¬º .047) were more frequent in whites than in blacks, but these differences were not observed when Type A and Type B acute aortic dissection cohorts where considered Table 3 218 207 138 137 54 19 210 3.0 (93.2%) (83.1%) (72.4%) (58.2%) (30.6%) (43.2%) (6.7%) (11.3%) (23.6%) (20.8%) (12.3%) (13.1%) (5.2%) (1.8%) (20.1%) (0.9-12.9) Black P Value 180 147 139 122 79 85 15 8 (97.3%) (82.6%) (76.8%) (67.8%) (44.6%) (50.0%) (9.1%) (4.5%) .033 .865 .220 .016 <.001 .096 .254 .006 39 (25.7%) 23 (14.0%) 27 (14.8%) 9 (5.3%) 4 (2.4%) 0 (0.0%) 13 (7.6%) 3.5(1.3-13.8) .575 .045 .361 .004 .110 .094 <.001 .123 separately. There was no signiÔ¨Åcant difference in overall in-hospital mortality between the 2 groups (14.3% in blacks vs 19.1% in whites, P ¬º .110). However, in the medically treated type A cohort, the in-hospital mortality rate was lower in blacks than in whites (59.8% vs 20.0%; Chest Radiography, Electrocardiography, and Diagnostic Imaging Studies Black P Value (78.9) (24.9%) (40.6%) (33.1%) (13.9%) (90.0) (28.8%) (45.2%) (20.1%) (10.9%) (6.6%) (5.4%) 151 33 63 61 17 171 29 79 69 7 6 1 (79.9) (23.2%) (46.3%) (46.6%) (13.2%) (90.5) (17.2%) (49.7%) (44.2%) (4.6%) (4.0%) (0.7%) .752 .661 .211 .003 .820 .853 .002 .291 <.001 .018 .217 .012 960 768 185 93 (83.0%) (66.3%) (16.1%) (8.3%) 165 114 39 17 (87.3%) (60.3%) (20.7%) (9.1%) .136 .111 .114 .714 783 211 30 19 4.5 (75.1%) (20.2%) (2.9%) (1.8%) (3.8-5.4) 137 25 3 3 4.0 (81.5%) (14.9%) (1.8%) (1.8%) (3.6-4.7) Variable Overall White Chest radiograph Normal Widened mediastinum Abnormal aortic contour Pleural effusion Electrocardiogram Normal NSST LVH Prior infarct ST elevation or new infarct Low voltage Diagnostic imaging Computed tomography Echocardiography (TEE or TTE) Aortography Magnetic resonance imaging First diagnostic test Computed tomography Echocardiography (TEE or TTE) Aortography Magnetic resonance imaging Widest diameter of ascending aorta (median [Q1-Q3]) 1070 259 408 335 132 1220 327 515 258 109 68 51 (79.0) (24.7%) (41.4%) (34.9%) (13.8%) (90.1) (27.2%) (45.8%) (23.5%) (10.0%) (6.3%) (4.7%) 919 226 345 274 115 1049 298 436 189 102 62 50 1125 882 224 110 (83.6%) (65.4%) (16.8%) (8.4%) 920 236 33 22 4.5 (76.0%) (19.5%) (2.7%) (1.8%) (3.7-5.4) .068 .104 .609 1.000 <.001 LVH ¬º left ventricular hypertrophy; NSST ¬º nonspeciÔ¨Åc ST-T changes; Q1-Q3 ¬º Ô¨Årst quartile-third quartile; TEE ¬º transesophageal echocardiogram; TTE ¬º transthoracic echocardiogram. Bossone et al Table 4 Racial Differences in Acute Aortic Dissection 913 Treatment Strategies, Complications, and Mortality Variable DeÔ¨Ånitive treatment Type A Surgical Medical Endovascular Hybrid Type B Surgical Medical Endovascular Hybrid In-hospital complications (pre-op/med Rx or post-op) All neurological deÔ¨Åcits Coma Myocardial infarction/ischemia Mesenteric ischemia/infarction Acute kidney failure Hypotension Cardiac tamponade Limb ischemia Overall in-hospital mortality Type A Surgical Medical Type B Surgical Medical Endovascular Follow-up survival (3-year Kaplan-Meier survival analysis estimates) Proportion of follow-up available, n (%) Overall Type A e survival estimate (SE) Surgical e survival estimate (SE) Medical e survival estimate (SE) Type B e survival estimate (SE) Surgical e survival estimate (SE) Medical e survival estimate (SE) Endovascular e survival estimate (SE) Overall n (%) White n (%) Black n (%) n ¬º 795 645 (81.1%) 107 (13.5%) 24 (3.0%) 19 (2.4%) n ¬º 557 96 (17.2%) 343 (61.6%) 112 (20.1%) 6 (1.1%) n ¬º 705 576 (81.7%) 97 (13.8%) 16 (2.3%) 16 (2.3%) n ¬º 458 80 (17.5%) 289 (63.1%) 83 (18.1%) 6 (1.3%) n ¬º 90 69 (76.7%) 10 (11.1%) 8 (8.9%) 3 (3.3%) n ¬º 99 16 (16.2%) 54 (54.5%) 29 (29.3%) 0 (0.0%) 290 24 139 99 315 305 124 172 250 190 111 60 60 17 27 16 P Value .250 .488 .001 .713 .755 .113 .012 .380 (22.8%) (2.3%) (10.9%) (7.7%) (24.4%) (23.7%) (9.7%) (13.4%) (18.5%) (23.8%) (17.2%) (56.1%) (10.8%) (17.7%) (7.9%) (14.3%) 247 24 120 82 242 278 114 144 223 172 101 58 51 15 24 12 (22.6%) (2.6%) (10.9%) (7.4%) (21.7%) (25.1%) (10.3%) (13.1%) (19.1%) (24.3%) (17.5%) (59.8%) (11.1%) (18.8%) (8.3%) (14.5%) 43 0 19 17 73 27 10 28 27 18 10 2 9 2 3 4 (23.8%) (0.0%) (10.7%) (9.5%) (41.0%) (15.3%) (5.6%) (15.7%) (14.3%) (20.0%) (14.5%) (20.0%) (9.1%) (12.5%) (5.6%) (13.8%) .731 .062 .926 .327 <.001 .004 .047 .331 .110 .364 .527 .020 .552 .729 .782 1.000 460 (41.7%) 411 85.7% 91.5% 93.6% 62.5% 80.4% 81.3% 79.9% 81.1% (43.6%) 49 78.0% 93.8% 92.3% ‚Äì 70.9% 50.0% 57.4% ‚Äì (30.2%) .001 0.224 .803 .815 .270 .317 .059 .120 .135 P ¬º .02). Survival estimates for 3-year postdischarge follow-up was similar in blacks and whites (22.0% vs 14.3%, P ¬º .224). DISCUSSION To our knowledge, this is the Ô¨Årst study to provide insights into the racial heterogeneity in etiology, clinical features, diagnostic Ô¨Åndings, treatments, and outcomes (including long-term mortality) of blacks compared with whites with acute aortic dissection. Our data indicate signiÔ¨Åcant variation in etiology and clinical presentation between black and white patients with acute aortic dissection. Most notably, (.021) (.020) (.135) (.029) (.084) (.035) (.070) (.061) (.074) (.082) (.204) (.132) blacks were more likely to be admitted with type B dissection as compared with whites. Genetic factors such as Marfan syndrome or bicuspid aortic valve as etiology of acute aortic dissection were less common, whereas hypertension and cocaine abuse Ô¨Ågured predominantly as a cause of acute aortic dissection in blacks. The younger age among blacks with acute aortic dissection may have accounted for a higher likelihood of having symptomatic (chest pain, back pain, abdominal pain) rather than asymptomatic (painless) presentation for their acute aortic dissection, the lower likelihood of having had prior surgery for acute aortic dissection or aortic aneurysm, and the more frequent incidence of abnormal ECG at presentation. Most of the other 914 The American Journal of Medicine, Vol 126, No 10, October 2013 differences in symptoms, signs, and Ô¨Åndings at presentation between blacks and whites were accounted for by the higher incidence of type B dissection in blacks. Thus, back or abdominal pain at presentation were more common, whereas a murmur of aortic regurgitation, syncope and hypotension, shock, or tamponade or low voltage on ECG at presentation were less common among blacks with acute aortic dissection. In-hospital complications also reÔ¨Çected these differences in dissection type between the 2 cohorts, with shock, cardiac tamponade, and coma being less common in blacks, whereas acute kidney injury was more common in blacks. Diagnostic modalities were utilized similarly in the 2 racial groups. Invasive treatment of type A dissection (the most desirable approach for these patients) was similar for the 2 groups. This was not surprising given that the participating centers were aortic centers of excellence, and disparity in treatment of disease is less likely at such institutions, particularly surrounding a care area where urgent cardiac surgery is the strongly preferred treatment. While the majority of type B patients were managed medically (as recommended by the American College of Cardiology, American Heart Association, and European Society of Cardiology guidelines),12 blacks with type B acute aortic dissection were more likely to be managed with aggressive invasive approaches (and less likely to be managed medically). Given that hypertension was more frequent in blacks, it is possible that blacks with type B acute aortic dissection had more ongoing symptoms, hemodynamic instability, rapid extension of dissection, expansion of aneurysm, more contained or eminent rupture (periaortic hematoma), or simply poorly controlled hypertension requiring more invasive therapy as recommended by guidelines. Overall inhospital mortality was similar between blacks and whites for type A as well as type B dissection despite some heterogeneity in treatment strategies. This lends credence to the fact that blacks and whites with acute aortic dissection were treated with strategies that were tailored for best outcomes based on presentation, progression, and complications, but not based on race. To our knowledge, no prior investigation involving a large number of patients has provided insight into the differences between blacks and whites with acute aortic dissection in their etiology, presentation, or outcomes. This is despite the fact that aortic dissection may be more common among blacks than among whites.13 The rarity of acute aortic dissection in general accounts for the paucity of such information in the literature. Only one small case series has described acute aortic dissection in 6 black Congolese patients, suggesting diagnoses at young age (mean age 40.2 √Ü 10.6 years, range 17 to 49 years).14 All patients had hypertension, and 2 patients who were operated on survived, whereas the 4 unoperated patients died. No comparison group was provided. The only other publication with regards to race and acute aortic dissection in the literature is in reference to higher incidence of cocaine-induced aortic dissection in black patients (not to be confused with a higher incidence of acute aortic dissection in blacks compared with whites who are cocaine abusers).15 Our Ô¨Åndings may have clinical implications for the prevention, diagnosis, and treatment of acute aortic dissection in blacks. Given that blacks may have a higher incidence of acute aortic dissection, a syndrome with high mortality, awareness and prevention of this condition would be helpful for not only decreasing the overall incidence of aortic dissection, but also decreasing cardiovascular mortality in blacks. However, as signiÔ¨Åcant differences exist in the etiology/predisposing factors for acute aortic dissection between groups, a preventive approach should be tailored to target these risk factors. Sustained and effective treatment of hypertension will not only be more effective in primary prevention, but may have an even greater potential for the secondary prevention of acute aortic dissection in blacks given that almost 90% of the black population in this study had history of hypertension. Whether different antihypertensive agents have differential effectiveness for controlling hypertension and preventive acute aortic dissection among blacks and whites remains to be established. The present goal for prevention of acute aortic dissection should be directed at control of blood pressure using diet and lifestyle, as well as a treatment regimen effective at reducing the blood pressure. Additionally, major efforts directed at educating and decreasing the use of cocaine also is likely to reduce the incidence of acute aortic dissection in blacks. With regards to the diagnosis of acute aortic dissection, it should be recognized that blacks are more likely to present with type B than with type A aortic dissection. As such, a high level of suspicion for acute aortic dissection in blacks presenting with back or abdominal pain may help promote an early use of imaging and diagnosis of this syndrome and improve survival in blacks with acute aortic dissection. While this article illustrates signiÔ¨Åcant differences in history and clinical presentation between black and white patients with acute aortic dissection, it is important to note that these discrepancies did not impact overall survival. Therefore, this lends credence to the idea that clinicians are appropriately diagnosing and treating these patients with management strategies that are both guideline-based and tailored for individual comorbidities. It is noteworthy that neither ethnic cohort had clinical histories or presenting symptoms that deviated signiÔ¨Åcantly from the known contributors to and indicators of aortic disease. We do not advocate ignoring the well-documented predictors of aortic dissection in favor of a narrow patient proÔ¨Åle determined solely by race. However, the differences noted herein do provide some insight and direction to clinicians interested in targeting their preventative efforts toward the patient population, such as treatment of baseline hypertension. Likewise, any factors that may enhance the level of suspicion in patients with this urgent condition, including the increased prevalence of abdominal and back pain in black patients, can only serve to improve our treatment of acute aortic dissection. Bossone et al Racial Differences in Acute Aortic Dissection 915 Limitations References All patients in the current study were black or white patients with acute aortic dissection admitted to centers of aortic excellence, and our results may not be applicable to those with chronic stable aortic dissection, those with acute aortic dissection treated at small centers, or those with acute aortic dissection of other ethnic origin. Differences in prehospital or pretransfer survivorship in blacks versus whites cannot be ascertained from IRAD. Data in IRAD are collected retrospectively through voluntary participation of sites and are subject to missing or incomplete information, particularly with regards to long-term outcomes. While this is the largest study of racial differences in aortic dissection to date, the relatively low numbers of black patients analyzed herein may impact the statistical power of the analysis. Treatment strategies or use of secondary prevention strategies for acute aortic dissection were not protocol driven and were determined by the treating physicians at the enrolling sites in accordance with guidelines. Thus, inference about the effectiveness of various strategies on the outcomes of blacks versus whites should be made with caution. 1. Sonel AF, Good CB, Mulgund J, et al; CRUSADE Investigators. Racial variations in treatment and outcomes of black and white patients with high-risk non-ST-elevation acute coronary syndromes: insights from CRUSADE (Can Rapid Risk StratiÔ¨Åcation of Unstable Angina Patients Suppress Adverse Outcomes With Early Implementation of the ACC/AHA Guidelines?). Circulation. 2005;111(10):1225-1232. 2. Peterson ED, Wright SM, Daley J, Thibault GE. Racial variations in cardiac procedure use and survival following acute myocardial infarction in the Department of Veteran Affairs. JAMA. 1994;271(15): 1175-1180. 3. Mehta RH, Marks D, Califf RM, et al. Differences in the clinical and angiographic features and outcomes In African American and Caucasians with acute myocardial infarction. Am J Med. 2006;119(1): 70e1-70e8. 4. Barnato AE, Lucas FL, Staiger D, Wennberg DE, Chandra A. Hospitallevel racial disparities in acute myocardial infarction treatment and outcomes. Med Care. 2005;43(4):308-319. 5. Bibbins-Domingo K, Pletcher MJ, Lin F, et al. Racial differences in incident heart failure among young adults. N Engl J Med. 2009;360: 1179-1190. 6. Hagan PG, Nienaber CA, Isselbacher EM, et al. The International Registry of Acute Aortic Dissection (IRAD): new insights into an old disease. JAMA. 2000;283:897-903. 7. Bossone E, Rampoldi V, Nienaber CA, et al. Usefulness of pulse deÔ¨Åcit to predict in-hospital complications and mortality in patients with acute type A aortic dissection. Am J Cardiol. 2002;89:851-855. 8. Mehta RH, Suzuki T, Hagan PG, et al. Predicting death in patients with acute type A aortic dissection. Circulation. 2002;105:200-206. 9. Suzuki T, Mehta RH, Ince H, et al. Clinical proÔ¨Åles and outcomes of acute type B aortic dissection in the current era: lessons from the International Registry of Acute Aortic Dissection (IRAD). Circulation. 2003;108(Suppl II):II312-II317. 10. Tsai TT, Evangelista A, Nienaber CA, et al. Long-term survival in patients presenting with type A acute aortic dissection: insights from the International Registry of Acute Aortic Dissection (IRAD). Circulation. 2006;114(Suppl I):I-350-1-356. 11. Tsai TT, Fattori R, Trimarchi S, et al. Long-term survival in patients presenting with type B acute aortic dissection: insights from the International Registry of Acute Aortic Dissection. Circulation. 2006;114: 2226-2231. 12. Erbel R, Alfonso F, Boileau C, et al. Task Force on Aortic Dissection, European Society of Cardiology. Diagnosis and management of aortic dissection. Eur Heart J. 2001;22:1642-1681. 13. Mukherjee D, Eagle KA. Aortic dissection‚Äîan update. Curr Probl Cardiol. 2005;30(6):287-325. 14. Bouramou√© C, Kimbally-Kaky G, Nkoua JL, Le Feuvre C, Ekoba J, Vacheron A. Aortic dissection among blacks. Report of 6 Congolese cases [French]. Ann Cardiol Angeiol (Paris). 2001;50(3):133-141. 15. Hsue PY, Salinas CL, Bolger AF, Benowitz NL, Waters DD. Acute aortic dissection related to crack cocaine. Circulation. 2002;105: 1592-1595. CONCLUSIONS Our study provides important insights into the clinical characteristics and outcomes of patients with acute aortic dissection. While suggesting a few important differences in type, etiology, and clinical presentation of blacks and whites with acute aortic dissection, our data indicate more similarities than differences, especially in mortality for both type A and B acute aortic dissection in the 2 cohorts. These Ô¨Åndings may have implications in terms of risk factors modiÔ¨Åcation for decreasing the incidence of acute aortic dissection in blacks. Hypertension and other modiÔ¨Åable risk factors for acute aortic dissection need to be addressed in both white and black populations. Furthermore, the level of awareness for typical and atypical symptoms of dissection needs to be raised regardless of race. Future studies are needed involving a large number of black patients with acute aortic dissection to evaluate prevention and treatment strategies for reducing and improving outcomes of black patients. ˇ˛Acute aortic syndromes Ibrahim Akin Rachel Clough Tim C Rehders Christoph A Nienaber Abstract Acute aortic syndromes comprise acute aortic dissection, intramural haema- toma, symptomatic penetrating aortic ulcers and traumatic aortic dissection. These conditions result primarily from disruption of the outer aortic layer, and involve thinning of the aortic wall, increased wall stress, progressive dilata- tion, evolution of intramural haemorrhage, and possible dissection and rupture. Chronic hypertension and connective tissue disorders are often implicated. Echocardiography, contrast-enhanced CT, dynamic MRI and aortography are currently used to confirm the diagnosis. Aortic dissection is primarily classified according to anatomical characteristics; those with and without ascending aortic involvement are distinguished for prognostic and therapeutic reasons. In general, open surgery is indicated when dissec- tion involves the ascending aorta, whereas medical management or endovas- cular stent-graft implantation are reserved for cases where the ascending aorta is spared. Pathology involving the aortic arch may be treated using a hybrid approach combining debranching of the head and neck vessels and interventional stent-graft implantation. Stent-graft induced remodelling of dissected aorta seems to have long-term benefits both in complicated and so-called uncomplicated distal dissection. In addition, long-term medical therapy to control hypertension is of paramount importance in all patients who have survived aortic dissection to reduce late complications, including recurrent dissection, aneurysm formation or late extension or rupture. Keywords Acute aortic syndrome; aortic dissection; hybrid approach; intramural haematoma; penetrating aortic ulcer; peripheral vascular dis- ease; stent-graft Disease is very old, and nothing about it has changed. It is we who change as we learn to recognise what was formerly imperceptible Jean Martin Charcot Ibrahim Akin MD is a Physician at University Hospital Rostock, Germany. Competing interests: none declared. Rachel Clough MD is a Clinician-Scientist at King s College London, Car- diovascular Imaging Department, UK. Competing interests: none declared. Tim C Rehders MD is Consultant Physician at the University Hospital Rostock, Germany. He qualified from Hamburg University and Universita¨ tsklinikum Hamburg-Eppendorf, Germany. Competing interests: none declared. Christoph A Nienaber MD FACC FESC is Head Cardiologist at the University Hospital Rostock, Germany. Competing interests: Professor Nienaber has received funds for research in the area of aortic diseases and fees for speaking about this topic at meetings; he has served as an expert witness in the John Ritter case in Glendale, California. Acute aortic syndrome is an established term that includes aortic dissection, intramural haematoma (IMH) and symptomatic penetrating aortic ulcer (PAU), as well as traumatic aortic lac- erations. First described by Morgagni more than 200 years ago,1 acute aortic dissection requires a tear in the aortic intima that is commonly preceded by medial wall degeneration or cystic medial necrosis. Although early studies2 highlighted the high mortality rate and infrequency of ante mortem diagnosis,2 knowledge regarding the incidence of aortic dissection in the general population is limited. Studies suggest an incidence of 2.6 e3.5 cases per 100,000 person/year3 and the prevalence is 0.2 e0.8% in large autopsy series. Peak incidence is in the sixth and seventh decade of life and men are affected twice as commonly as women.4 Aortic dissection in individuals under the age of 40 years is most common in those with Marfan s syndrome and during pregnancy. Pathology Aortic dissection is believed to begin with formation of a tear in the aortic intima, which exposes an underlying diseased medial layer directly to the driving force of intraluminal pulsatile blood. Penetration cleaves the diseased medial layers and progressively dissects the aortic wall. Driven by persistent intraluminal pres- sure, the dissection may progress antegradely (and sometimes retrogradely) along the aortic wall from the site of initial intimal tear, thereby forming a false lumen. Shear forces may lead to further tears in the intimal flap (the inner portion of the dissected aortic wall) to create additional entry or exit sites into the false lumen. Distension and systolic pressure within the false lumen may lead to dynamic compression of the true lumen and distal malperfusion. Aortic IMH is considered a precursor of dissection and origi- nates from ruptured vasa vasorum within the medial wall layers, resulting in aortic wall infarction that may provoke a secondary tear and classic aortic dissection. IMH is usually located in the descending aorta, typically associated with hypertension, and may extend, progress or resorb.5,6 Deep penetrating aortic atherosclerotic plaques can lead to IMH, aortic dissection, or perforation. Non-invasive imaging has recently elucidated this disease process, which often further complicates IMH. Classification The prime classification is based on the anatomical location of the intimal tear and degree of propagation of the false lumen. Intimal tears occur at points of presumed greatest haemodynamic stress; namely, the lateral wall of the ascending aorta and just distal to the ligamentum arteriosum of the descending thoracic aorta. Overall, 65% of intimal tears occur in the ascending aorta, 20% in the descending aorta, 10% in distal aortic arch and 5% at abdominal level. Two anatomical classification systems pre- dominate (Figure 1) e in both, dissections with and without ascending aortic involvement are distinguished for prognostic and therapeutic reasons.7,8 Aortic dissection may also be classified according to the timing of diagnosis relative to the onset of symptoms: acute within 2 weeks, subacute within 2e8 weeks, or chronic beyond 8 weeks. About one-third of patients with aortic dissection fall into AORTIC DISEASE MEDICINE 42:9 532 ! 2014 Elsevier Ltd. All rights reserved. Figure 1 the latter category; both mortality and the risk of progression decrease over time, with implications for the most appropriate treatment strategy.9,10 Aetiology and pathogenesis Any disease process that undermines the integrity of the elastic or muscular components of the media predisposes the aorta to dissection, and degeneration of this layer is the major predis- posing factor in most non-traumatic aortic dissection (Table 1). Degeneration of the media as a result of enhanced apoptosis is a feature of several hereditary connective tissue defects, notably Marfan s and EhlerseDanlos syndromes. In the absence of Marfan s syndrome, medial degeneration is usually minor in most cases of aortic dissection but is nevertheless qualitatively and quantitatively greater than that expected as part of the ageing process, possibly as a result of increased apoptosis secondary to hypertension. A bicuspid aortic valve is associated with acute aortic syndrome in 7e14% of patients. Other congenital cardiovas- cular abnormalities may predispose to dissection, including coarctation of the aorta and giant cell arteritis. Direct trauma to the aorta may cause dissection, whereas blunt trauma tends to cause localized tears, haematomas or transection; iatrogenic trauma (e.g. during cardiac catheterization, cardiac surgery or insertion of an intra-aortic balloon pump) may induce dissection, probably as a result of direct trauma to the aortic intima.11 Clinical features Diagnosis and effective clinical management of acute aortic syndromes require a high level of clinical suspicion and prompt action. The differential diagnosis in acute aortic dissection in- cludes acute coronary syndrome, pulmonary embolism, pneu- mothorax, pneumonia, musculoskeletal pain, acute cholecystitis, MEDICINE 42:9 533 ! 2014 Elsevier Ltd. All rights reserved. AORTIC DISEASE Commonly used classification systems in aortic dissection De Bakey Type I Type II Type III Type A Acuity: Acute: <2 weeks after onset Subacute: 2 8 weeks after onset Chronic: >8 weeks after onset Anatomic location: Ascending aorta: Ascending and descending aorta: Stanford Type A, De Bakey Type I Descending aorta: Stanford Type B, De Bakey Type III Pathophysiology: Type A Type B Stanford Type A, De Bakey Type II Class 1: Classical aortic dissection with initimal flap between true and false lumen Class 2: Aortic intramural hematoma without identifiable intimal flap Class 3: Intimal tear without hematoma (limited dissection) Class 4: Atherosclerotic plaque rupture with aortic penetrating ulcer Class 5: Iatrogenic or traumatic aortic dissection (intra-aortic catheterization, high-speed deceleration injury, blunt chest trauma) Conditions associated with increased risk of aortic dissection Long-standing arterial hypertension Smoking, dyslipidaemia, cocaine/crack Connective tissue disorders Hereditary fibrillinopathies Marfan s syndrome EhlerseDanlos syndrome Hereditary vascular diseases Bicuspid aortic valve Coarctation Vascular inflammation Giant cell arteritis Takayasu s arteritis Behc et s disease Syphilis Ormond s disease Deceleration trauma Car accident Fall from height Iatrogenic factors Cardiac catheterization Valvular/aortic surgery Side- or cross-clamping/aortotomy Graft anastomosis Patch aortoplasty Cannulation site Aortic wall fragility Table 1 oesophageal spasm or rupture, acute pancreatitis and acute pericarditis. The most common initial symptom is severe migrating back or chest pain.12 Amongst 464 patients reported from the International Registry of Acute Aortic Dissection (IRAD), 95% reported pain that was of abrupt onset in 85%.4 Those without pain are usually found to have chronic dissec- tion. The pain is typically stabbing, tearing or ripping and often described as the worst pain ever experienced e it is clearly different from the pain of myocardial infarction. The pain of aortic dissection tends to migrate from its point of origin to other sites, generally following the path of the dissection as it extends through the aorta. Such migratory pain is described by about 20% of patients and may be helpful in pinpointing the anatom- ical location of the dissection. Aortic dissection is a clinical chameleon . The clinical mani- festations in more than one-third of the patients may be explained by impaired perfusion of key organs as a result of side- branch obstruction.4 However, physical findings are variable and may be subtle or absent, even in the presence of extensive dissection. Aortic regurgitation with resultant congestive heart failure may occur and patients may exhibit pulse deficits and neurological manifestations, such as headache, cerebrovascular accidents (as a consequence of carotid involvement), syncope (caused by cardiac tamponade) and paraparesis/paraplegia sec- ondary to spinal ischaemia. Natural history Despite major advances in the non-invasive diagnosis of aortic dissection and its treatment, up to 28e55% of patients die without correct diagnosis.4 Registry data demonstrate that med- ical management of type A dissection is associated with a mor- tality of nearly 24% after 1 day, 29% after 2 days, 44% after 1 week and 50% after 2 weeks. Less than 10% of untreated pa- tients live for 1 year and almost all patients die within 10 years.4 Type B dissection has a better prognosis with survival rates of 89% at 1 month, 84% at 1 year, and 80% at 5 years.4 However, patients with complications such as renal failure, visceral ischaemia or contained rupture fare less well with a mortality of 20% at day 2 and 25e50% at 1 month, and often require urgent repair. Population-based data show a trend to higher incidences of the disease and higher mortality figures with increasing age.13 Diagnosis A low threshold of clinical suspicion is vital for timely diagnosis and management of this catastrophic condition. Chest radiography is often non-specific and does not have a definitive role.14 Possible findings include widening of the aortic contour, displaced calcification, aortic kinking, and opacification of the aortopulmonary window. Transthoracic/transoesophageal echocardiography, contrast-enhanced CT, MRI and aortography (Table 2) may all be used to confirm suspected dissection. Each modality has advantages and disadvantages in terms of conve- nience, risk, cost and local availability. Diagnostic accuracies are similar, but all differ in their ability to detect associated com- plications. The ideal imaging modality should: confirm or refute the diagnosis of aortic dissection determine whether dissection involves the ascending aorta or is confined to the descending aorta or arch reveal the anatomic features of the dissection, including its extent, the site(s) of entry and re-entry, the presence of thrombus in the false lumen, and branch vessel involvement. Unfortunately, no single modality provides all of this infor- mation. In the IRAD registry, the first diagnostic step was transthoracic and transoesophageal echocardiography (TTE/ TEE) in 33%, CT in 61%, MRI in 2% and angiography in 4%. Two or three of these investigations were required in many patients. Management All patients in whom acute aortic dissection is strongly suspected should be managed in an acute-care setting where blood pres- sure, cardiac rhythm and urine output can be monitored. Initial therapeutic aims to include elimination of pain and reduction of systolic blood pressure to 120 mmHg or less (or the lowest level ensuring adequate vital organ perfusion), using intravenous b- blockers (e.g. labetalol or esmolol) and vasodilators.11,12 Surgical treatment Definitive therapy for aortic dissection has evolved over the last two decades. The need for surgery should be determined whenever possible at the time of initial clinical evaluation. Sur- gical risk is increased by advancing age, comorbid disease (particularly coronary artery disease and emphysema), partial or MEDICINE 42:9 534 ! 2014 Elsevier Ltd. All rights reserved. AORTIC DISEASE AORTIC DISEASE Imaging modalities in the evaluation of suspected aortic dissection28,29 Diagnostic performance Sensitivity Specificity Site of intimal tear Presence of thrombus Presence of aortic regurgitation Pericardial effusion Branch vessel involvement Coronary artery involvement Advantages Readily available Quickly performed Bedside evaluation Non-invasive Intravenous contrast Cost Angiography CT ˛˛ ˛˛˛ ˛˛˛ ˛˛˛ ˛˛ ˛ ˛˛˛ ˛˛ ˛˛˛ ˛˛ ˛˛˛ ˛˛ ˛˛ Fairly Quite Fairly Quite No No No Yes Yes Yes High Reasonable MRI Transoesophageal echocardiography ˛˛˛ ˛˛˛ ˛˛˛ ˛˛˛ ˛˛˛ ˛˛ ˛˛˛ ˛ ˛ ˛˛˛ ˛˛˛ ˛˛˛ ˛˛ ˛ ˛˛ Fairly Very Fairly Very No Yes Yes Yes No/yes No Moderate Reasonable ˛˛˛, Excellent; ˛˛, good; ˛, fair; , not detected. Table 2 complete aortic rupture, hypotension (systolic blood pressure 100 mmHg), shock/cardiac tamponade and compromise of vital organs. Limb ischaemia, new neurological deficits and congestive heart failure may further complicate the outcome.15 Preoperative mortality in acute dissection ranges from 3% in stable patients who undergo surgery without delay to 30% or more when evaluation is prolonged or the case more complex. These data reinforce the need for prompt diagnosis and surgical repair to prevent even minimal progression of the dissection with associated complications. The aim of surgery in type A dissection is prevention of rupture or the development of pericardial effusion, which may lead to cardiac tamponade and death. Similarly, the sudden onset of aortic regurgitation or obstructed coronary flow requires ur- gent surgery to resect the region involved with subsequent replacement using a composite or interposition graft (provided the aortic valve leaflets are intact or resuspendable).16 Resection of the entire intimal flap may not be possible if the dissection extends to the aortic arch or descending aorta, when partial or total arch replacement may be necessary. Acute type A dissection in a previously ectatic proximal aorta (usually encountered in Marfan s syndrome) requires a different approach, entailing use of a composite graft (aortic tube graft with integrated valve) and coronary reimplantation.17 Valve-sparing operations are delicate endeavours in the emergency setting and best undertaken by expert operators in specialist centres. Endovascular aortic repair Percutaneous stent-grafts (Figure 2) are an established treatment alternative in type B dissection.18 The aim of the procedure is to close entry to the false lumen using a tube-like nitinol grid covered with a Dacron shell and thereby induce thrombosis of the false lumen with subsequent healing of the aortic wall (Figure 3). Early results describing this approach are encouraging.19,20 Data from the IRAD registry indicate better survival rates in patients after endovascular treatment of complicated type B dissection in comparison with open surgical repair.21,22 Accepted indications for the procedure include vital organ ischaemia, intractable pain, rapid expansion of the false Concept of interventional reconstruction of the dissected descending aorta with sealing of the proximal entries, depressurization of the false lumen and initiation of false lumen thrombosis Proximal entry Perfused false lumen Left carotid artery Left subclavian artery Thrombus Stent-graft MEDICINE 42:9 535 ! 2014 Elsevier Ltd. All rights reserved. Figure2 lumen (diameter > 55 mm), and signs of imminent rupture.3,18 Successful stent-graft treatment of the descending thoracic aorta has also been recently described;23 pre-emptive stent grafting and remodelling lead to long-term prognostic benefits and better 5-year survival than does medical management alone.24,25 The window of opportunity for best endovascular results seems to be 3 months in elective cases of type B dissec- tion.26 Technological advances and increased operator expertise make it likely that endovascular techniques will soon become first-line therapy for most patients presenting with anatomically suitable thoracic and thoraco-abdominal aortic pathology, regardless of initial clinical symptoms. Long-term therapy and follow-up Continued follow-up is essential for medically treated patients and those who have undergone open surgery or endovascular aortic repair. The primary aim of long-term surveillance is maintenance of blood pressure control and early detection of changes or un- stable aortic lesions that require subsequent open surgery or endovascular intervention (e.g. rapid expansion of a new or existing aneurysm, progression or recurrence of dissection, aortic regurgitation, peripheral vascular compromise). CT angiography and/or dynamic MRI are non-invasive and the preferred tech- niques for monitoring these patients e MRI also avoids repeated radiation exposure and provides excellent anatomical detail that may be helpful in evaluating serial changes.27 Patients are at greatest risk immediately following hospital discharge and during the first 2 years thereafter, so early follow-up is warranted; for example, patients may be seen at 3 and 6 months initially, then every 6 months for 2 years, after which the follow-up interval will be determined by individual patient risk. A MEDICINE 42:9 536 ! 2014 Elsevier Ltd. All rights reserved. AORTIC DISEASE Figure 3 CT scan demonstrating a type B aortic dissection in a 48-year-old man; note dynamic obstruction of the true lumen in the acute phase. After stent-graft placement across the proximal thoracic entry site the entire true lumen of the thoracic aorta is reconstructed with complete healing of the dissected aortic wall and shrinking of the completely thrombosed false lumen. TL, true lumen; FL, false lumen; TH, thrombus. MEDICINE 42:9 537 ! 2014 Elsevier Ltd. All rights reserved. AORTIC DISEASE ˇ˛Acute epididymo-orchitis Emma J Street Janet D Wilson Abstract Acute epididymo-orchitis is a clinical syndrome consisting of pain, swelling and inflammation of the epididymis that can extend into the testis. It is primarily caused by local extension of infection either from the urethra or the bladder. In men aged under 35 years, it is mostly caused by sexually transmitted infections (STIs), especially Chlamydia tra- chomatis or Neisseria gonorrhoeae. In men aged over 35 years, STIs can occur but acute epididymo-orchitis is more commonly a complication of a urinary tract infection. Mumps has become an important alternative aeti- ology of epididymo-orchitis following the recent epidemic in non-immune adults in 2005/2006. The main and most important differential diagnosis is torsion of the spermatic cord. Epididymo-orchitis typically presents with subacute onset of unilateral scrotal pain and swelling. Investigations are required for C. trachomatis, N. gonorrhoeae, and urinary tract infections. Management involves scrotal support and non-steroidal anti-inflammatory drugs (NSAIDs). Empirical antibiotics are chosen on the basis of sexual history, near- patient tests if available, urological past history and age. Keywords Chlamydia trachomatis; enteric organisms; epididymitis; epididymo-orchitis; Neisseria gonorrhoeae; orchitis; sexually transmitted infection Aetiology Evidence for the aetiology of epididymo-orchitis is many decades old. In men aged under 35 years, it is usually caused by sexually transmitted bacteria, commonly Chlamydia trachomatis or Neis- seria gonorrhoeae.1,2 In men aged over 35 years, sexually trans- mitted infections (STIs) can occur but acute epididymo-orchitis is more commonly a complication of a urinary tract infection (UTI) caused by a non-sexually transmitted Gram-negative enteric or- ganism such as Escherichia coli.1,2 Epididymo-orchitis in this age group is often a result of functional or anatomical urinary tract abnormalities causing urine to reflux via the ejaculatory ducts and vas deferens, or can follow urinary tract instrumentation. Men engaging in unprotected insertive anal intercourse are also at increased risk of epididymo-orchitis secondary to infection with enteric organisms.3 It is essential to take a good sexual history in all cases as this may guide later therapy. Mumps epididymo-orchitis should be considered an impor- tant alternative aetiology of epididymo-orchitis, particularly in those born in the 1980s.4 The prevalence of mumps remains high Emma J Street MB ChB MRCP is a Consultant in Genitourinary Medicine at Calderdale and Huddersfield NHS Trust, UK. Conflict of interests: none declared. Janet D Wilson MB ChB FRCP is a Consultant in Genitourinary Medicine at Leeds Teaching Hospitals NHS Trust, UK. Conflict of interests: none declared. in this birth cohort due to low uptake of MMR (measles, mumps and rubella) vaccination, despite a reduction since the peak in 2004. This complication can occur in up to 25% of affected post- pubertal males. Mycobacterium tuberculosis is a rarer infective cause of acute epididymo-orchitis, though it usually results in a more chronic presentation. It should be particularly considered in patients from high-prevalence countries, those with previous tuberculosis and patients with immunodeficiency such as HIV.5 Mycoplasma genitalium and Ureaplasma urealyticum have also been impli- cated, though evidence for these as common causes is lacking.6 Non-infective epididymo-orchitis may occur in Behc et s dis- ease7 and with amiodarone treatment.8 The main and most important differential diagnosis of epididymo-orchitis is torsion of the spermatic cord. Clinical features The anatomy of the scrotum is shown in Figure 1. Epididymo- orchitis typically presents with unilateral scrotal pain and swelling of relatively acute onset (Figure 2). Preceding dysuria and urethral discharge may suggest an STI, but such symptoms are often absent. There may be symptoms suggesting the presence of a UTI (e.g. urinary frequency, urgency, dysuria, haematuria). On examination, there is usually tenderness of the affected side, particularly of the epididymis. There is also usually palpable swelling of the epididymis, starting with the tail at the lower pole of the testis and spreading towards the head of the epididymis. There may be generalized swelling of the testicle, with oedema and erythema of the overlying scrotal skin. There may be a visible urethral discharge. Systemic features such as fever and rigors are most commonly seen in Gram-negative in- fections. There may be an associated hydrocoele. With mumps epididymo-orchitis, the patient typically pre- sents with a headache and fever before parotid swelling, fol- lowed 4e8 days later by unilateral testicular/epididymal swelling (bilateral in 15e30%). Scrotal involvement can occur without systemic symptoms. Specific to tuberculosis, the patient typically presents with a subacute or chronic onset of painless/painful scrotal swelling and may also manifest systemic symptoms of tuberculosis, a scrotal sinus or thickened scrotal skin. Diagnosis Epididymo-orchitis is a clinical diagnosis based on symptoms and signs. The history, genitourinary symptoms and risk of STIs BACTERIAL INFECTIONS What s new? C Mumps and tuberculosis are increasingly seen as causes of epididymo-orchitis in the UK C Quinolones are not recommended for treatment of Neisseria gonorrhoeae-associated epididymo-orchitis because of resis- tance. Ceftriaxone should be used where a sexually transmitted pathogen is likely C A high proportion of older men with epididymo-orchitis have functional or anatomical urethral obstruction MEDICINE 42:6 338 ! 2014 Elsevier Ltd. All rights reserved. BACTERIAL INFECTIONS Anatomy of the scrotum Spermatic cord Epididymis Testicle Scrotum Figure 1 From MayoClinic.com. Epididymis. Available from http://www. mayoclinic.org/condition/epididymitis/multimedia/scrotum-testicle-and- epididymis/IMG-20006290. Used with permission of Mayo Foundation for Medical Education and Research, all rights reserved. Investigations The following investigations are recommended1: Urethral microscopy e microscopy of a Gram-stained urethral smear can be used to diagnose both gonococcal and non-specific urethritis. Urethritis is diagnosed by the presence of five or more neutrophils per high-power field ( 1000). The absence of ure- thritis does not exclude the diagnosis of epididymo-orchitis. Typical Gram-negative intracellular diplococci may be seen in N. gonorrhoeae infection. Culture for N. gonorrhoeae is helpful where rapid transport to the laboratory is assured or where the swab can be directly inoculated onto the culture plate, but nucleic acid amplification tests (NAATs) are better tests for excluding gonorrhoea (see Laboratory diagnosis on pages 310e313 of this issue).9 NAATs for C. trachomatis and N. gonorrhoeae e on urine (the first 15e30 ml of urine should be collected after urine has been held for at least one hour) or a urethral swab. Mid-stream urine e for microscopy and culture. Urinalysis e for leucocytes and nitrites. This is particularly helpful for excluding urinary tract infection. If there is any doubt about the risk of testicular torsion urgent surgical exploration of the scrotum should be performed. A colour Doppler ultrasound can be useful in patients in whom there is diagnostic uncertainty. Further investigations that may be considered are tests for other sexually transmitted infections including blood-borne vi- ruses, mumps serology, early morning urine for tuberculosis culture and chest X-ray. All patients with a urinary tract pathogen-confirmed epididymo-orchitis should be referred to an urologist for investigations for structural abnormalities and uri- nary tract obstruction.1 Management Oral and written information and advice should be given. NSAIDs and a scrotal support are recommended for symptom relief. Empirical antibiotics should be started in all patients before NAAT/culture results are available dependent on the most likely causative organism on the basis of immediate tests (urethral smear, urinalysis) as well as age, sexual history (including insertive anal intercourse), any recent urethral instrumentation, catheterization or known urinary tract abnormalities. Choice of antibiotics will depend on local prevalence of gonorrhoea and local resistance patterns6 (Table 1). Partner notification and treatment should be implemented in all patients with a diagnosis of epididymo-orchitis where a sexually transmitted cause is suspected. Figure 2 Right-sided epididymo-orchitis with scrotal swelling and ery- thema of the overlying scrotal skin. alongside examination findings and preliminary investigations will suggest the most likely aetiology and guide empiric antibiotics. Testicular torsion is the main differential diagnosis. Table 1 Empirical antibiotic regimens: C Epididymo-orchitis where sexually transmitted pathogen likely (e.g. age <35 years): ceftriaxone 500 mg IM single dose ˛ doxycycline 100 mg orally 12-hourly for 10e14 days C Epididymo-orchitis where enteric organisms likely (e.g. age !35 years): ofloxacin 200 mg orally 12-hourly for 14 days (this has good anti- chlamydial activity as well) C Epididymo-orchitis where likely aetiology unknown: ceftriaxone 500 mg IM single dose ˛ ofloxacin 200 mg orally 12-hourly for 14 days (there have been no trials of the efficacy of this regimen but it would cover gonorrhoea, chlamydia and most urinary pathogens) MEDICINE 42:6 339 ! 2014 Elsevier Ltd. All rights reserved. Patients should be advised to abstain from sexual intercourse until they and their partner(s) have completed treatment and follow-up. Patients should be reviewed at 3 days. If there is no improve- ment the diagnosis should be re-assessed and therapy re-evaluated. Further follow-up is recommended at 2 weeks to assess compli- ance, partner notification and improvement of symptoms. A BACTERIAL INFECTIONS MEDICINE 42:6 340 ! 2014 Elsevier Ltd. All rights reserved. ˇ˛Acute exacerbations of chronic obstructive pulmonary disease: treatment and prevention Terence Seemungal Jadwiga A Wedzicha Abstract An acute exacerbation of chronic obstructive pulmonary disease (COPD) is sustained worsening of dyspnoea and sputum production in patients with COPD. They may be managed in the community with oral steroids and antibiotics but hospital referral is required where there is doubt about the diagnosis or if there are features of severity such as confu- sion, respiratory distress or haemodynamic instability. Regular review is required as failure to improve should prompt consideration of another diagnosis. In the emergency department, nebulized ≤2 agonists and an- ticholinergic bronchodilators should be given and arterial blood gases assessed. Patients with an arterial pH of 7.35 or less should be assessed for non-invasive ventilation. Patients who are stable and are not in type 2 respiratory failure should be considered for discharge if there is ad- equate home support. Warded patients should be discharged if they are stable for 24 hours and if both patient and doctor are confident that they can manage at home with outpatient follow-up at 4 to 6 weeks. About 25% of COPD patients may not have recovered to baseline lung function at this time. Keywords acute exacerbations; antibiotics; chronic obstructive pulmonary disease; nebulizer; oral steroids; respiratory distress Exacerbations of chronic obstructive pulmonary disease (COPD) are an important cause of morbidity and mortality in the condi- tion, and their incidence increases with its severity. Some patients suffer frequent exacerbations leading to hospital admission, with Terence Seemungal PhD is Senior Lecturer in Chest and Internal Medicine at the University of the West Indies (Trinidad Campus). He qualified from the University of the West Indies and trained in general medicine and chest medicine in Birmingham and London, UK. His research interests include the causes and mechanisms of exacerbations of chronic obstructive pulmonary disease, and their management. Competing interests: Dr Seemungal has accepted funding to attend international meetings from GSK, Astrazeneca, Boehringer Ingelheim, and Pfizer. Jadwiga A Wedzicha MD is Professor of Respiratory Medicine at University College London and Consultant Physician at the Royal Free Hospital, London, UK. She qualified from the University of Oxford and St Bartholomew s Hospital Medical School. Her research interests include exacerbations of chronic obstructive pulmonary disease, home ventilatory support and long-term oxygen therapy. Competing interests: none declared. considerable impact on their quality of life and activities of daily living. COPD exacerbations are associated with physiological deterioration and increased airway inflammatory changes caused by factors such as viruses, bacteria and, possibly, common pol- lutants. Current evidence suggests that appropriate manage- ment and prevention of exacerbations may modify the long-term course of COPD. Aetiology Table 1 lists the common microbial agents associated with exac- erbations of COPD. Respiratory viruses have been shown to cause prolonged exacerbations and are associated with increased lower airway inflammation. There is evidence that chronic infec- tion or prolonged carriage after acute infection may occur with adenovirus and respiratory syncytial virus. The role of bacteria is complex. The common bacteria identified in COPD are: " Streptococcus pneumoniae " Haemophilus influenzae " Moraxella catarrhalis In severe COPD, gram-negative organisms are also of importance, especially Pseudomonas aeruginosa. Patients with bacteria in their sputum are more likely to suffer exacerbations and experience pro- longed recovery and greater severity. Many studies have shown no change in bacterial species during exacerbations, though changes may be seen in the bacterial strains obtained from patients. Diagnosis and differential diagnosis COPD has four principal symptoms: " dyspnoea " cough " sputum volume " sputum purulence. The diagnosis of an acute exacerbation is based on an acute sus- tained deterioration in sputum purulence and increased volume, and dyspnoea.1,2 These may be associated with symptoms of an upper respiratory tract infection (cold or sore throat). The differential diagnosis includes: " pneumonia " pulmonary embolism " pneumothorax " pleural effusion " cardiac disease (congestive cardiac failure, arrhythmias) COPD Micro-organisms associated with COPD exacerbations Viruses Bacteria " Rhinovirus " Coronavirus " Respiratory syncytial virus " Influenza virus " Parainfluenza virus " Non-typable Haemophilus influenzae " Streptococcus pneumoniae " Moraxella catarrhalis " Pseudomonas " Staphylococcus aureus MEDICINE 36:4 223 © 2008 Elsevier Ltd. All rights reserved. Table 1 " rib fracture " inappropriate use of sedatives. The chest radiograph is not required for the diagnosis of a COPD exacerbation but it is useful in the establishment of many of the above. Briefly, patchy mainly air space shadowing chiefly in a lobar distribution is indicative of a community-acquired pneumo- nia, unilateral absence of lung markings at the periphery of the lung field together with a visible lung border in the apex indi- cates a pneumothorax, bat s wing shadowing with cardiomegaly indicates congestive cardiac failure. Where the chest radiograph is equivocal, brain-type natriuretic peptide, a useful marker of cardiac failure, is indicated. A rib fracture may be seen on a radiograph to be coincident with an area of localized chest wall tenderness. The differentiation between a pulmonary embolism and a COPD exacerbation may always be difficult but is more so in severe COPD. In severe COPD, the chest radiograph may show lit- tle more than hyperexpansion with large pulmonary arteries. The ECG shows a right ventricular strain pattern in both conditions. However, a low systolic blood pressure and an inability to elevate the arterial oxygen tension above 8 kPa (60 mmHg) in spite of use of high flow oxygen favours an embolic phenomenon. Severity A staging system for COPD exacerbations has not yet been agreed. However, it is generally accepted that the following features in the history are suggestive of a worse prognosis: " severity of underlying COPD " history of more than 3 exacerbations per year or 1 or more previous hospitalizations for COPD-related illness in the previous year " presence of comorbidities.1 Signs of severity are use of accessory muscles, haemodynamic instability, worsening or new-onset central cyanosis, peripheral oedema, signs of cor pulmonale and, importantly, acute confusion. Community management Initial treatment of an exacerbation of COPD comprises more frequent use of bronchodilators (e.g. salbutamol, ipratropium bromide) with an inhaler or in nebulized form temporarily. Antibiotics are indi- cated in patients with any two of increased dyspnoea, increased spu- tum volume and increased sputum purulence. There are insufficient comparative studies to guide us on the choice of antibiotic; hence clinical judgement according to local sensitivities is suggested. Community-based care is suggested where there is no evi- dence of decompensated type 2 respiratory failure and where there is supportive, usually nurse-led, care. Use of oral corticosteroids has been related to more rapid physiological recovery from exacerbations in the community and improved lung function, and may decrease the risk of relapse.3,4 The mechanism of these effects in COPD is largely unknown, and relatively short courses (about 2 weeks) are effective. However, the beneficial effects must be balanced against the side effects. Current guidelines suggest that oral corticosteroids (prednisolone, 30 40 mg once daily) should be used, but in patients already tak- ing oral corticosteroids, the dose is reduced over a longer period. Failure to improve should prompt consideration of an alterna- tive diagnosis but it is also to be noted that a significant proportion of patients may not recover completely even after 1 month of treatment, and the authors suggest a review of all patients treated in the community regularly and at 2 weeks after diagnosis. Referral to tertiary centre Regular assessment is required after initiation of therapy in the community. Referral to hospital should be considered when there is no improvement or the patient deteriorates. Referral should also be considered for those with a severe exacerbation (see above) and when any of the following criteria apply: " presence of a new cardiac arrhythmia " insufficient home support " elderly patient " uncertain diagnosis. Hospital management Accident and emergency department Initial management rapid management of these patients may prevent complications requiring intensive care. Airway, breathing and circulation should be assessed immediately on arrival. Controlled oxygen should be given via a Venturi mask to patients with an arterial oxygen saturation (SaO2) of less than 90% or an arterial partial pressure of oxygen (PaO2) of less than 8 kPa. Because carbon dioxide retention occurs insidiously with rising PaO2 in these patients, the aim is to maintain SaO2 at 92 94%. Arterial blood gases should be rechecked at 30 minutes to assess the pH. While this is performed, a postero-anterior chest radiograph and ECG are requested. Spirometry may be helpful; a forced expiratory volume in 1 second (FEV1) of less than 1.0 litre is considered an index of a severe exacerbation. Haemoglobin concentration and blood potassium levels should be measured. While the patient is in the A&E department, nebulized bron- chodilator therapy with both ≤2-agonists and anticholinergic agents should be started, oral corticosteroids administered and aminophylline considered. Most studies of intravenous amino- phylline have demonstrated only minor benefits in COPD exac- erbations, though the sample size in these studies has been relatively small. Antibiotics may be prescribed as above. Further treatment early discharge from the A&E depart- ment with out-patient follow-up may be considered in patients who suffer an uncomplicated exacerbation without severe symp- toms, in those with no respiratory failure, and when adequate home and community nurse support is available. Otherwise, the patient should be admitted to the ward. Non-invasive positive-pressure ventilation (NIV) should be considered in patients with an arterial pH of less than 7.35 on max- imal medical therapy. NIV has been shown to improve mortality and reduce the need for intubation and duration of hospital stay in these patients. In the absence of NIV or delay in its application, doxapram (a respiratory stimulant that may be helpful in acute respiratory failure) may be considered, with cardiac monitoring. A Cochrane Review suggests that doxapram may improve blood gas exchange in the short term .5 Patient monitoring is crucial at this stage because mortality increases once the pH declines below 7.26; ideally, patients should be cared for in a high-dependency unit. Referral to an intensive care unit is required for those who deteriorate further with increasing hypoxia and worsening pH. MEDICINE 36:4 224 © 2008 Elsevier Ltd. All rights reserved. COPD In-hospital management: continued monitoring of arterial blood gases is required while respiratory failure is still present, and attention to fluid status is vital. In-hospital treatment of acute COPD exacerbations with antibiotics is associated with reduced treatment failures and mortality. Fluid management and fluid balance are essential. Deep vein thrombosis prophylaxis is con- sidered in the immobilized, polycythaemic or dehydrated patient. Physiotherapy to induce sputum clearance is recommended. As the patient s symptoms improve, medication should be tapered back to the levels used before admission. Spirometry is used to confirm the severity of disease before discharge. Serial peak flow measurements are not helpful during exacerbations; they may vary little, and the median decrease at exacerbation is small. Discharge the median duration of stay in hospital is about 1 week. Discharge is considered when: " the patient has been clinically stable for 24 hours, and under- stands how to administer his or her treatment " home and follow-up care arrangements are completed " the patient, family and physician are confident that the patient can manage successfully at home. Out-patient follow-up Hospital assessment at 4 6 weeks is recommended for all patients, to consider their coping skills and measure FEV1. About 75% of patients have fully recovered by this time. Inhaler technique and the long- term need for nebulizers are assessed, and the patient may be given advice on how to recognize symptoms of a COPD exacerbation.6 There is evidence that an appropriate home-care package with spe- cialist nurse advice in the outpatient setting reduces readmission.1,2 Prevention of exacerbations Frequent COPD exacerbations are associated with poor quality-of- life scores, greater levels of hospitalization, greater health burden, more rapid decline in lung function and greater airway inflam- mation. Figure 1 shows the significant proportion of the quality- of-life scores accounted for an exacerbation frequency of greater than 3 per year. Measures to reduce the risk of exacerbation are therefore important. Viral infections are important possible therapeutic targets, though there are currently few appropriate interventions. Influenza vaccine is strongly recommended for all COPD patients, and patients are advised to avoid upper respira- tory tract infections. The pneumococcal vaccine, though recom- mended for patients with chronic lung disease, has been shown to be effective only in some subsets of COPD patients in randomized controlled studies. Systematic reviews have not supported the use of oral bacterial extracts as immunostimulants in COPD. Both inhaled corticosteroids and long-acting anticholinergic agents have been shown to reduce the frequency of exacerba- tions. Mucolytic agents are used in several countires world-wide but the current weight of evidence does not support their use in exacerbation prevention. Several non-pharmacologica therapies have been shown to be useful in exacerbation prevention and, of these, pulmonary rehabilitation has received much attention. Both long-term oxygen therapy and non-invasive ventilation reduce COPD-related hospitalization. The proportion of the variability in quality of life measured by the total St George s Respiratory questionnaire total score explained by variables explored in the London COPD Study 14% 6% 7% 19% 54% Ex Freq Age MRC Grade Other Wheeze Source: Seemungal T, Donaldson GC, Paul EA, Bestall JC, Jeffries DJ, Wedzicha JA. Effect of exacerbation on quality of life in patients with chronic obstructive pulmonary disease. Am J Respir Crit Care Med 1998; 157: 1418 22. Practice points " Acute exacerbations of COPD are defined by symptoms " Cardiac failure is an important differential diagnosis " Antibiotics are indicated when there are sputum changes " Oral steroids hasten recovery and decrease the risk of relapse " Patients with acute exacerbations treated in the emergency department must be treated proactively to prevent rapid deterioration ˇ˛Abstract European Journal of Internal Medicine 16 (2005) 209 210 Brief report Acute hepatitis related to prednisolone www.elsevier.com/locate/ejim Herman M.A. Hofstee*, Prabath W.B. Nanayakkara, Coen D.A. Stehouwer Department of Internal Medicine, VU University Medical Center, P.O. Box 7057, 1007 MB Amsterdam, The Netherlands Received 19 July 2004; accepted 5 October 2004 A 46-year-old woman, known to have multiple sclerosis (MS), presented with repeated episodes of elevated liver enzymes due to high- dose methylprednisolone therapy. D 2005 Elsevier B.V. All rights reserved. Keywords: Hepatitis; Liver enzymes; Methylprednisolone; Corticosteroids 1. Introduction Acute hepatitis and even some fatal cases of liver failure related to prednisolone therapy have been described [1 3]. Although no other cause for the hepatitis could be identified in these cases, no dechallenge rechallenge relationship between prednisolone and acute hepatitis has been described thus far. By chance, we saw a woman with episodes of elevated liver enzymes after repeated high doses of methylprednisolone. 2. Case report A 46-year-old woman diagnosed as having multiple sclerosis (MS) since 1993 and with normal liver enzymes at the time of diagnosis received her first course of high-dose methylprednisolone (500 mg i.v. for 5 days) in 1994 because of an exacerbation. In 1996, she had her second exacerbation and was treated again with high-dose methylprednisolone (1000 mg i.v. for 3 days). The response was less than expected and a trial of interferon-h was planned. However, the interferon treatment was cancelled due to the fact that the patient s liver enzymes, which were measured according to a * Corresponding author. Tel.: +31 204444444x98360; fax: +31 204440505. E-mail address: hma.hofstee@vumc.nl (H.M.A. Hofstee). 0953-6205/$ - see front matter D 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.ejim.2004.10.018 pre-treatment protocol 6 weeks after prednisolone therapy, were elevated (g glutamyltransferase 156 U/l, aspartate aminotransferase 755 U/l, alanine aminotransferase 1095 U/ l, alkaline phosphatase 140 U/l, lactate dehydrogenase 473 U/ l). The liver enzymes eventually returned to normal after 4 months, but a relationship between prednisolone and acute hepatitis was not recognized. 2500 2000 1500 1000 500 0 ASAT ALAT Fig. 1. Relation between transaminases and methylprednisolone (arrow indicates course of methylprednisolone). 6/1/1993 10/6/1996 17-6-1996 15-7-1996 2/9/1996 27-5-1998 24-6-1998 22-7-1998 22-5-2003 5/6/2003 1/7/2003 4/8/2003 210 H.M.A. Hofstee et al. / European Journal of Internal Medicine 16 (2005) 209 210 In 1998 and 2003, the patient received two other courses of high-dose methylprednisolone (1000 mg for 3 days), which again were accompanied by a rise in her liver enzymes after a few weeks and a normalization of the values within 4 months (Fig. 1). During all of the episodes, the patient refrained from taking any prescribed or recreative hepatotoxic drugs; she also denied any alcohol use. Repeated serological tests for hepatitis (hepatitis A, B, C), hepatropic viruses (CMV, EBV) and syphilis were negative. Smooth muscle, mitochondrial, liver membrane and nuclear autoantibodies were not detected. There were no signs of storage diseases (iron, copper). Repeated ultrasound scans of the liver and bile ducts were normal. A liver biopsy was not performed. Finally, after 5 years, a relationship between methylprednisolone and hepatitis was recognized and future prednisolone administration was stopped. 3. Discussion We report a case involving episodes of hepatitis after repeated courses of high-dose prednisolone therapy. In all three episodes, in spite of extensive evaluation, we could not find an alternative explanation for this phenomenon. To date, a dechallenge rechallenge relationship between pre- dnisolone and hepatitis has not been described. This patient probably had an idiosyncratic reaction to methylpredniso- lone; fortunately, repeated exposure did not leave any permanent sequelae. Although relatively rare, it is important to highlight this potentially lethal side effect because of the widespread use of prednisolone. ÔªøEuropean Journal of Internal Medicine 17 (2006) 217 ‚Äì 219 www.elsevier.com/locate/ejim Brief report Acute interstitial nephritis associated with salmonellosis Jo Caers a,*, Patrick Peeters b, Kaat Vanden Houte c, Jacques Sennesael b, Patricia Van der Niepen b, Dierik Verbeelen b a Department of Internal Medicine, Academic Hospital-Vrije Universiteit Brussel, Laarbeeklaan 101, 1090 Brussels, Belgium b Department of Nephrology, Academic Hospital-Vrije Universiteit Brussel, Laarbeeklaan 101, 1090 Brussels, Belgium c Department of Clinical Pathology, Academic Hospital-Vrije Universiteit Brussel, Laarbeeklaan 101, 1090 Brussels, Belgium Received 17 June 2005; received in revised form 21 September 2005; accepted 30 September 2005 Abstract Acute interstitial nephritis (AIN) is a common cause of acute renal failure. We report a case of AIN, confirmed by renal biopsy, that developed in a patient with typhoid fever due to a Salmonella hadar infection. AIN secondary to Salmonella infection is a rare complication that has only been described twice in the literature. Salmonella should be added to the list of possible causes of AIN. D 2006 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. Keywords: Acute interstitial nephritis; Salmonellosis; Corticosteroid therapy 1. Introduction Acute interstitial nephritis (AIN) was initially described as a post-infectious acute inflammatory renal disease associated with diphtheria and scarlet fever and characterized by cellular infiltration and fluid exudate within the renal interstitium. The introduction of effective antibiotic therapy decreased infection-related cases, and nowadays drugs are a common cause of interstitial nephritis. In some cases, it may be difficult to assess whether the nephritis may be attributed to the infection itself or to the drug given to treat the infection. AIN may also occur idiopathically or in association with malignancies and systemic diseases [1]. Our group recently reported that acute renal failure is observed more frequently in salmonellosis than in other causes of infectious diarrhea and is possibly caused by a toxic injury [2]. The spectrum of renal complications in typhoid fever includes mild to severe glomerular involve * Corresponding author. Vrije Universiteit Brussel, Department of Haematology and Immunology, Laarbeeklaan 103, B-1090 Brussels, Belgium. Tel.: +32 2 477 44 04; fax: +32 2 477 44 05. E-mail address: jcaers@vub.ac.be (J. Caers). ment, acute renal failure due to rhabdomyolysis, acute tubular necrosis, or interstitial nephritis. Salmonellosis leading to AIN has been described twice in the literature [3,4]. However, in these reports, the tubulointerstitial nephritis was diagnosed on clinical grounds and no biopsy was performed to confirm the diagnosis. This is the first report of salmonellosis associated with biopsy-proven AIN in adults. 2. Case report A 72-year-old man was admitted to a Moroccan hospital with a history of fever, malaise, vomiting, and diarrhea. Physical examination showed severe dehydration, a fever of 40-C, hypotension, and hyperperistalsis. On admission, hemoglobin was 12.8g/dl, hematocrit 31.2%, white blood cell count 12,500/mm3 with 74.4% granulocytes, and platelet count 99,000/mm3. Erythrocyte sedimentation rate was 38mm/h, PT 60%, and aPTT 37s. Serum sodium was 135mEq/l, potassium 4.2mEq/l, ureum 70mg/dl, and creatinine 2.12mg/dl. The patient was rehydrated and treated with ciprofloxacin for a presumed gastroenteritis. Initially, his fever dropped, but diarrhea persisted and on the 0953-6205/$ -see front matter D 2006 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.ejim.2005.09.026 218 J. Caers et al. / European Journal of Internal Medicine 17 (2006) 217 ‚Äì 219 5th day fever relapsed. Ciprofloxacin therapy was stopped and intravenous antibiotherapy with ceftriaxone was started. At this time, ureum was 180mg/dl and creatinine 7.0mg/dl. Aspartate aminotransferase (AST) and alanine aminotransferase (ALT) were 59U/l (N <37U/l) and 40U/l (N <40U/l), respectively. Platelet count remained low (75,000/mm3) and hemoglobin level dropped to 10.7g/dl. A renal ultrasound showed kidneys with a length of 10.7cm without evidence of hydronephrosis. On the 10th day, the patient was transferred to our institution. On admission to our hospital, he had 57,000/ mm3 platelets, 1.325 103 neutrophils and a hemoglobin of 10.6g/dl. Coagulation assays had normalized and serum creatinine was 7.43mg/dl. Serum lactic acid dehydrogenase (LDH), alkaline phosphatase, and gamma-glutamyltransferase were 853U/l (N 241‚Äì549U/l), 133U/l (N 43‚Äì114U/l), and 112U/l (N 13‚Äì73U/l), respectively. Serum transaminases were within normal limits. Proteinuria was 0.41g/L. There was microscopic hematuria and sterile pyuria; eosinophils were absent. A kidney biopsy, performed on day 14, revealed an interstitial edema and lymphocytic infiltrate with plasma cells and rare granulocytes (Fig. 1). Eosinophils were absent. The glomeruli appeared enlarged, while renal vessels were normal. Immunofluorescence showed the absence of immunoglobulin deposits. No seroconversion was seen for the following pathogens: hepatitis A, B or C, Yersinia, human immunodeficiency virus, Cytomegalovirus, Epstein‚ÄìBarr virus, toxoplasmosis, Legionella, or syphilis. Markers of systemic disease (antinuclear factor, ANCA, Waaler Rose) were negative. Fecal cultures, taken in our hospital, showed the presence of Salmonella hadar infection. A diagnosis of AIN secondary to Salmonella infection was made. The patient was treated with methylprednisolone 32mg once daily for 7days, with gradual tapering. Serum Fig.1.Kidneybiopsyrevealinganinterstitialedemaandlymphocyticinfiltratewithnumerousplasmacellsandraregranulocytes.Thetubulesshowcysticdilatationandattenuationoftheepithelium.Focally,theinflammatorycellsinfiltratethetubuleswithpartialdestructionoftheepithelium(PAS,25). creatinine improved to 2.96mg/dl on day 35 and the patient could be discharged. When seen at follow-up 6months later and after completing the steroid treatment, the serum creatinine level had almost normalized (1.45mg/dl). 3. Discussion Acute renal failure is caused in 2‚Äì3% of the cases by AIN. In kidney biopsy series of acute renal failure, AIN ranges from 3% to 15%. It is likely that this presence is substantially underestimated because the disease is often mild and reversible and, therefore, under-recognized. As mentioned earlier, deterioration of renal function is reported to occur more frequently in typhoid fever than in other causes of diarrhea. In our population, 36% of all patients with salmonellosis had a temporary decrease in kidney function [2]. The patient described in this report presented with fever and diarrhea and developed acute, but prolonged, renal failure. Renal biopsy showed characteristics of AIN. This patient presented with typical manifestations of a typhoid syndrome (high fever, diarrhea, dehydration, the development of anemia, transient leukocytosis followed by leukopenia and/or neutropenia, disturbed liver and coagulation tests), while fecal cultures showed the presence S. hadar. This clinical picture convinced us that the renal involvement must be seen as a complication of the primary infection. However, the contemporaneous antibiotic therapy (in this case ciprofloxacin) confuses the etiology. Ciprofloxacin has previously been described as a cause of AIN, elicited by a hypersensitivity reaction [5]. In our patient, different clinical arguments plead against a drug-induced AIN. First, renal function was disturbed before any drug was given. Second, there were no other signs of hypersensitivity, such as rash or myalgia. Third, in blood, urine, and renal biopsies, eosinophils were absent. Finally, it seems unlikely that the antibiotic therapy disturbed liver, kidney, bone marrow, and coagulation function at the same time. In most cases of AIN, renal function recovers spontaneously. Management of this disorder includes stopping all suspected drugs and appropriate treatment of the underlying infection. When recovery is slow or incomplete, renal replacement therapy may be necessary. Some investigators have advocated corticosteroid therapy for patients requiring hemodialysis or with prolonged renal insufficiency. However, there is no controlled study available defining the place of corticosteroids. It appears that corticosteroids might lead to a more rapid remission and more complete resolution of AIN, especially in patients who fail to show early recovery [6]. Therefore, for those with slow or absent recovery, current evidence supports the use of a short course of oral prednisolone (40‚Äì60mg/day) for 1week, followed by a rapid taper. Treatment may be delayed for 10days in order to allow a period for spontaneous recovery. In our case, the decision to start steroids was based on the J. Caers et al. / European Journal of Internal Medicine 17 (2006) 217 ‚Äì 219 219 prolonged renal failure and the tendency towards progressive renal failure. While a drop in serum creatinine was seen with the initiation of methylprednisolone, this may have been a natural consequence of treating the infection. ˇ˛Acute leukaemias Charles Craddock Abstract acute myeloid leukaemia (amL) and acute lymphoblastic leukaemia (aLL) are clonal malignancies of haematopoietic stem and progenitor cells. in recent years considerable progress has been made in defining the molecular basis of leukaemogenesis. These basic scientific insights have allowed the development of accurate risk stratification in both diseases, which can be used to inform therapeutic decision making. However, whilst there have been considerable advances in the treatment of children, the majority of adults with both amL and aLL are still des- tined to die of resistant disease. allogeneic stem cell transplantation is increasingly recognized as the most effective anti-leukaemic therapy in both diseases. Strategies which increase the number of patients eligible for allografting, whether this be the development of novel conditioning regimens or the increased availability of umbilical cord blood as a stem cell source, are therefore critical if the outcome of adults with acute leukaemia is to be improved. Keywords acute lymphoblastic leukaemia; acute myeloid leukaemia; allogeneic stem cell transplantation; graft-versus-leukaemia effect; leukaemic stem cell; minimal residual disease Acute myeloid leukaemia (AML) and acute lymphoblastic leukae- mia (ALL) typically present with bone marrow failure (see Prac- tice points). Until 40 years ago both diseases were, almost without exception, rapidly fatal. Progress in the treatment of ALL and AML has been based on the integration of molecular and cytogenetic abnormalities into a risk stratification score and the development of methodologies allowing the quantitation of minimal residual disease (MRD). Coupled with the development of combination chemother- apy regimens and advances in allogeneic stem cell transplantation, these developments have led to considerable improvement in the outcome of children and younger adults with acute leukaemia. Curative options still, however, remain limited in older patients. Epidemiology and pathogenesis of acute leukaemias The incidence of AML increases with patient age. In children and adults under the age of 50 the incidence is in the region of 3:100,000, but rises rapidly with increasing age and approaches 20:100,000 in adults in their eighth decade. ALL is less common Charles Craddock BBM BCh DPhil FRCP FRCPath is Director of the BMT Unit at the Queen Elizabeth Hospital, Birmingham, UK. Competing interest: none declared. than AML and is characterized by a bimodal incidence with one peak in children aged 4 14 years and another in adults over 50 years of age.1 The great majority of cases of both AML and ALL appear to arise sporadically, and no precipitating factors can be identified. However, epidemiological studies have identified a number of congenital syndromes and environmental factors which appear to predispose towards the development of acute leukaemia (Table 1). It is now proposed that AML occurs as the result of the acqui- sition within haematopoietic stem/progenitor cells of mutations in two distinct classes.2,3 Class I mutations result in an abnor- mal proliferative signal and typically occur as mutations in genes such as the receptor tyrosine kinase such as FLT3 or the RAS oncogene. Class II mutations result in a block in differentiation of leukaemic progenitors and are typically caused by non-ran- dom chromosomal translocations that result in the generation of mutated transcription factors and the inhibition of haematopoi- etic gene expression. Clinical presentation of acute leukaemias The majority of patients with acute leukaemia present with symptoms consequent upon bone marrow failure. These include Leukaemia What s new? " acquired mutations in genes such as FLT3 and NPm1 are of prognostic significance in patients with amL " aTRa improves survival in patients with aPmL " Demethylating agents such as azacitidine demonstrate activity in amL " allogeneic stem cell transplantation increases disease-free and overall survival in amL and aLL " Results with reduced intensity conditioning regimens in older patients with high-risk amL are promising Predisposing factors for the development of acute leukaemias Genetic disorders " Down s syndrome " Fanconi anaemia " Neurofibromatosis " kostman syndrome Chemical exposure " Benzene " Pesticides " Cigarette smoking Previous chemotherapy " Topoisomerase-ii inhibitors " alkylating agents, e.g. melphalan meDiCiNe 37:4 190 © 2009 elsevier Ltd. all rights reserved. Table 1 fatigue, dyspnoea or, in patients with underlying ischaemic heart disease, angina. Thrombocytopenia can result in bruising or mucosal bleeding and neutropenia results in fever consequent upon bacteraemia. Alternative presentations in AML include gum infiltrates, often associated with monocytic leukaemia, and bleeding associated with disseminated intravascular coagulation (DIC) a common complication of acute promyelocytic leukae- mia (APML). In ALL symptoms related to bone marrow failure are the most common presenting features, but up to 50% of patients can present with lymphadenopathy, splenomegaly or in patients with T cell ALL, a mediastinal mass. In contrast to AML, where central nervous system (CNS) involvement is very rare, approximately 10% of newly diagnosed patients with ALL have evidence of CNS involvement as demonstrated by the presence of leukaemic blasts in the cerebrospinal fluid (CSF). However, pre- sentation with CNS-related symptoms such as headache, vomit- ing or cranial nerve palsies is rare. In both AML and ALL, a small proportion of patients present with hyperleucocytosis defined as a presentation white blood cell (WBC) count >100 ◊ 109/ litre. This can result in dyspnoea or CNS symptoms such as con- fusion or obtundation. In patients with such symptoms, urgent leukapharesis is indicated. Diagnosis In both AML and ALL circulating blasts may be seen in the peripheral blood at presentation. A definitive diagnosis is made by bone marrow aspirate and in the rare patient with a dry tap (no aspirable bone marrow cells), by bone marrow trephine. Both myeloblasts and lymphoblasts are characterized by a high nuclear:cytoplasmic ratio and the presence of prominent nucle- oli (Figure 1). Myeloblasts sometimes contain azurophilic Auer rods which are pathognomonic of AML. Confident discrimina- tion between AML and ALL is dependent on immunophenotypic analysis, which demonstrates characteristic patterns of antigen expression on the surface of myeloid and lymphoid blasts. Diag- nostic work up of all acute leukaemias should also include cyto- genetic and molecular analysis. Figure 1 Leukaemic blasts in the bone marrow of patients with a newly diagnosed patient with amL. Note the azurophilic auer rod a pathognomonic feature of amL. Predictors of outcome in acute leukaemias Whilst treatment options for patients with AML and ALL have undoubtedly improved, it remains the case that the majority of adults are destined to relapse and die of resistant leukaemia. The identification, at presentation or shortly after commencement of treatment, of risk factors that predict outcome is therefore of paramount importance in identifying patients in whom treatment should be intensified. At the same time it is increasingly recog- nized that intensive chemotherapy may be inappropriate in older patients whose outcome with conventional chemotherapy would be poor. The most important predictors of outcome in both AML and ALL are presentation karyotype, age and response to initial ther- apy (Table 2). In AML presentation cytogenetics can be used to define three prognostic subgroups.4 In the good risk group, which includes patients with the chromosomal translocations t(15;17), t(8;21) and inversion 16, survival rates in the region of 60% are achievable in younger patients with intensive chemo- therapy. This contrasts with survival rates of less than 20% in patients with an adverse risk karyotype, characterized by abnor- malities of chromosome 5 or 7, an abnormality of chromosome 3 or a complex karyotype (defined as the presence of more than 4 cytogenetic abnormalities). Patients with normal cytogenetics form an intermediate risk group. In patients with normal cytoge- netics it is now apparent that mutations in genes such as FLT3, NPM1 and CEBPA also predict the likelihood of achieving a dura- ble response to chemotherapy5 and it is clear that increasingly sophisticated methods of molecular subclassification, such as microarray analysis, will become important in risk stratification Leukaemia Predictors of outcome in AML and ALL Risk factor age Presentation cytogenetics molecular profile initial response to chemotherapy Presentation white count Presence of co-morbidities Impact on overall survival Overall survival decreases with increasing age Good and adverse risk cytogenetic groups identifiable in amL and aLL mutations in genes such as FLT3 and NPm1 predict overall survival in amL BCR-aBL gene associated with decreased overall survival in aLL Failure to achieve complete remission associated with decreased overall survival in amL and aLL WBC >30 ◊ 109/litre and WBC >100 ◊ 109/litre associated with decreased overall survival in aLL and amL, respectively Concurrent cardiac or respiratory disease associated with decreased overall survival meDiCiNe 37:4 191 © 2009 elsevier Ltd. all rights reserved. Table 2 in the future.6 In ALL the presence of the Philadelphia chromo- some caused by a t(9;22) translocation predicts a high relapse rate and a less than 10% chance of overall survival in patients treated with chemotherapy alone.7 The prospect of long-term survival in adults with AML and ALL declines with increasing age. This is likely to be explained by changes in underlying disease biology as patients age, with older patients more likely to present with adverse risk cytogenetics.8 In older patients, it is increasingly recognized that chronological age is a poor predictor of a patient s ability to tolerate intensive chemotherapy and the concept of a frailty index has been devel- oped by a number of groups in order to help identify patients in whom intensive chemotherapy is likely to be poorly tolerated. The response to the first course of induction chemotherapy is a strong predictor of long-term survival in both AML and ALL. In patients with AML, the presence of >15% blasts after their first course of induction chemotherapy is associated with a much poorer outcome.9 Sensitive technologies, utilizing immunophe- notypic or molecular methods to detect MRD in patients who have achieved a morphological remission, are increasingly used for risk stratification.10 PCR-based strategies can now detect MRD with a sensitivity in the region of 1:104 cells in ALL and now play a central role in the identification of patients whose outcome with chemotherapy will be poor and in whom allogeneic stem cell transplantation should be performed if possible. Principles of treatment using intensive chemotherapy in acute leukaemia The general principle of treatment in both AML and ALL is to first achieve a morphological complete remission (CR) defined as the presence of fewer than 5% leukaemic blasts in the bone marrow accompanied by a normal neutrophil and platelet count using induction chemotherapy. This is followed by further blocks of intensive chemotherapy as consolidation therapy. Patients with ALL receive less intensive oral maintenance chemotherapy for up to 2 years. It is increasingly recognized that patients deemed to be at a high risk of relapse require early allogeneic stem cell transplantation in order to maximize delivery of anti-leukaemic chemotherapy. Treatment of AML In adults with newly diagnosed AML, 70 90% will achieve CR if treated with one or two courses of induction chemotherapy utilizing cytosine arabinoside and an anthracycline. Patients with good risk cytogenetics receive a further two courses of consolidation chemotherapy incorporating cytosine arabinoside, an anthracycline and etoposide. In patients with intermediate or adverse risk cytogenetics, patients with an human leukocyte antigen (HLA)-identical sibling under the age of 40 years of age are considered for allogeneic transplantation, given the substan- tial risk of disease relapse if these patients are treated with che- motherapy alone. One area of major advance has been the treatment of APML, where the use of all-trans-retinoic acid (ATRA) has dramati- cally improved the outcome of this molecular subgroup. ATRA works by stimulating differentiation of the leukaemic blasts and serves both to treat bleeding complications at presentation caused by DIC and, when administered with anthracycline che- motherapy, dramatically reduces the risk of disease relapse. As a consequence, disease-free survival rates in excess of 80% can now be reliably achieved in newly diagnosed patients with APML who are treated with a combination of ATRA and anthracycline chemotherapy. Treatment of ALL Induction chemotherapy regimens in ALL primarily utilize vin- cristine and corticosteroids and consolidation regimens add cyclophosphamide, etoposide and asparaginase. In patients with Philadelphia (Ph) positive ALL concurrent treatment with the tyro- sine kinase inhibitor imatinib may improve survival. Given the high incidence of CNS disease at presentation, CNS-directed treatment is a central feature of ALL treatment regimens. This takes the form of intrathecal methotrexate and systemic high-dose methotrexate followed by cranial irradiation in patients for whom transplanta- tion is not planned. Patients under the age of 45 with an available HLA identical sibling are considered for allogeneic transplant, and in patients with a high white count or adverse cytogenetics, an unrelated donor transplant should be considered. Stem cell transplantation Acute leukaemias are the most common indication for alloge- neic stem cell transplantation worldwide. Improved supportive care, the increased availability of matched unrelated donors and the growing realization of a potent graft-versus-leukaemia (GVL) effect, which can be exploited in older patients using reduced intensity conditioning regimens have led to a very substantial increase in allograft activity, which looks set to continue in the coming decade. In contrast, autologous stem cell transplantation, which was once commonly performed in AML, is now indicated only in a minority of patients in second CR. Transplantation in AML Allogeneic stem cell transplantation using a myeloablative con- ditioning regimen represents the most effective form of anti-leu- kaemic therapy in AML. This is consequent upon the curative potential of high-dose chemoradiotherapy and the genesis of an immunologically mediated GVL effect exerted by donor T lymphocytes. Patients in first CR demonstrate improved rates of disease-free and overall survival after transplantation from an HLA-identical sibling compared with patients treated with chemotherapy alone.11 However, the toxicity of myeloablative conditioning regimens (which typically incorporate high-dose cyclophosphamide and fractionated total body irradiation) increases with patient age and as a consequence the benefit of transplantation using an ablative preparative regimen is largely limited to patients under the age of 40 years. Donor versus no-donor analyses have also shown that only patients in first CR with intermediate or adverse risk cytogenetics benefit from a sibling transplant. Because of their extremely poor outcome with conventional chemotherapy, patients with adverse risk cytoge- netics and a suitable unrelated donor should also be considered for a myeloablative allogeneic transplant in first CR. Patients with favourable cytogenetics should not be transplanted in first CR. Stem cell transplantation is also indicated in relapsed patients who achieve a second CR with salvage chemotherapy. In most of these patients, allogeneic transplantation using a sibling or matched unrelated donor will be the treatment of choice, but meDiCiNe 37:4 192 © 2009 elsevier Ltd. all rights reserved. Leukaemia in patients with APML and t(8;21) and inversion (16), a second CR autologous transplantation can result in impressive rates of disease-free survival.12,13 Recent data also suggest that younger patients with primary refractory AML may benefit from a sibling or unrelated donor allograft and allografting should be consid- ered in all patients who fail to achieve a CR after two courses of REFEREnCES Leukaemia blood as a source of allogeneic stem cells, also hold considerable hope as a strategy for improving patient outcome.19 ∆% induction chemotherapy.14 Until recently allogeneic transplantation was restricted to patients with AML under the age of 45 50 years because of the excessive toxicity associated with the use of a myeloabla- tive conditioning regimen. The recent demonstration that the use of reduced intensity, often non-myeloablative, conditioning regimens secures donor engraftment with markedly reduced tox- icity has permitted the extension of a potentially curative GVL effect to older patients who would previously have been consid- ered untransplantable on the grounds of age or co-morbidity.15 Encouraging Phase II data now demonstrates survival rates in the region of 50% for patients with high-risk AML in first or second CR using reduced-intensity preparative regimens.16 Transplantation in ALL Transplantation using an HLA-identical sibling improves over- all survival and disease-free survival in adults with ALL.17 In addition, there is now a compelling case for performing a mye- loablative unrelated donor transplant in patients with features at presentation that predict a poor outcome if treated with che- motherapy alone. These include a high WBC at presentation, unfavourable cytogenetics or evidence of MRD after induction chemotherapy. As observed in AML the toxicity of myeloabla- tive conditioning regimens limits the benefit of allogeneic trans- plantation to younger patients. There is therefore interest in the possibility that in ALL, as in AML, reduced-intensity condition- ing regimens may allow a potentially curative GVL effect to be extended to older patients whose outcome with conventional chemotherapy would be poor. Future advances The characterization of dysregulated signalling pathways in both AML and ALL has led to the development of molecularly tar- geted therapies. These include FLT3 inhibitors, which appear to have little anti-leukaemic activity as monotherapy in AML but are currently being studied in conjunction with conventional chemotherapy, and imatinib, which has significant activity in Ph positive ALL. Recent Phase II studies have also demonstrated significant activity of epigenetic agents such as demethylating agents and histone deacetylase inhibitors in AML, particularly when these agents are administered in combination.18 The challenge for the coming decade will be to identify whether these agents can deliver durable remissions. The encouraging Phase II data reported from a number of groups using reduced-intensity conditioning regimens suggest that this will be the other major area of therapeutic advance in AML. The most promising chemotherapy developments in the treatment of ALL are the use of monoclonal antibodies such as anti-CD20 and the optimization of strategies for the deliv- ery of standard drugs such as corticosteroids and asparaginase. Developments in allogeneic transplantation, particularly the use of reduced-intensity conditioning regimens and umbilical cord Leukaemia Practice points " acute leukaemias typically present with symptoms or signs consequent upon bone marrow failure " age, presentation cytogenetics and initial response to chemotherapy are important predictors of outcome in patients with amL and aLL " immunophenotypic or molecular methods of mRD detection assists in disease stratification " allogeneic stem cell transplantation using an HLa-identical sibling improves disease-free survival in patients with amL associated with intermediate or adverse risk cytogenetics " allogeneic stem cell transplantation using an HLa-identical sibling improves disease-free and overall survival in adults with aLL " unrelated-donor allogeneic stem cell transplantation is indicated in patients with high-risk aLL defined by high presentation WBC, cytogenetics or mRD status meDiCiNe 37:4 194 © 2009 elsevier Ltd. all rights reserved. ˇ˛European Journal of Internal Medicine 18 (2007) 166 Letter to the Editor www.elsevier.com/locate/ejim Acute myocardial infarction during intravenous immunoglobulin (IVIG) therapy T. Eliasberg, W.R. SalibaN , M. Elias Department of Internal Medicine C, Ha'emeK Medical Center, Afula 18101, Affiliated to the Technion-Israel Institute of Technology, Faculty of Medicine, Haifa, Israel Received 8 August 2006; accepted 19 September 2006 Keywords: Immunoglobulin; IVIG; Myocardial infarction Immunoglobulins (IVIG) are often used in the treatment of patients with hypogammaglobulinemia in a wide range of autoimmune disorders. Mild adverse events are common, while severe adverse events occur in less than 4% of patients. Acute myocardial infarction is rarely reported. A 43-year-old man suffering from antiphospholipid syn- drome with steroid-dependent thrombocytopenia was also known to have several other risk factors for ischemic heart disease including obesity, a positive family history of coronary heart disease, and a smoking habit (50 pack years of cigarettes). He had been admitted to another hospital one week earlier with chest pain, electrocardiographic evidence of ST segment depression and T-wave inversion in leads AVL, V2 V6, and positive cardiac enzymes. A diagnosis of non-ST elevation myocardial infarction of the anterior wall was made; however, in light of his low platelet count, he did not receive antiplatelet or anticoagulant therapy but rather was treated with nitrates and beta blockers alone with an improvement in symptoms. He was transferred to our hospital so that coronary angiography could be performed, and the decision was made to administer IVIG therapy, 400 mg/kg, due to a low platelet count of 29,000/mm3 to prevent bleeding during the procedure. One hour after ini- tiation of the infusion, the patient experienced severe retro- sternal chest pain accompanied by dyspnea and perspiration. ECG revealed ST segment elevation of 3 mm in the anterior leads. Serum creatinine kinase was 1940 IU/L with an MB fraction of 18%. A diagnosis of myocardial reinfarction of the anterior wall was made, but in view of the severe thrombocy- topenia (29,000/mm3), the decision was made to withhold thrombolytic, antithrombotic, and antiplatelet therapy. The planned coronary angiography was not performed for fear of N Corresponding author. Tel.: +972 46495143; fax: +972 46495134. E-mail address: salibuss@yahoo.com (W.R. Saliba). complications, and the patient subsequently recovered from the acute episode, although echocardiography demonstrated resid- ual reduced left ventricular function. Treatment with IVIG is associated with several mild adverse reactions including anxiety, headache, fever, backache, nausea, and abdominal pain. Severe adverse events are reported in less than 4% of treated patients and include hypotension, renal failure, aseptic meningitis, pulmonary edema, anaphylaxis, and hemolytic anemia [1]. Lately, there have been an increasing number of thromboembolic events reported in association with IVIG use, with an estimated frequency of 3 5%, including DVT proximal to the site of infusion, pulmonary embolism, cerebrovascular accident, and fatal hepatic veno-occlusive disease. Acute myocardial infarction in the setting of IVIG therapy is very rare [2]. Thromboembolic events including myocardial infarction occur most often in patients with under- lying multiple cardiac risk factors, prior stroke, and hyperco- agulability and hyperviscosity states. The pathogenesis of the thromboembolic phenomena may be related to plasma expansion and increased viscosity. In addition, IVIG enhances platelet activation and this effect may be amplified by the increase in platelet count in patients treated with IVIG. Furthermore, IVIG has procoagulant activity, especially due to the presence of FXIa, which could lead to thrombin generation even when present in small amounts. Therefore, in patients with underlying risk factors for thromboembolic events, caution is advised before starting IVIG therapy. ˇ˛Abstract European Journal of Internal Medicine 18 (2007) 109 117 Original article www.elsevier.com/locate/ejim Acute phase proteins in patients with acute coronary syndrome: Correlations with diagnosis, clinical features, and angiographic findings Natale Daniele Brunetti N , Michele Correale, Pier Luigi Pellegrino, Andrea Cuculo, Matteo Di Biase Cardiology Department, University of Foggia, Italy Received 19 January 2006; received in revised form 18 May 2006; accepted 6 July 2006 Background: C-reactive protein (CRP) plasma levels increase in patients with acute coronary syndrome (ACS). The role and implications of increased plasma concentrations of other acute phase proteins (APPs), such as alpha-1-antitrypsin (A1AT), alpha-1 glycoprotein (A1GP), haptoglobin (HG), ceruloplasmin (CP), and C3c and C4 complement fraction, in patients with ACS are still not completely defined. Methods: A total of 218 consecutive patients with ACS were included in the study, 185 with acute myocardial infarction (AMI) and 33 with unstable angina (UA). In all patients, A1AT, A1GP, HG, CP, C3c and C4 complement fraction, and CRP were evaluated within 12 h after the onset of symptoms. Sixty-two patients with AMI underwent coronary angiography. Results: APPs showed a significant correlation with CRP concentrations. Patients with AMI had higher concentrations of A1AT and HG than UA patients. Cholesterol levels were correlated with APPs in patients with AMI. Patients with three coronary vessel disease or LAD disease had significantly higher C3c concentrations. Coronary collateral flow was associated with higher A1GP and CP concentrations, and total coronary occlusion with A1AT and CP. Conclusions: APPs were correlated with CRP concentrations in subjects with ACS. The increase in APPs in patients with ACS seems to be linked to the entity of myocardial damage and coronary atherosclerotic burden. © 2007 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. Keywords: Acute phase proteins; Acute coronary syndrome; Coronary atherosclerosis 1. Introduction Inflammation plays a role in the development of athero- sclerosis and coronary heart disease [1], and the inflammatory process is linked to ischemic myocardial damage and necrosis [2]. A number of studies have reported that C-reactive protein (CRP) levels increase during acute myocardial infarction (AMI) and unstable angina (UA) [3,4]. Fewer investigations have been published seeking to identify the relevance of other markers of inflammation, such as alpha-1-antitrypsin (A1AT), alpha-1 glycoprotein (A1GP), haptoglobin (HG), ceruloplas- N Corresponding author. Tel.: +39 3389112358; fax: +39 0881745424. E-mail address: dr.natale.daniele.brunetti@hotmail.it (N.D. Brunetti). min (CP), and C3c and C4 complement fraction in patients with acute coronary syndrome (ACS). A1GP is an acute phase protein (APP) that is elevated in inflammatory processes, pregnancy, rheumatoid arthritis, malignancy, and pneumonitis [5,6]. A1AT protects lungs from neutrophil elastase, which normally digests damaged or aging cells and bacteria in order to provide for healing. It can increase to three or four times the normal level to protect tissues in times of infection, pregnancy, or other situations that subject the body to an increase in neutrophil elastase. The primary function of HG is hemoglobin capture to prevent both iron loss and kidney damage during hemolysis. As one of the major acute phase reactants, HG is involved in the processes of inflammation, infection, and malignancy. 0953-6205/$ - see front matter © 2007 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.ejim.2006.07.031 110 N.D. Brunetti et al. / European Journal of Internal Medicine 18 (2007) 109 117 Table 1 Clinical characteristics of the patients included in the Mean Male gender 69.68% Age (years) 66.22 Hypertension 56.62% Dyslipidemia 57.99% study SD assess the difference in markers of inflammation levels in patients with ST elevation myocardial infarction (STEMI), non-ST elevation myocardial infarction (NSTEMI), and UA and their correlations with coronary angiography Diabetes Smoking Previous Previous Previous Previous Pre-infarction angina 37.90% Pain duration (min) 183.68 LVEF (%) 44.83 Lown's class 2.40 In-hospital complications 47.27% Troponin I peak (ng/mL) 48.71 CKMB peak (ng/mL) 104.35 CK peak (U/L) 1070.14 LDH peak (U/L) 477.73 Total cholesterol (mmol/L) 5.24 HDL cholesterol (mmol/L) 1.06 LDL cholesterol (mmol/L) 3.22 Admission CRP (mg/L) 32.1 Peak CRP (mg/L) 88.1 C3c (g/L) 1.12 C4 (g/L) 0.29 Alpha 1 glycoprotein (mg/L) 987.4 Alpha 1 antitrypsin (ºmol/L) 30.13 11.86 findings. 2. Methods A total of 218 consecutive patients with ACS who had been admitted to our intensive care unit were recruited for this study. Their clinical characteristics are given in Table 1 and angiographic data in Table 2. STEMI was defined as follows: 1.57 1. Biochemical alterations: 32.88% habit 36.53% CABG 5.94% PCI 5.94% MI 20.55% angina 30.59% Ceruloplasmin (mg/L) Haptoglobin (g/L) 8.45 320 82.5 1.7 0.74 HG possesses immunoregulative properties, inhibits prosta- glandin synthesis, and protects against harmful oxidation processes [7,8]. CP is a globulin with multiple functions: it is believed to control the release of iron into plasma from cells in which the metal is stored in the form of ferritin. CP is also a prominent serum antioxidant, catalyzes the oxidation of ferrous ion to ferric ion, and prevents the oxidation of polyunsaturated fatty acids. Finally, it modulates the inflammatory response and can regulate the concentration of various serum biogenic amines. CP is elevated in a variety of circumstances, including pregnancy or other conditions with high estrogen concentrations, infections, cirrhosis, malignancies, and hyperthyroidism [9]. The third comple- ment component, C3, is a multifunctional protein produced mainly by the liver, adipose tissue, and activated macro- phages. Apart from exerting several important functions in the immune system, C3 also behaves as an APP: it is synthesized by hepatocytes in response to interleukin [IL]- 1≤, which is secreted by activated macrophages at the site of inflammation. Clinical evidence suggests a link between ACS and auto-immune inflammatory diseases, commonly charac- terized by increased plasma concentrations of APP, as mentioned above [10 13]. Correlations between APP, myocardial damage, and coronary atherosclerosis still need to be completely defined. In this study, we sought to Table 2 Angiographic data Ruptured plaque Total occlusion Collateral flow (Rentrop N 0) LAD severe stenosis LCx severe stenosis RCA severe stenosis Number of stenotic vessels Normal angio-coronary findings Indication to PCI Indication to CABG Mean SD 0% 33.87% 39.34% 48.39% 43.55% 47.54% 1.42 0.92 4.84% 54.1% 29% 165.50 7.89 96.83 144.45 1529.80 353.82 a. Maximal concentration of troponin I exceeding the decision limit (99th percentile of the values for a reference control group) on at least one occasion during the first 24 h after the index clinical event. Maximal value of CK-MB exceeding the 99th percentile of the values for a reference control group on two successive samples, or maximal value exceeding twice the upper limit of normal for the specific institution on one occasion during the first hours after the index clinical event. Values for CK-MB 1.23 b. 0.27 1.15 70.2 115.3 0.25 0.09 375.2 should rise and fall. 2. New or presumed new ECG alterations: ST segment elevation at the J point in two or more contiguous leads with the cut-off points e"0.2 mV in leads V1, V2, or V3 and e"0.1 mV in other leads. 3. New or presumed new echocardiographic alterations: cardiac wall motion abnormalities. 4. Typical chest pain. Criteria for the diagnosis of NSTEMI were: 1. Biochemical alterations: a. Maximal concentration of troponin I exceeding the decision limit (99th percentile of the values for a reference control group) on at least one occasion during the first 24 h after the index clinical event. b. Maximal value of CK-MB exceeding the 99th percentile of the values for a reference control group on two N.D. Brunetti et al. / European Journal of Internal Medicine 18 (2007) 109 117 111 Table 3 Clinical characteristics of STEMI, NSTEMI, and UA patients UA n 33 NSTEMI n 62 p UA n 33 STEMI n 123 p NSTEMI STEMI p Mean SD Mean SD Mean SD Mean SD Mean Mean Male gender 63.64% Age 66.15 Hypertension 78.79% Dyslipidemia 57.58% Diabetes 36.36% Smoking 30.30% Previous CABG 15.15% Previous PCI 18.18% Previous MI 33.33% Previous angina 60.61% Pre-infarction angina 12.12% LVEF 48.61 Lown class 2.41 In-hospital 21.% complications Troponin I peak 0.41 CKMB peak 2.48 CK peak 110.16 LDH peak 217.38 Total cholesterol 4.98 HDL cholesterol 1.03 LDL cholesterol 2.73 Admission CRP 21.1 Peak CRP 51.7 67.74% 67.45 13.48 56.45% 53.23% 32.26% 32.26% ns 63.64% ns 66.15 b 0.05 78.79% ns 57.58% ns 36.36% ns 30.30% ns 15.15% b 0.05 18.18% ns 33.33% b 0.01 60.61% b 0.001 12.12% ns 48.61 ns 2.41 b 0.001 21.% b 0.001 0.41 b 0.001 2.48 b 0.001 110.16 b 0.01 217.38 ns 4.98 ns 1.03 ns 2.73 ns 21.1 ns 51.7 71.54% 65.71 11.25 50.41% 60.16% 31.71% 39.84% ns ns b 0.001 ns ns ns b 0.001 b 0.01 ns b 0.001 b 0.01 b 0.001 ns b 0.001 b 0.001 b 0.001 b 0.001 b 0.001 ns 5.33 ns 1.09 b 0.05 3.08 3.41 ns 37.7 31.9 ns 76.1 104.5 10.56 10.99 1.58 0.79 2.63 107.62 76.62 1.27 0.27 0.74 33.4 66.4 8.06% 3.23% 19.35% 30.65% 50.00% 46.78 2.38 56.% 18.51 55.86 589.83 384.58 5.33 1.09 3.08 6.28 1.59 21.60 69.05 635.86 271.37 1.40 0.33 1.29 10.56 10.99 1.58 0.79 2.63 107.62 76.62 1.27 0.27 0.74 33.4 66.4 2.44% 4.07% 17.89% 22.76% 39.02% 42.99 2.41 50.% 76.61 156.15 1564.57 594.69 5.26 1.05 3.41 31.9 104.5 7.03 1.59 120.42 167.30 1823.60 384.54 1.14 0.24 1.16 57.9 127.5 67.74% 67.45 56.45% 53.23% 32.26% 32.26% 8.06% 3.23% 19.35% 30.65% 50.00% 46.78 2.38 56.% 71.54% 65.79 50.41% 60.16% 31.71% 39.84% 2.44% 4.07% 17.89% 22.76% 39.02% 42.99 2.41 50.% ns ns ns ns ns ns ns ns ns ns ns b 0.001 ns ns b 0.001 b 0.001 b 0.001 b 0.001 ns ns ns ns ns 37.7 98.8 76.1 106.6 18.51 76.61 156.15 1564.57 594.69 5.26 1.05 55.86 589.83 384.58 successive samples, or maximal value exceeding twice the upper limit of normal for the specific institution on one occasion during the first hours after the index clinical event. Values for CK-MB should rise and fall. 2. New or presumed new ECG alterations (no ST segment elevation): a. ST segment depression. b. T-wave abnormalities (in two or more contiguous leads). 3. New or presumed new echocardiographic alterations: cardiac wall motion abnormalities. 4. Typical chest pain. UA was diagnosed as follows: 1. Typical chest pain and ECG modifications, such as ST depression or T-wave inversion in e"2 leads. 2. Absence of biochemical alterations (maximal value of troponin I not exceeding the 99th percentile of the values for a reference control group on two successive samples, Table 4 Correlations between plasma levels of acute phase proteins and CRP or maximal value not exceeding twice the upper limit of normal for the specific institution on one occasion during the first hours after the index clinical event). 3. Absence of new or presumed new echocardiographic alterations: cardiac wall motion abnormalities. According to these criteria, 123 patients were discharged with a STEMI diagnosis, 62 with an NSTEMI diagnosis, and 33 with an UA diagnosis. All of the patients underwent blood sampling within 12 h after the onset of symptoms in order to evaluate plasma A1AT, A1GP, HG, CP, C3c and C4 complement fraction, and CRP concentrations. Plasma concentration of CRP was Fig. 1. Mean alpha-1-antitrypsin plasma concentrations in UA, NSTEMI, and STEMI patients. All pts A1GP 0.57N N N A1AT 0.59N N N CP 0.28N N N HG 0.48N N N C3c 0.31N N N C4 0.43N N N N N N p b 0.001, N N p b 0.01, AMI pts 0.56N N N 0.59N N N 0.28N N 0.49N N N 0.26N 0.41N N N N p b 0.05. STEMI pts 0.77N N N 0.69N N N 0.47N N N 0.67N N N 0.28N 0.37N N NSTEMI UA pts 0.42N 0.58N N N 0.32N 0.38N 0.37N N N 112 N.D. Brunetti et al. / European Journal of Internal Medicine 18 (2007) 109 117 Fig. 2. Mean haptoglobin plasma concentrations in UA, NSTEMI, and STEMI patients. determined using a particle-enhanced turbidimetric immuno- assay (PETIA) technique with an assay range of 2 120 mg/L (Dade Behring, Dimension, Flex). A1AT, A1GP, HG, CP, and C3c and C4 complement fraction plasma concentrations were determined using an immuno-turbidimetric technique: A1AT (AAT, AUTO PLUS, ITC Diagnostics), A1GP (QUANTEX A1AGP, BIOKIT), HG (AUTO PLUS, ITC Diagnostics), CP (CER, PARKER LAB), and C3c and C4 (QUANTEX, BIOKIT). The normal range for plasma concentrations was 17.48 32.20 ºmol/L for A1AT, 500 1200 mg/L for A1GP, 0.5 3.2 g/L for HG, 200 400 mg/L for CP, 0.9 1.8 g/L for C3, and 0.1 0.4 g/L for C4. Total cholesterol plasma concentrations were determined with the colorimetric enzymatic method (OLYMPUS OSR 6516), LDL and HDL cholesterol plasma concentrations with a direct two-step redox reaction and photometry (BECK- MANN COULTER). Sixty-two AMI patients (69.35% with STEMI, 30.65% with NSTEMI) were submitted to coronary angiography in our department cath-lab in order to determine Fig. 3. Mean alpha-1 glycoprotein plasma concentrations in UA, NSTEMI, and STEMI patients. Fig. 4. Correlation between C3 and LDL in patients with STEMI. the presence of severe coronary stenosis (N 70%), coronary collateral flow, total coronary occlusion, ruptured plaque, and the number of coronary vessels presenting luminal narrowing greater than 70%. Philips Medical System Integris H3000 software was used to assess the degree of stenosis. In the statistical analysis, continuous variables were expressed as mean value ± standard deviation and categorical variables as a percentage. Differences in percentages were analyzed with the chi-square test. Mean values were compared with Student's t-test for variables with a normal distribution or with the Mann Whitney non-parametric U- test for variables with a non-normal distribution. Correlation between continuous variables was performed using Pearson's correlation test. A multivariate regression model was used to analyze correlations adjusted for potential confounders. A p value below 0.05 was considered statistically significant. 3. Results Patients with STEMI, NSTEMI, and UA showed differ- ences in the percentage of patients with hypertension, previous PCI, previous CABG, left ventricle ejection fraction (LVEF), and cardiac enzyme plasma peak concentrations (Table 3). Fig. 5. Correlation between C4 and LDL in patients with STEMI. Fig. 6. Correlation between alpha-1 glycoprotein and HDL in patients with STEMI. A1GP, A1AT, HG, CP, and C3c and C4 complement fractions showed a statistically significant correlation with CRP concentrations at admission (A1GP CRP r 0.57, A1AT CRP r 0.59, CP CRP r 0.28, HG CRP r 0.48, C3c CRP r 0.31, C4 CRP r 0.43; p b 0.001 in all cases). A significant correlation was also present in the entire group of patients with AMI, with STEMI, and with NSTEMI UA (Table 4). 3.1. STEMI NSTEMI UA comparison Patients with AMI had higher plasma concentrations of A1AT than patients with UA (30.86±8.64 vs 25.79± 5.59 ºmol/L, p b 0.01), both those with STEMI (31.57± 9.26 ºmol/L, p b 0.01) and those with NSTEMI (29.27± 6.91 ºmol/L, p b 0.05; Fig. 1). HG showed different plasma concentrations in patients with UA, NSTEMI, and STEMI (UA 1.25 ± 0.63 g/L vs NSTEMI 1.59 ± 0.62 g/L, p b 0.05; UA 1.25 ± 0.63 g/L vs STEMI 1.87 ± 0.77 g/L, p b 0.001; NSTEMI 1.59 ± 0.62 g/L vs STEMI 1.87 ± 0.77 g/L, p b 0.05; UA 1.25 ± 0.63 g/L vs all AMI 1.79 ± 0.73 g/L, p b 0.01; Fig. 2). A1GP plasma concentrations of UA patients showed differences of Fig. 7. Correlation between alpha-1-antitrypsin and HDL in patients with STEMI. Fig. 8. Correlation between haptoglobin and HDL in patients with STEMI. borderline significance in comparison with patients with STEMI (862.2 ± 386 vs 1584.4 ± 408.2 mg/L, p 0.05; Fig. 3). 3.2. Correlations with biochemical data Plasma concentrations of C3c and C4 fractions correlated significantly with LDL cholesterol concentrations in patients with STEMI (C3c LDL: r 0.31, p b 0.05, C4 LDL: r 0.358 p b 0.05); among the same patients, A1GP, A1AT, and HG correlated inversely with HDL cholesterol plasma concentra- tions (A1GP HDL: r"0.27, p b 0.05; A1AT HDL: r"0.3, p b 0.05; HG HDL: r " 0.31, p b 0.01). Correlations re- mained significant even after correcting for age, gender, and LVEF (Figs. 4 8). 3.3. Correlations with angiographic data Patients with AMI and severe (N 70%) stenosis in three coronary vessels had a significantly higher C3c fragment plasma concentration (1.44 ± 0.31 g/L) than patients with severe stenosis in two coronary vessels (1.15 ± 0.11 g/L, Fig. 9. C3 plasma concentrations and number of stenotic vessels in patients with STEMI. N.D. Brunetti et al. / European Journal of Internal Medicine 18 (2007) 109 117 113 114 N.D. Brunetti et al. / European Journal of Internal Medicine 18 (2007) 109 117 Table 6 Acute phase protein and coronary total occlusion in patients with AMI No total occlusion Total occlusion Alpha 1 glycoprotein % A1GP N n.v. Alpha 1 Mean 966.8 21.95% 28.61 SD 315.1 6.41 Mean 1108.1 23.81% 33.31 SD p 411.4 n.s. antitrypsin % A1AT N n.v. Ceruloplasmin 293.6 68.1 334.8 % CP N n.v. 8.33% 36.84% Haptoglobin 1.7 0.68 1.88 % HG N n.v. 0.00% 5.26% b 0.05 97.2 n.s. b 0.01 0.95 n.s. n.s. 25.00% 50.00% 12.08 n.s. b 0.05 Fig. 10. Mean C3 plasma concentrations in patients with and without stenosis N 70% in the LAD. Mann Whitney U-test p b 0.05), one coronary vessel (1.10 ± 0.19 g/L, p b 0.01) or with no angiographic evidence of luminal narrowing (1.06 ± 0.17 g/L, p b 0.05; Fig. 9). Patients with AMI and severe (N 70%) stenosis in the left anterior descending (LAD) coronary vessel had significantly higher C3c plasma concentrations than patients without stenosis in the LAD (1.26 ± 0.26 g/L vs 1.10 ± 0.19, p b 0.05; Fig. 10). Patients with AMI and angiographic evidence of coronary collateral flow had significantly higher plasma concentra- tions of A1GP (1145.8 ± 414.5 vs 924.6 ± 285.5 mg/L, p b 0.05) and CP (343.4 ± 83.1 vs 285.9 ± 72.3 mg/L, p b 0.01) than patients with no evidence of collateral flow (Table 5). Patients with AMI and total occlusion of a coronary vessel had a significantly higher concentration of A1AT (33.31± 12.08 vs 28.61 ± 6.41 ºmol/L, p b 0.05) and included a higher percentage of subjects with CP plasma values above normal levels (N 400 mg/L; 36.84% vs 8.33%, p b 0.01) than patients without coronary total occlusion (Table 6). 4. Discussion The present study demonstrates, for the first time, a clear increase in some APPs in subjects with ACS and significant differences in plasma concentrations of these APPs among patients with STEMI, NSTEMI, or UA in the very early phase of ACS. Plasma concentrations of A1AT and HG are increased in proportion to the severity of ACS, reaching the highest levels in patients with STEMI. Previous prospective studies have reported an increased incidence of MI or cardiovascular deaths (CVD) in apparently healthy individuals with elevated levels of inflammation- sensitive plasma proteins (ISPs) [14 20]. The Malmo Preven- tive Study has shown that the incidence of MI and stroke is increased in men with elevated plasma levels of fibrinogen, A1GP, A1AT, HG, or CP. Furthermore, the risk was gradually increased with a cumulative number of elevated ISPs [14,21,22]. Engstrom et al. [23] demonstrated how men who had been exposed to a low-grade inflammation many years earlier had more fatal future coronary events over a mean follow-up of 19 years. Among men who subsequently had acute coronary events, the number of elevated ISPs at baseline showed significant associations with the proportions of fatal outcomes (first day or within 28 days). The same authors [24] identified a link between both APP (HG, CP, A1AT, A1GP) levels in the top quartile and the number of APPs in the top quartile and an increased risk of cardiac events and cardiovas- cular mortality. In another paper [14] evaluating plasma cholesterol and five ISPs (fibrinogen, A1AT, HG, CP, and A1GP) in 6063 healthy men, it was reported that, after a mean 18.7-year follow-up, men with hypercholesterolemia and high ISP levels (2 to 5 ISPs in the top quartile) had a significantly higher risk of CVD [relative risk (RR)=2.4; CI 1.8 3.3] and cardiac events (RR=2.3; CI 1.8 3.0) than men with normal cholesterol and low ISP levels. Available data are still controversial since, in a report from the MRFIT study [25] and in the Angina Prognosis Study in Stockholm (APSIS), which comprised 809 patients with stable angina pectoris, A1GP did not carry any independent prognostic information [26]. Our study results, however, aside from the previously reported prognostic relevance of APP activation, seem to point towards a close link between plasma concentration of markers of inflammation and the severity of ACS, expressed by ECG presentation. Different pathogenic subsets underlying STEMI, NSTEMI, and UA seem to enhance the early activation of different APPs. Table 5 Acute phase protein and coronary collateral flow in patients with AMI No collateral flow Mean 924.6 285.5 Collateral flow Mean SD 1145.8 414.5 p b 0.05 b 0.05 n.s. b 0.05 b 0.01 b 0.05 n.s. n.s. SD Alpha 1 glycoprotein % A1GPNn.v. 13.51% 37.50% Alpha 1 28.48 6.53 32.49 antitrypsin % A1ATNn.v. 21.62% 50.00% Ceruloplasmin 285.9 72.3 343.4 % CPNn.v. 8.82% 33.33% Haptoglobin 1.62 0.66 2 % HGNn.v. 0.00% 4.76% 11.45 83.1 0.92 One possible explanation for this increased inflammatory activation could be the link between increased APP plasma concentrations and the severity of coronary atherosclerosis, as reported by Gabor et al. [27]. In a study evaluating C3 levels in 266 patients with pre-existing severe coronary artery disease (CAD) who had undergone aorto-coronary bypass graft surgery, C3 concentrations were more elevated in the patients with severe CAD than in 182 healthy controls. Pathological C3 levels (C3 e" 1.8 g/L) were able to predict major complications of atherosclerosis (death by cardiac events, new AMI, stroke, carotid surgery, and peripheral arterial disease) that developed during the follow-up period only in women (OR: 4.1, 95% C.I. 1.23 13.61, p 0.0249), independent of other risk factors for atherosclerosis. Other studies have already pointed out the presence of elevated concentrations of complement components in the circulation of patients with atherosclerosis [28] and the presence of activated complement components in atheroscle- rotic lesions [29]. A link between increased levels of HG phenotype 2 2 and severity of coronary atherosclerosis, expressed as the presence of a number of bypass grafts greater than three, was demonstrated in 765 male subjects with previous CABG and this phenotype by Delanghe et al. [30]. Another possible explanation may be found in the relation between APPs, serum lipids, and coronary risk factor co- morbidity. Several cardiovascular risk factors (smoking, hypertension, cholesterol, and diabetes) are associated with increased ISPs levels [31,32]. Serum C3 is strongly correlated with fasting insulin [33], insulin resistance [34], cholesterol, blood glucose, body mass index, systolic blood pressure, and a sedentary lifestyle [33,34]. On the other hand, increased APP response could reflect plaque instability. Men with high ISPs could have a predisposition for developing unstable coronary plaques with infiltration of inflammatory cells, which could cause coronary events of a greater severity [23]. Lipid- and macrophage-rich vulnerable plaques could reasonably represent a direct or indirect source of C3 production [35]; thus, C3 would be a marker of the presence of vulnerable plaques. Other earlier studies have already pointed out the presence of elevated concentrations of complement components in the circulation of patients with atherosclerosis [28,36 38], but also the presence of activated complement components in atherosclerotic lesions [29]. Muscari et al. [39,40] indicated that high C3 levels can be considered a strong and independent predictor of the progres- sion of atherosclerotic vascular disease. C5b-9 complex, the terminal product of complement activation, has been demon- strated in atherosclerotic plaques [41 43]. Activated comple- ment has cell-damaging and chemotactic properties [44,45] and, as such, may favor the development of intimal lesions and/or monocyte recruitment at the site of atheroma formation. Muscari et al. found that patients with severe atherosclerotic lesions had higher serum levels of complement components and IgA than selected control subjects; C3 C4 levels were higher in subjects with severe, angiographically assessed peripheral atherosclerosis who had already been affected by multiple ischemic events than in selected controls [39]. However, in this retrospective study, only serum C4 was found to be independently associated with atherosclerotic disease by multivariate analysis. Atheroma development might be associ- ated with a chronic activation of the complement system since the occurrence of complement activation has been well docu- mented in human atherosclerotic lesions [41 43]; in the hyper- cholesterolemic rabbit, it has been found to precede monocyte infiltration and foam cell formation [46]. However, little evi- dence has been published reporting that complement activation in the arterial wall is reflected in circulating complement levels, and no correlation has previously been demonstrated between serum complement components and the extent of atheroscle- rotic disease [47]. Elevation of serum C3 might indicate a condition of coronary instability of the inflammatory type, capable of triggering coronary thrombosis [48]. In the present study, we found significant correlations not only between APPs and features of coronary atherosclerosis, but also between APPs and coronary collateral flow. Hochberg et al. [49] previously pointed out involvement of HG phenotypes in neo-angiogenesis and the presence of collateral coronary flow in diabetic patients. In their work, these authors hypothesized a possible role for HG in enhancing collateral vessel development, due to its antiox- idant properties preserving pericyte, a cell type that has been demonstrated to be critical for the maturation of nascent capillary endothelial cells into mature blood-carrying struc- tures [50]. Our results, in contrast, seem to suggest a possible link between the increase in A1GP, A1AT, and CP and the presence of jeopardized or viable myocardium, associated with a total coronary occlusion or a collateral coronary flow, as reported by Gwechenberger et al. [51] in a canine model. In fact, even non-infarcted myocardium seems to signifi- cantly contribute to inflammatory marker production, as demonstrated by Ono et al. 1 week after an acute episode of myocardium ischemia in a rat model [52]. Further investigations are required to completely clarify links between increased APP plasma concentrations and ACS, possibly involving larger cohorts of patients. Yet, the existence, clinical relevance, and coronary angiographic features of these links may be assumed to be definitive. 5. Learning points APPs correlate with CRP concentrations in subjects with ACS. APPs levels in patients with ACS seem to correlate with ECG presentation, cholesterol levels, and coronary atherosclerosis. ˇ˛Addison s disease Antonia M Brooke John P Monson Abstract Addison s disease or primary adrenocortical failure is a rare condition, most commonly caused in the UK by autoimmune destruction of the adrenal glands. The insidious onset of symptoms over many months means there is often a delay in diagnosis and patients can first present in adrenal crisis. The diagnosis is made by measurement of a low serum cortisol at 9 a.m. in the presence of an elevated adrenocorticotro- phic hormone (ACTH) or by a poor cortisol response to exogenous ACTH on provocation testing. There are many possible causes of adrenal failure and once the diagnosis of adrenal insufficiency has been made these should be differentiated further. Replacement with hydrocortisone and fludrocortisone should approximate physiological levels as closely as possible and be monitored closely. Additional adrenal androgen replace- ment can be added, particularly if the patient has a poor quality of life, although the long-term effects of this are unknown. Keywords Addison s; adrenal; autoimmune; dehydroepiandrosterone; fludrocortisone; hydrocortisone; tuberculosis Addison s disease (AD) denotes primary adrenocortical failure and was first described by Thomas Addison in 1855.1 Most of the original cases described were caused by tuberculosis (TB), but the most common aetiology is now autoimmunity. AD remains rare; the prevalence is about 120/million.2 It is important to diagnose the condition promptly, to prevent the patient suffering a life-threatening crisis. A survey from the US National Adrenal Disease Foundation revealed that 60% of patients with AD had sought medical attention from two or more physicians before the diagnosis was considered; the reason for this delay was the insidious onset of the disease. Anatomy and pathophysiology The adrenals are small Y-shaped glands (each limb measuring <5 mm) located at the superior poles of the kidneys. They comprise a cortex (90%) surrounding the medulla (10%) (Figure 1). The cortex secretes: ˇ€‹ the glucocorticoid cortisol from the zona fasciculata ˇ€‹ androgens (e.g. dehydroepiandrosterone, DHEA) from the zona reticularis Antonia M Brooke MD MRCP MA MBBS is a Specialist Registrar in Endocrinology at St Bartholomew s Hospital, London, UK. Competing interests: none declared. John P Monson MD FRCP FRCPI is Emeritus Professor of Clinical Endocrinology at St Bartholomew s and The Royal London Hospital, London, UK and Consultant Physician at the London Clinic Centre for Endocrinology. Competing interests: none declared. ˇ€‹ mineralocorticoids (aldosterone) from the zona glomerulosa, predominantly under the control of the renineangiotensin system (although 5e10% of total aldosterone production is mediated by adrenocorticotrophic hormone, ACTH). AD involves all three zones of the adrenal cortex. Overt symptoms do not usually appear until more than 90% of the gland has been destroyed. Glucocorticoids, mineralocorticoids and androgens are all reduced, in contrast to the situation in secondary adrenal failure (ACTH deficiency), in which mineral- ocorticoid secretion is relatively preserved. Pathological exami- nation of the adrenals in cases of immune destruction reveals fibrosis with a mononuclear cell infiltrate, occasional plasma cells and, rarely, germinal centres. Aetiology In developed countries, about 75e80% of cases of AD are caused by autoimmune destruction.3 TB is the second most common cause. Other causes are rare (Table 1).4 Patients with autoim- mune AD are at a 50e60% risk of developing another autoim- mune disorder (10% of patients develop type 1 diabetes mellitus). There is an association with human leukocyte antigen (HLA) DR3 and HLA DR4. A family history of autoimmune disease can be associated with multiple organ-specific autoimmunity. Determination of the cause may be guided by age at presen- tation and gender. ˇ€‹ At birth, adrenal haemorrhage (from anoxia) is the most common cause. ˇ€‹ In young females, an autoimmune basis is more likely (three times more common than in males). ˇ€‹ Infection and metastases should be considered in males and the elderly. ˇ€‹ Although the adrenal glands are a relatively common site of metastases, hormonal insufficiency is much less common. ˇ€‹ Fungal infections are more common in the immunocompro- mised; up to 5% of patients with AIDS have adrenal insuffi- ciency at a late stage. Clinical features Patients can present with an insidious onset of symptoms or during an adrenal crisis, depending on the acuteness of the hormonal deficit and any intercurrent illness.5 The most common symptoms are non-specific and therefore diagnosis is often ADRENAL What s new? C Addison s disease has recently been shown to be associated with a twofold increased risk of mortality compared to the general population C A new slow-release once-daily hydrocortisone preparation is under investigation and may help with compliance and mimic normal daytime physiology more closely C Additional androgen treatment, usually in the form of DHEA, offers psychological benefit to some patients although long- term data are unavailable MEDICINE 37:8 416 O 2009 Published by Elsevier Ltd. ADRENAL Causes of primary adrenal deficiency Adrenal destruction Autoimmune C Isolated C Autoimmune polyglandular syndrome 1 (autosomal recessive, equally common in males and females, chronic mucocutaneous candidiasis, acquired hypoparathyroidism (90%), Addison s disease (60%)) C Autoimmune polyglandular syndrome 2 (autosomal recessive, autosomal dominant and polygenic, more common in females, Addison s disease (100%), autoimmune disease of the thyroid (Schmidt s syndrome), immune-mediated diabetes (Carpenter s syndrome)) Infections C Tuberculosis C Fungal (histoplasmosis, Cryptococcus) C Opportunistic (cytomegalovirus in AIDS) Metastases C Lung, breast, kidney C Lymphoma Haemorrhage C WaterhouseeFriderichsen syndrome (meningococcal septicaemia) Infiltrations C Amyloidosis, haemochromatosis Others C Adrenoleukodystrophy Adrenal dysgenesis C Congenital adrenal hypoplasia C Mutations in SF1 Impaired steroidogenesis C Congenital adrenal hyperplasia Mitochondrial disorderseIatrogenic C Adrenal suppressors (e.g. ketoconazole, etomidate) C Enzyme inducers (e.g. phenytoin, rifampicin) Figure 1 Cross-section of the adrenal gland. c, capsule; a, arteriole; n, nerve fibres; gc, ganglionic cells; zg, zona glomerulosa; s, sinusoids; zf, zona fasciculata; ma, medullary artery; i, isolated islets of chromaffin cells; zr, zona reticularis; m, medulla. delayed (Table 2). Hyperpigmentation (from elevation of mela- nocyte-stimulating hormone and ACTH) is often seen in the buccal mucosa, nail-beds and areas that are exposed to light and pressure, suggesting primary rather than secondary adrenal failure. Investigations and diagnosis In patients who present acutely, the diagnosis must be made quickly and treatment initiated.6 Biochemical abnormalities Patients are often hyponatraemic, hyperkalaemic and acidotic at presentation. Mineralocorticoid deficiency leads to sodium depletion, reduced extracellular fluid volume and hypotension. Reduced renal water clearance compounds the hyponatraemia. Hyperkalaemia develops as a result of reduced renal potassium and hydrogen ion excretion, and type IV renal tubular acidosis ensues. Low serum glucose (from reduced glycogen stores) is common, although severe hypoglycaemia is rare in adults. Table 1 Serum cortisol The diagnosis is made by documenting low serum cortisol in the presence of elevated plasma ACTH. A serum cortisol level of less than 100 nmol/litre at 9 a.m. is diagnostic of deficiency and a cortisol above 550 nmol/litre makes the diagnosis unlikely. Elevated 9 a.m. plasma ACTH in the presence of apparently normal serum cortisol is also suggestive of primary adrenal failure. Tetracosactrin test The diagnosis is confirmed by a suboptimal cortisol response to synthetic ACTH; tetracosactrin, 250 mg i.m. or i.v., is given at 9 a.m. and serum cortisol measured at 0, 30 and 60 minutes. A normal response is a peak of more than 550 nmol/litre; occasionally, false- negative results are obtained. Depot tetracosactrin can also be used to distinguish primary from secondary adrenal failure; 1 mg is MEDICINE 37:8 417 O 2009 Published by Elsevier Ltd. Table 2 given intramuscularly and serum cortisol measured at 0, 6 and 24 hours. In AD, there is no increase after 6 hours, whereas in secondary adrenal failure a continuous increase is seen. Investigations for the cause Once primary adrenal insufficiency has been diagnosed, the cause should be established. Adrenal antibodies to enzymes of the adrenal cortex (e.g. 21-OH, P450scc, 17-OH) are found in more than 90% of patients with recent-onset adrenal autoim- munity. In patients in whom adrenal antibodies have been identified (as a screening measure in those with other autoim- mune diseases), the earliest sign of impending adrenal failure is an increase in recumbent plasma renin activity. The raised renin is caused by a failing zona glomerulosa with low-normal or low aldosterone levels, and consequent sodium depletion (reduced extracellular and circulating volume). Clinical evidence of failure of the zona fasciculata (and cortisol deficiency) may be delayed for several years. However, of patients with adrenal antibodies, only about 20% per year develop adrenal insufficiency. If there is a clinical suspicion of TB, chest and abdominal radiography should be performed, looking particularly for apical shadowing and adrenal calcification. Computed tomography of the adrenals is necessary only when metastasis, an infiltrative process or TB is suspected. Adrenal failure may be associated with modest and reversible elevation of serum thyroid-stimu- lating hormone, which simulates mild hypothyroidism. Management The aims of treatment are to replace the deficient hormones and treat any reversible causes of adrenal disease. Glucocorticoids Glucocorticoid replacement is usually given three times daily, with the largest dose (10e20 mg) before getting out of bed to mimic the physiological peak just before waking, followed by 5 mg at midday and 5 mg at 6 p.m. The cortisol can be measured in the serum, allowing adjustment of the dose to achieve physio- logical levels. A new once-daily modified-release hydrocortisone formulation is under investigation, with the aim of mimicking physiological levels even more tightly and helping to improve compliance.7 Dexamethasone or prednisolone are less satisfac- tory because individuals vary in their sensitivity to synthetic glucocorticoids, and monitoring of serum concentration is not straightforward. An increase in the dose of hydrocortisone is usually required in pregnancy, and is made on clinical grounds (average 25e50% increase). Mineralocorticoids The aim of treatment with fludrocortisone is to achieve normal sodium homeostasis as shown by normal or slightly elevated renin levels. This can normally be achieved with a dose of 50e100 mg b.d. Measurement of plasma renin activity is mandatory to avoid excessive mineralocorticoid replacement. Over-treatment may result in hypertension and, rarely, oedema. Adrenal androgens There has recently been interest in replacing adrenal androgens such as DHEA in patients with AD. Patients with Addison s report a reduced quality of life compared to the general pop- ulation, and DHEA replacement may improve self-esteem, mood and fatigue scores, and libido (particularly in women),8 which may be testosterone-driven. The long-term benefits remain to be established. Management during intercurrent illness Endogenous adrenal secretion is increased in critically ill and perioperative patients with healthy adrenal glands. In those with adrenal insufficiency, glucocorticoid doses should be doubled during significant febrile illnesses.9 Patients should keep an ampoule of hydrocortisone for intramuscular injection in their refrigerator, in case of diarrhoea and vomiting at home. In a crisis, an isotonic saline infusion and hydrocortisone, 100 mg i.m. 6-hourly, should be administered. In critical illness, an MEDICINE 37:8 418 O 2009 Published by Elsevier Ltd. ADRENAL Symptoms, signs and investigations Symptoms Anorexia Weight loss (>90% of patients) Fatigue Muscle weakness and myalgia (generalized) Weight loss Gastrointestinal (abdominal pain, diarrhoea, vomiting) Dizziness Headache Depression and behavioural changes Reduced libido, reduced axillary hair Sweating Salt craving ACTH, adrenocorticotrophic hormone. Signs Hyperpigmentation Postural hypotension Dehydration Vitiligo ˇ€‹ goitre Pyrexia of unknown origin (occasionally) Investigations Hyponatraemia (90%) Hyperkalaemia (65%) Hypoglycaemia Low 9 a.m. cortisol and response to ACTH stimulation Elevated 9 a.m. ACTH Elevated renin and low/normal aldosterone Elevated urea Elevated thyroid-stimulating hormone Eosinophilia, lymphocytosis, raised ESR, hypercalcaemia, normochromic anaemia Adrenal autoantibodies Chest and abdominal radiography, CT (for calcification) intravenous infusion of hydrocortisone, 1e3 mg/hour, maintains a constant serum cortisol level of 600e800 nmol/litre. It is unnecessary to replace mineralocorticoids when giving high doses of hydrocortisone because there is a significant mineralo- corticoid effect as a result of saturation of renal 11b-hydroxy- steroid dehydrogenase type 2 activity by hydrocortisone. Patient education Every patient should wear a MedicAlert bracelet (or necklace), carry a steroid card and be aware of the need to increase gluco- corticoid doses in the event of severe intercurrent illness. Many patients find support from other patients and support groups useful.10 Patients should be reviewed annually and if appropriate screened for the development of other autoimmune diseases. AD has recently been associated with a twofold increase in standardized mortality ratio, mainly due to cardiovascular, malignant and infectious diseases.11 While the cause for this is unknown, optimizing the replacement therapy to as near physi- ology as possible should always be a key goal. Practice points C Addison s disease is rare. The most common cause in the UK is autoimmune destruction of the adrenals, followed by infiltra- tion of the glands by tuberculosis C Patients with autoimmune Addison s disease have a 60% chance of developing another autoimmune disease C Symptoms of cortisol deficiency are insidious in onset but skin hyperpigmentation can help differentiate between primary and secondary adrenal failure C Addison s disease should be considered in any patient with unexplained hyponatraemia or hyperkalaemia and a 9 a.m. cortisol should be measured C Treatment with hydrocortisone and fludrocortisone should be monitored closely to ensure physiological levels, and the doses should be increased in times of severe illness or pregnancy ˇ˛Adenovirus infections Phyllis Flomenberg Abstract Adenoviruses are common causes of febrile illness in early childhood. They are most often associated with respiratory tract disease, but can also cause gastrointestinal, ophthalmological, cardiac, neurological and genitourinary manifestations. Viral culture is a sensitive method for detecting most adenoviruses. Most adenoviral diseases are self-limiting, and treatment is supportive. However, fatal invasive disease can occur in immunocompromised patients and, occasionally, in healthy children and adults. Monitoring blood samples for adenoviral DNA by quantitative polymerase chain reaction analysis is useful for early detection of infec- tion in haematopoietic stem cell recipients. Although there have been no controlled trials, a number of reports suggest that cidofovir has clinical activity against adenoviral disease. Treatment of invasive infection in haematopoietic stem cell recipients by adoptive transfer of adenovirus- specific T cells is also under investigation. Keywords adenovirus infections; adoptive immunotherapy; cidofovir; haematopoietic stem cell transplantation; infectious disease trans- mission; keratoconjunctivitis; pharyngoconjunctival fever; pneumonia, viral; polymerase chain reaction; viral vaccines Adenoviruses are ubiquitous DNA viruses that are common causes of febrile illness in early childhood.1 They are most often associated with respiratory tract disease, but can also cause gastrointestinal, ophthalmological, cardiac, neurological and genitourinary manifestations. Although most adenoviral diseases are self-limiting, fatal invasive disease can occur in immuno- compromised patients and, occasionally, in healthy children and adults.2 Epidemiology Adenovirus infections occur worldwide throughout the year. There are more than 50 different adenovirus serotypes, classified into six groups (AeF). Group C serotypes 1, 2 and 5 are primarily associated with respiratory disease and are those most commonly isolated from the general population. Transmission occurs via aerosol particles, the faeco-oral route or contaminated fomites. Adenoviruses are non-enveloped and survive for long periods on surfaces in the environment. They are resistant to lipid disinfectants, but are inactivated by heat, formaldehyde and chlorine. Adenoviruses cause 5e10% of all febrile illnesses in infants and young children. Most individuals exhibit serological evidence of adenovirus infection by the age of 10 years. The Phyllis Flomenberg MD is an Infectious Disease Consultant at Thomas Jefferson University, Philadelphia, USA. Competing interests: none declared. viruses are common in households and day-care centres, and nosocomial transmission has been documented. Adenoviruses cause persistent infections, and prolonged faecal shedding is common. Reactivation of persistent infection can play a role in adenovirus disease in immunocompromised patients. Many epidemics of adenoviral infections have been described, including: ˇ€‹ serious outbreaks of acute respiratory disease in military recruits3 ˇ€‹ pharyngoconjunctival fever in summer camps and in associ- ation with public swimming pools and lakes ˇ€‹ haemorrhagic keratoconjunctivitis in medical facilities.4 Clinical features The clinical manifestations of adenovirus disease vary with age and the presence of immune dysfunction (Table 1). Respiratory tract: adenovirus is the virus most commonly iso- lated from young children with febrile respiratory illness. The usual duration of illness is 5e7 days, but symptoms can persist for up to 14 days. ˇ€‹ Primary diagnoses are pharyngitis, coryza and otitis media. ˇ€‹ It is common to find exudative tonsillitis and cervical adenopathy, which are clinically indistinguishable from group A streptococcal disease. ˇ€‹ Fever and other systemic manifestations (e.g. malaise, headache, myalgia, abdominal pain) are common. ˇ€‹ Pharyngitis is commonly associated with conjunctivitis, laryngo-tracheitis, bronchitis or pneumonia. Less often, adenovirus can cause a pertussis-like syndrome, bronchioli- tis, coryza without fever, or an exanthem. Pneumonia e adenovirus (most commonly group B types 3, 7 and 21) causes about 10% of cases of pneumonia in children. The pneumonia is more severe in infants than in older children and may be associated with lethargy, diarrhoea, vomiting, pharyngitis and conjunctivitis. Radiology reveals diffuse, bilat- eral pulmonary infiltrates similar to those seen in other viral pneumonias. Pathological changes include necrotizing bron- chitis, bronchiolitis, and pneumonia with mononuclear cell infiltration, hyaline membranes and necrosis. Extrapulmonary complications include meningo-encephalitis, hepatitis, myocar- ditis, nephritis, neutropenia and disseminated intravascular coagulation. In children, there is a high incidence of pulmonary sequelae, particularly bronchiectasis. Adenovirus group E type 4 and group B type 7 (and, recently, type 14) have been associated with outbreaks of severe acute respiratory disease in military recruits. VIRAL INFECTIONS What s new? C Detection of adenovirus viremia by quantitative PCR assay facilitates early diagnosis of disease in transplant recipients C Cidofovir has potential benefit for treatment of invasive adenoviral disease MEDICINE 37:12 676 O 2009 Elsevier Ltd. All rights reserved. Adenovirus infections Diseases caused by adenoviruses Infants Pharyngitis, corzya, pneumonia Otitis media Diarrhoea Children Pharyngoconjunctival fever Pneumonia Diarrhoea, mesenteric adenitis Haemorrhagic cystitis Myocarditis Young adults Acute respiratory disease Adults Epidemic keratoconjunctivitis Immunocompromised patients Pneumonia Gastroenteritis, hepatitis Haemorrhagic cystitis, nephritis Meningoencephalitis Predominant serotypes 1, 2, 5 1, 2, 5 40, 41 3, 7 3, 7, 21 40, 41, 2 11, 21 1, 2, 5 4, 7, 14 8, 19, 37 1, 2, 5 5 11, 34, 35 1, 2, 5 Table 1 Eyes: group B types 3 and 7 can cause benign follicular conjunctivitis, which can occur alone or with pharyngitis (pharyngo-conjunctival fever). Epidemic keratoconjunctivitis causes more serious disease and is associated primarily with group D types 8, 19 and 37; it is characterized by bilateral conjunctivitis associated with pre-auricular adenopathy, fol- lowed by development of painful corneal opacities that can cause prolonged impaired vision. Gastrointestinal tract: in young children, 5e15% of acute diar- rhoeal illness is caused by the enteric group F adenovirus types 40 and 41. Less commonly, other adenovirus serotypes have been associated with mesenteric adenitis, which may mimic appendicitis and occasionally cause intussusception. Hepatitis is a common complication of adenovirus pneumonia in infants. In patients who are immunocompromised, adenovirus can cause gastroenteritis; hepatitis can also occur, caused particularly by group C type 5. Genitourinary tract: adenovirus group B types 11 and 21 have been associated with the uncommon syndrome of haemorrhagic cystitis in boys. In adults, an oculogenital syndrome caused by adenovirus type 19 has occasionally been described. In immu- nocompromised patients, adenovirus group B types 11, 34 and 35 can cause haemorrhagic cystitis, with or without nephritis. Heart: adenoviruses are a common cause of acute viral myocarditis in children and have been associated with graft loss following cardiac transplantation.5 Nervous system: meningitis and encephalitis have been reported occasionally in association with adenovirus infection. Neurological involvement can be a primary manifestation or be found in association with severe pneumonia, particularly type 7. Investigations Viral culture is a sensitive method for detecting most adenovi- ruses. Most serotypes cause a characteristic cytopathic effect in human epithelial and fibroblast cell lines after a few days to weeks. In the more rapid shell-vial culture assay, samples are centrifuged directly on to cell monolayers, incubated for 1e2 days, and examined for adenovirus proteins by immunofluores- cence assay. Adenovirus can be cultured from throat, naso- pharynx, conjunctiva, sputum, urine, stool, cerebrospinal fluid, blood or tissue specimens. However, the enteric adenovirus group F types 40 and 41 do not grow in routine cell lines, and positive stool cultures for other serotypes can represent asymp- tomatic faecal shedding. Detection of adenoviral antigens in clinical samples by enzyme- linked immunosorbent assay (ELISA) or immunofluorescence assay is less sensitive than viral culture, but more rapid. ELISA is particularly useful for the fastidious serotypes 40 and 41. Detection of adenoviral DNA by polymerase chain reaction (PCR) analysis is more sensitive and rapid than viral culture. Monitoring blood samples for adenoviral DNA by quantitative PCR analysis is useful for early detection of adenovirus disease in haematopoietic stem cell recipients6,7 and can be used to follow the response to treatment.8,9 Histopathology: adenovirus disease can be confirmed by char- acteristic intranuclear inclusions in tissue sections (Figure 1). Cells develop small eosinophilic inclusions soon after infection. During the later stages of infection, basophilic inclusions appear that initially may be surrounded by a clear halo within the nucleus. When these intranuclear inclusions enlarge and obscure the nuclear membrane, the cells are termed smudge cells . On electron microscopy, icosahedral virions are seen in the nucleus and typically form large, paracrystalline aggregates. In situ DNA and immunohistochemical assays are also available. Serology: the most widely available assays (complement-fixation antibodies and ELISA) detect antibodies that react against all adenovirus serotypes. Because of the high prevalence of anti- bodies in the general population, a four-fold or greater increase in titre between paired sera is needed to confirm recent infection. Management and prevention Most adenoviral diseases are self-limiting, and treatment is supportive. Bacterial superinfection can occur. A number of reports suggest that cidofovir, a broad-spectrum monophosphate nucleotide antiviral drug, has clinical activity against adenoviral disease.8,9 However, there have been no controlled trials, and cidofovir is not licensed for this use. Treatment of invasive infection in haematopoietic stem cell recipients by adoptive transfer of adenovirus-specific T cells is also under investigation.10,11 An enteric-coated live vaccine with adenovirus group E type 4 and group B type 7 was successful in preventing epidemics of MEDICINE 37:12 677 O 2009 Elsevier Ltd. All rights reserved. VIRAL INFECTIONS Figure 1 Histopathology of fulminant adenovirus type 5 hepatitis in a bone marrow transplant recipient. Multiple cells with intranuclear inclusions at different stages are scattered throughout the hepatic parenchyma (arrows). An area of necrosis is present in the lower part of the section. acute respiratory disease in military training camps but is no longer being manufactured. New outbreaks have prompted development of a new vaccine.3 Replication-deficient adenovirus is being investigated as a vaccine vector for immunization against other infectious pathogens, including HIV-1, malaria and Ebola virus, and for treatment of a variety of tumours. VIRAL INFECTIONS Practice points C Adenoviruses are common causes of febrile illness in infants and young children C Most infections are self-limited and treatment is supportive C Fatal disease can occur in immunocompromised hosts and occasionally in healthy children and adults CLINICAL RESEARCH STUDY Adherence to Drugs That Prevent Cardiovascular Disease: Meta-analysis on 376,162 Patients Sayed H. Naderi, BMedSci, Jonathan P. Bestwick, MSc, David S. Wald, MD Wolfson Institute of Preventive Medicine, Barts and the London School of Medicine, Queen Mary University of London, Charterhouse Square, London. ABSTRACT OBJECTIVE: Combination therapy, specifically with aspirin, cholesterol and blood pressure-lowering drugs, substantially reduces the risk of coronary heart disease, but the full preventive effect is only realized if treatment continues indefinitely. Our objective was to provide a summary estimate of adherence to drugs that prevent coronary heart disease, according to drug class and use in people who have had a myocardial infarction (secondary prevention) and people who have not (primary prevention). METHODS: A meta-analysis of data on 376,162 patients from 20 studies assessing adherence using prescription refill frequency for the following 7 drug classes was performed: aspirin, angiotensinconverting enzyme inhibitors, angiotensin receptor blockers, beta-blockers, calcium-channel blockers, thiazides, and statins. Meta-regression was used to examine the effects of age, payment, and treatment duration. RESULTS: The summary estimate for adherence across all studies was 57% (95% confidence interval [CI], 50-64) after a median of 24 months. There were statistically significant differences in adherence between primary and secondary prevention: 50% (CI, 45-56) and 66% (CI, 56-75), respectively (P œ≠ .012). Adherence was lower for thiazides (42%) than for angiotensin receptor blockers (61%) in primary prevention (P œ≠ .02). There were no other statistically significant differences between any of the drug classes in primary or secondary prevention studies. Adherence decreased by 0.15% points/month (P œ≠ .07) and was unrelated to age or whether patients paid for their pills. CONCLUSION: Adherence to preventive treatment is poor and little related to class of drug, suggesting that side effects are not the main cause. General, rather than class-specific, measures at improving adherence are needed. ¬© 2012 Elsevier Inc. All rights reserved. ‚Ä¢ The American Journal of Medicine (2012) 125, 882-887 KEYWORDS: Adherence; Cardiovascular disease; Prevention SEE RELATED EDITORIAL AND RELATED ARTICLE pp. 841 and 888 Coronary heart disease is the leading cause of death in both developed and developing countries, accounting for approximately one third of all deaths worldwide.1 Combination therapy specifically with aspirin, blood pressure, and serum cholesterol-lowering drugs is effective in pre- Funding: None. Conflict of Interest: None. Authorship: All authors had access to the data and played a role in writing this manuscript. Requests for reprints should be addressed to David S. Wald, FRCP, Queen Mary University of London, Wolfson Institute of Preventive Medicine, Barts and the London School of Medicine, Charterhouse Square, London EC1M 6BQ, UK. E-mail address: d.s.wald@qmul.ac.uk 0002-9343/$ -see front matter ¬© 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2011.12.013 vention, reducing risk by an estimated 80%, with complete adherence to treatment.2-4 There is evidence that patients stop taking treatment that is intended to be taken lifelong or take it less than prescribed (nonadherence), reducing the potential preventive effect.5 The most widely reported method for studying adherence uses electronic pharmacy data to assess the extent to which patients refill prescriptions over a specified time period.6 Patients possessing medication more than approximately 75% of the time are judged to be adherent.7 Provided that patients obtain their medication from a closed pharmacy system (eg, Health Maintenance Organization), the method provides a reasonably objective measure of adherence that can be simply applied in large studies. Naderi et al Adherence to Drugs That Prevent Cardiovascular Disease 883 Studies reporting adherence using pharmacy refill data A random effects meta-analysis29 was used to combine data have yielded conflicting results, with estimates ranging from each study to give a summary estimate of adherence from 30% to 80%.8,9 There is uncertainty over the true across all studies. Stratified analyses were performed to prevalence. Differences in the tolerability of specific classes of determine adherence in primary and secondary prevention drug, the motivation of the population studied (eg, patients and for each drug class (aspirin, angiotensin-converting with or without a history of myocarenzyme inhibitors, angiotensin redial infarction), and whether paceptor blockers, beta-blocker, caltients pay for their medication are cium-channel blockers, thiazides, CLINICAL SIGNIFICANCE possible explanations for the differand statins), and according to ent estimates. There has been no method of payment. ‚óè Approximately one third of patients quantitative summary of the prevaMeta-regression was used to exwith a history of myocardial infarction lence of adherence across studies. amine possible sources of heteroand approximately one half without do This prompted us to carry out a geneity, including age, gender, not adhere to effective cardiovascular meta-analysis to quantify the prevmethod of payment, and duration of preventive treatment. alence of adherence to cardiovascutreatment. All analyses were perlar preventive medications accordformed using Stata version 10 ‚óè Nonadherence is not greatly influenced ing to the class of drug, patient (StataCorp LP, College Station, by the class of drug prescribed (aspirin, group, and method of payment. Tex).30 blood pressure-lowering drugs, or stat- MATERIALS AND METHODS ins), suggesting that side effects are not the main cause. We searched medical databases (PubMed) for studies that assessed the extent to which individuals remained on coronary heart disease preventive drug therapy over time. The search terms used were [adherence, persistence, compliance, or concordance] and [drug treatment, aspirin, angiotensin-converting enzyme inhibitors, angiotensin receptor blockers, beta-blockers, calcium channel blockers, thiazide diuretics, or statin]. We included studies of patients with and without a diagnosis of coronary heart disease receiving drugs for the prevention of a coronary heart disease event from the above 7 classes. Studies were included that measured adherence using pharmacy prescription refill data and reported the number of patients who had at least 75% of days covered by a specified drug over a defined time period. We also sought information on the method of payment for medication: self-payment (medication cost borne completely by patient), state payment (cost borne completely by the health care provider), or copayment (cost borne jointly by patient and healthcare provider). If this was not reported in the published articles, the authors were contacted, and in cases where there was no response, the method of payment generally adopted in the countries concerned was applied.10,11 The primary search generated 2433 citations, which were reduced to 80 on inspection of the titles and review of the abstracts and finally to 17 on review of the published reports. An additional 3 articles were identified through hand-searching the citation lists of relevant studies and review articles, yielding 20 studies for inclusion in the analysis.7-9,12-28 Data were abstracted independently by 2 investigators, and the datasets were cross-checked. For each study, an average estimate of adherence was calculated from the average reported adherence to each drug class weighted by the number of patients taking each drug. RESULTS Table 1 shows details of the studies in the meta-analysis, which included data on 376,162 patients (mean age 64 years, 49% were male). Eleven studies were on patients without a history of coronary heart disease (eg, patients being treated for hypertension) who were receiving drugs for the primary prevention of coronary heart disease, and 9 studies were on patients with a diagnosis of coronary heart disease (eg, after a myocardial infarction) who were receiving drugs for the secondary prevention of recurrent events. Table E1 (available online) shows the prevalence of adherence, according to each drug class in each study. Figure 1 shows a meta-analysis plot of adherence to all classes of drug in each study, stratified according to their use in primary or secondary prevention and ranked in order of increasing effect. The results yield an overall summary adherence of 57% (95% confidence interval [CI], 50-64) over a median treatment period of 24 months. The results were 50% (CI, 45-56) and 66% (CI, 56-75) for primary and secondary prevention studies, respectively (P œ≠ .012 for the difference). There was considerable heterogeneity between studies (P œΩ .001). Figure 2 shows adherence on a class-specific basis, separately for primary and secondary prevention studies. In primary prevention, there was a statistically significant lower adherence for thiazides compared with angiotensin receptor blockers (P œ≠ .02). There were no statistically significant differences between any of the other drug classes in primary or secondary prevention. Meta-regression analysis showed a 0.15 percentage point decrease in adherence per month of follow-up (P œ≠ .07). There was no effect on adherence of gender, age, or method of payment. DISCUSSION The results of this analysis show that approximately two thirds of patients with a history of coronary heart disease 884 The American Journal of Medicine, Vol 125, No 9, September 2012 Table 1 Details of Studies Included in the Meta-Analysis First Author, Year Primary Prevention Benner, 20028 Degli Esposti, 200212 Van Wijk, 200513 Simons, 200814 Marentette, 200215 Breekveldt-Postma, 200816 Perreault, 200517 Bloom, 199818 Elliott, 200719 Avorn, 199820 Grant, 200421 Secondary Prevention Kramer, 20067 Blackburn, 200526 Wei, 200422 Gislason, 200627 Kopjar, 200323 Simpson, 200324 Roughead, 201025 Ho, 200828 Rasmussen, 20079 Patient Group Studied Location Method of Payment No. of Subjects Mean Age (y) High cholesterol Hypertension Hypertension Hypertension Hypertension Hypertension Hypertension Hypertension Hypertension High cholesterol High cholesterol USA Italy Netherlands Australia Canada Netherlands Canada USA USA Canada/USA USA Copayment Copayment Copayment Copayment Copayment Copayment Copayment Copayment Copayment Copayment Copayment 231 16,783 2214 48,690 16,413 81,735 19,704 21,723 60,685 2650 4405 74 56 60 68 61 55 58 56 58 75 53 20 44 36 44 44 40 38 44 48 18 62 120 12 120 33 18 24 36 12 12 12 12 USA Canada State payment Self-payment 17,035 400 60 58 71 77 12 60 angina Scotland Denmark USA Canada Australia Copayment Copayment Copayment Copayment Copayment 386 2976 8768 14,507 9635 66 68 65 74 78 59 63 100 57 74 12 60 18 12 61 PCI, USA Copayment 15,767 66 70 48 Canada State payment 31,455 76 55 12 Myocardial infarction Myocardial infarction, PCI, CABG Myocardial infarction Myocardial infarction Myocardial infarction, Myocardial infarction Myocardial infarction, or IHD Myocardial infarction, CABG Myocardial infarction angina, angina Male (%) Follow-up (mo) PCI œ≠ percutaneous coronary intervention; CABG œ≠ coronary artery bypass grafting. adhere to drugs prescribed to prevent a second coronary heart disease event and approximately half of those without coronary heart disease adhere to drugs prescribed to prevent a first event. Previous reviews on adherence to cardiovascular preventive treatments were descriptive and did not combine results within or across studies to provide a quantitative summary of effect.31,32 This is the first meta-analysis to do this and shows that after allowing for the 15 percentage point difference in adherence between primary and secondary prevention, there were small differences in adherence between different classes of drug. This suggests that specific drug properties (eg, side effects and frequency of dosing) have a minor influence on whether patients remain on treatment long-term. General factors, such as the frequency and intensity of clinical review by the prescriber,26 whether prescriptions are initiated in primary or secondary care,27 and the level of patient education,8,28 may be more important determinants of adherence. Heterogeneity between studies was observed and is expected because of such variations in practice, but there was insufficient information on these factors in the study reports for a quantitative assessment across studies. We restricted our analysis to studies that measured adherence by prescription refills because this is the most widely used method in the literature and can be simply applied to large populations using a closed pharmacy system. However, the method has certain limitations.6,33,34 It relies on patients obtaining their medications from a single provider (eg, a single health insurance plan), which may not always be the case. Also, evidence that a prescription has been refilled is not evidence that it has been taken, so estimates are likely to be overestimates. Other methods for assessing adherence to medication include electronic medication monitors, pill counts, and measurements of physiologic markers (eg, blood pressure).35 The use of different methods in different studies limits the extent to which results from different studies can be combined. The simplest method is to ask patients whether they are still taking their medication, although this approach is open to bias and possible overestimation. Our PubMed search, for example, identified 7 studies that assessed adherence in this way (all in patients who had experienced a myocardial infarction), including 49,791 patients over a median follow-up of 14 months.36-42 The summary estimate for adherence in these studies was 90% (95% CI, 87-92). Perhaps the most reliable method for assessing adherence is direct analysis of the drug or drug metabolite. If such tests could be made simple and affordable, they would provide a useful standard method for assessing adherence in both clinical studies and practice. Naderi et al Adherence to Drugs That Prevent Cardiovascular Disease 885 Figure 1 Percent adherence to all classes of drug in each study, stratified according to their use in primary or secondary prevention. CI œ≠ confidence interval. There are approximately 400,000 deaths from coronary heart disease in the United States each year.43 Combination therapy reduces the risk of a first or second event by approximately 80%,2-4 so with broad coverage of the population at risk (all patients who have a history of cardiovascular disease and all others aged œæ 50 years) and full adherence to treatment, approximately 320,000 deaths could be avoided each year. Applying the overall adherence rates from our study (and assuming an equal proportion of deaths are first and second events) indicates that nonadherence leads to approximately 130,000 (41% of 320,000) avoidable deaths each year in the United States. These estimates are intended to give an approximate indication of the consequences of nonadherence on coronary heart disease death and do not take into account the number of stroke deaths each year (œ≥130,000) or nonfatal cardiovascular disease events. Even if these estimates were half as great, the cost of nonadherence is substantial. Few interventions to improve adherence to cardiovascular drugs have been successfully implemented, and those that have shown a benefit have tended to be complex, costly, and difficult to sustain.44-46 There is a need for a simple and effective method that can be combined with each new prescription. Text messaging has been used successfully to improve the use of sunscreen and other treatments.47,48 In a randomized trial of 80 patients, a single text message sent weekly doubled the rate of sunscreen use among those receiving text messages compared with controls.47 Mobile phones are widely used, and text messaging may be a useful way of encouraging adherence. CONCLUSIONS Our results show that approximately one third of patients who have had a myocardial infarction and approximately one half of those who have not had a myocardial infarction do not adhere to effective cardiovascular preventive treatment long-term. Adherence is not greatly dependent on the class of drug prescribed, suggesting that interventions to improve adherence need to be broadly applied. 886 The American Journal of Medicine, Vol 125, No 9, September 2012 Primary prevention Drug class Secondary prevention Number of studies Adherence (%, 95% CI) Number of studies Adherence (%, 95% CI) Aspirin 0 - 2 65 (53, 77) ACE inhibitors 9 56 (49, 64) 6 70 (66, 75) ARB‚Äôs 6 61 (51, 70) 0 - Beta blockers 6 44 (38, 51) 7 62 (49, 76) CCB‚Äôs 8 48 (38, 58) 2 76 (69, 82) Diuretics 7 42 (34, 50) 0 - Statins 4 57 (51, 64) 7 76 (70, 82) 0 10 20 30 40 50 60 70 80 90 100 Percent Adherent 0 10 20 30 40 50 60 70 80 90 100 Percent Adherent Figure 2 Percent adherence according to drug class and use in primary and secondary prevention. CI œ≠ confidence interval; ACE œ≠ angiotensin-converting enzyme; ARB œ≠ angiotensin receptor blocker; CCB œ≠ calcium channel blocker. References 1. World Health Organization. The top 10 causes of death. Available at: http://www.who.int/mediacentre/factsheets/fs310/en/index.html. Accessed February 2, 2012. 2. Wald NJ, Law MR. A strategy to reduce cardiovascular disease by more than 80%. BMJ. 2003;326:1419-1423. 3. Law MR, Wald NJ, Morris JK, Jordan RE. Value of low dose combination treatment with blood pressure lowering drugs: analysis of 354 randomised trials. BMJ. 2003;326:1427-1431. 4. Law MR, Wald NJ, Rudnicka AR. Quantifying effect of statins on low density lipoprotein cholesterol, ischaemic heart disease and stroke: systematic review and meta-analysis. BMJ. 2003;326:1423-1427. 5. Medicines Adherence: Involving Patients in Decisions About Prescribed Medicines and Supporting Adherence. Full Guideline January 2009. Hyde Park, London: National Collaborating Centre for Primary Care, The Royal College of General Practitioners; 2009. 6. Steiner JF, Prochazka AV. The assessment of refill compliance using pharmacy records: methods, validity and applications. J Clin Epidemiol. 1997;50:619-625. 7. Kramer JM, Hammill B, Anstrom KJ, et al. National evaluation of adherence to beta-blocker therapy for 1 year after acute myocardial infarction in patients with commercial health insurance. Am Heart J. 2006;152:454e.1-454e.8. 8. Benner JS, Glynn RJ, Mogun H, Neumann PJ, Weinstein MC, Avorn J. Long-term persistence in use of statin therapy in elderly patients. JAMA. 2002;288:455-461. 9. Rasmussen JN, Chong A, Alter DA. Relationship between adherence to evidence based pharmacotherapy and long-term mortality after acute myocardial infarction. JAMA. 2007;297:177-186. 10. Expat Focus. Available at: http://www.expatfocus.com/expatriatenetherlands-holland-healthcare-medical. Accessed February 2, 2012. 11. Socialist Health Association. Available at: http://www.sochealth. co.uk/news/prescriptions.htm. Accessed February 2, 2012. 12. Degli Esposti L, Valpiani G, Saragoni S, Capone A. A retrospective, population based analysis on persistence with antihypertensive drug therapy in primary care practice in Italy. Clin Ther. 2002;24: 1347-1357. 13. Van Wijk BLG, Klungel OH, Heerdink ER, de Boer A. Rate and determinants of 10 year persistence with antihypertensive drugs. J Hypertens. 2005;23:2101-2107. 14. Simons LA, Ortiz M, Calcino G. Persistence with antihypertensive medication: Australia wide experience, 2004-2006. Med J Aust. 2008; 188:224-227. 15. Marentette MA, Gerth WC, Billings DK, Zarnke KB. Antihypertensive persistence and drug class. Can J Cardiol. 2002;18:649-656. 16. Breekveldt-Postma NS, Penning-van Beest FJA, Siiskoven SJ, et al. The effect of discontinuation of antihypertensives on the risk of acute myocardial infarction and stroke. Curr Med Res Opin. 2008;24:121127. 17. Perreault S, Lamarre D, Blais L, et al. Persistence with treatment in newly treated middle-aged patients with essential hypertension. Ann Pharmacother. 2005;39:1401-1408. 18. Bloom BS. Continuation of initial antihypertensive medication after 1 year of therapy. Clin Ther. 1998;20:671-681. 19. Elliott WJ, Plauschinat CA, Skrepnek GH, Gause D. Persistence, adherence, and risk of discontinuation associated with commonly prescribed antihypertensive drug monotherapies. J Am Board Fam Med. 2007;20:72-80. 20. Avorn J, Monette J, Lacour A, et al. Persistence of use of lipid lowering medications: a cross national study. JAMA. 1998;279:14581462. 21. Grant RW, O‚ÄôLeary KM, Weilburg JB, Singer DE, Meigs JB. Impact of concurrent medication use on statin adherence and refill persistence. Arch Intern Med. 2004;164:2343-2348. 22. Wei L, Flynn R, Murray GD, MacDonal TM. Use and adherence to beta-blockers for secondary prevention of myocardial infarction: who is not getting the treatment? Pharmacoepidemiol Drug Saf. 2004;13: 761-766. 23. Kopjar B, Sales AEB, Pineros S, Sun H, Li YF, Hedeen AN. Adherence with statin therapy in secondary prevention of coronary heart disease in veterans administration male population. Am J Cardiol. 2003;92:1106-1108. 24. Simpson E, Beck C, Richard H. Drug prescriptions after acute myocardial infarction: dosage, compliance and persistence. Am Heart J. 2003;145:438-444. 25. Roughead EE, Vitry AI, Preiss AK, et al. Assessing overall duration of cardiovascular medicines in veterans with established cardiovascular disease. Eur J Cardiovasc Prev Rehabil. 2010;17:71-76. 26. Blackburn DF, Dobson RT, Blackburn JL, et al. Adherence to statins, beta blockers and angiotensin converting enzyme inhibitors following Naderi et al 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. Adherence to Drugs That Prevent Cardiovascular Disease a first cardiovascular event: a retrospective cohort study. Can J Cardiol. 2005;21:485-488. Gislason GH, Rasmussen JN, Abildstrom SZ, et al. Long-term compliance with beta-blockers, angiotensin-converting enzyme inhibitors and statins after acute myocardial infarction. Eur Heart J. 2006;27: 1153-1158. Ho PM, Magid DJ, Shetterly SM, et al. Medication nonadherence is associated with a broad range of adverse outcomes in patients with coronary artery disease. Am Heart J. 2008;155:772-779. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7:177-188. StataCorp. Stata Statistical Software: Release 10. College Station, TX: StataCorp LP; 2007. Ho PM, Bryson CL, Rumsfeld JS. Medication adherence its importance in cardiovascular outcomes. Circulation. 2009;119:3028-3035. Patient compliance with statins. Available at: www.medicine.ox.ac.uk/ bandolier/booth/cardiac/patcomp.html. Accessed February 2, 2012. Hamilton RA, Briceland LL. Use of prescription-refill records to assess patient compliance. Am J Hosp Pharmacol. 1992;49:1691-1696. Tamblyn R, Lavoie G, Petrella L, et al. The use of prescription claims databases in pharmacoepidemiological research: the accuracy and comprehensiveness of the prescription claims databases in Quebec. J Clin Epidemiol. 1995;48:999-1009. Osterberg L, Blaschke T. Adherence to medication. N Engl J Med. 2005;353:487-497. Gehi AK, Ali S, Na B, Whooley MA. Self-reported medication adherence and cardiovascular events in patients with stable coronary heart disease. The Heart and Soul Study. Arch Intern Med. 2007;167: 1798-1803. Wald DS, Morton G, Walker K, et al. Long-term continuation on cardiovascular drug treatment in patients with coronary heart disease. Ann Pharmacother. 2007;41:1644-1647. 887 38. Melloni C, Alexander KP, Ou F, et al. Predictors of early discontinuation of evidence based medicine after acute coronary syndrome. Am J Cardiol. 2009;104:175-181. 39. Eagle KA, Kline-Rogers E, Goodman SG, et al. Adherence to evidence-based therapies after discharge for acute coronary syndromes: an ongoing prospective, observational study. Am J Med. 2004;117:73-81. 40. Sud A, Kline-Rogers EM, Eagle KA, et al. Adherence to medications by patients after acute coronary syndromes. Ann Pharmacother. 2005; 39:1792-1797. 41. Newby LK, La Poiante NMA, Chen AY, et al. Long-term adherence to evidence based secondary prevention therapies in coronary artery disease. Circulation. 2006;113:203-212. 42. Amar J, Ferrieres J, Cambou JP, et al. Persistence of combination of evidence based medical therapy in patients with acute coronary syndromes. Arch Cardiovasc Dis. 2008;101:301-306. 43. Roger VL, Go AS, LIoyd-Jones DM, et al. Heart disease and stroke statistics‚Äî2011 update: a report from the American Heart Association. Circulation. 2011;123;e18-e209. 44. Haynes RB, McDonald H, Garg AX, Montague P. Interventions for helping patients follow prescriptions for medications. Cochrane Database Syst Rev. 2002;2:CD000011. 45. Piette JD. Interactive voice response systems in the diagnosis and management of chronic disease. Am J Manage Care. 2000;6:817-827. 46. Murray MD, Young J, Hoke S, et al. Pharmacist intervention to improve medication adherence in heart failure: a randomised trial. Ann Intern Med. 2007146:714-725. 47. Armstrong AW, Watson AJ, Makredes M, et al. Text-message reminders to improve sunscreen use: a randomized, controlled study using electronic monitoring. Arch Dermatol. 2009;145:1230-1236. 48. Free C, Knight R, Robertson S, et al. Smoking cessation support delivered via mobile phone text messaging (txt2stop): a single blind, randomised trial. Lancet. 378;9785:49-55. 887.e1 Table E1 The American Journal of Medicine, Vol 125, No 9, September 2012 Number of Subjects Assessed (n) and Percent Adherence (%) According to Specified Drug Class in Each Study First Author Primary Prevention Benner8 Degli Esposti12 Van Wijk13 Simons14 Marentette15 BreekveldtPostma16 BreekveldtPostma16 Perreault17 Bloom18 Elliott19 Avorn20 Grant21 Secondary Prevention Kramer7 Blackburn26 Wei22 Gislason27 Kopjar23 Simpson24 Roughead25 Ho28 Rasmussen9 Aspirin n (%) Angiotensin Calcium Channel Angiotensin-Converting Receptor Blockers Beta-blockers Blockers Thiazides Enzyme Inhibitors n (%) n (%) n (%) n (%) n (%) ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî 4986 342 27,266 4826 11,893 ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî 5693 (58) 5842 (67) 21,138 (65) ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî 77 ‚Äî ‚Äî ‚Äî 943 ‚Äî ‚Äî 9134 (71) 6287 7073 (59) 6691 ‚Äî 10,021 ‚Äî ‚Äî ‚Äî (40) (62) (45) (55) (59) (42) (70) (69) (75) (78) ‚Äî 2459 ‚Äî 1152 16,068 (47) ‚Äî ‚Äî 2807 4497 (60) 34,884 317 (58) (38) (32) (50) (39) ‚Äî ‚Äî 4680 180 9251 2342 6396 (31) (53) (31) (52) (39) ‚Äî 4341 540 ‚Äî 6438 13,854 Statins n (%) 231 (32) (30) ‚Äî (38) ‚Äî ‚Äî (30) ‚Äî (46) ‚Äî ‚Äî ‚Äî 2628 (59) 3246 (57) 567 (71) 4994 (50) 9314 (69) ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî 2944 (58) 5094 (59) 11,520 (60) ‚Äî ‚Äî 5193 (48) ‚Äî 5226 (44) ‚Äî 18,713 (56) ‚Äî ‚Äî 2650 (64) ‚Äî 4405 (68) ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî 4191 (72) ‚Äî 9168 (79) ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî 17,035 116 386 2038 ‚Äî 7595 ‚Äî 11,865 24,319 (45) (50) (59) (55) (72) (71) (84) 10,211 (60) ‚Äî 207 (49) ‚Äî 503 8768 2970 7225 13,596 17,826 (83) (71) (80) (84) (74) (88) IMAGING FOR THE CLINICIAN SPECIAL SECTION CLINICAL RESEARCH STUDY Robert G. Stern, MD, Section Editor Adherence to PIOPED II Investigators‚Äô Recommendations for Computed Tomography Pulmonary Angiography Daniel M. Adams, MD,a Scott M. Stevens, MD,a,b Scott C. Woller, MD,a,b R. Scott Evans, PhD,c James F. Lloyd, BS,c Gregory L. Snow, PhD,d Todd L. Allen, MD,e Joseph R. Bledsoe, MD,e Lynette M. Brown, MD, PhD,f,b Denitza P. Blagev, MD,f Todd D. Lovelace, MD,g Talmage L. Shill, MD,g Karen E. Conner, MD, MBA,g Valerie T. Aston, RRT,f C. Gregory Elliott, MDa,b a Department of Medicine, Intermountain Medical Center, Murray, Utah; bDepartment of Medicine, University of Utah School of Medicine, Salt Lake City; cMedical Informatics, LDS Hospital, Salt Lake City, Utah; dMedical Statistics, LDS Hospital, Salt Lake City, Utah; eDepartment of Emergency Medicine, Intermountain Medical Center, Murray, Utah; fDivision of Pulmonary and Critical Care Medicine, Intermountain Medical Center, Murray, Utah; gDepartment of Radiology, Intermountain Medical Center, Murray, Utah. ABSTRACT BACKGROUND: Computed tomography (CT) pulmonary angiography use has increased dramatically, raising concerns for patient safety. Adherence to recommendations and guidelines may protect patients. We measured adherence to the recommendations of Prospective Investigation of Pulmonary Embolism Diagnosis (PIOPED II) investigators for evaluation of suspected pulmonary embolism and the rate of potential false-positive pulmonary embolism diagnoses when recommendations of PIOPED II investigators were not followed. METHODS: We used a structured record review to identify 3500 consecutive CT pulmonary angiograms performed to investigate suspected pulmonary embolism in 2 urban emergency departments, calculating the revised Geneva score (RGS) to classify patients as ‚Äúpulmonary embolism unlikely‚Äù (RGS ’Ö 10) or ‚Äúpulmonary embolism likely‚Äù (RGS œæ 10). CT pulmonary angiograms were concordant with PIOPED II investigator recommendations if pulmonary embolism was likely or pulmonary embolism was unlikely and a highly sensitive D-dimer test result was positive. We independently reviewed 482 CT pulmonary angiograms to measure the rate of potential false-positive pulmonary embolism diagnoses. RESULTS: A total of 1592 of 3500 CT pulmonary angiograms (45.5%) followed the recommendations of PIOPED II investigators. The remaining 1908 CT pulmonary angiograms were performed on patients with an RGS ’Ö 10 without a D-dimer test (n œ≠ 1588) or after a negative D-dimer test result (n œ≠ 320). The overall rate of pulmonary embolism was 9.7%. Potential false-positive diagnoses of pulmonary embolism occurred in 2 of 3 patients with an RGS ’Ö 10 and a negative D-dimer test result. CONCLUSIONS: Nonadherence to recommendations for CT pulmonary angiography is common and exposes patients to increased risks, including potential false-positive diagnoses of pulmonary embolism. ¬© 2013 Elsevier Inc. All rights reserved. ‚Ä¢ The American Journal of Medicine (2013) 126, 36-42 KEYWORDS: Computed tomography; Guidelines; Pulmonary embolism SEE RELATED EDITORIAL p. 3 The widespread use of computed tomography (CT) pulmonary angiography for evaluation of suspected pulmonary embolism raises concern over potential harm to patients.1 CT pulmonary angiography exposes patients to ionizing radiation and the risk of renal injury, soft tissue injury, or anaphylaxis from intravenous contrast.2,3 Furthermore, estimates sug- Presented at: the American Thoracic Society meeting, May 16, 2011, Denver, Colorado. Authorship: All authors had access to the data and played a role in writing this manuscript. Requests for reprints should be addressed to C. Gregory Elliott, MD, University of Utah School of Medicine, Department of Medicine, Intermountain Medical Center, 5121 S. Cottonwood, #307, Murray, UT 84107. E-mail address: greg.elliott@imail.org Funding: Supported in part by the North American Thrombosis Forum Traveling Fellowship. Conflict of Interest: None. 0002-9343/$ -see front matter ¬© 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.amjmed.2012.05.028 Adams et al Adherence to Recommendations for CT Pulmonary Angiography 37 gest that more false-positive diagnoses of pulmonary Enterprise Data Warehouse and identified consecutive patients embolism occur when CT pulmonary angiography is perwho had chest CT ordered in the emergency departments at formed on populations with a low prevalence (œΩ5%) of pulLDS Hospital or Intermountain Medical Center. The emermonary embolism.4 False-positive interpretations of CT pulgency department physician‚Äôs report was reviewed manually monary angiograms expose patients to risks of antithrombotic by one investigator (DA), who used a data abstraction form therapies, including fatal and nonfato verify that each CT pulmonary tal major bleeding, and repeat CT angiogram was performed to inpulmonary angiograms that further vestigate suspected pulmonary CLINICAL SIGNIFICANCE increase radiation exposure.5 embolism.13-16 We analyzed The nonspecific clinical presen3500 consecutive CT pulmonary ‚óè Current evidence-based recommendations tation of pulmonary embolism and angiograms ordered to evaluate and guidelines agree that assessment of the potential risks associated with clinically suspected pulmonary pretest probability is essential and that CT pulmonary angiography motiembolism. No CT pulmonary ancomputed tomography (CT) pulmonary vate physicians to seek optimal apgiograms were excluded. angiography is unnecessary when pretest proaches for the investigation of probability is low and a sensitive D-dimer suspected pulmonary embolism. Data Collection and test result is negative. Such approaches often take the Processing form of expert recommendations ‚óè Physicians often order CT pulmonary anData elements were collected and guidelines. In 2006, Prospective giography, although pulmonary emboboth by electronic query and by Investigation of Pulmonary Embomanual review of the electronic lism is unlikely and a D-dimer test result lism Diagnosis (PIOPED II) invesmedical record. Electronically is negative. tigators published recommendacaptured historical elements ‚óè Nonadherence to recommendations for CT tions for the diagnostic approach to were verified for each patient by patients with suspected pulmonary pulmonary angiography exposes patients manual review of the emergency embolism.6 In 2008, the European to increased risk, including false-positive department physician‚Äôs report by Society of Cardiology published diagnoses of pulmonary embolism. using a data abstraction form. To guidelines on the diagnosis of pulassess the accuracy of elements monary embolism,7 and in 2011, captured electronically, one inthe American College of Emervestigator (DA) performed an ingency Physicians published a clinical policy for evaluation of dependent manual review of data elements from 100 8 adults with suspected pulmonary embolism. consecutive records and confirmed 100% agreement. ElCurrent recommendations and guidelines to evaluate susements not captured electronically but apparent in the pected pulmonary embolism agree that assessment of pretest manual review (eg, surgery at an outside institution) were probability is an essential step before ordering a CT pulmonary recorded in the study database. Appendix 1 lists the 6-8 angiogram. When pulmonary embolism is unlikely, a negmethod of collection for each data element. ative highly sensitive D-dimer test result allows anticoagulants Pretest probability for pulmonary embolism was as6,9-12 to be withheld safely without CT pulmonary angiography. sessed using the revised Geneva score (RGS).17 This tool Furthermore, when pulmonary embolism is unlikely and a has been validated prospectively11 and is composed of obhighly sensitive D-dimer test result is negative, CT pulmonary jective data elements. Prior venous thromboembolism was angiography may lead to false diagnoses of pulmonary embocollected using a computer program developed at Interlism more often than true diagnoses of pulmonary embolism. mountain Healthcare with high sensitivity and specificity The primary objective of this study was to measure the for identification of prior venous thromboembolism.18 Reproportion of CT pulmonary angiograms that were concorcent surgery was extracted from the Operating Room Mandant with recommendations for the evaluation of suspected agement Information System. Recent fracture was identified pulmonary embolism.6 by International Classification of Diseases, Ninth Revision codes 800-829. Cancer was identified by International ClasMATERIALS AND METHODS sification of Diseases, Ninth Revision codes 140-239, and active cancer was verified by manual record review. UniStudy Site and Population lateral leg pain and hemoptysis were recorded manually Intermountain Medical Center is a university-affiliated terfrom the emergency physician‚Äôs report. If these were not tiary medical center in Murray, Utah, and LDS Hospital is described in the emergency physician‚Äôs report, we cona community hospital in Salt Lake City, Utah. In 2009 and cluded that they were absent. Heart rate was defined as the 2010, emergency department visits exceeded 83,000 and highest recorded value from the vital signs before CT pul25,000, respectively. The Intermountain Healthcare Institumonary angiography. Pain on palpation of the deep veins of tional Review Board approved this Health Insurance Portabilthe leg and unilateral edema (also accepted was the physiity and Accountability Act-compliant study and waived written cian‚Äôs interpretation that findings were consistent with deep informed consent. We queried the Intermountain Healthcare 38 vein thrombosis) were recorded from the physical examination component of the physician‚Äôs note, and if they were not recorded, they were considered absent. Both emergency departments used high-sensitivity D-dimer tests (Stago latex agglutination [Diagnostica Stago, Parsippany, NJ] and mini-VIDAS [bioMerieux, Durham, NC]). Values from 0 to 499 ng/mL were considered negative. All CT pulmonary angiograms were performed on 64 detector row CT scanners. Intermountain Medical Center used Toshiba Aquilion CT scanners (Toshiba America Medical Systems, Inc, Tustin, Calif) for CT pulmonary angiography. LDS Hospital used both Toshiba Aquilion and General Electric VCT scanners (GE Healthcare, Waukesha, Wis). AutomA software (GE Healthcare) was used to deliver tailored radiation exposure. The standard CT pulmonary angiogram protocol used 80 mL of iso-osmolar contrast with an iodine concentration of 370 mg/mL injected at a rate of 4 to 5 mL/s and timed for maximum pulmonary artery opacification through bolus tracking (Toshiba America Medical Systems, Inc) or 20 mL timing bolus (GE Healthcare). Every CT pulmonary angiogram was interpreted by an in-house, board-certified radiologist. Primary Outcome Measure The primary measurement for this study was concordance of CT pulmonary angiogram examinations with recommendations published by the PIOPED II investigators.6 We used the RGS17 to quantify the probability of pulmonary embolism as ‚Äúpulmonary embolism likely‚Äù (RGS œæ10) or ‚Äúpulmonary embolism unlikely‚Äù (RGS ’Ö 10). CT pulmonary angiograms were classified as concordant with PIOPED II investigator recommendations if the patient‚Äôs pretest probability classification was ‚Äúpulmonary embolism likely‚Äù or the pretest probability was ‚Äúpulmonary embolism unlikely‚Äù and a D-dimer test result was positive. CT pulmonary angiograms were not concordant with PIOPED II investigator recommendations if the patient‚Äôs pretest probability was ‚Äúpulmonary embolism unlikely‚Äù and no D-dimer test was performed or if the D-dimer test result was negative. Some clinical situations, such as pregnancy and recent trauma or surgery, decrease the likelihood of a negative D-dimer test result and may influence clinicians to forego D-dimer testing. To assess the impact of such circumstances, we recorded the number of patients who were classified as ‚Äúpulmonary embolism unlikely‚Äù and did not have a D-dimer test performed but were pregnant or had recent (within 1 month) surgery or trauma. The rate of pulmonary embolism on CT pulmonary angiogram (‚Äúdiagnostic yield‚Äù) was determined from the original interpretations of all CT pulmonary angiograms (number of positive pulmonary embolism interpretations/number of CT pulmonary angiograms). The rate of potential false-positive pulmonary embolism diagnoses was determined by an independent (blinded to D-dimer testing and initial interpretation of the CT pulmonary angiogram) interpretation of 482 CT pulmonary angiograms (all 241 CT pulmonary angiograms reported to show The American Journal of Medicine, Vol 126, No 1, January 2013 acute pulmonary embolism on the initial interpretation and an equal number, 241, of randomly selected CT pulmonary angiograms reported negative for pulmonary embolism on initial interpretation). The independent interpretation was performed by 1 of 3 fellowship-trained, board-certified chest radiologists (TL, TS, and KC). Kappa coefficients for agreement of the 3 chest radiologists were calculated by overlapping 20% of the sample and were found to be 1.0, 0.94, and 0.94. Data Analysis We descriptively analyzed the patients‚Äô demographics and outcome measurements. Ninety-five percent confidence intervals were calculated for the proportion of CT pulmonary angiograms that were concordant or nonconcordant with PIOPED II investigator recommendations. The rate of pulmonary embolism with 95% confidence intervals was calculated for patients grouped by concordance with PIOPED II investigators‚Äô recommendations. Median values and interquartile range were calculated for the RGS for the entire patient group and patient subgroups defined by recommendation concordance. The rate of potential false-positive pulmonary embolism diagnoses was determined according to subgroup on the basis of concordance with PIOPED II investigator recommendations. RESULTS Sample and Patient Characteristics The reports of 5220 consecutive chest CT scans ordered from the 2 emergency departments were reviewed to identify 3500 consecutive CT pulmonary angiograms ordered between May 22, 2009, and June 30, 2010, for suspected pulmonary embolism. A total of 2755 (78.7%) CT pulmonary angiograms were performed at Intermountain Medical Center, and 745 (21.3%) CT pulmonary angiograms were performed at LDS Hospital. Table 1 shows the clinical characteristics for all patients and subgroups prespecified by concordance with PIOPED II investigator recommendations. The ‚Äúpulmonary embolism likely‚Äù subgroup differed from the ‚Äúpulmonary embolism unlikely‚Äù subgroup in variables used to calculate the RGS. Patients in the ‚Äúpulmonary embolism unlikely‚Äù (RGS ’Ö10) group (n œ≠ 1588) for whom D-dimer testing was not ordered were more likely to have had recent surgery, trauma, or fracture; active cancer; or prior venous thromboembolism than patients in the ‚Äúpulmonary embolism unlikely‚Äù group (n œ≠ 1745) who had D-dimer levels measured. Other variables (eg, pregnancy or unilateral leg pain) were observed at similar rates for the patients in the ‚Äúpulmonary embolism unlikely‚Äù group who did or did not have a D-dimer test ordered. Concordance With PIOPED II Investigators‚Äô Recommendations A total of 1592 of 3500 (45.5%) CT pulmonary angiograms were concordant with the PIOPED II investigators‚Äô recom- Adams et al Table 1 Adherence to Recommendations for CT Pulmonary Angiography 39 Patient Characteristics* RGS ’Ö 10 and All N Age, y, mean (SD) Female, n (%) Prior VTE, n (%) Surgery,‚Ä† n (%) Fracture,‚Ä† n (%) Trauma, n (%) Cancer,‚Ä° n (%) Pregnancy, n (%) Hemoptysis, n (%) Leg pain, n (%) Leg edema, n (%) Heart rate minœ™1, mean (SD) 3500 52 2204 682 399 54 117 244 70 69 249 231 94.0 (19.0) (63.0) (19.5) (11.4) (1.5) (3.3) (7.0) (2.0) (2.0) (7.1) (6.6) (20.5) RGS œæ 10 D-dimer Positive D-dimer Negative D-dimer Done D-dimer Not Done 167 56 (19.6) 101 (60.5) 97 (58.1) 33 (19.8) 7 (4.2) 10 (6.0) 29 (17.4) 3 (1.8) 6 (3.6) 115 (68.9) 133 (79.6) 104.0 (17.1) 1425 54 (19.3) 928 (65.1) 153 (10.7) 117 (8.2) 13 (0.9) 34 (2.4) 59 (4.1) 29 (2.0) 17 (1.2) 48 (3.4) 36 (2.5) 93.0 (20.5) 320 44 200 50 12 2 8 9 2 6 17 7 91.0 1745 52 1128 203 119 15 42 68 31 23 65 43 92.7 1588 51 (18.8) 975 (61.4) 382 (24.1) 237 (14.9) 32 (2.0) 64 (4.0) 147 (9.3) 36 (2.3) 40 (2.5) 69 (4.3) 55 (3.5) 94.4 (20.8) (15.9) (62.5) (15.6) (3.8) (0.6) (2.5) (2.8) (0.6) (1.9) (5.3) (2.2) (18.5) (19.1) (64.6) (11.6) (6.8) (0.9) (2.4) (3.9) (1.8) (1.3) (3.7) (2.5) (20.2) RGS œ≠ revised Geneva score;17 SD œ≠ standard deviation; VTE œ≠ venous thromboembolism. *Characteristics of the patient at the time of each CT pulmonary angiogram were included. ‚Ä†Within 2 mo. ‚Ä°Currently treated or cure œΩ 1 y. mendations. Of these, 167 (4.8%) were ‚Äúpulmonary embolism likely‚Äù and 1425 (40.7%) were ‚Äúpulmonary embolism unlikely‚Äù with a positive highly sensitive D-dimer test result. The remaining 1908 (54.5%) CT pulmonary angiograms were not concordant with the PIOPED II investigators‚Äô recommendations. Of these, 1588 (45.4%) were ‚Äúpulmonary embolism unlikely‚Äù without a D-dimer test, and 320 (9.1%) were ‚Äúpulmonary embolism unlikely‚Äù with a negative D-dimer test. Of the 1588 patients who were classified as ‚Äúpulmonary embolism unlikely‚Äù without D-dimer tests, 319 (20.1%) had a clinical situation in which the clinician might forego D-dimer testing (pregnancy, recent surgery, or trauma). If these cases were considered concordant with recommendations of the PIOPED II investigators, the rate of CT pulmonary angiogram concordance with recommendations of the PIOPED II investigators increased from 45.5% to 54.6%. Table 2 Diagnostic Yield Table 2 shows the diagnostic yield and RGS data related to concordance with the recommendations of the PIOPED II investigators. On the basis of the initial interpretations of 3500 CT pulmonary angiograms, the overall pulmonary embolism rate was 9.7%. The pulmonary embolism rate for patients in the ‚Äúpulmonary embolism likely‚Äù (RGS œæ10) group was 37.1%, and the pulmonary embolism rate for patients in the ‚Äúpulmonary embolism unlikely‚Äù (RGS ’Ö 10) group with a negative D-dimer test result was 0.9%. The pulmonary embolism rate (10.0%) for patients in the ‚Äúpulmonary embolism unlikely‚Äù (RGS ’Ö 10) group with a positive D-dimer test result was higher than the pulmonary embolism rate (8.4%) for patients in the ‚Äúpulmonary embolism unlikely‚Äù (RGS ’Ö 10) group in whom D-dimer was not tested. CT pulmonary angiograms classified as concordant with recommendations were positive for pulmonary embo- Recommendation Concordance, Revised Geneva Score, and Diagnostic Yield Concordant subgroups PE likely PE unlikely and positive D-dimer Nonconcordant subgroups PE unlikely and no D-dimer PE unlikely and negative D-dimer Revised Geneva Score Diagnostic Yield‚Ä† n % 95% CI Median (IQR)* n % PE 95% CI 167 1425 4.8 40.7 4-6 39-42 12 (11-13) 5 (3-6) 62 142 37.1 10.0 29.8-44.5 8.4-11.5 1588 320 45.4 9.1 44-47 8-10 5 (4-7) 5 (3-5) 134 3 8.4 0.9 7.1-9.8 0.2-2.5 CI œ≠ confidence interval; IQR œ≠ interquartile range; PE œ≠ pulmonary embolism; RGS œ≠ revised Geneva score.17 *Median RGS and IQR are for all patients in the group, not just those with a positive CT pulmonary angiogram. ‚Ä†Diagnostic yield œ≠ number of CT pulmonary angiogram examinations with PE/number of CT pulmonary angiogram examinations. 40 The American Journal of Medicine, Vol 126, No 1, January 2013 Table 3 Rate of Potential False-Positive Computed Tomography Pulmonary Angiogram Interpretations Concordant Potential false œ© rate (95% CI) Not Concordant RGS œæ 10 RGS ’Ö 10 and Positive D-dimer RGS ’Ö 10 and no D-dimer RGS ’Ö 10 and Negative D-dimer 0.00 (0.00-0.06) 0.06 (0.03-0.09) 0.03 (0.01-0.05) 0.67 (0.23-0.96) CI œ≠ confidence interval; RGS œ≠ revised Geneva score.17 lism 12.8% of the time (204/1592) compared with 7.2% of the nonconcordant CT pulmonary angiograms (137/1908). Potential False Diagnoses of Pulmonary Embolism Independent interpretation of 241 consecutive CT pulmonary angiograms interpreted initially to show acute pulmonary embolism identified 13 potential false-positive interpretations. The potential false-positive rate of initial CT pulmonary angiogram interpretation was high (0.67) for diagnoses of acute pulmonary embolism when RGS was ’Ö 10 with a negative D-dimer test result (Table 3). DISCUSSION We found that more than half of 3500 CT pulmonary angiograms performed to investigate clinically suspected acute pulmonary embolism were not concordant with recommendations of the PIOPED II investigators. There are several possible explanations for the disparity we observed between practice and published recommendations. First, 95% of patients who underwent CT pulmonary angiography for suspected pulmonary embolism had a pretest probability of ‚Äúpulmonary embolism unlikely‚Äù (RGS ’Ö 10). Most emergency physicians assess pretest probability by gestalt rather than using a standardized tool. Although this has been reported to be safe,6 gestalt assessment may lead to additional testing because it seems less reliable for less-experienced clinicians and has poor interobserver agreement compared with standardized prediction tools.19,20 Furthermore, gestalt is less specific than validated pretest probability scores.21 Clinicians, using a gestalt assessment of pretest probability, may be more likely to perform CT pulmonary angiography on the basis of their gestalt of intermediate pretest probability of pulmonary embolism, either bypassing or overriding D-dimer tests. Second, there are several risk assessment scores. It is possible some physicians used a different risk assessment score, which may have led to a different classification of pretest probability than the RGS. However, the most widely used of these, the Wells score,22 classifies more patients in the low category than the RGS.23 Therefore, if clinicians used the Wells score, we would expect even higher use of D-dimer assays. Third, the evidence is less compelling that a highly sensitive D-dimer test excludes pulmonary embolism safely when the RGS is 4 to 10.8 This led the writing committee of the American College of Emergency Physicians to advise caution in excluding pulmonary embolism when the RGS is 4 to 10 and the D-dimer test result is negative.8 It is possible that some physicians shared the perspective that strong evidence to support the use of D-dimer testing to exclude pulmonary embolism is lacking for patients with an intermediate pulmonary embolism risk.8 This perspective could have led physicians to avoid D-dimer tests when their suspicion for pulmonary embolism was intermediate, leading to nonconcordance with recommendations of the PIOPED II investigators. Fourth, it is possible that physicians were influenced by clinical variables not represented in the RGS. Courtney et al24 recently showed that non‚Äì cancer-related thrombophilia, pleuritic chest pain, and a family history of venous thromboembolism increased the probability of pulmonary embolism. Consideration of these variables may have led physicians to bypass D-dimer testing. Although this possibility does not alter our measurement of concordance with recommendations of the PIOPED II investigators, it may lead some to conclude that we underestimated the rate of appropriate use of CT pulmonary angiography. Finally, some physicians may simply choose to perform CT pulmonary angiography in all patients with suspected pulmonary embolism, because they may be unfamiliar with recommendations and guidelines, find that assessing pretest probability is complex, or believe that CT pulmonary angiography protects them from malpractice claims. Variation exists between recommendations. In June 2011 (after the patient encounters in our study), the American College of Emergency Physicians Clinical Policies Subcommittee issued recommendations on the diagnosis of suspected pulmonary embolism,8 which differ from the recommendations of the PIOPED II investigators. Differences include acceptance of gestalt to determine pretest probability of pulmonary embolism and a low (level C) recommendation for the use of a sensitive D-dimer test to exclude pulmonary embolism in patients with an intermediate (RGS 4-10) pretest probability of pulmonary embolism.8 Irrespective of which recommendation is invoked, our data demonstrate nonconcordant use of CT pulmonary angiography for patients with a low pretest probability of pulmonary embolism (RGS 0-3) and negative sensitive D-dimer test results. Substantial research has focused on reducing the overuse of CT pulmonary angiography for suspected pulmonary Adams et al Adherence to Recommendations for CT Pulmonary Angiography embolism because CT pulmonary angiograms are costly and inconvenient, and engender risks, such as contrast dye injuries,25,26 radiation exposure,27-29 and overdiagnosis of pulmonary embolism.30,31 A combination of assessment of pretest probability and use of a sensitive D-dimer assay has been shown to identify a group of patients with suspected pulmonary embolism for whom the diagnosis can be excluded and anticoagulants can be withheld safely without CT pulmonary angiography.10,11,23 However, there are barriers to performing pretest probability assessment and sensitive D-dimer testing in emergency departments, including complexity of formalized pretest probability scoring, imperatives for shorter emergency department evaluations, and concern regarding the risks of excluding suspected pulmonary embolism without imaging. Our work confirms the safety of excluding pulmonary embolism on the basis of an objective pretest probability score of ‚Äúpulmonary embolism unlikely‚Äù (RGS ’Ö 10) and a negative D-dimer test result. We demonstrate that the diagnostic yield decreases when CT pulmonary angiography use is not concordant with recommendations of the PIOPED II investigators. Study Limitations Our study has several limitations. First, data were obtained retrospectively from medical records. Historical data were taken from the emergency physician‚Äôs report and electronic data. It is possible that there were elements that the physician did not include in the report. We tried to compensate for this limitation by using objective and well-defined data elements. Second, because we assessed only patients who underwent CT pulmonary angiography, our methods did not identify patients in whom pulmonary embolism was suspected, but deep vein thrombosis was diagnosed by venous ultrasonography, without CT pulmonary angiography. Our approach may have underestimated concordance with recommendations by excluding these patients. However, our cohort included a similar proportion of patients with pain on palpation and unilateral edema (6.6%) as in the studies by Douma et al23 (5.8%) and van Belle et al10 (5.7%), in which a clinical prediction rule was determined prospectively and venous ultrasonography was not performed. Thus, we believe the diagnosis of deep vein thrombosis without CT pulmonary angiography when pulmonary embolism was suspected had little influence on our measurement of concordance with recommendations for CT pulmonary angiography. Finally, our study included 2 institutions in the same metropolitan area staffed by the same emergency physicians and radiologists. Therefore, our results may not be representative of other settings. CONCLUSIONS CT pulmonary angiogram examinations are often not concordant with expert recommendations and guidelines for investigation of suspected pulmonary embolism. Emergency department physicians order many CT pulmonary angiograms when pulmonary embolism is unlikely and sen- 41 sitive D-dimer test results are negative. This practice lowers the diagnostic yield of CT pulmonary angiograms and exposes some patients to risks associated with CT pulmonary angiography, including false-positive test results and unnecessary treatment. References 1. Amis ES Jr, Butler PF, Applegate KE, et al. American College of Radiology white paper on radiation dose in medicine. J Am Coll Radiol. 2007;4:272-284. 2. Einstein AJ, Henzlova MJ, Rajagopalan S. Estimating risk of cancer associated with radiation exposure from 64-slice computed tomography coronary angiography. JAMA. 2007;298:317-323. 3. Lee CI, Haims AH, Monico EP, et al. Diagnostic CT scans: assessment of patient, physician, and radiologist awareness of radiation dose and possible risks. Radiology. 2004;231:393-398. 4. Corwin MT, Donohoo JH, Partridge R, et al. Do emergency physicians use serum D-dimer effectively to determine the need for CT when evaluating patients for pulmonary embolism? Review of 5,344 consecutive patients. AJR Am J Roentgenol. 2009;192:1319-1323. 5. Kline JA, Courtney DM, Beam DM, et al. Incidence and predictors of repeated computed tomographic pulmonary angiography in emergency department patients. Ann Emerg Med. 2009;54:41-48. 6. Stein PD, Woodard PK, Weg JG, et al. Diagnostic pathways in acute pulmonary embolism: recommendations of the PIOPED II investigators. Am J Med. 2006;119:1048-1055. 7. Torbicki A, Perrier A, Konstantinides S, et al. Guidelines on the diagnosis and management of acute pulmonary embolism: the Task Force for the Diagnosis and Management of Acute Pulmonary Embolism of the European Society of Cardiology (ESC). Eur Heart J. 2008;29:2276-2315. 8. Fesmire FM, Brown MD, Espinosa JA, et al. Critical issues in the evaluation and management of adult patients presenting to the emergency department with suspected pulmonary embolism. Ann Emerg Med. 2011;57:628-652 e675. 9. Anderson DR, Kahn SR, Rodger MA, et al. Computed tomographic pulmonary angiography vs ventilation-perfusion lung scanning in patients with suspected pulmonary embolism: a randomized controlled trial. JAMA. 2007;298:2743-2753. 10. van Belle A, Buller HR, Huisman MV, et al. Effectiveness of managing suspected pulmonary embolism using an algorithm combining clinical probability, D-dimer testing, and computed tomography. JAMA. 2006;295:172-179. 11. Righini M, Le Gal G, Aujesky D, et al. Diagnosis of pulmonary embolism by multidetector CT alone or combined with venous ultrasonography of the leg: a randomised non-inferiority trial. Lancet. 2008;371:1343-1352. 12. Perrier A, Roy PM, Sanchez O, et al. Multidetector-row computed tomography in suspected pulmonary embolism. N Engl J Med. 2005; 352:1760-1768. 13. Gilbert EH, Lowenstein SR, Koziol-McLain J, et al. Chart reviews in emergency medicine research: where are the methods? Ann Emerg Med. 1996;27:305-308. 14. Badcock D, Kelly A-M, Kerr D, Reade T. The quality of medical record review studies in the international emergency medicine literature. Ann Emerg Med. 2005;45:444-447. 15. Worster A, Bledsoe RD, Cleve P, et al. Reassessing the methods of medical record review studies in emergency medicine research. Ann Emerg Med. 2005;45:448-451. 16. Lowenstein SR. Medical record reviews in emergency medicine: the blessing and the curse. Ann Emerg Med. 2005;45:452-455. 17. Le Gal G, Righini M, Roy PM, et al. Prediction of pulmonary embolism in the emergency department: the revised Geneva score. Ann Intern Med. 2006;144:165-171. 18. Evans RS, Sharp JH, Linford LH, et al. Risk of symptomatic DVT associated with peripherally inserted central catheters. Chest. 2010; 138:803-810. 42 19. Chunilal SD, Eikelboom JW, Attia J, et al. Does this patient have pulmonary embolism? JAMA. 2003;290:2849-2858. 20. Rodger MA, Maser E, Stiell I, et al. The interobserver reliability of pretest probability assessment in patients with suspected pulmonary embolism. Thromb Res. 2005;116:101-107. 21. Lucassen W, Geersing GJ, Erkens PM, et al. Clinical decision rules for excluding pulmonary embolism: a meta-analysis. Ann Intern Med. 2011;155:448-460. 22. Wells PS, Anderson DR, Rodger M, et al. Derivation of a simple clinical model to categorize patients probability of pulmonary embolism: increasing the models utility with the SimpliRED D-dimer. Thromb Haemost. 2000;83:416-420. 23. Douma RA, Mos IC, Erkens PM, et al. Performance of 4 clinical decision rules in the diagnostic management of acute pulmonary embolism: a prospective cohort study. Ann Intern Med. 2011;154:709-718. 24. Courtney DM, Kline JA, Kabrhel C, et al. Clinical features from the history and physical examination that predict the presence or absence of pulmonary embolism in symptomatic emergency department patients: results of a prospective, multicenter study. Ann Emerg Med. 2010;55:307-315 e301. 25. Kooiman J, Klok FA, Mos IC, et al. Incidence and predictors of contrastinduced nephropathy following CT-angiography for clinically suspected acute pulmonary embolism. J Thromb Haemost. 2010;8:409-411. 26. Lieberman PL, Seigle RL. Reactions to radiocontrast material. Anaphylactoid events in radiology. Clin Rev Allergy Immunol. 1999;17: 469-496. 27. Berrington de Gonzalez A, Mahesh M, Kim KP, et al. Projected cancer risks from computed tomographic scans performed in the United States in 2007. Arch Intern Med. 2009;169:2071-2077. 28. Smith-Bindman R, Lipson J, Marcus R, et al. Radiation dose associated with common computed tomography examinations and the associated The American Journal of Medicine, Vol 126, No 1, January 2013 lifetime attributable risk of cancer. Arch Intern Med. 2009;169:20782086. 29. Shrimpton PC, Hillier MC, Lewis MA, Dunn M. National survey of doses from CT in the UK: 2003. Br J Radiol. 2006;79:968-980. 30. Ranji SR, Shojania KG, Trowbridge RL, Auerbach AD. Impact of reliance on CT pulmonary angiography on diagnosis of pulmonary embolism: a Bayesian analysis. J Hosp Med. 2006;1:81-87. 31. Cronin P, Kelly AM. Influence of population prevalences on numbers of false positives: an overlooked entity. Acad Radiol. 2011;18:1087-1093. Appendix 1 Revised Geneva Score and Other Data Elements and Method of Collection Data Element Age Prior VTE Surgery or fracture in prior month Active malignant condition Unilateral leg pain Hemoptysis Heart rate Pain on palpation and unilateral edema Trauma Pregnancy VTE œ≠ venous thromboembolism. Electronic Collection X X X X Manual Review and Collection X X X X X X X X X X CLINICAL RESEARCH STUDY Adherence to Secondary Prevention Medications and Four-year Outcomes in Outpatients with Atherosclerosis Dharam J. Kumbhani, MD, SM, MRCP,a Ph. Gabriel Steg, MD,b Christopher P. Cannon, MD,c,d Kim A. Eagle, MD,e Sidney C. Smith, Jr., MD,f Elaine Hoffman, PhD,d Shinya Goto, MD,g E. Magnus Ohman, MD,h Deepak L. Bhatt, MD, MPH,c,d,i; on Behalf of the REduction of Atherothrombosis for Continued Health Registry Investigators* a Division of Cardiology, University of Texas Southwestern Medical Center, Dallas, Tex; bINSERM U-698, Paris, France; Universit√© Paris Diderot, Paris, France; and Assistance Publique-H√¥pitaux de Paris, Paris, France; cBrigham and Women‚Äôs Hospital and Harvard Medical School, Boston, Mass; dTIMI Study Group, Boston, Mass; eUniversity of Michigan Cardiovascular Center, Ann Arbor, Mich; fCenter for Cardiovascular Science and Medicine, University of North Carolina at Chapel Hill, Chapel Hill, NC; gTokai University School of Medicine, Isehara, Japan; hDivision of Cardiology, Duke University, Durham, NC; iVA Boston Healthcare System, Boston, Mass. ABSTRACT BACKGROUND: Although nonadherence with evidence-based secondary prevention medications is common in patients with established atherothrombotic disease, long-term outcomes studies are scant. We assessed the prevalence and long-term outcomes of nonadherence to secondary prevention (antiplatelet agents, statins, and antihypertensive agents) medications in stable outpatients with established atherothrombosis (coronary, cerebrovascular, or peripheral artery disease) enrolled in the international REduction of Atherothrombosis for Continued Health registry. METHODS: Adherence with these medications in eligible patients at baseline and 1-year follow-up was assessed. The primary outcome was a composite of cardiovascular death, myocardial infarction, or stroke at 4 years. RESULTS: A total of 37,154 patients with established atherothrombotic disease were included. Adherence rates with all evidence-based medications at baseline and 1 year were 46.7% and 48.2%, respectively. Nonadherence with any medication at baseline (hazard ratio, 1.18; 95% conÔ¨Ådence interval, 1.11-1.25) and at 1 year (hazard ratio, 1.19; 95% conÔ¨Ådence interval, 1.11-1.28) were both signiÔ¨Åcantly associated with an increased risk of the primary end point. The risk of all-cause mortality was similarly elevated. Corresponding numbers needed to treat were 31 and 25 patients for the composite end point and total mortality, respectively. This also was true for each disease-speciÔ¨Åc subgroup. Patients who were fully adherent at both time points had the lowest incidence of adverse outcomes, whereas patients who were nonadherent at both time points had the worst outcomes (P < .01). CONCLUSIONS: Our analysis of a large international registry demonstrates that nonadherence with evidence-based secondary prevention therapies in patients with established atherothrombosis is associated with a signiÔ¨Åcant increase in long-term adverse events, including mortality. √ì 2013 Elsevier Inc. All rights reserved. The American Journal of Medicine (2013) 126, 693-700 KEYWORDS: Atherosclerosis; Cardiovascular disease; Compliance/adherence; Registry; Secondary prevention Funding: See last page of article. ConÔ¨Çict of Interest: See last page of article. Authorship: See last page of article. *A complete list of the REACH Registry Investigators appears in Bhatt DL, Steg PG, Ohman EM, et al. International prevalence, recognition, and treatment of cardiovascular risk factors in outpatients with atherothrombosis. JAMA. 2006;295:180-189. Requests for reprints should be addressed to Deepak L. Bhatt, MD, MPH, FACC, VA Boston Healthcare System, 1400 VFW Parkway, Boston, MA 02132. E-mail address: dlbhattmd@post.harvard.edu 0002-9343/$ -see front matter √ì 2013 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.amjmed.2013.01.033 Patients with established atherothrombosis, namely, coronary, cerebrovascular, and peripheral artery disease, constitute a high-risk cohort and have an elevated risk of future ischemic events. The mainstay of management in these patients is the appropriate use of evidence-based secondary prevention measures, such as antiplatelet, lipid-lowering, and antihypertensive agents.1 However, adherence to these medications in practice remains low.2 In an effort to narrow the gap between evidence and practice, the last few years have seen the launch 694 The American Journal of Medicine, Vol 126, No 8, August 2013 of several large-scale quality-improvement initiatives in only, without documented evidence of atherothrombosis at patients with established atherosclerosis. Although these baseline (Supplemental Figure, online). initiatives have resulted in signiÔ¨Åcant improvements in adherence with these measures over the past decade,3-7 they Follow-up focused almost exclusively on in-hospital and short-term use Detailed patient and medication information was collected at of these agents. Long-term adherence remains suboptimal. baseline, with annual follow-up at 1, 2, 3, and 4 years. The The REduction of Atheroinitial follow-up was planned for thrombosis for Continued Health 2 years, with an additional 2-year CLINICAL SIGNIFICANCE (REACH) Registry is one of the extension proposed shortly before largest contemporary outpatient Aspirin, statins, and antihypertensive that period ended. Not all counregistries that was initiated to evatries and sites that were in the agents have class I indications in patients luate patients who would represent 2-year follow-up cohort elected to with established atherosclerosis. the entire spectrum of stable continue participation in the reg Less than 50% of patients are fully atherosclerotic clinical syndromes: istry, although the majority did from those with risk factors (but adherent with these medications. elect to continue. who are asymptomatic) to those Nonadherence is associated with worse with established atherosclerotic Adherence to Secondary outcomes at 4 years. arterial disease within any circulaPrevention Therapies tory bed. We studied adherence Starting these medications is associated Data relating to medication use rates with evidence-based secwith a lowering in risk, whereas stopwere patient self-reported at the ondary prevention measures at ping these medications is associated time of the study visit and collected baseline and 1 year in patients with with an elevation in risk; not taking centrally via use of a standardized established coronary, cerebrovasthem at all is associated with the international case report form. For cular, or peripheral artery disease highest risk. the purpose of this analysis, an enrolled in the REACH registry. antiplatelet agent was deÔ¨Åned as We also compared 4-year clinical the use of aspirin or other similar outcomes (death, myocardial infarction, stroke) in patients on medications; lipid-lowering agent was deÔ¨Åned as the use of the basis of their adherence with these secondary prevention a statin or other lipid-lowering therapies; and antihypertenmeasures. sive agent was deÔ¨Åned as the use of any of the following: angiotensin-converting enzyme inhibitors, angiotensin II receptor antagonists, beta-blockers, calcium-channel MATERIALS AND METHODS blockers, diuretics, nitrates, or other antihypertensive agents. The use of the latter was assessed in eligible patients only, Data Source that is, those with a history of hypertension. For patients with The methods of the REACH Registry have been pub8-12 established cerebrovascular and peripheral artery disease, BrieÔ¨Çy, patients aged at least 45 years with !3 lished. adherence was deÔ¨Åned as the use of an antiplatelet agent, risk factors for atherosclerosis and patients with established a lipid-lowering agent, and any 1 antihypertensive agent in coronary, cerebrovascular, or peripheral artery disease were eligible patients. For patients with coronary artery disease, in enrolled. The multiple risk factors category consisted of addition to the above, the concomitant use of a beta-blocker diabetes, diabetic nephropathy, ankle-brachial index 0.9, also was included in the deÔ¨Ånition. Nonadherence, which asymptomatic carotid stenosis of !70%, carotid intimawas deÔ¨Åned as failure to adhere with any of these eligible media thickness at least 2 times that at adjacent sites, medications, was assessed at 2 time points: at baseline (time systolic blood pressure !150 mm Hg despite treatment, of study entry) and 1 year after enrollment. hypercholesterolemia treated with medication, current smoking of !15 cigarettes per day, and age !65 years for men or !70 years for women. Ascertainment of Outcomes The primary outcome was a composite of cardiovascular death, myocardial infarction, or stroke over 4 years. Study Population Secondary end points studied were all-cause mortality and At the time of Ô¨Ånal database lock in April 2009, 69,055 cardiovascular mortality. Cardiovascular mortality included patients from 5587 physician practices in 44 countries fatal stroke, fatal myocardial infarction, or other cardiovasbetween December 2003 and June 2004, and followed up cular death (other death of cardiac origin; pulmonary until 2008, had been enrolled. From this, we excluded 1167 embolism; any sudden death, including unobserved and patients without baseline information and 23,009 patients unexpected death unless proven otherwise by autopsy; who were enrolled in centers that did not participate in the death after a vascular operation, vascular procedure, or 4-year follow-up. Of the 45,227 patients eligible for analamputation; death attributed to congestive heart failure; ysis, we further excluded 8073 patients who had risk factors death after a visceral or limb infarction; and any other death Kumbhani et al Adherence to Secondary Prevention Medications that could not be deÔ¨Ånitely attributed to a nonvascular cause or hemorrhage). Any myocardial infarction or stroke followed by a death whatever the cause in the next 28 days was considered to be a fatal myocardial infarction or fatal stroke. End points were not adjudicated, but based on physician reporting at the time of follow-up. Statistical Analysis Mean (standard deviation) and percentages are reported for continuous and categoric variables, respectively. Cumulative incidence curves were constructed using the KaplaneMeier approach. Multivariate Cox regression analyses were conducted with time to cardiovascular death, myocardial infarction, or stroke as the primary outcome variable and nonadherence with any evidence-based measure as the primary independent variable. Hazard ratios (HRs) and their 95% conÔ¨Ådence intervals (CIs) were calculated. Adherence at baseline and 1 year was separately analyzed. For the 1-year analysis, patients with events between baseline and up to 1 year were excluded. Other variables included in these models have all been shown to be signiÔ¨Åcant independent predictors of the primary outcome at 4 years in a prior analysis.8 These include gender, age, current smoker, history of diabetes, body mass index <20 (calculated as weight in kilograms divided by height in meters squared), timing of ischemic event ( 1 year or >1 year), polyvascular disease versus single vascular disease, congestive heart failure, atrial Ô¨Åbrillation/Ô¨Çutter, and Eastern Europe, Middle East, or Japan versus other regions. Geographic regions were collapsed into higher-risk (Eastern Europe and Middle East) and lower-risk (Japan/Australia) locations. Next, we subdivided the study population into 4 groups: consistent adherers (fully adherent at baseline and at 1 year); negative converters (fully adherent at baseline, but not at 1 year); positive converters (nonadherent at baseline, but fully adherent at 1 year); and consistent nonadherers (nonadherent at both baseline and 1 year). Patients incurring events between baseline and 1 year were again excluded from this analysis. Finally, we studied the impact of partial nonadherence on the primary end point. All statistical analyses were performed using SAS version 9.3 (SAS Institute, Cary, NC). All P values were 2-tailed, with statistical signiÔ¨Åcance set at .05. All CIs were calculated at the 95% level. RESULTS A total of 37,154 patients with established atherothrombosis were included, of whom 19,799 (53.3%) had coronary artery disease alone, 7746 (20.9%) had cerebrovascular disease alone, 2420 (6.5%) had peripheral artery disease alone, and 720 (1.9%) had evidence of coronary, cerebrovascular, and peripheral artery disease; the rest had 2 disease territories involved. At the time of enrolment, aspirin was the most commonly used antiplatelet agent (84.0%), statins were the most commonly used lipid-lowering agents (93.3%), and beta-blockers (55.2%), angiotensin-converting 695 enzyme inhibitors (51.3%), and diuretics (45.2%) were the most commonly used antihypertensive agents. Adherence with all evidence-based secondary prevention measures was noted in 46.7% of patients at baseline and 48.2% at 1 year after enrollment (Table 1). Adherence with at least 2 eligible measures was observed in 89.4% of the patients at baseline. Characteristics of the study population at the time of study enrollment/baseline are presented in Table 2. Nonadherent patients were more likely than adherent patients at baseline to be older and female, and to have a history of congestive heart failure and atrial Ô¨Åbrillation. Japanese and Australian patients also were more likely to be nonadherent. On the other hand, nonadherent patients were less likely to have a history of hypertension or hypercholesterolemia, and less likely to be current smokers. Adherence at Baseline and Clinical Outcomes at 4 Years (n [ 37,154) After multivariate adjustment, patients who were nonadherent with any evidence-based secondary prevention medication at baseline demonstrated an 18% higher hazard of the primary outcome of cardiovascular death/myocardial infarction/stroke at 4 years compared with patients who were adherent (17.4% vs 13.4%; HR, 1.18; 95% CI, 1.111.25) (Table 3). Number needed to treat was 31 patients for the composite end point, 25 patients for all-cause mortality, and 41 patients for cardiovascular mortality. The proportional hazards assumption for adherence status (adherent vs nonadherent) was met for the multivariate Cox model. Table 1 Adherence to Evidence-based Measures at Baseline and 1-year Follow-up All patients Disease-speciÔ¨Åc Coronary artery disease Stable angina Unstable angina Previous MI Post-angioplasty Post-CABG Cerebrovascular disease TIA Stroke Peripheral arterial disease Measure-speciÔ¨Åc Antiplatelet agent Lipid-lowering agent Antihypertensive agent Beta-blockers* Baseline (n ¬º 37,154) 1 Year (n ¬º 34,485) 46.7 48.2 47.8 46.3 51.7 52.2 55.7 52.8 49.7 56.7 46.9 57.3 49.1 47.4 52.7 53.4 56.3 55.3 50.3 56.4 47.4 59.5 84.9 72.5 99.0 62.6 85.3 73.5 97.8 64.1 CABG ¬º coronary artery bypass grafting; MI ¬º myocardial infarction; TIA ¬º transient ischemic attack. All values are percentages. *Only in patients with established coronary artery disease. 696 Table 2 The American Journal of Medicine, Vol 126, No 8, August 2013 Characteristics of the Study Population at Time of Study Enrollment/Baseline Characteristic Sociodemographic Age (y) Men Region North America/Latin America/Western Europe/Asia Eastern Europe/Middle East Japan/Australia Diabetes mellitus Hypercholesterolemia Hypertension Obesity (BMI !30 kg/m2) Current smoker Heart failure Atrial Ô¨Åbrillation Ischemic event 1 y Laboratory values Serum creatinine (mg/dL) Fasting blood glucose (mg/dL) Fasting total cholesterol (mg/dL) Fasting triglycerides (mg/dL) Medication history Aspirin Statin ACE inhibitor ARB Beta-blocker Diuretic Calcium-channel blocker Nitrate/other antianginal medication Other antihypertensive Diabetic patients with !1 antidiabetic drug NSAIDs Fully Adherent* (n ¬º 17,336) Nonadherent (n ¬º 19,818) P Value 66.8 (9.8) 68.5 69.5 (10.0) 67.5 <.001 .03 <.001 80.7 71.7 13.8 5.6 37.3 93.0 82.5 28.6 45.8 14.3 7.3 41.7 11.8 16.5 36.4 46.5 76.8 23.0 43.2 16.3 14.6 38.1 .07 <.001 <.001 <.001 <.001 <.001 <.001 <.001 1.1 (0.6) 117.0 (41.7) 191.7 (57.8) 165.3 (96.3) 1.1 (0.7) 116.7 (42.5) 194.2 (44.0) 150.4 (89.1) .25 .58 <.001 <.001 85.6 94.0 50.6 20.5 80.2 40.0 30.5 29.3 8.1 88.4 9.3 58.7 44.6 41.1 20.6 26.2 37.5 38.5 29.5 8.8 86.3 9.0 <.001 <.001 <.001 .81 <.001 <.001 <.001 .76 .02 .003 .38 ACE ¬º angiotensin-converting enzyme; ARB ¬º angiotensin receptor blocker; BMI ¬º body mass index; NSAID ¬º nonsteroidal anti-inÔ¨Çammatory drug. Numbers represent mean √Ü standard deviation for continuous variables and % for binary or categoric variables. P values were obtained with Student t test for continuous variables and chi-square test for categoric variables. *Fully adherent with all evidence-based medications (antiplatelet agent, lipid-lowering agent, and antihypertensive for patients with cerebrovascular and peripheral arterial disease; also includes beta-blockers for patients with coronary artery disease). Adherence at 1-year Follow-up and Clinical Outcomes at 4 Years (n [ 34,485) After multivariate adjustment, patients who were nonadherent with any evidence-based secondary prevention medication at 1-year follow-up demonstrated a 19% higher hazard of the primary outcome of cardiovascular death/ myocardial infarction/stroke at 4 years compared with patients who were adherent (12.6% vs 9.8%; HR, 1.19; 95% CI, 1.11-1.28) (Table 3). Adherence at Baseline and 1-year Follow-up and Clinical Outcomes at 4 Years Cumulative hazard curves for the primary end point and allcause mortality stratiÔ¨Åed by adherence at baseline and at 1 year are demonstrated in Figure 1. Patients who were fully adherent at both time points had the lowest incidence of cardiovascular death/myocardial infarction/stroke, as well as all-cause mortality, whereas patients who were nonadherent at both time points had the worst outcomes, especially for allcause mortality (P < .0001 for both end points). On multivariate adjusted analysis, there was an increase in hazard for the primary end point in positive converters (HR, 1.19; 95% CI, 1.03-1.38), negative converters (HR, 1.36; 95% CI, 1.17-1.57), and consistent nonadherers (HR, 1.21; 95% CI, 1.12-1.31) compared with consistent adherers. Partial Nonadherence at Baseline Nonadherence with antiplatelet agents (HR, 1.08; 95% CI, 1.001-1.17) and lipid-lowering agents (HR, 1.43; 95% CI, 1.34-1.52) at baseline was associated with a signiÔ¨Åcant increase in the risk of the primary end point at 4 years. Likewise, nonadherence with the use of at least 2 secondary prevention medications at baseline also was associated with Kumbhani et al Adherence to Secondary Prevention Medications 697 Table 3 Adjusted Multivariate Hazard Ratios for 4-year Clinical Outcomes for Patients Nonadherent with Evidence-based Secondary Prevention Measures at Baseline and 1 Year Versus Those Who Were Adherent End Point Baseline All patients CAD only CVD only PAD only 1y All patients CAD only CVD only PAD only CV Death/MI/Stroke HR (95% CI); P Value All-cause Mortality HR (95% CI); P Value CV Mortality* HR (95% CI); P Value 1.18 1.12 1.25 1.14 (1.11-1.25); (1.02-1.22); (1.10-1.42); (0.89-1.47); <.001 .02 <.001 .29 1.30 1.26 1.35 1.35 (1.22-1.40); (1.14-1.39); (1.15-1.60); (1.05-1.73); <.001 <.001 <.001 .02 1.31 1.27 1.33 1.27 (1.20-1.42); (1.12-1.45); (1.07-1.65); (0.91-1.76); <.001 <.001 .01 .16 1.19 1.18 1.27 1.41 (1.11-1.28); (1.06-1.32); (1.08-1.50); (1.04-1.89); <.001 .004 .004 .02 1.28 1.28 1.39 1.60 (1.18-1.40); (1.13-1.45); (1.13-1.70); (1.19-2.15); <.001 <.001 .002 .002 1.31 1.34 1.44 1.51 (1.18-1.46); (1.14-1.58); (1.09-1.90); (1.03-2.22); <.001 <.001 .01 .04 CAD ¬º coronary artery disease; CI ¬º conÔ¨Ådence interval; CV ¬º cardiovascular; CVD ¬º cerebrovascular disease; HR ¬º hazard ratio; MI ¬º myocardial infarction; PAD ¬º peripheral arterial disease. *Cardiovascular mortality deÔ¨Åned as fatal stroke, fatal myocardial infarction, or other cardiovascular death (see text for details). an increase in the risk of the primary end point at 4 years (HR, 1.28; 95% CI, 1.17-1.40). DISCUSSION Our analysis of a large international observational registry of 37,154 patients with established atherothrombotic disease demonstrates that only 46.7% of patients at baseline and 48.2% at 1 year post-entry were fully adherent with guideline-recommended secondary prevention medications, such as antiplatelet agents, lipid-lowering therapies, and antihypertensive agents. Nonadherence with any of these medications at baseline and 1-year follow-up was independently associated with an increased risk of adverse clinical outcomes at 4 years, including all-cause mortality and cardiovascular mortality. This also was true individually for patients with coronary, cerebrovascular, and peripheral artery disease. Patients who were fully adherent at both baseline and 1 year had the lowest hazard of adverse clinical outcomes at 4 years, whereas negative converters (fully adherent at baseline but not at 1-year follow-up) and consistent nonadherers (nonadherent at baseline and 1 year) seemed to have the highest hazard independently of other clinical variables known to inÔ¨Çuence 4-year outcomes signiÔ¨Åcantly.8 This is one of the largest prospective studies on long-term adherence rates and their association with long-term clinical outcomes in patients with stable atherothrombotic disease. Medication adherence is a growing concern for clinicians, healthcare systems, and other stakeholders (eg, payers) because of increasing evidence of its association with both short- and long-term adverse outcomes and higher costs of care.13-22 In this analysis, we report long-term adherence rates that are similar to those in other studies in patients with coronary artery disease;2,13,17,18 data assessing long-term adherence in patients with cerebrovascular disease are limited20,23 and virtually nonexistent in patients with peripheral artery disease.24 In recognition of the close relationship between in-hospital adherence and outcomes,16 and in-hospital adherence and long- term adherence,14,17,25,26 a number of national initiatives, such as the Get With The Guidelines program and the Guidelines Applied in Practice (GAP) initiative, have been introduced in the United States to improve in-hospital and at-discharge adherence with evidence-based therapies in patients with established coronary and cerebrovascular diseases,27-29 and have resulted in signiÔ¨Åcant improvements.3-7 However, there are currently few measures that are targeted speciÔ¨Åcally toward improving long-term adherence. The US Department of Health and Human Services recently launched ‚ÄúMillion Hearts,‚Äù an initiative that aims to prevent 1 million heart attacks and strokes over the next 5 years.30 A key component of this campaign is improving compliance with the same measures that we report in this study: the use of aspirin, effective lowering of low-density lipoprotein cholesterol, and blood pressure control. Our study thus helps to deÔ¨Åne the magnitude of the problem and suggests that this strategy is likely to be cost-effective, with only 25 patients needed to be treated appropriately to prevent 1 death and 31 patients needed to be treated to prevent 1 composite end point. Further, because REACH is an observational registry, it is instructive to note that adherence rates at baseline and 1 year were essentially identical. Thus, in the absence of active interventions, patient adherence is unlikely to improve significantly. Moreover, patients who were negative converters, that is, switched from being fully adherent at baseline to nonadherent at 1 year, had 4-year outcomes that were similar to those who were consistently nonadherent, whereas those who were positive converters, that is, switched from being nonadherent at baseline to fully adherent at 1 year, had 4-year outcomes approaching those for patients who were adherent at both time points. Stated simply, this suggests that initiation of these medications is associated with a lowering in risk, whereas stopping these mediations is associated with an elevation in risk; not taking them at all is associated with the highest risk. Our analysis also suggests that patient subsets who are known to have poorer long-term outcomes are paradoxically more likely to be nonadherent, such as those who are older 698 The American Journal of Medicine, Vol 126, No 8, August 2013 Figure 1 Cumulative hazard curves for the primary end point of cardiovascular death/myocardial infarction/stroke (A) and all-cause mortality (B) based on adherence at baseline and 1 year (P < .001 for both curves by log-rank test). CVD ¬º cerebrovascular disease; MI ¬º myocardial infarction. Kumbhani et al Adherence to Secondary Prevention Medications and female, and who have greater comorbidities, such as atrial Ô¨Åbrillation and congestive heart failure. Targeted interventions to improve adherence in these high-risk subsets may be especially beneÔ¨Åcial.22 One of the biggest strengths of the REACH registry is that it provides high-quality data with high follow-up rates and from diverse patient types and environments. Limitations of the REACH data are those inherent to registries, such as selection bias and the presence of unmeasured confounders.31 For this analysis, selection bias would tend to overestimate adherence rates, with likely an even greater impact on long-term outcomes than currently described. A unique form of unmeasured confounding known as ‚Äúhealthy adherer effect‚Äù could be operational; the lower risk of adverse outcomes associated with adherence may be a surrogate marker for overall healthy behavior, such as healthy eating and regular exercise. However, its impact in similar settings has been recently disputed.21 Likewise, placebo adherence studies have demonstrated lower mortality in adherent compared with nonadherent patients.32,33 However, a signiÔ¨Åcant impact of this effect seems to be due to loss of adherence closer to a terminal illness/event.32 In the current study, exposure (adherence) and outcomes were assessed at least 3 years apart. Thus, although it is unlikely that a similar ‚Äúplacebo adherence effect‚Äù is operational, it cannot be completely excluded. Moreover, our Ô¨Åndings are in concordance with those noted by carefully conducted randomized controlled trials. Medication use was assessed by patient selfreport using detailed questionnaires/case report forms without external validation. Although other measures such as pharmacy prescription reÔ¨Ålls and electronic medication monitors can be more accurate, patient self-report has been shown to be the most useful method in the clinical setting;15 the use of questionnaires also has good correlation with various electronic measures.34 Information regarding side effects and contraindications to these medications also was not available. Finally, this analysis is unable to separate physician nonadherence (due to nonprescription) from patient nonadherence. The former is a more surmountable problem and may be alleviated by systems-based interventions, including improved physician education. CONCLUSIONS Our analysis of a large international contemporary registry in stable outpatients with atherothrombosis demonstrates that long-term adherence with evidence-based secondary prevention therapies remains suboptimal. Lack of adherence at baseline and 1 year were both associated with an elevated risk of adverse clinical outcomes at 4 years, with negative converters at 1 year and consistent nonadherers having the highest risk compared with consistent adherers. Targeted interventions to improve long-term adherence with evidence-based measures in the outpatient setting may help to realize the full potential of these treatments in this high-risk patient population. ACKNOWLEDGEMENTS The authors thank Amarachi Umez-Eronini, MPH, from the TIMI Study Group, for statistical input. 699 References 1. Smith SC Jr, Allen J, Blair SN, et al. AHA/ACC guidelines for secondary prevention for patients with coronary and other atherosclerotic vascular disease: 2006 update: endorsed by the National Heart, Lung, and Blood Institute. Circulation. 2006;113:2363-2372. 2. Naderi SH, Bestwick JP, Wald DS. Adherence to drugs that prevent cardiovascular disease: meta-analysis on 376,162 patients. Am J Med. 2012;125:882-887. 3. Roger VL, Go AS, Lloyd-Jones DM, et al. Heart disease and stroke statisticse2011 update: a report from the American Heart Association. Circulation. 2011;123:e18-e209. 4. Schwamm LH, Fonarow GC, Reeves MJ, et al. Get With the Guidelines-Stroke is associated with sustained improvement in care for patients hospitalized with acute stroke or transient ischemic attack. Circulation. 2009;119:107-115. 5. Eagle KA, Montoye CK, Riba AL, et al. Guideline-based standardized care is associated with substantially lower mortality in Medicare patients with acute myocardial infarction: the American College of Cardiology‚Äôs Guidelines Applied in Practice (GAP) Projects in Michigan. J Am Coll Cardiol. 2005;46:1242-1248. 6. Kumbhani DJ, Fonarow GC, Cannon CP, et al. Predictors of adherence to performance measures in patients with acute myocardial infarction. Am J Med. 2013;126:74 e1-9. 7. Vaishnava P, Eagle KA. Almost getting with the guidelines. Am J Med. 2013;126:4-5. 8. Bhatt DL, Eagle KA, Ohman EM, et al. Comparative determinants of 4-year cardiovascular event rates in stable outpatients at risk of or with atherothrombosis. JAMA. 2010;304:1350-1357. 9. Bhatt DL, Steg PG, Ohman EM, et al. International prevalence, recognition, and treatment of cardiovascular risk factors in outpatients with atherothrombosis. JAMA. 2006;295:180-189. 10. Steg PG, Bhatt DL, Wilson PW, et al. One-year cardiovascular event rates in outpatients with atherothrombosis. JAMA. 2007;297:1197-1206. 11. Ohman EM, Bhatt DL, Steg PG, et al. The REduction of Atherothrombosis for Continued Health (REACH) Registry: an international, prospective, observational investigation in subjects at risk for atherothrombotic events-study design. Am Heart J. 2006;151:786 e1-10. 12. Alberts MJ, Bhatt DL, Mas JL, et al. Three-year follow-up and event rates in the international REduction of Atherothrombosis for Continued Health Registry. Eur Heart J. 2009;30:2318-2326. 13. Ho PM, Magid DJ, Shetterly SM, et al. Medication nonadherence is associated with a broad range of adverse outcomes in patients with coronary artery disease. Am Heart J. 2008;155:772-779. 14. Ho PM, Bryson CL, Rumsfeld JS. Medication adherence: its importance in cardiovascular outcomes. Circulation. 2009;119:3028-3035. 15. Osterberg L, Blaschke T. Adherence to medication. N Engl J Med. 2005;353:487-497. 16. Peterson ED, Roe MT, Mulgund J, et al. Association between hospital process performance and outcomes among patients with acute coronary syndromes. JAMA. 2006;295:1912-1920. 17. Newby LK, LaPointe NM, Chen AY, et al. Long-term adherence to evidence-based secondary prevention therapies in coronary artery disease. Circulation. 2006;113:203-212. 18. Goyal A, Alexander JH, HaÔ¨Çey GE, et al. Outcomes associated with the use of secondary prevention medications after coronary artery bypass graft surgery. Ann Thorac Surg. 2007;83:993-1001. 19. Kumbhani DJ, Cannon CP, Fonarow GC, et al. Association of hospital primary angioplasty volume in ST-segment elevation myocardial infarction with quality and outcomes. JAMA. 2009;302:2207-2213. 20. Asberg S, Henriksson KM, Farahmand B, et al. Ischemic stroke and secondary prevention in clinical practice: a cohort study of 14,529 patients in the Swedish Stroke Register. Stroke. 2010;41:1338-1342. 21. Rasmussen JN, Chong A, Alter DA. Relationship between adherence to evidence-based pharmacotherapy and long-term mortality after acute myocardial infarction. JAMA. 2007;297:177-186. 22. Heidenreich PA. Patient adherence: the next frontier in quality improvement. Am J Med. 2004;117:130-132. 700 23. Glader EL, Sjolander M, Eriksson M, Lundberg M. Persistent use of secondary preventive drugs declines rapidly during the Ô¨Årst 2 years after stroke. Stroke. 2010;41:397-401. 24. Lardizabal JA, Deedwania PC. BeneÔ¨Åts of statin therapy and compliance in high risk cardiovascular patients. Vasc Health Risk Manag. 2010;6:843-853. 25. Butler J, Arbogast PG, BeLue R, et al. Outpatient adherence to betablocker therapy after acute myocardial infarction. J Am Coll Cardiol. 2002;40:1589-1595. 26. Jackevicius CA, Li P, Tu JV. Prevalence, predictors, and outcomes of primary nonadherence after acute myocardial infarction. Circulation. 2008;117:1028-1036. 27. LaBresh KA, Ellrodt AG, Gliklich R, Liljestrand J, Peto R. Get with the guidelines for cardiovascular secondary prevention: pilot results. Arch Intern Med. 2004;164:203-209. 28. LaBresh KA, Gliklich R, Liljestrand J, Peto R, Ellrodt AG. Using ‚ÄúGet With The Guidelines‚Äù to improve cardiovascular secondary prevention. Jt Comm J Qual Saf. 2003;29:539-550. 29. Mehta RH, Montoye CK, Gallogly M, et al. Improving quality of care for acute myocardial infarction: The Guidelines Applied in Practice (GAP) Initiative. JAMA. 2002;287:1269-1276. 30. Frieden TR, Berwick DM. The ‚ÄúMillion Hearts‚Äù initiativeepreventing heart attacks and strokes. N Engl J Med. 2011;365:e27. 31. Bhatt DL. Advancing the care of cardiac patients using registry data: going where randomized clinical trials dare not. JAMA. 2010;303: 2188-2189. 32. Padula AM, Pressman AR, Vittinghoff E, et al. Placebo adherence and mortality in the Heart and Estrogen/Progestin Replacement Study. Am J Med. 2012;125:804-810. 33. Horwitz RI, Viscoli CM, Berkman L, et al. Treatment adherence and risk of death after a myocardial infarction. Lancet. 1990;336:542-545. 34. Garber MC, Nau DP, Erickson SR, Aikens JE, Lawrence JB. The concordance of self-report with other measures of medication adherence: a summary of the literature. Med Care. 2004;42:649-652. Funding: The REduction of Atherothrombosis for Continued Health (REACH) Registry is sponsored by SanoÔ¨Å-Aventis, Bristol-Myers Squibb, and the Waksman Foundation (Tokyo, Japan). The REACH Registry is endorsed by the World Heart Federation. The sponsors did not review this The American Journal of Medicine, Vol 126, No 8, August 2013 manuscript. The statistical analyses by the TIMI Study Group were funded with a grant from SanoÔ¨Å-Aventis. ConÔ¨Çict of Interest: DJK: honoraria from American College of Cardiology, Somahlutions, Inc. PhGS: research grant from Servier; participated in consultancy or advisory board for Eisai, Amgen, Astellas, Bayer, Boehringer Ingelheim, BMS, Daiichi-Sankyo-Lilly, GSK, Merck, PÔ¨Åzer, Roche, The Medicines Company, AstraZeneca, SanoÔ¨Å-Aventis, and Servier; and a stockholder in Aterovax. CPC: research grants from Intekrin Therapeutics, Accumetrics, AstraZeneca, GlaxoSmithKline, Merck, and Takeda; honoraria from PÔ¨Åzer and AstraZeneca; participated in consultancy or advisory board for Bristol-Myers Squibb/SanoÔ¨Å, Novartis, and Alnylam; and ownership interest in Automedics Medical Systems. KAE: grant/research support from Bristol-Myers Squibb, Blue Cross Blue Shield of Michigan, National Institutes of Health, SanoÔ¨Å-Aventis, the Mardigian Foundation Varbedian Fund, GORE, and the Hewlett Foundation; and a consultant for the National Institutes of Health, the National Heart, Lung, and Blood Institute, SanoÔ¨Å-Aventis, and the Robert Wood Johnson Foundation. SCS: none. EH: none. SG: research grants from SanoÔ¨ÅAventis, Eisai, and Boehringer Ingelheim; and participated in consultancy or advisory board for Eisai, SanoÔ¨Å-Aventis, and Otsuka. EMO: research grants from Bristol-Myers Squibb, CV Therapeutics, Daiichi Sankyo, Datascope, Eli Lilly, Marquet, SanoÔ¨Å-Aventis, Schering-Plough, and The Medicines Company; and consulting or other services for Abiomed, AstraZeneca, CV Therapeutics, Datascope, Gilead Sciences, Liposcience, Marquet, Northpoint Domain, Pozen, Response Biomedical, SanoÔ¨ÅAventis, The Medicines Company, and WebMD (theheart.org). DLB: Advisory Board of Medscape Cardiology; Board of Directors of Boston VA Research Institute, Society of Chest Pain Centers; Chair of American Heart Association Get With The Guidelines Science Subcommittee; Honoraria from American College of Cardiology (Editor, Clinical Trials, Cardiosource), Duke Clinical Research Institute (clinical trial steering committees), Slack Publications (Chief Medical Editor, Cardiology Today Intervention), WebMD (CME steering committees); Senior Associate Editor of Journal of Invasive Cardiology; research grants from Amarin, AstraZeneca, Bristol-Myers Squibb, Eisai, Ethicon, Medtronic, SanoÔ¨ÅAventis, and The Medicines Company; and unfunded research for FlowCo, PLx Pharma, and Takeda. Authorship: All authors had access to the data and played a role in writing this manuscript. Kumbhani et al Adherence to Secondary Prevention Medications Enrolled in international REACH registry (n=69,055) Excluded (n=1,167) ‚ô¶ Not meeting inclusion criteria (n=1,109) ‚ô¶ Withdrew consent (n= 58) Baseline data available (n=68,236) ‚ô¶ Included in baseline publication (n=67,888) ‚ô¶ Data provided by center at 1 year follow-up (n= 348) Excluded (n=23,009) ‚ô¶ Enrolled at centers not participating in 4year follow-up (n=20,980) st ‚ô¶ Lost to follow-up after 1 visit (n= 2,029) Eligible for 4-year follow-up (n=45,227) Excluded (n=8,073) ‚ô¶ No atherothrombotic disease at baseline Included in the final analysis (n=37,154) Excluded (n=2,669) ‚ô¶ Death, myocardial infarction or stroke between baseline and 1 year Included in the 1 year analysis (n=34,485) Supplemental Figure CONSORT Ô¨Çow diagram of study participants. REACH ¬º REduction of Atherothrombosis for Continued Health. 700.e1 Adult-Onset Still Disease Manifestations, Treatment, Outcome, and Prognostic Factors in 57 Patients Mathieu Gerfaud-Valentin, MD, Delphine Maucort-Boulch, MD, Arnaud Hot, MD, Jean Iwaz, PhD, Jacques Ninet, MD, PhD, Isabelle Durieu, MD, PhD, Christiane Broussolle, MD, PhD, and Pascal S√®ve, MD, PhD Abstract: We conducted a retrospective observational study to describe a cohort and identify the prognostic factors in adult-onset Still disease (AOSD). Patients enrolled in this retrospective chart review fulÔ¨Ålled either Yamaguchi or Fautrel criteria. Candidate variables were analyzed with logistic unadjusted and adjusted regression models. Fifty-seven patients were seen in the internal medicine (75%) and rheumatology (25%) departments over a mean period of 8.4 years. The median time to diagnosis was 4 months. The course of AOSD was monocyclic in 17 patients, polycyclic in 25, and chronic in 15. The assessment of glycosylated ferritin (GF) in 37 patients was correlated with early diagnosis. Nine 18F-Ô¨Çuorodeoxyglucose positron emission tomography (18FDG-PET) scans identiÔ¨Åed the lymph nodes and glands as the main sites of hypermetabolism. Complications were frequent (n = 19), including reactive hemophagocytic syndrome (n = 8). None of the 3 deaths could be attributed to AOSD. Corticosteroid dependence, as predicted by a low GF level, occurred in 23 patients (45%). A quarter of the patients received tumor necrosis factorŒ± blockers or anakinra with good tolerance. Fever >39.5¬∞C was predictive of monocyclic AOSD, while arthritis and thrombocytopenia were associated with chronic and complicated AOSD, respectively. The youngest patients had the highest risks of resistance to Ô¨Årst-line treatments. AOSD remains difÔ¨Åcult to diagnose. Mortality is low despite frequent complications. GF and 18FDG-PET scans were of value in the diagnostic approach. The condition in highly symptomatic patients evolved to systemic AOSD, whereas more progressive patterns with arthritis predicted chronic AOSD. (Medicine 2014;93: 91‚Äì99) Abbreviations: AE = adverse event, AOSD = adult-onset Still disease, BM = bone marrow, CI = conÔ¨Ådence interval, CRP = C-reactive protein, CSs = corticosteroids, CT = computed tomography, DMARDs = disease-modifying antirheumatic drugs, ESR = erythrocyte sedimentation rate, 18FDG-PET = 18F-Ô¨Çuorodeoxyglucose positron emission tomography, GF = glycosylated ferritin, IL = interleukin, IVIg = polyvalent intravenous immunoglobulins, MTX = From the Hospices Civils de Lyon, Department of Internal Medicine (MGV, CB, PS), Croix-Rousse University Hospital, Lyon and Universit√© Lyon I, Villeurbanne; Hospices Civils de Lyon, Service de Biostatistiques (DMB, JI), Lyon, CNRS UMR 5558, Equipe Biostatistique Sant√©, Pierre-B√©nite, and Universit√© Lyon I, Villeurbanne; Hospices Civils de Lyon, Department of Internal Medicine (AH, JN), Edouard Herriot University Hospital, Lyon; and Hospices Civils de Lyon, Department of Internal Medicine (ID), Centre Hospitalier Lyon Sud, Pierre-B√©nite; France. Financial support and conÔ¨Çicts of interest: The author listed here has received Ô¨Ånancial support (personal or institutional) from the listed companies, unrelated to the present work: PS: PÔ¨Åzer, LFB, and GlaxoSmithKline. No funding source or sponsor has been involved in the current study. The authors have no conÔ¨Çicts of interest to disclose. Correspondence: Pascal S√®ve, MD, PhD, Department of Internal Medicine, H√¥pital de la Croix-Rousse, 103 Grande Rue de la Croix-Rousse, F-69317 Lyon Cedex 04, France (e‚Äêmail: pascal.seve@chu-lyon.fr). Copyright ¬© 2014 by Lippincott Williams & Wilkins ISSN: 0025-7974 DOI: 10.1097/MD.0000000000000021 Medicine ‚Ä¢ Volume 93, Number 2, March 2014 methotrexate, NSAIDs = nonsteroidal antiinÔ¨Çammatory drugs, OR = odds ratio, PMN = polymorphonuclear neutrophils, RA = receptor antagonist, RHS = reactive hemophagocytic syndrome, SD = standard deviation, SF = serum ferritin, TNF-Œ± = tumor necrosis factor Œ±. INTRODUCTION First described in 1971 by Bywaters,2 adult-onset Still disease (AOSD) is an uncommon systemic inÔ¨Çammatory disorder of unknown etiology. Its prevalence is estimated to be less than 1 case per 100,000 people, and it affects predominantly young adults.30 The precise pathogenesis of this disease remains unknown, but it seems that genetically predisposed hosts develop autoinÔ¨Çammatory disorders triggered by macrophage cell activation and TH1 cytokines such as interleukin (IL)-1, IL-2, IL-6, IL-18, tumor necrosis factor (TNF)-Œ±, and interferon Œ≥.31 The main features of AOSD are a high spiking fever, evanescent rash, sore throat, polyarthralgia, lymphadenopathy, hepatosplenomegaly, serositis, and leukocytosis, as well as elevated liver enzymes, polymorphonuclear neutrophils (PMN), erythrocyte sedimentation rate (ESR), and serum ferritin (SF). Despite the diagnostic value attributed to high SF associated with low glycosylated fraction of ferritin (<20%), the diagnosis of AOSD remains one of exclusion.16 The clinical course of the disease may have 1 of 3 patterns: a self-limiting or monocyclic systemic course, an intermittent or polycyclic systemic course, and a chronic articular course.41 The treatment of AOSD remains empirical. Nonsteroidal antiinÔ¨Çammatory drugs (NSAIDs); corticosteroids (CSs); diseasemodifying antirheumatic drugs (DMARDs), such as methotrexate (MTX); and polyvalent intravenous immunoglobulins (IVIg) are usually used.11 The recent use of biologic agents in AOSD has been shown capable of improving considerably the condition of different subgroups of patients.10,36 Data have shown a most impressive response with anakinra in patients with systemic disease, whereas TNF-Œ± blockers and tocilizumab had better results in chronic AOSD.35 Within this context, determining whether the clinical and laboratory features found at diagnosis can predict the outcome of AOSD would be of great value in patient management. Thus, we conducted the present study to describe a cohort of AOSD patients, identify the baseline prognostic factors that inÔ¨Çuence the clinical course of the disease, and monitor the response to treatment and the appearance of complications. PATIENTS AND METHODS Patients This retrospective study received institutional review board approval. All patients registered as having AOSD from 1998 to 2010 were identiÔ¨Åed through a review of the database of the Medical Information Department of Hospices Civils de Lyon. Patients were included if they fulÔ¨Ålled either Yamaguchi42 or www.md-journal.com Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 91 Gerfaud-Valentin et al Fautrel16 criteria. The exclusion criteria were an onset of the disease before 16 years old and insufÔ¨Åcient medical record data. Medicine ‚Ä¢ Volume 93, Number 2, March 2014 The analyses were performed with R software (R Foundation for Statistical Computing, Vienna, Austria; http://cran.r-project.org/). RESULTS Data Clinical and laboratory data were collected and analyzed by the same investigator (MGV) using a standardized form. The collected data included the following: 1) clinical features; 2) laboratory features including blood cell count, serum electrolytes, coagulation parameters, SF and glycosylated ferritin (GF), ESR, C-reactive protein (CRP), liver enzymes, triglycerides, rheumatoid factor, antinuclear autoantibody, serum protein electrophoresis; 3) pathology laboratory Ô¨Åndings; and, 4) imaging features: computed tomography (CT) and 18F-Ô¨Çuorodeoxyglucose positron emission tomography (18FDG-PET). All prescribed treatments were chronologically listed with their doses, durations, results, adverse events (AEs), and, when available, the reasons for treatment failure. The doses of CSs were expressed in milligrams of equivalent prednisone/day. The disease was considered CS dependent when a patient suffered from AOSD recurrence despite a maintenance dose >15 mg/d without tapering.41 The study considered 3 clinical AOSD courses: 1) monocyclic, deÔ¨Åned as a single episode that faded subsequently and was followed by persistent good health after 1 year or more of follow-up;41 2) polycyclic, deÔ¨Åned as a complete remission followed by 1 or more exacerbations; and 3) chronic, deÔ¨Åned as persistently active disease, usually associated with polyarthritis.8,37 Controlled disease was deÔ¨Åned as clinically asymptomatic AOSD with no biologic inÔ¨Çammatory syndrome at 1-year follow-up.36 Complicated AOSD included 1 or more of the following conditions: acute fulminant hepatitis, disseminated intravascular coagulation, thrombotic microangiopathy, reactive hemophagocytic syndrome (RHS) as deÔ¨Åned by HLH-2004 criteria,19 shock, multiple organ failure, myocarditis, complicated pericarditis (tamponade, followed by restrictive pericarditis), severe sepsis, acute respiratory distress syndrome in adults, or AA amyloidosis.36 Statistical Analysis The sociodemographic, clinical, laboratory, and other variables were expressed as numbers and percentages when considered as categorical data and as means and standard deviations when considered as continuous data. The variables were Ô¨Årst displayed per type of clinical course (Table 1). The differences between the 3 subgroups were tested overall. In order to limit the inÔ¨Çation of alpha level with such a sample size, a few relevant variables were tested using the Fisher exact test for categorical variables and the Kruskal-Wallis test for continuous variables. Then, continuous variables were standardized before regression analyses. To identify the variables that predict the progression from 1 clinical course to another, we selected a set of candidate variables stemming from the initial visit: arthritis; fever >39.5¬∞C; 1 of the following, lymphadenopathy, splenomegaly, or hepatomegaly; serositis; elevated liver enzymes; PMN; and GF. The risk of having a monocyclic, polycyclic, or chronic form at the end of follow-up was quantiÔ¨Åed, and the odds ratios (ORs) of each compared to the chronic form for monocyclic and polycyclic forms and compared to the monocyclic form for chronic forms were estimated by logistic regression. Unadjusted and adjusted models Ô¨Årst were Ô¨Åtted. Then, a Ô¨Ånal adjusted regression model was built with the variables found signiÔ¨Åcant in the unadjusted and adjusted analyses together with a few relevant confounding variables. The results were expressed in ORs and 95% conÔ¨Ådence intervals (95% CIs). 92 www.md-journal.com Clinical Features Fifty-seven patients (27 men and 30 women) with AOSD were identiÔ¨Åed: 53 fulÔ¨Ålled the Fautrel16 classiÔ¨Åcation criteria and 55 fulÔ¨Ålled the Yamaguchi42 criteria. These patients were followed in an internal medicine department (43/57, 75%) or in a rheumatology department (14/57, 25%) over a mean period of 8.4 years (range, 1.2‚Äì48.6 yr). Most patients were of European (44/57) or North African (7/57) origin. The median age at AOSD diagnosis was 36 years (range, 16‚Äì75 yr) (see Table 1). The median delay to diagnosis was 4 months (range, 1‚Äì312 mo). At AOSD diagnosis, 10 patients (18%) were aged over 55 years. The disease-onset manifestations were fever (95%), weight loss (44%), articular symptoms (arthralgia and/or arthritis; 95%), radiologic joint erosions (7%), rash (77%), sore throat or pharyngitis (53%), lymphadenopathy (60%), splenomegaly (30%), hepatomegaly (21%), pleurisy (18%), pericarditis (19%), or myalgia (44%). Two patients developed symptoms consistent with Still disease when aged between 16 and 18 years old. The clinical courses in the 57 AOSD patients were monocyclic in 17 patients (30%), polycyclic in 25 (44%), and chronic in the remaining 15 (26%). Among the cases of chronic AOSD, 12 were chronic arthritis, 2 were isolated inÔ¨Çammatory syndromes, and 1 was chronic hepatitis not induced by the treatment (see Table 1). Laboratory Features Leukocytosis (white blood cell count ‚â•10,000/mm3) was present at diagnosis in 41 (72%) patients (mean ¬± SD, 13.93 ¬± 7.05/mm3) and was composed of ‚â•80% PMN in 44 (77%) (11.98 ¬± 6.14/mm3). Anemia occurred in 65% of patients, and thrombocytosis (>400,000/mm3) in 26%. Fibrinogen was increased (>4 g/L) in 79%, and CRP in 98% (mean ¬± SD, 183.4 ¬± 108.5 mg/L). SF was increased in 82% of patients (mean ¬± SD, 8745 ¬± 19,867 Œºg/L; range, 80‚Äì130,000). GF was measured at diagnosis in 37 patients (65%) and found ‚â§20% of SF in 28 of them (76%). During the remission of 5 patients, the GF normalized in 4 (up to 31%; 40%; 45%, and 50%) but remained low at 19% in the Ô¨Åfth. Thirty-three patients among the 50 (66%) who had liver function tests assessed at diagnosis exhibited abnormalities. Twenty-seven had cytolysis with or without cholestasis (mean aspartate aminotransferase, 362 IU/L; and mean alanine aminotransferase, 528 IU/L). Rheumatoid factor was absent in all patients. Fifty-two patients were tested for antinuclear autoantibodies: 4 patients were positive but none for soluble nuclear antigen and anti-double-stranded DNA antibodies. Triglycerides were high in 14 of the 22 patients in whom they were assayed. Bone marrow aspiration was carried out in 33 patients, including 9 with hemophagocytosis; in 8 of the latter, the diagnostic criteria for RHS were met (see Table 1). Forty-Ô¨Åve pathology examinations were performed: 19 on bone marrow, 9 on liver biopsies; 7 on lymph node biopsies; 5 on skin biopsies; and 1 each on pericardial, pleural, lung, synovial, and salivary gland biopsies. The repeatedly found results were nonspeciÔ¨Åc inÔ¨Çammatory reactions on 9 bone marrow biopsies, hemophagocytosis on 6 bone marrow biopsies, and 5 normal bone marrow biopsies. Of the 22 thoraco-abdomino-pelvic CT scans available, 7 were normal; 5 showed pulmonary involvement (pneumonia, ventilatory disorders, micronodules); 7 deep lymph node enlargement (subdiaphragmatic in 4 cases, mediastinal in 5 cases); ¬© 2014 Lippincott Williams & Wilkins Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine ‚Ä¢ Volume 93, Number 2, March 2014 Prognostic Factors in Adult-Onset Still Disease TABLE 1. Patient Characteristics at Time of Diagnosis According to the Clinical Course of the Disease Characteristic* General and clinical characteristics Male Female Median age at diagnosis (yr) [range] Median delay to diagnostic (mo) [range] Fever Arthralgia/arthritis‚Ä† Rash Lymphadenopathy Sore throat/pharyngitis Weight loss Myalgia Splenomegaly Hepatomegaly Pericarditis‚Ä† Pleurisy‚Ä† Abdominal pain Laboratory characteristics White blood cells ‚â•10,000/mm3 Polymorphonuclear cells ‚â•80%‚Ä° Anemia‚Ä° Platelets >400,000/mm3‚Ä° Mean C-reactive protein (mg/L) Elevated ESR‚Ä° Mean serum ferritin (Œºg/L) [range] Glycosylated ferritin ‚â§20%‚Ä° Elevated liver enzymes‚Ä° Hemophagocytosis¬ß Reactive hemophagocytic syndrome¬∂ Negative for rheumatoid factor‚Ä° Positive for antinuclear antibodies‚Ä° Complications Monocyclic Polycyclic Chronic All Courses (n = 17) (n = 25) (n = 15) (n = 57) 8 (47) 9 (53) 36 [19‚Äì75] 3 [1‚Äì12] 16 (94) 14 (82) 13 (76) 11 (65) 14 (82) 5 (29) 8 (47) 7 (41) 3 (18) 5 (29) 4 (24) 2 (12) 13 (52) 12 (48) 36.5 [16‚Äì72] 4 [1‚Äì312] 25 (100) 25 (100) 18 (72) 14 (56) 20 (80) 13 (52) 9 (36) 5 (20) 5 (20) 6 (24) 6 (24) 6 (24) 6 (40) 9 (60) 36 [18‚Äì63] 3 [0‚Äì276] 14 (93) 15 (100) 13 (87) 8 (53) 10 (67) 7 (47) 8 (53) 3 (20) 4 (27) 0 0 2 (13) 27 (47) 30 (53) 36 [16‚Äì75] 4 [0‚Äì312] 54 (95) 54 (95) 44 (77) 34 (60) 30 (53) 25 (44) 25 (44) 17 (30) 12 (21) 11 (19) 10 (18) 10 (18) 10 (59) 11/16 (69) 9/16 (56) 2/15 (13) 163 5/6 8935 [138‚Äì41,181] 13/15 (87) 12/17 (71) 3 (18) 3 (18) 14/14 (100) 2/16 (13) 5 (29) 20 (80) 19/23 (83) 8/22 (36) 7/21 (33) 198 10/10 10,860 [100‚Äì130,000] 12/17 (71) 17/23 (74) 4 (16) 4 (16) 23/23 (100) 1/23 (4) 11 (44) 9/12 (75) 8/12 (67) 6/11 (55) 3/11 (27) 185 9/9 (100) 3342 [84‚Äì22,410] 3/5 (60) 4/11 (36) 2 (18) 1 (9) 12/12 (100) 1/11 (9) 3 (20) 39/54 (72) 39/50 (78) 32/49 (65) 12/47 (26) 183 24/25 (96) 8745 [84‚Äì130,000] 28/37 (76) 27/50 (54) 9 (16) 8 (14) 49/49 (100) 4/52 (8) 19 (33) *Number (%) unless stated otherwise. ‚Ä†Only statistically signiÔ¨Åcant differences among the subgroups of patients: arthralgia/arthritis (p = 0.04) and serositis (p = 0.009). ‚Ä°No. positive/no. tested (percentage). ¬ßHemophagocytosis on bone marrow aspirations. ¬∂As deÔ¨Åned by HLH-2004 criteria (19). and 5 showed splenomegaly and/or hepatomegaly. Six serous effusions were identiÔ¨Åed on CT. Among the 9 18FDG-PET scans performed, 2 were normal; 4 showed hypermetabolic lymph nodes; 3 showed hypermetabolism of the salivary glands, 1 of the testis, 1 of the pericardium, 1 of the gastric fundus, and 1 of the liver. The median standardized uptake value was 4 (range, 1.8‚Äì8.0). Complications Nineteen patients had complications: 8 patients had RHS (of whom 6 had RHS as presenting syndrome at AOSD diagnosis), 4 had myocarditis, 4 had acute respiratory distress syndrome, 3 had cardiac tamponade, 2 had cardiopulmonary shock, 2 had multiple organ failure, 2 had fulminant hepatitis, and 1 patient had disseminated intravascular coagulation. Although complications seemed ¬© 2014 Lippincott Williams & Wilkins to be more prevalent in polycyclic AOSD, the difference between groups was not statistically signiÔ¨Åcant (see Table 1). Three deaths occurred, but none could be directly attributed to AOSD or its complications: 1 myocardial infarction, 1 accidental death, and 1 death due to a non-Hodgkin lymphocytic lymphoma that occurred 20 years after AOSD onset during disease remission. Treatment Three patients had monocyclic AOSD that regressed spontaneously without treatment, and 54 patients received a Ô¨Årst-line treatment. The different lines of treatment are detailed in Table 2, and treatments that led to remission of the disease are described in Table 3. Twenty-eight patients received NSAIDs: indomethacin in 6 patients, ketoprofen in 9, aspirin in 7, diclofenac in 3, and www.md-journal.com Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 93 Medicine ‚Ä¢ Volume 93, Number 2, March 2014 Gerfaud-Valentin et al other drugs in 3. NSAIDs were used as Ô¨Årst-line treatment in 22 patients and controlled the disease in only 5 (18%). Among the 6 patients who experienced gastrointestinal AEs, 4 were prescribed proton pump inhibitors. Of 51 patients treated with CSs, 49 received them as a Ô¨Årstor second-line treatment. CSs were Ô¨Årst intravenous in 11 patients. The initial dose of oral CSs was 1 mg/kg per day in 36 patients (70%), <1 mg/kg per day in 12 patients (24%), and >1 mg/kg per day in 3 patients (6%). Thirty-eight of 51 patients (75%) developed various AEs such as Cushing syndrome (n = 19), osteoporosis (n = 8), aseptic osteonecrosis (n = 5), CS-induced diabetes (n = 4), high blood pressure (n = 4), cataract (n = 3), psychiatric disorders (n = 3), and infectious diseases (n = 2). Twenty-three of 51 patients (45%) developed a CS-dependent disease. In the present study, the use of <1 mg/kg per day initial dose was not found to be associated with dependence on high-dose CSs, complicated AOSD, resistance to treatment, or progression to a more severe disease course (data not shown). Thirty-three patients (58%) required MTX as DMARD. Eleven developed AEs (33%), mainly elevated liver enzymes (n = 5), low blood cell counts (n = 3), or cough (n = 2). MTX was the last-line treatment in 14 of 33 patients (42%). IVIg was sufÔ¨Åcient to control the disease in 4 of 23 cases. It was more frequently prescribed in polycyclic or chronic AOSD (n = 20) and in complicated AOSD (n = 12). One patient developed an acute renal failure after IVIg infusion. One woman received IVIg as Ô¨Årst-line treatment during pregnancy without any complication. Treatment with biologic agents was prescribed 23 times in 15 of 57 patients (26%). These 15 patients had polycyclic or chronic AOSD, complicated in 7 of 15 patients and/or CSdependent in 7 of 15 patients. The characteristics of these patients were not signiÔ¨Åcantly different from those of patients who did not receive these targeted therapies (data not shown). Of 17 prescriptions for TNF-Œ± blockers, 8 (47%) succeeded in controlling the disease: inÔ¨Çiximab (n = 4/8), etanercept (n = 3/8), and adalimumab (n = 1/1). TNF-Œ± blockers were effective in 5 of 9 (56%) chronic AOSD cases and 2 of 6 (33%) cases of polycyclic AOSD, but this difference in effectiveness was not signiÔ¨Åcant (p = 0.6). Among 3 switches between TNF-Œ± blockers, only 1 was efÔ¨Åcient. Anakinra, an IL-1-receptor antagonist (IL-1RA), was prescribed 6 times and was always the last line of treatment; it led to remission in 5 of 6 patients after a mean follow-up of 27.8 months (range, 14‚Äì36 mo). Among the 4 patients with polycyclic AOSD who received anakinra, 2 had experienced RHS. The treatment was effective in all but 1 who suffered from recurrence of the disease after 11 months on anakinra and required drug discontinuation. Two patients with chronic AOSD were effectively treated with anakinra. In all 6 patients, there was no AE other than a local inÔ¨Çammatory reaction. In this small sample, there was no signiÔ¨Åcant difference in the effectiveness of anakinra between the polycyclic and the chronic course of AOSD. Prognostic Factors Patients with GF levels assessed during the diagnostic process had signiÔ¨Åcantly shorter times to diagnosis compared to those without (OR, 3.33; p = 0.039). Furthermore, univariate analysis showed that GF testing was correlated only with the monocyclic course of AOSD (OR, 5.97; 95% CI, 1.14 ‚Äì 60.65; p = 0.018). In the unadjusted analysis, a short delay to diagnosis (‚â§2 mo) was rather predictive of a monocyclic course of AOSD compared with other severe courses (OR, 5.31; 95% CI, 1.386 ‚Äì 22.676; p = 0.007). However, this was not conÔ¨Årmed by multivariate analysis. The unadjusted analysis of the variables collected at diagnosis revealed 2 determinants of a monocyclic course of AOSD: fever >39.5¬∞C and abnormal liver function tests, whereas the presence of arthritis was negatively correlated with a monocyclic course. The adjusted multivariate analysis conÔ¨Årmed the predictive role of high fever in the occurrence of monocyclic AOSD (Table 4). In the unadjusted analysis, the prognostic determinants of a polycyclic course were also fever >39.5¬∞C and abnormal liver function tests, whereas arthritis was a protective factor. In the adjusted multivariate analysis, only arthritis as a protective factor remained (see Table 4). When these 2 systemic courses of the disease were compared, arthritis was rather negatively correlated with a polycyclic course, whereas white blood cell count appeared higher in patients with progression to a polycyclic course than in patients with progression to a monocyclic course (see Table 4). In the unadjusted analysis, the absence of serositis and normal liver function tests seemed to predict a chronic course of AOSD. The mean initial SF was 3342 Œºg/L in patients who experienced chronic AOSD and 9893 Œºg/L in those who experienced a systemic‚Äîmonocyclic or polycyclic‚Äîcourse; the difference was hardly signiÔ¨Åcant (p = 0.053). However, these differences were not signiÔ¨Åcant in the adjusted analysis. Unadjusted and TABLE 2. Distribution and Adverse Events of Treatment Lines in 57 Patients With AOSD Drug NSAIDs (n = 28) CS (n = 51) CS dependence IVIg (n = 23) MTX (n = 33) TNF-Œ± blockers (n = 17) Anakinra (n = 6) Other‚Ä° (n = 10) Line 1 Line 2 Line 3 Line 4 Line 5 Line 6 (n = 54) (n = 46) (n = 32) (n = 22) (n = 9) (n = 5) Adverse Events* 22 31 ‚Äî 1 ‚Äî ‚Äî ‚Äî ‚Äî 2 18 ‚Äî 7 15 ‚Äî 1 3 3 1 ‚Äî 7 12 4 ‚Äî 5 1 1 ‚Äî 8 3 8 ‚Äî 1 ‚Äî ‚Äî ‚Äî ‚Äî 3 3 2 1 ‚Äî ‚Äî ‚Äî ‚Äî ‚Äî 2 3 ‚Äî 6 (21)‚Ä† 38 (75) 23 (45) 1 (4) 11 (33) 1 (6) 0 ‚Äî *Number (percentage). ‚Ä†Among the 6 patients who had gastrointestinal adverse effects of NSAIDs, 4 were prescribed proton pump inhibitors. ‚Ä°Other: colchicine (n = 4), hydroxychloroquine (n = 4), azathioprine (n = 1), salazopyrine (n = 1). 94 www.md-journal.com ¬© 2014 Lippincott Williams & Wilkins Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine ‚Ä¢ Volume 93, Number 2, March 2014 Prognostic Factors in Adult-Onset Still Disease TABLE 3. Treatments in Use at AOSD Remission Treatment Monocyclic AOSD (n = 17) CS CS and MTX IVIg TNF-Œ± blockers Anakinra Other Polycyclic AOSD (n = 25) 9 3 1 0 0 1 (NSAIDs) 5 12* 1 3 4 0 Chronic AOSD (n = 15) Total (n = 54) 2 4 0 5 2 1 (MTX alone) 16 20 2 8‚Ä† 6‚Ä† 2 *Six of 12 patients were prescribed transient IVIg therapy. ‚Ä†All patients receiving biotherapy but 1 were treated with CS. adjusted analyses indicated that the presence of arthritis at diagnosis was a strong predictor of a chronic disease course (see Table 4). The same was true for radiologic joint erosion (OR, 14.21; 95% CI, 1.236‚Äì763.542; p = 0.004 in the unadjusted analysis only). Only thrombocytopenia remained a signiÔ¨Åcant predictor of occurrence of complications in the adjusted regression (OR, 8.72; 95% CI, 1.42‚Äì53.66; p = 0.020). Response to Treatment Not surprisingly, steroid dependence was higher in polycyclic and chronic AOSD (OR, 8.01; 95% CI, 0.79‚Äì137.84; p = 0.007). In the multivariate adjusted analysis, low GF seemed to be predictive of high-dose CS dependence (OR, 0.3, 95% CI, 0.07‚Äì1.26; p = 0.10 per each SD increment in GF level, SD = 14.16; mean GF, 16.7%). In the unadjusted analysis only, splenomegaly seemed to protect from steroid dependence (OR, 0.28; 95% CI, 0.79‚Äì8.87; p = 0.074), and an age at diagnosis more than the average age (38 yr) appeared predictive of resistance to the Ô¨Årst 2 lines of treatment (OR, 0.6; 95% CI, 0.34‚Äì1.06; p = 0.078). DISCUSSION In the present retrospective study, we analyzed the empirical treatment of AOSD, the prognostic factors of the disease, and new data such as 18FDG-PET scan characteristics in 57 patients over an important mean follow-up of 8.4 years. Compared to previous studies, the present study is original because it included mostly white patients, was carried out recently, and used biologic agents. The current Ô¨Åndings from white AOSD patients are consistent with those previously reported in the main literature.3,4,6,16,27,33,37,44 The delay to Ô¨Ånal diagnosis was highly variable, which indicates possible diagnostic difÔ¨Åculties. Weight loss was more common in the current study (44%) than in previous studies (18% for Pay et al;33 19% for Cagatay et al3). This TABLE 4. Prognostic Factors of Outcome in 57 Patients With AOSD Unadjusted Analysis Prognostic Factor Monocyclic vs. chronic course Fever >39.5¬∞C Abnormal liver function tests Arthritis Glycosylated ferritin assessment Polycyclic vs. chronic course Fever >39.5¬∞C Abnormal liver function tests Arthritis Chronic vs. monocyclic course Fever > 39.5¬∞C Abnormal liver function tests Arthritis Polycyclic vs. monocyclic course Arthritis White blood cells count Complicated vs. uncomplicated Thrombocytopenia Adjusted Analysis* Odds ratio [95% CI] p Odds ratio [95% CI] p 10.83 [1.79‚Äì65.55] 3.85 [0.76‚Äì19.47] 0.19 [0.04‚Äì0.97] 5.97 [1.14‚Äì60.65] 0.009 0.103 0.046 0.018 8.62 [1.34‚Äì55.34] 3.76 [0.50‚Äì27.97] 0.17 [0.02‚Äì1.46] ‚Äì 0.023 0.196 0.106 ‚Äì 6.00 [1.12‚Äì32.29] 4.96 [1.06‚Äì23.16] 0.08 [0.02‚Äì0.38] 0.037 0.042 0.001 3.53 [0.54‚Äì23.05] ‚Äì 0.102 [0.02‚Äì0.51] 0.188 ‚Äì 0.005 0.09 [0.01‚Äì0.56] 0.26 [0.05‚Äì1.31] 5.14 [1.033‚Äì25.60] 0.009 0.103 0.046 0.26 [0.03‚Äì1.94] 0.27 [0.04‚Äì1.98] 5.87 [0.69‚Äì50.23] 0.187 0.196 0.106 0.189 [0.059‚Äì0.607] 1.806 [0.993‚Äì3.285] 0.005 0.053 0.177 [0.049‚Äì0.632] 2.041 [1.080‚Äì3.859] 0.008 0.028 7.36 [1.23‚Äì90.33] 0.025 8.72 [1.42‚Äì53.66] 0.020 *Variables tested in the adjusted analysis: fever >39.5¬∞C; arthritis; serositis; hepatomegaly or splenomegaly; abnormal liver function tests; serum ferritin; glycosylated ferritin; white blood cells count. ¬© 2014 Lippincott Williams & Wilkins www.md-journal.com Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 95 Gerfaud-Valentin et al may be the result of underreporting, but also of higher proportions of systemic AOSD in the current study compared to others (74% here vs. 60% on average in the literature). This also stands for polycyclic AOSD (44% here vs. 23% on average). These differences may also be due to a referral bias and longer follow-up times in the present cohort. In fact, most patients in the current study were seen in the internal medicine department (not in rheumatology as usual) for fever of unknown origin. This might have increased the number of systemic polycyclic patterns. In addition, the long follow-up times enabled us to see more relapses; thus, more polycyclic courses than in other studies. The interest in SF and GF is mainly supported by a retrospective study by Fautrel et al.14 However, in a series of 14 patients, Vignes et al39 assumed that high SF could be a marker of activity in the acute phase of AOSD (because it rapidly normalized with treatment and remission), whereas GF remained low at 3 years (at 16% on average), which could be used to diagnose AOSD under treatment or during remission. This could not be conÔ¨Årmed in the current study because only 1 of the 5 GF assayed during remission within the Ô¨Årst 3 years of follow-up remained <20% (precisely, 19%). We describe here the largest series, to our knowledge, of 18 FDG-PET scans on 9 AOSD patients and report that the lymph nodes and glands were the main sites of hypermetabolism. This involvement of lymphoid tissues may force to rule out lymphoma. In case of fever of unknown origin, 18FDG-PET may support the diagnosis of AOSD by allowing the rejection of hypotheses of localized infection, solid tumor, or other noninfectious inÔ¨Çammatory diseases.7,25 Moreover, 18FDG-PET may be useful in monitoring the disease progression and responses to treatment,5 although this has to be conÔ¨Årmed by prospective cohorts. The treatment of AOSD remains largely empirical, relying so far on a few prospective or retrospective studies and not on double-blinded randomized trials with suitable sample sizes. Some considerations about treatment are listed below. First, although NSAIDs are traditionally considered as the Ô¨Årst-line therapy, they were effective in controlling the disease in 18% of the present cohort patients. Indomethacin was deemed more effective than salicylates but was prescribed in less than a quarter (21%) of patients. Moreover, AEs, mainly gastrointestinal, were prevalent (21%). Thus, as suggested by Franchini et al,18 NSAIDs constitute a supportive treatment that could be administered during the diagnostic process before conÔ¨Årmation of AOSD and then reserved for use if CSs are insufÔ¨Åcient. Second, in the current study neither the dose nor the route of administration had an impact on the therapeutic response to CS and the outcome of AOSD. This contrasts with the report of Kong et al,27 where patients treated with prednisone (or its equivalent) ‚â•40 mg (0.8 mg/kg) achieved quicker remission and had fewer relapses compared with those who received a lower dosage. As in the literature, the tolerance in the current study was poor, with 75% of AEs during treatment. In addition, we conÔ¨Årm the high prevalence of steroid dependence in AOSD (45% here and 42% reported by Kim et al24). As one would expect and as already described,17,24,43 CS dependence was significantly higher in the polycyclic and chronic forms than in monocyclic AOSD, and is an additional factor of poor prognosis in the former patterns. However, in contrast to the results reported by Kim et al,24 in the current study splenomegaly appeared to be a protective factor against steroid dependence. We have shown that young patients were at risk of resistance to the Ô¨Årst 2 lines of treatment, which often include CS, and that a low GF level would predict steroid dependence. Accordingly, 96 www.md-journal.com Medicine ‚Ä¢ Volume 93, Number 2, March 2014 it makes sense to pay particular attention to age, splenomegaly, elevated ESR, and low GF before an early introduction of sparing treatments such as MTX. Third, regarding MTX, in 1999 Fautrel et al13 published a retrospective study on 26 AOSD patients highlighting the beneÔ¨Åcial role of MTX as a second-line treatment in case of CS dependence and as a sparing treatment. As in most of studies on AOSD, MTX was the most prescribed DMARD and had good tolerance. Fourth, 23 of 57 patients (40%) received transient IVIg therapy. Tolerance was excellent but remission of the disease was obtained in only 4 cases (17%). Data on IVIg in AOSD are scarce. In 2 open-label studies, IVIg was shown to be effective in 8 of 14 patients early in the disease course.34,40 For Kim et al,24 IVIg treatment had no consequence on the course or the prognosis of AOSD. As with patients with RHS,12 IVIg may be useful in life-threatening situations. Fifth, a growing body of literature has documented the efÔ¨Åcacy of several biologic agents in the treatment of CS- and DMARD-refractory AOSD. In the current series, 15 patients (26%) did not achieve satisfactory control of the disease with CSs and DMARDs and were thus given various biologic agents, which is comparable to the study of Franchini et al.18 TNF-Œ± blockers were efÔ¨Åcient in controlling the disease in only 47% of our prescriptions. TNF-Œ± blockers seemed to be more effective in chronic than in polycyclic courses of the disease.35 Previous results on TNF-Œ± blockers were retrospective and rather heterogeneous. In the largest previous study15 that we know of, 25 prescriptions of anti-TNF-Œ± (inÔ¨Çiximab and etanercept) were administered to 20 AOSD patients resistant to conventional DMARDs. After a mean follow-up of 15 months, only 5 prescriptions led to remission, whereas 4 were ineffective, and 4 led to AEs. Altogether, more than 120 patients were treated with a TNF-Œ± blocker and, as in the present study, complete remissions were irregular and often transient.15,21,24,33 Switching from 1 TNF-Œ± blocker to another was beneÔ¨Åcial in only a few cases; in the current study, this was observed only once. Finally, these treatments may cause worsening of the disease because of RHS occurrence.23 In the present study, the use of anakinra resulted in a rapid and complete remission in 5 of 6 patients. Several case reports and 2 small case series have shown spectacular effects of anakinra in AOSD,22,28 which has been conÔ¨Årmed by 2 prospective clinical trials. In the Ô¨Årst, Laskari et al28 reported on 25 AOSD patients, 9 of whom were given anakinra alone, and 16, anakinra and a DMARD. As in our patients, the clinical efÔ¨Åcacy was rapid (0.2 mo) and frequent (84% at 3 mo; 80% at end of follow-up), with a favorable agent safety proÔ¨Åle.36 In the second, more recent trial, Nordstr√∂m et al32 compared the outcomes of 12 AOSD patients treated with anakinra and 10 patients treated with a DMARD (all taking prednisolone ‚â•10 mg/d) early in the course of AOSD to avoid the adverse effects of corticotherapy. With no statistically signiÔ¨Åcant differences in characteristics between the 2 groups, greater improvements were seen in the anakinra group. None of our patients received anti-IL-6 receptor, which has been considered for treating AOSD. A recent review of 35 patients reported that 86% of tocilizumab-treated patients experienced prompt articular improvement, and 96% experienced a disappearance of systemic symptoms.10 Regarding safety, tocilizumab was well tolerated, but severe effects such as macrophage activation syndrome39,40 are possible and require ongoing vigilance. In the present series, one-third of the patients experienced life-threatening complications of AOSD, none of which led to ¬© 2014 Lippincott Williams & Wilkins Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine ‚Ä¢ Volume 93, Number 2, March 2014 death. The most common complication was RHS. We report 8 of 57 (14%) cases of AOSD-associated RHS, which is similar to the incidence reported in 2 other series;1,20 however, in those series, the time relationship between RHS and AOSD was not reported. A complicated onset of AOSD is overrepresented in the current study compared with previous data (33% vs. 5%‚Äì 11%, respectively).3,6,38 This may be the consequence of 1) the high proportion of polycyclic AOSD; 2) the recruitment in internal medicine department compared with rheumatology departments in other series; 3) the recruitment in a university hospital, which often increases the proportion of severe cases; and 4) a longer follow-up period than in previous studies. None of the 3 deaths during the study period could be attributed to Still disease. This is in agreement with the benign and nonfatal nature of the disease. Nevertheless, authors of 2 recent studies in Asian patients24,43 reported high mortality rates (9.26% and 10%, respectively) due to infection and disease progression, but only a small proportion of the patients had received biologic agents. Multicentric studies are needed to check whether the prognosis of AOSD is signiÔ¨Åcantly inÔ¨Çuenced by patient ethnicity. In the present study, we sought the clinical and laboratory features of good prognosis (monocyclic disease) and poor prognosis (polycyclic, chronic articular disease, treatment failure, and complications). Arthritis at diagnosis was a signiÔ¨Åcant predictor of disease chronicity. This reinforces the few reports that confer a poor prognosis to the presence of polyarthritis.6,8 In addition, as already observed,6 joint erosion (as shown by standard radiography) seemed to predict chronicity. However, there could have been an overlap with seronegative rheumatoid arthritis. Moreover, we showed that high fever was a good predictor of a systemic‚Äîparticularly monocyclic‚Äîdisease course. To our knowledge, this has not been reported before. According to our data, thrombocytopenia at diagnosis predicted the occurrence of complications, but in fact this is confounding, because the most prevalent complication in our patients was RHS, in which thrombocytopenia is almost constant. It is noteworthy that our results suggest an early diagnosis improves the prognosis: indeed, in the unadjusted analyses, a ‚â§2-month delay was predictive of a monocyclic course. Moreover, GF level testing during the diagnostic process was significantly associated with earlier diagnoses and seemed to be predictive of a monocyclic course. However, that testing could be a confounding factor because GF may have been more frequently assessed in highly symptomatic patients who evolved spontaneously toward a monocyclic course. Thus, we may recommend early and frequent GF assays to help early diagnosis and early treatment. Several studies have suggested the use of SF as prognostic factor in AOSD.6,27,29,43 In our unadjusted analysis results, a mild increase in SF was associated with a chronic course of AOSD. In 2009, in a retrospective study of 61 Chinese AOSD patients, Zeng et al43 found that an elevated SF level was an unfavorable prognostic factor. Two more recent studies showed that high SF levels were signiÔ¨Åcantly associated with a high rate of relapse.6,27 In a study by Lee et al29 in 2009, only the middle range of the adjusted SF level (area under the curve divided by the number of hospitalization days; that is, 784.1‚Äì4120.0 ng/mL) had a signiÔ¨Åcant predictive value for disease chronicity. The present study results are consistent with this, which seems close to the clinical reality. Indeed, higher SF levels (>8000 or 10,000 ng/mL) do not seem predictive of a chronic course, but were associated with the ‚Äúhighly symptomatic‚Äù type of AOSD we describe below, which is monocyclic or associated with RHS.20 ¬© 2014 Lippincott Williams & Wilkins Prognostic Factors in Adult-Onset Still Disease Among other laboratory features, the white blood cell count seems to be an interesting predictor of patient outcome. Kong et al27 found that increased white blood cell counts (‚â•30,000/mm3) were associated with AOSD relapses. Other studies revealed that inÔ¨Çammation markers, such as elevated serum ESR or CRP, were signiÔ¨Åcantly associated with poor prognoses24 or higher relapse rates.27 The present study provides important information on the clinical, laboratory, therapeutic, and prognostic features of AOSD, but has a number of limitations. First, the study was retrospective, noncomparative, and had to contend with some degree of missing clinical and laboratory investigation data. A prospective study would provide much clearer and exhaustive information. The second limitation is that the study was carried out in a single university hospital and might have included a higher percentage of severe cases than would be expected in primary settings. Thus, the results are applicable only to secondary or tertiary AOSD practices and to AOSD cases seen in internal medicine departments. Moreover, AOSD management differs widely between countries and even physicians. Therefore, we await additional studies from other institutions in different countries to enrich and conÔ¨Årm the present results. Conclusion The results of this second retrospective study on French AOSD patients are consistent with previous results. AOSD diagnosis remains difÔ¨Åcult. GF level testing seems to be underused, whereas it would shorten the delay to diagnosis and help in predicting high-dose CS dependence (which is prevalent in AOSD). The hypermetabolism on 18FDG-PET scan focuses on the lymphoid tissues and glands. The usefulness of the treatment sequence CS, MTX, then IL-1-RA before TNF-Œ± blockers should be conÔ¨Årmed by prospective studies. The long mean follow-up period (8.4 yr) allowed us to study various prognostic factors. On the basis of the initial clinical and laboratory data, we suggest distinguishing 2 phenotypes of the disease, which progress quite differently: 1) a highly symptomatic onset with high fever, serositis, elevated liver enzymes, and very high SF with arthralgia but less commonly arthritis, which would evolve to systemic‚Äîmonocyclic or polycyclic‚ÄîAOSD; and 2) a more progressive onset with arthritis and radiologic joint erosions and poor systemic symptoms, which would progress to a chronic course. Overall, our results indicate that AOSD is a relatively benign disease and that most deaths may be related to the AEs of long-term treatment. Prospective studies in independent cohorts are needed to conÔ¨Årm these results. ACKNOWLEDGMENTS The authors thank Pr. R. Chapurlat, Dr. S. Dargent, Pr. P. Lantelme, Pr. G. Llorca, Pr. D. Peyramond, Pr. J. Tebib, Pr. E. Vignon, Pr. D. Vital-Durand, and Pr. F. Zoulim for their valuable comments. REFERENCES 1. Arlet J-B, Le Thi Huong D, Marinho A, Amoura Z, Wechsler B, Papo T, Piette J-C. Reactive haemophagocytic syndrome in adult-onset Still‚Äôs disease: a report of six patients and a review of the literature. Ann Rheum Dis. 2006;65:1596‚Äì1601. 2. Bywaters EG. Still‚Äôs disease in the adult. Ann Rheum Dis. 1971;30:121‚Äì133. 3. Cagatay Y, Gul A, Cagatay A, Kamali S, Karadeniz A, Inanc M, Ocal L, Aral O, Konice M. Adult-onset Still‚Äôs disease. Int J Clin Pract. 2007;63:1050‚Äì1055. www.md-journal.com Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 97 Gerfaud-Valentin et al Medicine ‚Ä¢ Volume 93, Number 2, March 2014 4. Chen P-D, Yu S-L, Chen S, Weng X-H. Retrospective study of 61 patients with adult-onset Still‚Äôs disease admitted with fever of unknown origin in China. Clin Rheumatol. 2012;31:175‚Äì181. 21. Husni ME, Maier AL, Mease PJ, Overman SS, Fraser P, Gravallese EM, Weinblatt ME. Etanercept in the treatment of adult patients with Still‚Äôs disease. Arthritis Rheum. 2002;46:1171‚Äì1176. 5. Choe J-Y, Chung DS, Park S-H, Kwon H-H, Kim S-K. Clinical signiÔ¨Åcance of 18F-Ô¨Çuoro-dexoxyglucose positron emission tomography in patients with adult-onset Still‚Äôs disease: report of two cases and review of literatures. Rheumatol Int. 2010;30:1673‚Äì1676. 22. Kalliolias GD, Georgiou PE, Antonopoulos IA, Andonopoulos AP, Liossis S-NC. Anakinra treatment in patients with adult-onset Still‚Äôs disease is fast, effective, safe and steroid sparing: experience from an uncontrolled trial. Ann Rheum Dis. 2007;66:842‚Äì843. 6. Colina M, Zucchini W, Ciancio G, Orzincolo C, Trotta F, Govoni M. The evolution of adult-onset Still disease: an observational and comparative study in a cohort of 76 Italian patients. Semin Arthritis Rheum. 2011;41:279‚Äì285. 7. Crouzet J, Boudousq V, Lechiche C, Pouget JP, Kotzki PO, Collombier L, Lavigne JP, Sotto A. Place of (18)F-FDG-PET with computed tomography in the diagnostic algorithm of patients with fever of unknown origin. Eur J Clin Microbiol Infect Dis. 2012;31:1727‚Äì1733. 8. Cush JJ, Medsger TA Jr, Christy WC, Herbert DC, Cooperstein LA. Adult-onset Still‚Äôs disease. Clinical course and outcome. Arthritis Rheum. 1987;30:186‚Äì194. 9. De Bandt M, Saint-Marcoux B. Tocilizumab for multirefractory adult-onset Still‚Äôs disease. Ann Rheum Dis. 2009;68:153‚Äì154. 10. De Boysson H, Fevrier J, Nicolle A, Auzary C, Geffray L. Tocilizumab in the treatment of the adult-onset Still‚Äôs disease: current clinical evidence. Clin Rheumatol. 2013;32:141‚Äì147. 23. Kaneko K, Kaburaki M, Muraoka S, Tanaka N, Yamamoto T, Kusunoki Y, Abe H, Endo H, Kawai S. Exacerbation of adult-onset Still‚Äôs disease, possibly related to elevation of serum tumor necrosis factor-alpha after etanercept administration. Int J Rheum Dis. 2010;13:e67‚Äì69. 24. Kim H-A, Sung J-M, Suh C-H. Therapeutic responses and prognosis in adult-onset Still‚Äôs disease. Rheumatol Int. 2012;32:1291‚Äì1298. 25. Kim YJ, Kim SI, Hong K-W, Kang MW. Diagnostic value of (18)F-FDG PET/CT in patients with fever of unknown origin. Intern Med J. 2012;42:834‚Äì837. 26. Kobayashi M, Takahashi Y, Yamashita H, Kaneko H, Mimori A. BeneÔ¨Åt and a possible risk of tocilizumab therapy for adult-onset Still‚Äôs disease accompanied by macrophage-activation syndrome. Mod Rheumatol. 2011;21:92‚Äì96. 27. Kong X-D, Xu D, Zhang W, Zhao Y, Zeng X, Zhang F. Clinical features and prognosis in adult-onset Still‚Äôs disease: a study of 104 cases. Clin Rheumatol. 2010;29:1015‚Äì1019. 11. Efthimiou P, Paik PK, Bielory L. Diagnosis and management of adult onset Still‚Äôs disease. Ann Rheum Dis. 2006;65:564‚Äì572. 28. Laskari K, Tzioufas AG, Moutsopoulos HM. EfÔ¨Åcacy and long-term follow-up of IL-1R inhibitor anakinra in adults with Still‚Äôs disease: a case-series study. Arthritis Res Ther. 2011;13:R91. 12. Emmenegger U, Frey U, Reimers A, Fux C, Semela D, Cottagnoud P, Spaeth PJ, Neftel KA. Hyperferritinemia as indicator for intravenous immunoglobulin treatment in reactive macrophage activation syndromes. Am J Hematol. 2001;68:4‚Äì10. 29. Lee S-W, Park Y-B, Song J-S, Lee S-K. The mid-range of the adjusted level of ferritin can predict the chronic course in patients with adult onset Still‚Äôs disease. J Rheumatol. 2009;36:156‚Äì162. 13. Fautrel B, Borget C, Rozenberg S, Meyer O, Le Loet X, Masson C, Koeger AC, Kahn MF, Bourgeois P. Corticosteroid sparing effect of low dose methotrexate treatment in adult Still‚Äôs disease. J Rheumatol. 1999;26:373‚Äì378. 30. Magadur-Joly G, Billaud E, Barrier JH, Pennec YL, Masson C, Renou P, Prost A. Epidemiology of adult Still‚Äôs disease: estimate of the incidence by a retrospective study in west France. Ann Rheum Dis. 1995;54:587‚Äì590. 14. Fautrel B, Le Moel G, Saint-Marcoux B, Taupin P, Vignes S, Rozenberg S, Koeger AC, Meyer O, Guillevin L, Piette JC, Bourgeois P. Diagnostic value of ferritin and glycosylated ferritin in adult onset Still‚Äôs disease. J Rheumatol. 2001;28:322‚Äì329. 31. Mavragani CP, Spyridakis EG, Koutsilieris M. Adult-onset Still‚Äôs disease: from pathophysiology to targeted therapies. Int J InÔ¨Çam. 2012;2012:879020. 15. Fautrel B, Sibilia J, Mariette X, Combe B. Tumour necrosis factor alpha blocking agents in refractory adult Still‚Äôs disease: an observational study of 20 cases. Ann Rheum Dis. 2005;64:262‚Äì266. 32. Nordstrom D, Knight A, Luukkainen R, van Vollenhoven R, Rantalaiho V, Kajalainen A, Brun JG, Proven A, Ljung L, Kautiainen H, Pettersson T. BeneÔ¨Åcial effect of interleukin 1 inhibition with anakinra in adult-onset Still‚Äôs disease. An open, randomized, multicenter study. J Rheumatol. 2012;39:2008‚Äì2011. 16. Fautrel B, Zing E, Golmard J-L, Le Moelle G, Bissery A, Riou C, Rozenberg S, Piette J-C, Bourgeois P. Proposal for a new set of classiÔ¨Åcation criteria for adult-onset Still disease. Medicine (Baltimore). 2002;81:194‚Äì200. 33. Pay S, Turkcapar N, Kalyoncu M, Simsek I, Beyan E, Ertenli I, Osturk MA, Duzgun N, Erdem H, Ozbalkan Z. A multicenter study of patients with adult-onset Still‚Äôs disease compared with systemic juvenile idiopathic arthritis. Clin Rheumatol. 2006;25:639‚Äì644. 17. Fraisse T-C, Degraeve F, Riviere S, Le Quellec A. [Evolution and prognosis of adult onset Still‚Äôs disease. A monocentric study of 17 patients]. Rev Med Interne. 2006;27:658‚Äì664. 34. Permal S, Wechsler B, Cabane J, Perrot S, Blum L, Imbert JC. [Treatment of Still disease in adults with intravenous immunoglobulins]. Rev Med Interne. 1995;16:250‚Äì254. 18. Franchini S, Dagna L, Salvo F, Aiello P, Baldissera E, Sabbadini MG. EfÔ¨Åcacy of traditional and biologic agents in different clinical phenotypes of adult-onset Still‚Äôs disease. Arthritis Rheum. 2010;62:2530‚Äì2535. 35. Pouchot J, Arlet J-B. Biological treatment in adult-onset Still‚Äôs disease. Best Pract Res Clin Rheumatol. 2012;26:477‚Äì487. 19. Henter J-I, Horne A, Arico M, Egeler RM, Filipovic AH, Imashuku S, Ladisch S, McClain K, Webb D, Winiarski J, Janka G. HLH-2004: diagnostic and therapeutic guidelines for hemophagocytic lymphohistiocytosis. Pediatr Blood Cancer. 2007;48:124‚Äì131. 20. Hot A, Toh M-L, Coppere B, Perard L, Girard-Madoux MH, Mausservey C, Desmurs-Clavel H, Ffrench M, Ninet J. Reactive hemophagocytic syndrome in adult-onset Still disease: clinical features and long-term outcome: a case‚Äìcontrol study of 8 patients. Medicine (Baltimore). 2010;89:37‚Äì46. 98 www.md-journal.com 36. Pouchot J, Fautrel B. Maladie de Still de l‚Äôadulte. In: Guillevin L, Meyer O, Sibilia J, eds. Traite des Maladies et Syndromes Systemiques. 5th ed. Paris: Flammarion; 2008: 1249‚Äì1263. 37. Pouchot J, Sampalis JS, Beaudet F, Carette S, Decary F, Salusinsky-Sternbach M, Hill RO, Gutkowski A, Harth M, Myhal D. Adult Still‚Äôs disease: manifestations, disease course, and outcome in 62 patients. Medicine (Baltimore). 1991;70:118 ‚Äì136. 38. Uppal SS, Al-Mutairi M, Hayat S, Abraham M, Malaviya A. Ten years of clinical experience with adult onset Still‚Äôs disease: is the outcome improving- Clin Rheumatol. 2007;26:1055‚Äì1060. ¬© 2014 Lippincott Williams & Wilkins Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine ‚Ä¢ Volume 93, Number 2, March 2014 39. Vignes S. Percentage of glycosylated serum ferritin remains low throughout the course of adult onset Still‚Äôs disease. Ann Rheum Dis. 2000;59:347‚Äì350. Prognostic Factors in Adult-Onset Still Disease 42. Yamaguchi M, Ohta A, Tsunematsu T, Kasukawa R, Mizushima Y, Kashiwagi H, Kashiwazaki S, Tanimoto K, Matsumoto Y, Ota T. Preliminary criteria for classiÔ¨Åcation of adult Still‚Äôs disease. J Rheumatol. 1992;19:424‚Äì430. 40. Vignes S, Wechsler B, Amoura Z, Papo T, Frances C, Huong DL, Veyssier P, Godeau P, Piette JC. Intravenous immunoglobulin in adult Still‚Äôs disease refractory to non-steroidal anti-inÔ¨Çammatory drugs. Clin Exp Rheumatol. 1998;16:295‚Äì298. 43. Zeng T, Zou Y-Q, Wu M-F, Yang C-D. Clinical features and prognosis of adult-onset Still‚Äôs disease: 61 cases from China. J Rheumatol. 2009;36:1026‚Äì1031. 41. Wouters JM, Van de Putte LB. Adult-onset Still‚Äôs disease; clinical and laboratory features, treatment and progress of 45 cases. Q J Med. 1986;61:1055‚Äì1065. 44. Zhu G, Liu G, Liu Y, Xie Q, Shi G. Liver abnormalities in adult onset Still‚Äôs disease: a retrospective study of 77 Chinese patients. J Clin Rheumatol. 2009;15:284 ‚Äì288. ¬© 2014 Lippincott Williams & Wilkins www.md-journal.com Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 99 CLINICAL RESEARCH STUDY Adult Primary Immune Deficiency: What Are We Missing? Bharat T. Srinivasa,a,b Reza Alizadehfar,a Martin Desrosiers,c Joseph Shuster,a Nitika Pant Pai,d Christos M. Tsoukasa,b a Division of Allergy and Clinical Immunology, bDepartment of Microbiology and Immunology, cDepartment of Otorhinolaryngology, and dDivision of Clinical Epidemiology, McGill University Health Centre, Montreal, Quebec, Canada. ABSTRACT BACKGROUND: More than 200 primary immune deficiencies have been described. In adults, their identification can be difficult. The lack of timely referrals, diagnostic facilities, and available expertise often delay appropriate treatment. Because an increasing number of adults are now diagnosed with immune deficiencies, there is a need to better understand the immune deficits in this age group. The study objective was to analyze the diagnostic spectrum of adults with primary immune deficiency and to determine the presumptive diagnostic accuracy of the referring physicians. METHODS: We conducted a retrospective chart review over a 10-year period of all individuals referred to a dedicated center for adults with primary immune deficiency. Suspected cases were confirmed using standard clinical criteria and state of the art immune assays. RESULTS: Of the 381 individuals studied, 244 were diagnosed as immune deficient. Of these, 210 had primary immune deficiency classified as novel, defined, and undefined. Forty-three patients had a prior diagnosis and were referred for follow-up care, and 201 patients were newly diagnosed. Most patients had common variable immune deficiency. Despite an apparent high index of suspicion in initiating the referrals, only one third of these patients had a prior quantitative assessment of serum immunoglobulins. CONCLUSIONS: In this first known analysis of a large cohort of adults with suspected immune deficiency using established diagnostic criteria, we confirmed the diagnosis in two thirds of all patients. Our findings highlight the wide spectrum of primary immune deficiency states seen in adult medical practices and the need for increased awareness of their existence. ¬© 2012 Published by Elsevier Inc. ‚Ä¢ The American Journal of Medicine (2012) 125, 779-786 KEYWORDS: Adult; Clinical; Common variable immune deficiency; Good‚Äôs syndrome; Hypogammaglobulinemia; Immune deficiency; Primary immune deficiency; T-cell immune deficiency Primary immune deficiencies are caused by inherited defects in the immune system. More than 100 molecular defects, predominantly in children, have been described recently.1,2 An increasing number of adults are being diagnosed with primary immune deficiency, and recent studies estimate that up to 1:1200 people in the United States are diagnosed with some form of primary immune deficiency.3 Immunoglobulin (Ig)-A, the most commonly diagnosed in Funding: Research Institute of the McGill University Health Centre, Anna Marie Solinas Laroche Award. Conflict of Interest: None. Authorship: All authors had access to the data and played a role in writing this manuscript. Requests for reprints should be addressed to Christos M. Tsoukas, A5-140, McGill University Health Center, 1650 Cedar Avenue, Montreal, Quebec, Canada, H3G 1A4. E-mail address: chris.tsoukas@muhc.mcgill.ca 0002-9343/$ -see front matter ¬© 2012 Published by Elsevier Inc. http://dx.doi.org/10.1016/j.amjmed.2012.02.015 the western world, is usually subclinical in presentation, with an incidence in Caucasians of up to 1:600.4 The most frequently diagnosed adult primary immune deficiency with clinical importance is common variable immune deficiency, with a prevalence of 1:25,000 among Caucasians.4 Usually, higher rates of primary immune deficiencies are observed in populations with high consanguinity rates or among genetically isolated populations.1 These deficiencies are a highly variable set of disorders with a diverse range of genotypes and clinical and immune phenotypes. In addition to the well-defined syndromes, subtle and often subclinical immune defects can occur in adults. These are expressed with variable penetrance and may go unrecognized in general clinical practice. Often, individuals are recognized as immune deficient after extensive investigations for unforeseen, unusual, recurrent, or severe infections. Because of the relative rarity, clinical heterogeneity, 780 The American Journal of Medicine, Vol 125, No 8, August 2012 and limited diagnostic tools, misdiagnoses in clinical set‚óè macrophage function assays; tings and misclassification of patients often occur. Delays in ‚óè quantitative serum complement and immunoglobulin isomaking specific diagnoses lead to late initiation of appropriate type quantitation assays of specific antibody responses to therapies and subsequent, often irreversible, complications.5 a variety of vaccine challenges; and Abnormalities in humoral immunity account for more ‚óè enzymatic and genetic assays as necessary. Institutional than 50% of primary immune deapproval for the study was obficiencies.6 Individuals with these tained from the McGill Univerdefects, among other problems, sity Health Centre. CLINICAL SIGNIFICANCE are susceptible to bacterial infecThe study included adults tions. T-cell anomalies account ‚óè A high proportion of adults referred to (aged œæ 18 years) at the initial evalfor fewer primary immune defius were confirmed to have an immune uation, where the reasons for referciencies but are considerably more deficiency. ral, clinical presentation, suspected 7 severe and often fatal. In adults, pre- and final postevaluation diag‚óè Despite the high index of suspicion, screendeaths may occur from opportunoses were available. Individuals ing immunoglobulin levels were carried out nistic infections or neoplasms. with human immunodeficiency viChildren may present with a failonly in a minority of those referred. rus infection or receiving immuno7 ure to thrive. suppressive therapy were excluded ‚óè Current guidelines for the diagnosis of A majority of adults with prifrom the study. immune deficiency could not classify all mary immune deficiencies are not our patients. diagnosed or treated early in their Statistical Analysis course.8 This is due to an overall ‚óè Older individuals with a history of reThe Mann-Whitney test was used paucity of prevailing knowledge current or severe infections may have a for comparing data sets that were of these illnesses in the health care humoral immune deficiency characternon-normally distributed. For norcommunity and to an insufficient ized by low levels of B cells. mally distributed data, the Student awareness of the existence of such t test was used. P values ’Ö.05 conditions in adults. were regarded as significant. In 2000, the Immune Deficiency Treatment Centre of the McGill University Health Centre established a clinic dedicated to the diagnosis and RESULTS care of adults with primary immune deficiencies. The aims Patient Demographics of the clinic are to improve the long-term management of A total of 381 individuals met the study entry criteria. young adults with primary immune deficiencies transitionFigure 1 outlines the diagnostic algorithm and classification ing to adult care; identify, diagnose, and manage adults with by the presence and type of immune deficiency. Table 1 primary immune deficiencies; and raise awareness in the lists the distribution of gender, age, and racial profiles for medical community of the existence of these adult each group of referred patients. disorders. The patients were almost exclusively Caucasian (97%), Our study objectives were to determine the spectrum of and the majority were female (61%). The median age did immune defects in this adult cohort and to determine the not differ by gender. The median age at the date of referral presumptive prereferral diagnostic accuracy of immune dewas 48 years in men and 45 years in women. Patients with ficiency by the referring physicians. primary and secondary immune deficiency did not differ significantly with respect to gender or ethnicity; however, MATERIALS AND METHODS men with secondary immune deficiency were older than We conducted a retrospective chart review of all referrals to those diagnosed with primary immune deficiency. Those our center with a suspected diagnosis of immune deficiency with autoinflammatory and autoimmune syndromes were during a 10-year period (January 1, 2000, to December 31, younger but did not differ in gender and ethnicity from 2009). Established clinical and laboratory diagnostic criteria those with immune deficiency. used to identify primary and secondary immune deficiencies Most referrals originated from medical subspecialists. were based on guidelines of the European Society for ImThe majority (55%) were from allergists and clinical immunodeficiencies and the Pan-American Group for Immumunologists, otorhinolaryngologists, and pulmonologists. nodeficiency (ESID/PAGID).9 Immune testing included but Only 17% came from family practitioners and internists. A was not limited to diagnosis of immune deficiency was confirmed in 244 individuals, representing 64% of the study patients (Table 1). ‚óè extensive immunophenotyping of T, B, and natural killer cells; Clinical Phenotypes ‚óè lymphocyte functional assays, such as lymphocyte proliferative responses to mitogens and recall antigens, cyPatients were grouped according to the presence or abtokine release assays, and in vivo anergy screening; sence of immune deficiency on the basis of a combination Srinivasa et al Adult Primary Immune Deficiency 781 Figure 1 Diagnostic algorithm and classification by presence and type of immune deficiency. ESID/ PAGID œ≠ European Society for Immunodeficiencies and the Pan-American Group for Immunodeficiency. of clinical and laboratory criteria. Those with immune deficiencies were classified as having confirmed or probable primary immune deficiency if they met the ESID/ PAGID diagnostic criteria and secondary immune deficiency if it was acquired as a result of a comorbid condition. Individuals with a clinical history and laboratory evidence of immune deficiency that did not meet ESID/PAGID diagnostic criteria and had no evidence of a secondary cause were categorized as novel or undefined cases (Figure 1). A wide spectrum of clinical phenotypes was noted. Of the 244 patients diagnosed with immune deficiency, 167 had ESID/PAGID defined primary immune deficiency, 43 had a novel or undefined deficiency, and the remaining 782 Table 1 The American Journal of Medicine, Vol 125, No 8, August 2012 Demographic Profiles of All Referred Patients for Assessment in the Immune Deficiency Clinic Immune Deficient Characteristics Gender Male Female Median age (y) Male Female Ethnicity Caucasian Black Asian Nonimmune Deficient Total Patients n œ≠ 381 Primary n œ≠ 210 (55%) Secondary n œ≠ 34 (9%) Autoinflammatory* and Autoimmune‚Ä† n œ≠ 21 (5%) Other/Nonimmune Deficient N œ≠ 116 (31%) 39% 61% 45 48 45 42% 58% 44 42 45 41% 59% 56 70 47 43% 57% 39 38 39 33% 67% 46 50 42 (149) (232) (18-89) (18-89) (18-88) 97% (372) 1% (3) 2% (6) (88) (122) (18-83) (18-80) (19-83) 97% (205) 2% (3) 1% (2) (14) (20) (19-79) (48-79) (19-75) 94% (33) 0% (0) 6% (1) (9) (12) (18-79) (19-64) (18-79) 95% (20) 0% (0) 5% (1) (38) (78) (18-89) (18-89) (18-88) 98% (114) 0% (0) 2% (2) Patients with primary and secondary immune deficiency did not differ significantly with respect to gender or ethnicity; however, men with secondary immune deficiency were older than those diagnosed with primary immune deficiency. Those with autoinflammatory and autoimmune syndromes were younger but did not differ in gender and ethnicity from those with immune deficiency. *Autoinflammation is defined as unprovoked inflammation without high-titer autoantibodies or antigen-specific T lymphocytes. The diagnoses were established using the International Union of Immunological Societies Primary Immunodeficiencies Expert Committee criteria.18 ‚Ä†Autoimmune refers to one of many clinical syndromes caused by the activation of T or B cells, or both, in the absence of an ongoing infection or other discernible cause.19 These syndromes included patients with systemic lupus erythematosus or rheumatoid arthritis and were diagnosed using criteria of the American College of Rhumatology.20 34 had a secondary immune deficiency. The most commonly observed primary immune deficiencies were common variable immune deficiency (43%, n œ≠ 72), IgA deficiency (15%, n œ≠ 25), hypogammaglobulinemia with infections that did not meet the diagnostic criteria for common variable immune deficiency (10%, n œ≠ 17), and idiopathic CD4œ© T lymphopenia (8%, n œ≠ 14). A total of 201 patients were newly diagnosed, resulting in a referral accuracy of 53%. The remaining 43 patients had an established primary immune deficiency and were transitioned from pediatric centers or relocated from other clinics. Overall, by using the ESID/PAGID diagnostic criteria and ruling out secondary immune deficiencies, we accurately classified 80% of the immune-deficient patients as having a primary cause. The remaining 20% were considered to have novel (14%) or undefined (6%) primary immune deficiency (Figure 1). By grouping individuals with a documented history of bacterial infections and humoral immune defects, we noted that some patients did not fit the ESID/PAGID criteria. These patients had hypogammaglobulinemia with less frequent or nonrecurrent and mild bacterial infections, none requiring IgG replacement therapy; and pronounced B-cell lymphocytopenia with severe or recurrent bacterial infections. The first group consisted of 17 individuals, mostly older than those with common variable immune deficiency but with similar levels of IgG (Figures 2 and 3). They were primarily referred because of a perceived increase in the frequency of respiratory symptoms or evidence of hypogammaglobulinemia. The second group consisted of 12 individuals with a median age of more than 65 years, a history of recurrent bacterial infections, low B-cell numbers (2 standard deviations below mean), and borderline normal serum levels of immunoglobulins (Figures 2 and 3). The infections were often severe and not confined to the respiratory system. In 1 case, a 74-year-old woman had 9 hospitalizations in 3 years for episodes of sepsis, with 6 of these admissions to an intensive care unit (Figures 2 and 3). In addition to patients with B-cell lymphocytopenia, other lymphocyte subset anomalies were noted. Idiopathic CD4œ© T-cell lymphocytopenia is a rare disease and almost always seen in adults.10 Fourteen individuals (7% of our patients) were diagnosed with this pathology, reflecting the unique nature of our adult cohort. Despite low and persistent absolute CD4œ© T-cell counts, only 1 of these patients developed opportunistic infections. Most of these patients were asymptomatic and diagnosed during T-cell phenotyping. Our patients differed from many individuals previously reported in the literature who were first diagnosed after the appearance of opportunistic infections. Presumably our patients with idiopathic CD4œ© T-cell cytopenia are at risk of developing opportunistic infections despite being asymptomatic. Six individuals had persistently low natural killer cells. All had a history of severe and recurrent viral infections, mostly with cytomegalovirus. The relative and absolute natural killer cell numbers are noted in Figure 4. Despite extensive investigation, 14 patients with recurrent or severe infections had a varied laboratory spectrum of immune abnormalities and lacked evidence for a secondary cause of immune deficiency. They did not meet the ESID/ PAGID criteria or match the above novel immune deficiency phenotypes. These individuals were therefore categorized as having undefined primary immune deficiency. Future availability of new diagnostic assays may allow for reclassification of these patients. Srinivasa et al Adult Primary Immune Deficiency 783 Serum IgG Mann Whitney test P values: < 0.0001 0.0047 Immunoglobulin IgG (g/l) 20 ns 0.0001 < 0.0001 15 ns 10 5 0 M edian n= Common Variable Immune Deficiency Hypogammaglobulinemia B-cell Lymphocytopenia IgA Deficiency Non Immunedeficient 4.650 26 5.500 17 8.000 12 9.130 11 10.50 47 Figure 2 Comparison of the serum IgG of each group of patients with humoral immune deficiencies, including those of the 2 novel groups. All serum samples were obtained before any immunoglobulin replacement therapy for patients with common variable immune deficiency. IgG levels of individuals with IgA deficiency were in the normal range (7-16 g/L) and similar to those of the nonimmune-deficient individuals. Median levels of IgG in the B-cell lymphopenia group were borderline normal and significantly higher than those with hypogammaglobulinemia. Significant differences were noted between the IgG levels of the nonimmune-deficient individuals and those with B-cell lymphopenia, as well as those with hypogammaglobulinemia. Ig œ≠ immunoglobulin. Five deaths were reported during the study period. All deceased patients were male with severe T-cell defects. These included a 21-year-old with dedicator of cytokinesis 8 deficiency who died of metastatic anal carcinoma and a 70-year-old with Good‚Äôs syndrome who died of progressive multifocal leukoencephalopathy. Of 2 individuals with severe T-cell abnormalities and multiple opportunistic infections, 1 died of leukemia and 1 died of infectious complications after bone marrow transplantation at 27 years of age. The last individual had idiopathic CD4 T-cell lymphopenia and died of overwhelming sepsis at 25 years of age. The reasons for referral were analyzed. A large number of patients had chronic sinusitis (n œ≠ 53, 14%), chronic bronchitis (n œ≠ 18, 5%), and recurrent pneumonia (n œ≠ 17, 5%). Because these clinical findings are frequently noted among healthy adults, we compared the relative distributions of each of these presentations in the nonimmunedeficient individuals with those with common variable immune deficiency. A significantly larger number of patients referred for chronic sinusitis and recurrent pneumonia had common variable immune deficiency (P œΩ.01). This was not the case for those referred with chronic bronchitis or bronchiectasis. Viral infections were significantly more common as reasons for referral (P œΩ.01) in the nonimmunedeficient patients. Other patients were referred for fatigue, chronic diarrhea, recurrent otitis, and chronic skin and fungal infections. Noninfectious systemic disorders also were present in 26 nonimmune-deficient patients and 6 common variable im- mune-deficient patients. In patients with common variable immune deficiency, associated diseases included idiopathic thrombocytopenic purpura or hemolytic anemia. Noninfectious dermatologic disorders were noted exclusively in the nonimmune-deficient group, and this finding was statistically significant when compared with those with common variable immune deficiency. Of the common variable immune-deficient patients, 78% (56) had sinopulmonary infections as their primary presenting phenotype. Of interest, only one third of those with common variable immune deficiency had a prereferral quantitative assessment of serum immunoglobulins. DISCUSSION With the exception of IgA deficiency, primary immune deficiencies are relatively uncommon adult diseases. Because no single clinical phenotype exists, diagnosing these deficiencies in general clinical practice is often perceived as difficult, and a low index of suspicion often exists. This perception may in part be reflected in the present study: Only 17% of all referrals came from primary care physicians. The majority of referrals to the clinic (53%) came from specialists in upper airway diseases. This was not surprising given that most of our patients with primary immune deficiency had humoral defects and presented with bacterial infections of the respiratory tract. In pediatric cohorts of primary immune deficiency, most patients are male.11-15 Of note, our patients were mostly 784 The American Journal of Medicine, Vol 125, No 8, August 2012 Age at first visit M ann Whitney test P values: 0.0073 0.0199 < 0.0001 100 ns 80 Age 60 40 20 0 Common Variable Immune Deficiency M edian n= 36.00 57 Hypogammaglobulinemia B-cell Lymphocytopenia 52.00 17 67.00 IgA Deficiency 48.00 11 12 Non Immunedeficient 46.00 47 Figure 3 Comparison of age distributions of groups of patients with humoral deficiencies with those without immune deficiency. B-cell lymphocytopenic patients were significantly older than individuals in all groups at their first clinic visit, with a median age of 67 years. The onset of symptoms that were attributed to the immune deficiency occurred later in life. In contrast, the common variable immune-deficient patients had the lowest median age at their first clinic visit, and the clinical onset of immune deficiency occurred in young adulthood. ns œ≠ not significant. female. This could be attributed to the nature of our study population, referral bias, and the fact that in a minority of cases, males with X-linked defects have fatal clinical manifestations in childhood. Given the spectrum of possible inherent defects, the clinical phenotype of immune deficiency diseases is diverse, reflecting the clinical expression of infections from a variety of pathogens, the frequency and severity of infection, or the expression of autoimmune processes. During the initial evaluation of patients with suspected primary immune deficiencies, the causative microbial agent should trigger a pattern of clinical investigations that follows a particular arm of the immune system. For instance, individuals with defects in humoral immunity are more prone to bacterial Percentage of Natural Killer cells Absolute Natual Killer cells M ann Whitne y te st P v alues: 0.0006 M ann Whitne y te st P v alues: 0.0183 600 80 Cells / ul % of Lymphocyte s 100 60 40 400 200 20 0 0 Natural killer lymphocytopenia M edian n= 4.000 6 Non Immunedeficient 11.00 47 Natural killer lymphocytopenia M e dian n= 56.00 5 Non Immunedeficient 180.0 47 Figure 4 Comparison of relative and absolute numbers of natural killer cells in patients with natural killer lymphocytopenia and in nonimmune-deficient patients. There was a highly significant difference between the groups in the percentage of circulating natural killer cells and in the absolute values. Srinivasa et al Adult Primary Immune Deficiency infections, whereas severe viral or fungal infections are frequently noted with T-cell defects. Most adult primary immune deficiencies observed in daily clinical practice reflect humoral immune defects, which are not generally fatal in the short term. The functioning T-cell arm of the immune system may compensate and offer partial protection in these individuals. On the other hand, T-cell defects and combined immune deficiencies are rarely seen in adults because they have a more severe, often fatal, childhood clinical phenotype. This was reflected in our findings, in which 72% of the 167 adults with defined primary immune deficiency had a humoral defect, 17% were diagnosed with a T-cell‚Äìmediated immune defect, and only 3% had combined deficiencies in both arms of the immune system. Indeed, all 5 patients who died during the study period were male with severe cellmediated immune defects. A majority of our patients were diagnosed with common variable immune deficiency (n œ≠ 72, 43%). IgA deficiency is generally seen more frequently than common variable immune deficiency and was underrepresented in our cohort (n œ≠ 25, 15%). Up to 85% to 90% of IgA-deficient patients are generally asymptomatic.16 Individuals with low serum IgA are thus not likely to be referred to our center for evaluation. It remains to be seen whether some of the noncommon variable immune-deficient patients with hypogammaglobulinemia will evolve clinically to fit the ESID/PAGID criteria for common variable immune deficiency. A total of 116 referred patients were classified as nonimmune deficient. We wanted to determine whether the characteristics in the clinical history of these immune-competent patients differed from those with primary immune deficiency. Because common variable immune deficiency was the most frequently encountered primary immune deficiency, we compared the profile of infections by site and by organ system of the nonimmune deficient with the common variable immune-deficient patients. A significantly larger number of patients referred with chronic sinusitis and recurrent pneumonia had common variable immune deficiency (P œΩ.01), underscoring the importance of these clinical syndromes in leading one to suspect the presence of immune deficiency disorders. Of note, only 35% of the 72 common variable immune-deficient patients had a quantitative immunoglobulin assessment before referral. There was an overall high number of individuals referred to us with a presumptive diagnosis of immune deficiency for those referred patients. It was thus surprising that quantitative immunoglobulin testing was not undertaken by the referring physicians despite the wide accessibility and affordability of the test. Having ruled out secondary causes of immune deficiency, and despite considerable efforts to determine the exact cause of recurrent infections in 14 patients (4% of the study population), we were unable to provide specific ESID/ PAGID-based diagnoses, and their syndromes remain as undefined primary immune deficiencies. 785 Even with increasing new knowledge and the ongoing identification of underlying molecular defects in the immune system, many individuals with primary immune deficiencies still experience delayed diagnoses and poor treatment.17 The J Project of the European Society for Immunodeficiency was set up in March 2004 in Eastern and Central Europe to ensure adequate transfer of knowledge among health professionals, scientific societies, and patient associations. To date, the project has had reasonable success in increasing awareness of the diseases among medical practitioners, community leaders, and the general population in Eastern Europe.17 CONCLUSIONS Our center is one of the few dedicated to the diagnosis and management of adults with primary immune deficiency. This study of the center‚Äôs initial cohort of patients, using state of the art immune testing, describes a wide spectrum of adult primary immune deficiencies. We recognized that not all patients could be defined on the basis of ESID/PAGID criteria. As such, we have identified and describe 2 novel groups of patients with humoral abnormalities. Our findings highlight the need for a better understanding of the nature of adult primary immune deficiencies. References 1. Notarangelo LD. 15. Primary immunodeficiencies. J Allergy Clin Immunol. 201;125(2 Suppl 2): S182-S194. 2. Notarangelo L, Casanova JL, Fischer A, et al. Primary immunodeficiency diseases: an update. J Allergy Clin Immunol. 2004;114:677687. 3. Boyle JM, Buckley RH. Population prevalence of diagnosed primary immunodeficiency diseases in the United States. J Clin Immunol. 2007;27:497-502. 4. Hammarstrom L, Vorechovsky I, Webster D. Selective IgA deficiency (SIgAD) and common variable immunodeficiency (CVID). Clin Exp Immunol. 2000;120;225-231. 5. Yarmohammadi H. Estrella L, Doucette J, Cunningham-Rundles C. Recognizing primary immune deficiency in clinical practice. Clin Vaccine Immunol. 2006;13:329-332. 6. Woroniecka M, Ballow M. Office evaluation of children with recurrent infection. Pediatr Clin North Am. 2000;47:1211-1224. 7. Buckley R. Primary cellular immunodeficiencies. J Allergy Clin Immunol. 2002;109:747-757. 8. Mansouri D, Adimi P, Mirsaedi M, et al. Primary immune deficiencies presenting in adults: seven years of experience from Iran. J Clin Immunol. 2005;25:385-391. 9. Conley ME, Notarangelo LD, Etzioni A. Diagnostic criteria for primary immunodeficiencies. Representing PAGID (Pan-American Group for Immunodeficiency) and ESID (European Society for Immunodeficiencies). Clin Immunol. 1999:93;190-197. 10. Walker UA, Warnatz K. Idiopathic CD4 lymphocytopenia. Curr Opin Rheumatol. 2006;18:389-395. 11. Rezaei N, Aghamohammadi A, Moin M, et al. Frequency and clinical manifestations of patients with primary immunodeficiency disorders in Iran: update from the Iranian Primary Immunodeficiency Registry. J Clin Immunol. 2006;26:519-532. 12. Grumach AS, Duarte AJ, Bellinati-Pires R, et al. Brazilian report on primary immunodeficiencies in children: 166 cases studied over a follow-up time of 15 years. J Clin Immunol. 1997;17:340-345. 786 13. Hayakawa H, Iwata T, Yata J, Kobayashi N. Primary immunodeficiency syndrome in Japan. I. Overview of a nationwide survey on primary immunodeficiency syndrome. J Clin Immunol. 1981;1: 31-39. 14. Stray-Pedersen A, Abrahamsen TG, Fr√∏land SS. Primary immunodeficiency diseases in Norway. J Clin Immunol. 2000;20:477-485. 15. Reda SM, Afifi HM, Amine MM. Primary immunodeficiency diseases in Egyptian children: a single-center study. J Clin Immunol. 2009;29: 343-351. 16. Yel L. Selective IgA deficiency. J Clin Immunol. 2010;30:10-16. The American Journal of Medicine, Vol 125, No 8, August 2012 17. Mar√≥di L, Casanova J. Primary immunodeficiency diseases: the J Project. Lancet. 2009;373:2179-2181. 18. Notarangelo LD, Fischer A, Geha RS, et al. Primary immunodeficiencies: 2009 update. J Allergy Clin Immunol. 2009;124:11611178. 19. Davidson A, Diamond B. Autoimmune disease. N Engl J Med. 2001; 345:340-350. 20. Hochberg MC. Updating the American College of Rheumatology revised criteria for the classification of systemic lupus erythematosus. Arthritis Rheum. 1997:40:1725. ˇ˛European Journal of Internal Medicine 11 (2000) 161 164 Original article Adult b-thalassemic patients with chronic hepatitis C: long-term efficacy of a-IFN treatment Chryssoula Labropoulou-Karatzaa ,*, Consantin Goritsasa, Helen Fragopanagoua, abc Panagiota Matsouka , Iris Spiliopoulou , Calypso Barbatis c Received 5 October 1999; received in revised form 27 January 2000; accepted 28 March 2000 a b Department of Internal Medicine, Patra University Hospital, Patra, Greece Department of Microbiology, Patra University Hospital, Patra, Greece Histopathology Department, Red Cross Hospital, Athens, Greece www.elsevier.com / locate / ejim Abstract Background: The purpose of this study was to assess the effectiveness of a-IFN in adult b-thalassemic patients with chronic hepatitis C. After a long-term follow-up, we describe the special pattern of biochemical and virological response of thalassemics. Methods: Thirty-two anti-HCV-positive adult thalassemic patients (19 female and 13 male, mean age 23.465.5 years) with biopsy-proven chronic hepatitis were treated with IFN a2b at a dose of 3 MU thrice weekly for 6 12 months. The patients were followed up until 45 62 months after the end of treatment. Results: A sustained response was obtained in eight patients (25%). Only two of the sustained responders (25%) normalized ALT during the first 3 months of treatment. Both early and late biochemical responders cleared HCV-RNA after 6 months of treatment. Eight patients (25%) responded with ALT normalization within 2 months of treatment but relapsed soon after stopping IFN. Sixteen patients (50%) did not respond to IFN. Conclusion: The response rate in multitransfused thalassemic patients with chronic hepatitis C treated with IFN is similar to that in non-thalassemics. The special feature of thalassemics is that early biochemical response does not predict a sustained response; on the contrary, patients who normalize ALT after 6 months of IFN treatment usually do not relapse. Ÿˆ 2000 Elsevier Science B.V. All rights reserved. Keywords: b-Thalassemia major; Hepatitis C; a-Interferon 1. Introduction Hepatitis C virus (HCV) infection is a common cause of chronic liver disease in multitransfused patients. More than 60% of b-thalassemic patients throughout the world are infected with HCV [1]. Alpha-interferon (a-IFN), in combination with ribavirin, is currently the first line treatment for patients with HCV-related chronic hepatitis. In thalassemic patients, ribavirin is relatively contraindi- cated and a-IFN is the main available therapeutic option [2]. Iron overload is the second cause of chronic liver disease in people with b-thalassemia major. Hepatic iron *Corresponding author, 46 Lykochorou Str., Patra 26000, Greece. Tel.: 130-61-224-737; fax: 130-61-993-982. overload in the liver acts as a cofactor in determining the severity of liver damage and the response to IFN therapy [3 5]. The aim of this study was to assess the long-term biochemical and virological efficacy of a-IFN in adult, multitransfused thalassemics with chronic hepatitis C. We evaluated the rate of prolonged remission and the special pattern of response to IFN. 2. Patients and methods 2.1. Patients We retrospectively studied all of the patients followed at 0953-6205/00/$ see front matter Ÿˆ 2000 Elsevier Science B.V. All rights reserved. PII: S0953-6205(00)00081-9 162 C. Labropoulou-Karatza et al. / European Journal of Internal Medicine 11 (2000) 161 164 our thalassemia center who satisfied the following con- ditions. 1. Patients received regular blood transfusions, according to a standardized protocol, to maintain a hemoglobin level of 10 13 g/dl. They also had regular therapy with desferroxamine. 2. Patient serum ALT was at least 1.5 times the normal value, and they were seropositive for antibodies to HCV for more than 6 months. 3. Patient percutaneous liver biopsy showed chronic hepatitis with or without cirrhosis. 4. Patients received treatment with IFN a2b at a dose of 3 MU three times weekly for a period longer than 6 months (nine patients for 6 months and 23 patients for 12 months). Seven of the 16 non-responders and seven of the eight responders who relapsed were offered a second course of 3 MU IFN a2b thrice weekly for 12 months. 5. The follow-up period from the end of IFN treatment was at least 12 months. The follow-up period was longer than 48 months for all patients who were considered complete and sustained responders. 2.2. Methods 2.2.1. Virologic and histologic evaluation Anti-HCV was tested by third-generation ELISA (Axsym Abbott) and confirmed by third-generation linked immunoassay (Sorin). Viral parameters were studied by means of the Amplicor HCV RNA qualitative test (Roche Diagnostic). HCV RNA was determined before treatment, at the end of treatment, and at the end of the follow-up period. Percutaneous liver biopsies were performed with Men- ghini needles. All of the specimens were examined by the same pathologist and evaluated according to the staging and grading system described by Ishak et al. [6]. Liver siderosis was graded on Perl s stained sections according to Rowe et al. [7]. 2.2.2. Statistical analysis Continuous variables were expressed as the mean and standard deviation (S.D.). The differences between parametric data were analyzed with Student s t-test and chi-square analysis was applied to categorical variables. Probability values below 0.05 were considered significant. A one-way ANOVA for interval data and the Kruskal-Wallis k-sample test for categorical data were used for comparison of clinical features between groups. A multivariate analysis logistic regression was employed to model the probability of response, adjusting for important prognostic variables. 3. Results 3.1. Features of patients studied Thirty-two adult thalassemics (19 females and 13 males) were included in the study. The mean age of the patients was 23.465.5 years (range 17 35 years). Cirrhosis was present in four patients (stage 6) and incomplete cirrhosis in seven patients (stage 5). The histologic degree of siderosis was mild (grade 1 2) in eight (25%), moderate (grade 3) in 14 (44%), and severe (grade 4) in 10 (31%) subjects. Twenty patients tested by PCR before starting therapy were all found to be HCV-PCR positive. The genotypes of the patients were distributed as follows: 1a: n54, 1b: n53, 2: n53, 3: n55, 4: n58. Organ dysfunction secondary to hemochromatosis was present in the following percentages: diabetes mellitus 56%, heart failure 34%, hypogonadism 47%, hypo- parathyroidism 65%, hypothyroidism 47%. 3.2. Pattern of response to IFN (Table 1) In eight subjects (25%) ALT normalized during treat- ment or within 3 months after stopping IFN, and they remained with normal ALT up until the end of the post- treatment follow-up period (mean 52.565 months, range 45 62 months). ALT normalization among the eight complete and sustained responders (CSR) was achieved within 3 months in two subjects (25%), at 6 months in one subject (12%), and at the end of 12 months of therapy in another five subjects (63%). Thus, only 25% of the CSR responded promptly to IFN. All sustained responders cleared HCV RNA after at least 6 months of treatment, and four of them after 12 months, and they remained HCV- RNA-negative during the follow-up period. The follow-up period for all CSR was longer than 48 months. At 3 months of treatment, HCV-RNA was available for six out of the eight sustained responders and it was positive for all of them. Eight subjects (25%) normalized ALT promptly after starting treatment (within 1 month) but contrary to what is usually observed in non-thalassemic patients, the bio- chemical response was not sustained after stopping IFN. According to the above, only two of 10 patients (20%) with prompt biochemical response to IFN sustained the biochemical normalization and virological negativity after the end of treatment. All eight patients with temporary response to IFN also became HCV-PCR-negative under Table 1 Pattern of response to IFN of adult thalassemic patients Early (,2 months) response no relapse Early (,2 months) response relapse Late ($6 months) response no relapse Virological response only No response n52 (6.25%) n58 (25%) n56 (18.75%) n54 (12.5%) n512 (37.5%) C. Labropoulou-Karatza et al. / European Journal of Internal Medicine 11 (2000) 161 164 163 therapy, but HCV-RNA reappeared in serum along with biochemical relapse. Sixteen subjects (50%) remained with abnormal ALT throughout the entire treatment and follow-up period. Four of these (12.5% of the non-responders) were HCV-RNA- negative at the end of follow-up. IFN treatment was well tolerated. No patient discon- tinued IFN because of side effects. The dose of IFN was reduced in two patients. The reason for reduction was musculoskeletal pain and increased frequency of blood transfusions (one patient each). Relapsers and non-respon- ders who had a second course of IFN were treated with the same daily dose because they could not tolerate higher doses. 3.3. Predictors of complete and sustained response (Table 2) Staging score was significantly lower in CSR than in non-responders (P50.002). Multivariate analysis including also age, ferritin value, grade, hepatic iron score, and organ dysfunction confirmed stage as a statistically significant variable associated with CSR. Ferritin values and Rowe score did not differ from those of non-responders. Patients with sustained response tended to have fewer organs affected by hemochromatosis, but the difference from non-responders was just short of statistical signifi- cance (P50.092). Only diabetes mellitus was significantly more frequent in non-responders than in CSR (66% vs. 25% P50.041, OR: 6, 95% CI: 1.07 33.5). CSR had significantly higher pretreatment values of SGPT than non-responders (2126119 vs. 129668, P5 0.021). Also, fewer patients with CSR were splenectom- ized (37.5% vs. 58%), although this difference also was not found to be statistically significant (P50.069). Total dose of IFN, as well as duration of treatment, was not different between CSR and non-responders (IFN dose Table 2 Characteristics of b-thalassemic patients with chronic hepatitis C treated with interferon a2ba 360694 MU vs. 376677 MU; duration of treatment 1062.7 months vs. 9.962.6 months). The genotypes of non-responders were: 1a (n52), 1b (n52), 2 (n51), 3 (n52), 4 (n54). Genotypes were not available for the four patients with virological response without biochemical response. Patients who responded initially and subsequently relapsed had the following genotypes: 1a (n51), 2 (n51), 3 (n53), 4 (n53). Complete and sustained responders had the genotypes 1a (n51), 1b (n51), 2 (n51) and 4 (n51). The role of viral genotypes could not be evaluated because of the small number of patients for whom genotype was available. 4. Discussion Patients with b-thalassemia and chronic hepatitis C form a cohort with special features that modify the natural history of the disease and the response to treatment. They are usually infected early in life by transfusions. Secondary hemochromatosis affects many organs whose dysfunction probably exerts an effect on the liver. Moreover, excessive liver iron is an important factor in determining the severity of liver damage [5]. Hepatic iron overload has also been shown to correlate with a poor response to IFN [3,4]. Our trial is in agreement with previous studies in b- thalassemic patients with chronic HCV infection, which reported a high rate of end-of-treatment response (50%), not different from that of non-thalassemics [8 10]. This study also shows that the relapse rate is not different from that of non-thalassemics (50% of the end-of-treatment responders). The unexpected finding in our patients was that most (80%) of those who normalized ALT within 3 months of starting IFN treatment relapsed as soon as IFN was stopped; most of them (7 / 8) also had the same pattern of response in subsequent courses of IFN. Unfortunately, HCV-PCR at 3 months after starting interferon was avail- able for only four of the eight patients in that group, and it was negative for three of them and positive for one. In contrast, no patient whose biochemical and/or viro- logical response was delayed for 6 or more months relapsed. HCV-PCR at 3 months of treatment was positive for all four patients for whom it was available, contrary to what is usual for complete and sustained responders who show an early virological response. All of these six patients had a follow-up of at least 4 years, and they are still in complete biochemical and virological response. There is no clear explanation for this pattern of respond- ing to IFN treatment, which is quite the opposite to what takes place in non-thalassemics [11]. We believe that it is not the small number of patients that creates a false impression, because none of the patients with delayed response relapsed during the follow-up. The amount of IFN given to the patients was absolutely the same among patients who relapsed and those with sustained response. Age (years) Gender (F/M, F%) Splenectomy (n, %) SGPT (IU/l) Ferritin (ng/ml) Rowe score (0 4) Stage (0 6) Grade (0 18) Diabetes (n, %) Gonads (n, %) Thyroid (n, %) Parathyroid (n, %) Heart (n, %) CSR n58 23.764.7 6/2 (75%) 2 (25%) 2126119 376562919 3.1360.64 2.1461.57 663.42 2 (25%) 5 (62.5%) 3 (37.5%) 4 (50%) 1 (12.5%) Non-responders P-value n524 23.365.9 NS 13 / 11 (54%) NS 15 (62.5%) P50.069 129668 P50.021 260561686 2.9660.95 3.9661.37 P50.002 7.1362.7 NS 16 (66%) P50.041 20 (83%) NS 12 (50%) 17 (70.8%) 10 (41%) NS NS NS NS NS a Values are expressed as means6S.D. 164 C. Labropoulou-Karatza et al. / European Journal of Internal Medicine 11 (2000) 161 164 We think that physicians taking care of thalassemic patients must bear in mind the special features of response to treatment in such patients, so that premature interruption of treatment will be avoided. In non-thalassemics there is generally agreement that treatment should be stopped after 3 months if there is no response [12,13]. The cause of the difference from non-thalassemics must be sought among the effects of long-term iron overload. Delayed response was also found in another study [14] where it was correlated to the iron load, but in this study there was a very low overall relapse rate and no distinction was made between early and late responders. Results of studies of response to a-IFN therapy suggest that a readily mobilized, chelatable iron pool in hepato- cytes may play an important role in mediating hepatic inflammation and/or in abrogating a response to IFN [15]. It is also shown that iron deposition in sinusoidal cells and portal tracts is significantly less frequent among patients with complete response to IFN than in those with poor or no response [16]. Of particular note is the finding that markers of lipid peroxidation were found predominantly in CD68-positive macrophages within portal tracts, cells that probably are among those that stain positively for iron in liver biopsy [17,18]. On the other hand, it is shown that the presence of non-transferrin-bound iron inhibits lymphocyte prolifer- ation, but this effect is more striking in Th1 CD41 cells and CD8 CTL, whereas the number of Th2 CD4 cells and suppressor-type CD81 cells may be enhanced [19]. That kind of immune response is associated with a failure to clear viral infections. Although a direct effect of iron in enhancing rates of viral replication seems possible, it has not yet been proven [20]. One study on the effect of IFN on HCV turnover in vivo [21] concluded that the pre- dominant antiviral effect of IFN is inhibition of de novo infection of susceptible cells, but IFN also exerts a dose- dependent effect on possible antigen-specific degradation of free virus. The combined effect of iron overload and infection early in life on the dynamics between the virus, the immune response, and IFN probably determines the observed special patterns of response to IFN treatment of adult b-thalassemic patients. In conclusion, follow-up of adult b-thalassemic patients with chronic hepatitis C for 4 years after the end of a-IFN treatment showed that IFN leads to sustained virologic and biochemical response in about 25% of the patients. Al- though the percentages of sustained responders do not differ between thalassemic and non-thalassemic patients, a special feature of thalassemics is that biochemical and virological response is often delayed, being evident only after the sixth month of treatment. ˇ˛Aetiopathology of rheumatoid arthritis Peter C Taylor Abstract Although the aetiology of rheumatoid arthritis remains unknown, there have been considerable advances in our understanding of its aetiopatho- genesis in the last two decades. The role of genetic and environmental interactions in giving rise to immune dysregulation and the key cellular and cytokine mediators, as well as the signalling pathways involved, have been extensively studied. These insights have led to major advances in therapeutic options in the pharmacological management of this condition. Keywords Aetiopathogenesis; angiogenesis; anti-citrullinated peptide antibodies; B cell; co-stimulation; cytokine; dendritic cells; rheumatoid arthritis; T cell Genetic factors Genetic factors became implicated in the aetiopathogenesis of rheumatoid arthritis (RA) because of the slight increase in the frequency of RA in first-degree relatives of RA patients, especially if seropositive for rheumatoid factor. In identical twins, concor- dance rates for disease are around 30% compared with 5% in non-identical twins. Such observations suggest that RA is a polygenic disease, and that non-inherited factors are also of great importance.1 Over 100 risk loci have been described for predisposition to RA, the largest contribution coming from those encoding particular class II human leucocyte antigens (HLA). This dis- covery arose with the observation that 60e70% of Caucasian patients with RA are HLA-DR4 positive compared with 20e25% of control populations. Furthermore, patients with more severe RA, especially those with extra-articular complications such as vasculitis and Felty s syndrome, are even more likely to be HLA- DR4 positive than patients with disease confined to joints. Class II HLA molecules are expressed on the surface of antigen-presenting cells. They play a key role in the presentation of processed linear peptide antigens to T cells. Antigen is bound to the HLA-binding cleft formed by the a and b chains of the HLA class II molecule. This tri-molecular HLAeantigen complex binds, in turn, to the variable portion of the T-cell receptor. Peter C Taylor MA PhD FRCP FRCPE is the Norman Collisson Professor of Musculoskeletal Sciences, University of Oxford and a Fellow of St. Peter s College Oxford, UK. His clinical expertise is in rheumatoid arthritis, ankylosing spondylitis and primary inflammatory arthritis. He is Head of the Clinical Trials Group at the Kennedy Institute of Rheumatology, Nuffield Dept of Orthopaedics, Rheumatology and Musculoskeletal Sciences, and Director of the Oxford University NIHR Musculoskeletal Clinical Trials Unit. Competing interests: Professor Taylor has received research grant support from UCB. He has served as a consultant for Takeda, Astra-Zeneca, UCB, Merck, Pfizer, Lilly and Celgene and a speaker for BMS, Janssen, Abbvie and Merck. The nucleotide sequence of HLA DR b1 exons coding the five amino acid residues 70e74 predicts susceptibility to RA and is associated with RA in 83% of Caucasian patients in the United Kingdom.2e4 This sequence is located in the a-helix forming the wall of the peptide-binding groove, and is likely to be involved in antigen binding and subsequent interaction with T-cell receptors. Possible molecular mechanisms accounting for susceptibility to RA include permissive binding of specific peptides, such as those on autoantigens or on environmental antigens, initiation of dis- ease by specific binding of superantigens to HLA molecules, or modulation of the T-cell repertoire by selection or tolerance. Autoantibodies Rheumatoid factors (RF) are autoantibodies directed against the Fc portion of IgG. They are found in 75e80% of RA patients at some time during the course of their disease. High-titre IgM RF is relatively specific for the diagnosis of RA in the context of a chronic polyarthritis and was, for decades, the sole serologic cri- terion widely used in the diagnosis of RA. It has little predictive value in the general population although the presence of RF, particularly at high levels, may antedate the clinical development of RA.5 There is now compelling evidence that in a particular genetic context, smoking is a potential trigger for rheumatoid arthritis. Furthermore, when these two factors occur together, they are associated with an autoantibody response directed against highly citrullinated peptides that may considerably ante- date the onset of clinical features of RA (Figure 1). Citrulline is derived from the amino acid arginine after peptide translation under the influence of the enzyme peptidyl arginine deiminase (PADI4). Genetic variants of this enzyme are associated with RA. Individuals carrying double alleles for the shared epitope, and who smoke, are at approximately 16-fold increased risk of seropositive rheumatoid arthritis, compared with non-smokers without the shared epitope genes.6,7 Identification of the true citrullinated target RA antigen(s) is of importance for under- standing aetiopathogenesis. Proposed candidates include citrulli- nated fibrinogen, citrullinated vimentin, citrullinated a-enolase and citrullinated type II collagen. Many patients have detectable autoantibody reactivity to several of these.8e13 Testing for RHEUMATOID ARTHRITIS What s new? C Biologic therapies are protein based and do not penetrate in- side cells. They have been developed to target extracellular or cell surface-related targets. Blockade of TNF, IL-6R, CD20 and co-stimulation molecules are all effective treatments for RA available in the clinic C An alternative approach to inhibition of the action of pro- inflammatory cytokines is the development of small molecular therapies that penetrate cells and prevent intracellular signal- ling through cytokine receptors. Many small molecules that target signalling enzymes, including inhibitors of p38 MAP ki- nase, SYK and JAK, have been investigated in trials. Of these, one JAK inhibitor has recently been approved for use in the clinic in North America by the FDA but has not received approval by EMEA for use in Europe as of late 2013 MEDICINE 42:5 227 ! 2014 Elsevier Ltd. All rights reserved. Induction Genes Host factors Environment Chronicity Effector cells Pathology Inflammation Pain Stiffness Swelling Tenderness Deformities Disability Tissue damage APC T cell B cell Cytokines Chemokines Growth factors Macrophage Synoviocyte Osteoclast RHEUMATOID ARTHRITIS The potential cause and progression of rheumatoid arthritis Figure 1 The potential cause and progression of RA. In genetically susceptible individuals, host and environment factors may result in activation of pathogenic T cells and a subsequent cascade of inflammatory events which fail to switch off. An example includes the increased occurrence of auto- antibodies to certain citrullinated peptides in genetically susceptible individuals who smoke. Cytokines are involved in co-ordinating the immune response, and recruitment and activation of effector cells that result not only in inflammation, but also local tissue damage in the RA joint. anti-citrullinated peptide antibodies (ACPA) has become common in the evaluation of patients for RA. Since ACPA is now recognized as a component of the new criteria for diagnosis of RA,14 mea- surement of IgM RF might be considered to be redundant. How- ever, several studies have demonstrated that a small proportion of patients who will be categorized as RA may be positive for RF alone in the early stages of presentation, such that the specificity and sensitivity of a diagnosis if RA is improved by measuring both RF and ACPA. Moreover, high-titre RF appears to be a better predictor of severe disease course, showing more striking corre- lations with extra-articular manifestations such as interstitial lung disease and vasculitis than appears to be the case for ACPA. Cellular pathology RA is characterized by chronic inflammation of synovial joints with synovial proliferation and infiltration by blood-derived cells, in particular, memory T cells, macrophages and plasma cells, all of which show signs of activation.15 Angiogenesis is a feature from the earliest stages of disease development. The synovial tissue becomes markedly hyperplastic and locally invasive at the inter- face of cartilage and bone, with progressive destruction of these tissues in the majority of cases. This invasive tissue is termed pannus and comprises mainly lining cells with the appearance of proliferating mesenchymal cells, with very little sublining lym- phocytic infiltration. The accompanying destruction of bone and cartilage is mediated by degradative enzymes, notably matrix metalloproteinases. Although RA has its principal manifestation in joints there is also evidence of systemic involvement, including up-regulation of acute-phase proteins and a variety of extra- articular features. These occur predominantly in patients who are RF positive and carry the HLA-DR4 gene. Presentation of self-antigen to T cells is thought to be central to the pathogenesis of RA. Various cell types are capable of antigen-presenting cell function. Dendritic cells in particular have attracted much interest in recent years, because of their potent ability to present antigen and, in particular, their unique capacity to activate na 1ve T cells.16 T cells are abundant in active RA, comprising 20e50% of cells in the inflamed synovium. Activated T cells contribute to the regulation of osteoclast activation and MEDICINE 42:5 228 ! 2014 Elsevier Ltd. All rights reserved. STEP 1: TCR recognizes MHC-peptide complex (signal 1) STEP 2: T cell activation follows engagement of an activating co-stimulatory pathway APC CD80/86 MHC APC Naive T cell MHC Signal 1 TCR CD28 Signal 2 TCR Outcome Figure 2 The T-cell receptor (TCR) binds to and recognizes components of both the MHC and peptide to elicit signal one. There are many different variants of both the T-cell receptor and MHC molecules. However, this is insufficient to activate the T cell. Activation follows engagement of a co- stimulatory pathway. CD80/86:CD28 is the best characterized co- stimulatory pathway. CD28 is constitutively expressed on T cells and binds to CD80/86. CD80/86:CD28 facilitates T-cell activation, proliferation, sur- vival and cytokine production. Activated T cell RHEUMATOID ARTHRITIS Pro-inflammatory Anti-inflammatory Figure 3 The concept of a cytokine dysequilibrium. Many cytokines are detectable in rheumatoid synovial tissues, including those with predominantly anti-inflammatory properties, but the net effect is a dominance of pro-inflammatory activity. ENA, epithelial neutrophil-activating peptide; FGF, fibroblast growth factor; GM-CSF, granulocyte macrophage colony stimulating factor; IL, interleukin; IL-1ra, interleukin-1 receptor antagonists; LIF, leukaemia inhibitory factor; M-CSF, macrophage colony stimulating factor; MCP, monocyte chemotactic protein; MIP, macrophage inflammatory protein; OSM, oncostatin M; PDGF, platelet-derived growth factor; RANTES, regulated on activation normal T expressed and secreted; sTNFR, soluble TNF receptors; TGF b, transforming growth factor b; TNF, tumour necrosis factor; VEGF, vascular endothelial growth factor. thus joint destruction.17 Successful T-cell activation requires multiple signals (Figure 2). One signal is provided by presenta- tion of an antigen, bound to cell surface MHC molecules on antigen-presenting cells, to a specific T-cell receptor. In the absence of further co-stimulatory signals, T cells become unre- sponsive and may be eliminated through apoptosis. B cells also play a major role in the pathogenesis of RA and this is the basis for the use of rituximab (a B cell-depleting agent) in the treatment of this disease.18,19 B cells are important for the generation of immunoglobulins, including rheumatoid factors and other autoantibodies. They are also highly efficient antigen- presenting cells that may contribute significantly to T-cell responses. The role of cytokines Cytokines are small, short-lived proteins and important media- tors of local intercellular communication. They play a key role in integrating responses to a variety of stimuli in inflammatory processes. Some cytokines, such as interleukin-1 (IL-1), IL-6 and tumour necrosis factor-a (TNF-a), are pro-inflammatory; others, such as interleukin-10 (IL-10) and transforming growth factor b (TGF-b), exert predominantly anti-inflammatory effects. Cytokines derived from macrophages and fibroblasts are abundant in the rheumatoid synovium. These include IL-1, TNF-a, granulocyte macrophage colony stimulating factor (GM-CSF), IL-6 and numerous chemoattractant cytokines known as chemokines.1 Many of these factors are of importance in regulating inflammatory cell migration and activation. By contrast, factors produced by T cells, for example, IFNg, interleukin (IL)-2 and IL-4, are surprisingly sparsely expressed. However, a number of cyto- kines, including IL-7, IL-12, IL-15 and IL-18, cause co-stimulation of T helper (Th) cells. There is a predominance of Th1 cell activity as defined by IFNg production, and low Th2 cell activity as defined by IL-4 production. An extensive range of pro-inflammatory cytokines can be detected in RA synovial samples, regardless of differences in donor disease duration, severity or even conventional (non-bio- logic) drug therapy. In support of the concept of a cytokine dysequilibrium within the chronic inflammatory situation in rheumatoid synovium is the observation that multiple anti- inflammatory mediators are also upregulated, but at a level insufficient to suppress synovitis. Examples include the abun- dant expression of IL-10, IL-13 and TGF-b, both in latent and active form. Naturally occurring cytokine inhibitors, such as interleukin-1 receptor antagonists (IL-1ra) and soluble TNF re- ceptors, the specific inhibitors of IL-1 and TNF-a respectively, are also upregulated in the rheumatoid joint (Figure 3). TNF-a is a pleiotropic cytokine with biological properties that include enhanced synovial proliferation, production of prosta- glandins and metalloproteinases as well as regulation of other pro-inflammatory cytokines. The predicted clinical success of anti-TNF therapy in RA was based on the demonstration of RA synovial tissue expression of TNF-a and its receptors, in vitro experiments employing dissociated synovial cell cultures and preclinical in vivo studies.20 Biologic therapies targeting TNF and IL-6, particularly when used in conjunction with methotrexate therapy, demonstrated clinical efficacy and confirmed the val- idity of these cytokines as molecular targets for therapy. Cytokines are able to effect a biological response in cells if these bear surface receptors that recognize the cytokine ligand, and if receptor engagement is followed by activation of an intracellular signalling pathway. Janus kinases (JAKs) are cyto- plasmic protein tyrosine kinases used by multiple cytokines involved in inflammatory responses. Key cytokines in RA that utilize JAK include IFNa and IFNb, IL-6, IL-7, IL-10, IL-12, IL-15, IL-21 and IL-23. Each cytokine receptor requires two associated JAKs (out of four JAK family members) in order to signal.21 The JAK1/3 inhibitor, tofacitinib, is a novel, oral JAK inhibitor MEDICINE 42:5 229 ! 2014 Elsevier Ltd. All rights reserved. GM-CSF M-CSF IL-8, MCP-1, MIP-1 RANTES, ENA78 sTNF-R IL-1ra IL-1 TNF± IL-6 FGF LIF PDGF OSM VEGF TGF IL-10, IL-13 ≤ IL-11, IL-16 IL-12, IL-15 IL-17, IL-18 proven in a phase III trial programme to ameliorate symptoms and signs of RA, and to limit structural damage. It has been approved for use in North America but was not approved in Europe in 2013. Spleen tyrosine kinase (SYK) is a cytoplasmic protein tyrosine kinase that binds to receptors containing immune-receptor tyrosine-based activation motifs such as the B cell and Fc receptors, through which many signalling pathways involved in inflammation become activated. Fostamatinib is a drug with selectivity for SYK, which has shown efficacy in clin- ical trials but at a magnitude that has been considered uncom- petitive with biologic therapies. For this reason it is not being developed as a therapeutic for RA. A RHEUMATOID ARTHRITIS Practice points C Understanding the important role of cytokines in the immuno- pathogenesis of RA has led to two potential approaches to cytokine modulation of rheumatoid synovitis: inhibition of dominant pro-inflammatory cytokines, such as TNF-a (infliximab, adalimumab) and IL-6 (anti-IL-6 receptor; tocilizumab); or augmentation of the inadequate anti-inflammatory activity of certain cytokines or naturally occurring cytokine inhibitors, for example, by administration of soluble TNF receptors (etanercept) C Understanding the important role of immune cells has led to new biologic therapies targeting B-cell subsets (rituximab) and the T cell antigen-presenting cell interaction (co-stimulatory molecule blockade; abatacept) C Understanding the role of intracellular signalling molecules has led to trials of many small molecular inhibitors in the treatment of RA. Some, such as p38 MAP kinase inhibitors, have been un- successful because of adverse events and low efficacy and others, such as fostamatinib will not be further developed for RA because of tolerability problems and relatively modest efficacy. The first oral JAK inhibitor, tofacitinib, is now FDA approved for use in the clinic and others are in development MEDICINE 42:5 230 ! 2014 Elsevier Ltd. All rights reserved. EBIOM-00219; No of Pages 9 EBioMedicine xxx (2015) xxx‚Äìxxx Contents lists available at ScienceDirect EBioMedicine journal homepage: www.ebiomedicine.com A Five-miRNA Panel IdentiÔ¨Åed From a Multicentric Case‚Äìcontrol Study Serves as a Novel Diagnostic Tool for Ethnically Diverse Non-small-cell Lung Cancer Patients Cheng Wang a,b,1, Meng Ding a,b,1, Mingde Xia c,1, Sidi Chen b, Anh Van Le d, Rafael Soto-Gil d, Yi Shen e, Nan Wang b, Junjun Wang a, Wanjian Gu f, Xiangdong Wang g, Yanni Zhang h, Ke Zen b,‚Åé, Xi Chen b,‚Åé, Chunni Zhang a,b,‚Åé‚Åé, Chen-Yu Zhang b,‚Åé a Department of Clinical Laboratory, Jinling Hospital, State Key Laboratory of Pharmaceutical Biotechnology, Nanjing Advanced Institute for Life Sciences, Nanjing University School of Life Sciences, Nanjing University, Nanjing, China b State Key Laboratory of Pharmaceutical Biotechnology, Nanjing Advanced Institute for Life Sciences, Nanjing University School of Life Sciences, Jiangsu Engineering Research Center for MicroRNA Biology and Biotechnology, Nanjing University, Nanjing, China c Johnson & Johnson Innovation Center Asia PaciÔ¨Åc, Shanghai, China d Ortho-Clinical Diagnostics, NJ, USA e Department of Thoracic Surgery, Jinling Hospital, Nanjing University School of Medicine, Nanjing University, Nanjing, China f Department of Clinical Laboratory, Jiangsu Province Hospital of TCM, Nanjing, China g Department of Clinical Laboratory, Nanjing Chest Hospital, Nanjing, China h Department of Medical Oncology, Cancer Hospital of Xuzhou, Xuzhou, China a r t i c l e i n f o Article history: Received 15 June 2015 Received in revised form 24 July 2015 Accepted 27 July 2015 Available online xxxx Keywords: Non-small-cell lung cancer Multiethnic Multicentric Serum microRNAs Biomarker Diagnosis a b s t r a c t Circulating microRNAs (miRNAs) are promising biomarkers for cancer detection. However, multiethnic and multicentric studies of non-small-cell lung cancer (NSCLC) are lacking. We recruited 221 NSCLC patients, 161 controls and 56 benign nodules from both China and America. Initial miRNA screening was performed using the TaqMan Low Density Array followed by conÔ¨Årming individually by RT-qPCR in Chinese cohorts. Finally, we performed a blind trial from an American cohort to validate our Ô¨Åndings. RT-qPCR conÔ¨Årmed that miR-483-5p, miR-193a-3p, miR-25, miR-214 and miR-7 were signiÔ¨Åcantly elevated in patients compared to controls. The areas under the curve (AUCs) of the ROC curve of this Ô¨Åve-serum miRNA panel were 0.976 (95% CI, 0.939‚Äì1.0; P b 0.0001) and 0.823 (95% CI, 0.75‚Äì0.896; P b 0.0001) for the two conÔ¨Årmation sets, respectively. In the blind trial, the panel correctly classiÔ¨Åed 95% NSCLC cases and 84% controls from the American cohort. Most importantly, the panel was capable of distinguishing NSCLC from benign nodules with an AUC of 0.979 (95% CI, 0.959‚Äì1.0) in the American cohort and allowed correct prediction of 86% and 95% stage I‚ÄìII tumors in the Chinese and American cohorts, respectively. This serum miRNA panel holds the potential for diagnosing ethnically diverse NSCLC patients. ¬© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction Tumors of the lung are one of the highest incidence of cancers and the leading cause of cancer deaths worldwide. According to the 2008 WHO epidemiological statistics, the incidence of lung cancer is more than 130 per 100,000 individuals, which accounts for 13% (1.6 million) of the total cancer cases and 18% (1.4 million) of cancer-related mortality; even worse, the number of new cases is expected to double worldwide in next decades (Jemal et al., 2011). The vast majority (85‚Äì 90%) of lung cancer cases are non-small-cell lung cancer (NSCLC) (D'Addario et al., 2010). Early-stage NSCLCs are symptom-free and difÔ¨Åcult to detect; therefore, the disease is often diagnosed at a locally advanced or metastatic stage (stage III or stage IV), at which point few clinical treatment options remain. Despite numerous developments relating to NSCLC therapies in recent years, the overall survival rate of NSCLC patients has only improved by approximately 15% over the last Abbreviations: miRNA, microRNA; NSCLC, non-small-cell lung cancer; RT-qPCR, quantitative reverse transcription polymerase chain reaction; TLDA, TaqMan Low-Density Array; Cq, quantiÔ¨Åcation cycle; AUC, the area under the ROC curve; TNM, tumor‚Äìnode‚Äìmetastasis. ‚Åé Corresponding authors at: State Key Laboratory of Pharmaceutical Biotechnology, School of Life Sciences, Nanjing University, 22 Hankou Rd., Nanjing 210093, China. ‚Åé‚Åé Correspondence to: C. Zhang, Department of Clinical Laboratory, Jinling Hospital, 305 East Zhongshan Rd., Nanjing 210002, China. E-mail addresses: kzen@nju.edu.cn (K. Zen), xichen@nju.edu.cn (X. Chen), zchunni27@hotmail.com (C. Zhang), cyzhang@nju.edu.cn (C.-Y. Zhang). 1 Cheng Wang, Meng Ding and Mingde Xia contributed equally to this work and all should be considered as Ô¨Årst authors. http://dx.doi.org/10.1016/j.ebiom.2015.07.034 2352-3964/¬© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Please cite this article as: Wang, C., et al., A Five-miRNA Panel IdentiÔ¨Åed From a Multicentric Case‚Äìcontrol Study Serves as a Novel Diagnostic Tool for Ethnically Diverse No..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.07.034 2 C. Wang et al. / EBioMedicine xxx (2015) xxx‚Äìxxx few decades (Spiro & Silvestri, 2005; Siegel et al., 2012). In comparison, the Ô¨Åve-year survival rate is approximately 33% for cases discovered at early stages (stage I or II) (Little et al., 2005). At present, diagnosis largely relies to a large extent on imaging methods, such as chest X-rays, CT scans, or PET‚ÄìCT scans, followed by bronchoscopy and biopsy (Oken et al., 2005; International Early Lung Cancer Action Program Investigators et al., 2006). While no blood marker for NSCLC is available at present, some blood-based protein markers such as carcinoembryonic antigen and squamous cell carcinoma antigen have been applied as candidate diagnostic markers for NSCLC in clinics. Unfortunately, the low sensitivity and speciÔ¨Åcity of these antigens limit the diagnostic accuracy and utility. For these reasons, the development of sensitive and reliable biomarkers remains a major challenge for researchers. MicroRNAs (miRNAs) are a family of small, single-stranded noncoding RNAs that are critical regulators of numerous diseases, including cancer, and their expression patterns have the potential to diagnose various types of cancer (Esquela-Kerscher & Slack, 2006; Kong et al., 2012; Calin & Croce, 2006). Recent studies from our group and others have demonstrated that human body Ô¨Çuid such as serum/plasma contains numerous stable miRNAs, potentially paving the way toward a novel class of cancer biomarkers (Chen et al., 2008; Mitchell et al., 2008; Eichelser et al., 2013). A number of studies have reported that circulating miRNA expression patterns differ between NSCLC patients and healthy controls, and some dysregulated serum miRNAs hold potential as promising biomarkers for the diagnosis and prognosis of NSCLC (Chen et al., 2008; Hu et al., 2010; Bianchi et al., 2011; Chen et al., 2012a; Heegaard et al., 2012; Wang et al., 2013). However, data produced by different groups on this topic are quite variable. Therefore, a systematical analysis of serum miRNA expression proÔ¨Åles from multiethnic and multicentric NSCLC cases is urgently needed. In the present study, we used a high-throughput TaqMan LowDensity Array (TLDA) (Applied Biosystems) scanning combined with an individual quantitative reverse transcription polymerase chain reaction PCR (RT-qPCR) conÔ¨Årmation to establish a panel of signiÔ¨Åcantly upregulated serum miRNAs in NSCLC patients using samples from three independent Chinese cohorts. A single-blind trial was then performed to assess the ability of the panel to diagnose NSCLC in an American cohort and to function as a reliable diagnostic indicator of NSCLC in patients of different ethnicities. 2. Materials and Methods 2.1. Participants and Study Design We totally enrolled 438 participants from both China and America, including 221 NSCLC patients, 161 normal controls and 56 benign nodules. To identify a miRNA signature associated with NSCLC, we Ô¨Årst used a discovery cohort consisting of 62 individuals (31 patients with NSCLC and 31 normal controls) from Jinling Hospital, China, to Ô¨Ånd serum miRNAs as surrogate markers. Subsequently, veriÔ¨Åcation was conducted using individual RT-qPCR assays to reÔ¨Åne the number of serum miRNAs included in the NSCLC signature by a 2-phase experimental procedure from two additionally independent, randomized Chinese cohorts: one consisted of 19 lung cancer cases (from Cancer Hospital of Xuzhou) and 19 normal controls (from Jinling Hospital) and the other consisted of 63 case samples (from Jinling Hospital, Jiangsu Province Hospital of TCM and Nanjing Chest Hospital) and 63 normal controls (from Jinling Hospital). Finally, we evaluated the serum samples from an American cohort, which included 108 NSCLC patients, 48 normal controls and 56 benign nodules (as non-cancer controls). The American cohort was selected from the Mayo Clinic (Mayo Validation Support Services, 3050 Superior Drive, NW, Rochester, Minnesota) in a blinded fashion (the investigators performing the analysis on the blood samples were blinded to the patients' clinical diagnosis and the medical sample providers conducted the data analysis) to validate the predictive power of the identiÔ¨Åed miRNA signature for NSCLC. The overall study design is shown in Fig. 1. The protocol was approved by the ethics committee board at each participating institution, and written informed consent was obtained from all participants. 2.2. Eligibility, Exclusion Criteria and Processing of Serum Histopathological analysis of the tumors was performed according to the WHO criteria. Patients were eligible if they had been diagnosed with a pathological NSCLC that met histological or cytological criteria. The tumor stage at the time of diagnosis was assessed according to the American Joint Committee on Cancer guidelines (http://www. cancerstaging.org/). Eligibility criteria for NSCLC patients both China cohorts and American cohort included: 1) if eighteen years of age old or older; 2) conÔ¨Årmed diagnosis of NSCLC; 3) not previously suffered from lung cancer and other types of cancers; 4) not previously with chemotherapy or radiotherapy; 5) and free of synchronous multiple cancers. For patients who underwent surgery, deÔ¨Ånitive tumor stage was determined based on the operative Ô¨Åndings. For those patients unsuitable for surgical therapy, tumors were assessed using histopathology or imaging technology such as bronchoscopy, dynamic computed tomography (CT), magnetic resonance imaging, and/or endoscopic ultrasonography. Eligibility criteria for benign nodules enrolled from American cohort (Mayo Clinic) included: 1) nodules detected on CT scan; 2) biopsy performed and conÔ¨Årmed negative for cancer or biopsy performed and nodules detected and were conÔ¨Årmed benign based on CT scan trial algorithm; 3) Ô¨Åve years of follow-up conÔ¨Årming no diagnosis of lung cancer; and 4) no previous history of cancer prior to enrollment. Individuals who showed no evidence of disease during the medical checkup were selected as normal controls. Eligibility criteria for normal controls in American cohort included: 1) no nodules detected on CT scan; 2) Ô¨Åve years of following up conÔ¨Årming no diagnosis of lung cancer; and 3) no previous history of cancer prior to enrollment. The recruitment of normal subjects in Chinese cohorts was conducted in the Healthy Physical Examination Center of the Jinling Hospital (Nanjing, China). The Ô¨Ånally selected controls underwent routine laboratory and imaging tests such as complete blood cell counts, sera biochemical proÔ¨Åle, baseline electrocardiogram, tumor markers (CEA and AFP), a type-B ultrasound for the abdomen and pelvis, and a chest X-ray. All the healthy individuals did not have any pulmonary pathology or any constitutional symptoms after consulting the physical examination records. Peripheral venous blood (3‚Äì4 ml) was collected in serum vacutainer tubes (Becton, Dickinson and Company, Franklin Lakes, New Jersey) with clot activator from participants 12 h overnight fasting, and serum was immediately separated by a 10-min centrifugation at 1500 g. Supernatant serum samples were then carefully transferred to a new Eppendorf tube and stored at ‚àí80 ¬∞C until analyze. In addition, we checked the hemolysis status of all the samples and only the qualiÔ¨Åed samples were used in our study. The demographics and clinical features of all NSCLC patients and controls are shown in Table 1. All blood samples from patients were gathered before treatments, including surgery, chemotherapy, radiotherapy, and tissue manipulation (such as biopsy, bronchoscopy, or Ô¨Åne needle aspiration). 2.3. RNA Isolation, TLDA and RT-qPCR Assays The serum expression proÔ¨Åling of miRNA was scanned using TLDA on an ABI PRISM 7900HT (Luo et al., 2013). A hydrolysis probe-based RT-qPCR assay was then used to validate candidate miRNAs following the manufacturer's instructions as described previously (Luo et al., 2013). The concentrations of serum miRNAs were normalized to let-7d/g/i trio and relative levels were evaluated using the comparative quantiÔ¨Åcation cycle (Cq) method (Chen et al., 2013). The details of RNA isolation, TLDA and RT-qPCR assays are provided in the Supplemental Methods. Please cite this article as: Wang, C., et al., A Five-miRNA Panel IdentiÔ¨Åed From a Multicentric Case‚Äìcontrol Study Serves as a Novel Diagnostic Tool for Ethnically Diverse No..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.07.034 C. Wang et al. / EBioMedicine xxx (2015) xxx‚Äìxxx 3 Fig. 1. Overall study design and numbers of patients with NSCLC included in the discovery, training, validation and testing sets. 2.4. Statistical Analysis The serum miRNA concentrations were represented as the mean (SE), and the other variables were expressed as the mean (SD). The nonparametric Mann‚ÄìWhitney U-test was performed to compare miRNA concentrations between the groups. The differences in other variables between the two groups were compared using Student's t-test or twosided œá2 test. A P-value b 0.05 was considered statistically signiÔ¨Åcant. We performed a risk score analysis to evaluate the associations between the concentrations of the serum miRNAs and NSCLC. BrieÔ¨Çy, the risk Please cite this article as: Wang, C., et al., A Five-miRNA Panel IdentiÔ¨Åed From a Multicentric Case‚Äìcontrol Study Serves as a Novel Diagnostic Tool for Ethnically Diverse No..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.07.034 4 C. Wang et al. / EBioMedicine xxx (2015) xxx‚Äìxxx score of each miRNA, denoted as s, was set to 1 if its concentration was higher than the upper 95% reference interval for the corresponding miRNA concentration in controls and to 0 otherwise. A risk score function to predict NSCLC was deÔ¨Åned according to a linear combination of concentration for each miRNA. For example, the RSF for sample i using information from Ô¨Åve miRNAs was: rsfi = ‚àë5j ‚àí 1Wj ¬∑ sij. In the above equation, sij is the risk score for miRNA j on sample i, and Wj is the weight of the risk score of miRNA j. To determine the Ws, Ô¨Åve univariate logistic regression models were Ô¨Åtted using the disease status with each of the risk scores. The regression coefÔ¨Åcient of each risk score was used as the weight to show the contribution of each miRNA to the RSF. We then constructed the ROC curves to evaluate the speciÔ¨Åcity and sensitivity of NSCLC prediction for each miRNA individually and for the combination of miRNAs. All analyses above were performed with the use of SPSS statistical software (version 16.0). Additional details regarding statistical methods are provided in the Supplemental Methods. 2.5. Funding This work was supported by the grants from the National Basic Research Program of China (973 Program) (2014CB542300), the Research Special Fund for Public Welfare Industry of Health of China (201302018), the National Natural Science Foundation of China (81171661, 81472021, 81401257 and 81301511), the Natural Science Foundation of Jiangsu Province (BK2011013, BK20140730), the Medical ScientiÔ¨Åc Research Foundation of Nanjing Military Command (12Z28), and the Research Funds of the Jinling Hospital (2014043). The funding agency has no role in the study design, date analysis, interpretation of results or writing of this manuscript. 3. Results 3.1. Serum miRNA Screening by TLDA We globally analyzed the miRNA expression patterns in two pooled serum samples from 31 NSCLC patients and 31 normal controls in Chinese cohort 1 using TLDA approach. A miRNA was considered upregulated if its Cq value was b25 in the NSCLC sample and its concentration showed at least a 2-fold increase in the NSCLC sample compared to the control sample. Class-comparison analysis of all 754 miRNAs revealed that 38 miRNAs (5%) were upregulated in the NSCLC sample (Supplemental Table 1), 16 of the upregulated miRNAs which have been reported to be conserved in lung tissue and involved in cell proliferation, migration, immune related or known as ‚Äúoncomir‚Äù were subsequently validated in the individual patients using RT-qPCR, while the newly identiÔ¨Åed miRNAs and miRNA-star miRNAs were excluded for further conÔ¨Årmation. 3.2. Increases in Serum miRNA Concentrations ConÔ¨Årmed by RT-qPCR The 16 miRNAs selected from the TLDA results were Ô¨Årst conÔ¨Årmed in a training set consisting of an independent Chinese cohort including 19 cases and 19 controls (the training set). Nine of the 16 miRNAs were found to be markedly dysregulated in sera from NSCLC cases. Of these 9 miRNAs, Ô¨Åve upregulated miRNAs, including miR-483-5p, miR-193a-3p, miR-214, miR-25, and miR-7, were chosen for further analysis because the difference between cases and controls was most signiÔ¨Åcant for this set (exhibiting a fold change N 1.5 and a P b 0.05) (Supplemental Table 2 and Fig. 2). The Ô¨Åve miRNAs were further examined in another Chinese cohort including 63 cases and 63 controls (the validation set). The changes in the levels of these Ô¨Åve miRNAs in the NSCLC group were consistent with those from the training set (fold change N 1.5 and P b 0.05) (Fig. 2), thus conÔ¨Årming stability of the production pattern of these miRNAs. 3.3. Assessment of Risk Score in Chinese Cohorts Next, we conducted ROC curve analyses on each of the individual Ô¨Åve serum miRNAs to assess the diagnostic value for discriminating between NSCLCs and normal subjects in the training set. The AUCs for these miRNAs ranged from 0.609 to 0.900 (Supplemental Table 3). To clarify the diagnostic capability of the combination of the Ô¨Åve miRNAs, we conducted a risk score analysis on the data set and used the results to predict NSCLC case and normal status. When the optimal cutoff value was 2.617, all of the NSCLC samples (100%) had a risk score N 2.617, whereas only 3 of the 19 controls exhibited a risk score N 2.617 (Table 2). The ROC curve for the panel demonstrated a remarkable accuracy of diagnosis, evidenced by an AUC of 0.976 (95% CI, Table 1 Demographic and clinical features of the NSCLC patients and controls of the Chinese cohorts and American cohort. Variable Chinese cohorts American cohort Cohort 1 Normal control Age ‚Äî yra ‚â§59 N59 Unknown Sex ‚Äî no. (%) Male Female Unknown Smoking status ‚Äî no. (%) Ever and current Never Tumor subtype ‚Äî no. (%) Adenocarcinoma Squamous cell carcinoma Other TNM stage ‚Äî no. (%) I II III IV Missing data a Cohort 2 Cohort 3 NSCLC Normal control NSCLC Normal control NSCLC Normal control Benign nodules NSCLC n = 31 n = 31 n = 19 n = 19 n = 63 n = 63 n = 48 n = 56 n = 108 59.8 ¬± 8.5 20 (64.5) 11 (35.5) 0 (0) 59.6 ¬± 11.5 13 (41.9) 18 (58.1) 0 (0) 62.1 ¬± 9.4 10 (52.6) 9 (47.4) 0 (0) 61.8 ¬± 12.7 8 (42.1) 11 (57.9) 0 (0) 59.7 ¬± 5.7 36 (57.1) 27 (42.9) 0 (0) 61.9 ¬± 9.5 25 (39.7) 38 (60.3) 0 (0) 58.5 ¬± 5.1 36 (75.0) 12 (25.0) 0 (0) 63.7 ¬± 6.7 15 (26.8) 40 (71.4) 1 (1.8) 67.2 ¬± 10.2 21 (19.4) 87 (80.6) 0 (0) 25 (80.6) 6 (19.4) 0 (0) 23 (74.2) 8 (25.8) 0 (0) 13 (68.4) 6 (31.6) 0 (0) 11 (57.9) 8 (42.1) 0 (0) 37 (58.7) 26 (41.3) 0 (0) 49 (77.8) 14 (22.2) 0 (0) 20 (41.7) 28 (58.3) 0 (0) 25 (44.6) 30 (53.6) 1 (1.8) 52 (48.1) 56 (51.9) 0 (0) 12 (38.7) 19 (61.3) 16 (51.6) 15 (48.4) 7 (36.8) 12 (63.2) 8 (42.1) 11 (57.9) 18 (28.6) 45 (71.4) 39 (61.9) 24 (38.1) 48 (100) 0 (0) 55 (98.2) 1 (1.8) 103 (95.4) 5 (4.6) ‚Äì ‚Äì 12 (38.7) 9 (29.0) 10 (32.3) ‚Äì ‚Äì ‚Äì 9 (47.3) 4 (21.1) 6 (31.6) ‚Äì ‚Äì ‚Äì 26 (41.3) 26 (41.3) 11 (17.4) ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì 52 (48.1) 27 (25.0) 29 (26.9) ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì 7 (22.6) 4 (12.9) 9 (29.0) 11 (35.5) 0 ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì 1 (5.3) 2 (10.5) 2 (10.5) 12 (63.2) 2 (10.5) ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì 3 (4.7) 8 (12.7) 9 (14.3) 41 (65.1) 2 (3.2) ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì 43 (39.8) 15 (13.9) 29 (26.9) 17 (15.7) 4 (3.7) The data are expressed as the mean (SD). Please cite this article as: Wang, C., et al., A Five-miRNA Panel IdentiÔ¨Åed From a Multicentric Case‚Äìcontrol Study Serves as a Novel Diagnostic Tool for Ethnically Diverse No..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.07.034 C. Wang et al. / EBioMedicine xxx (2015) xxx‚Äìxxx 0.939‚Äì1.0; P b 0.0001), which was much larger than the AUC values for each of the Ô¨Åve individual miRNAs (Supplemental Table 3 and Fig. 3A). Furthermore, at the optimal cutoff value, the diagnostic sensitivity and speciÔ¨Åcity of the Ô¨Åve-miRNA panel for NSCLC diagnosis were 100% and 84%, respectively (Fig. 3A). We further evaluated the diagnostic values of the Ô¨Åve individual miRNAs and the Ô¨Åve-miRNA panel in the validation set. As shown in Supplemental Table 4, the AUCs for each of the Ô¨Åve miRNAs ranged from 0.663 to 0.835. At the same cutoff value (2.617), 55 of the 63 NSCLC samples (87%) had a risk score N 2.617 (Table 2). And the discriminate sensitivity and speciÔ¨Åcity of the Ô¨Åve-miRNA panel for NSCLC were 89% and 68%, respectively, and the AUC was 0.823 (95% CI, 0.750‚Äì0.896; P b 0.0001) (Supplemental Table 4 and Fig. 3B). 3.4. Evaluation of the Predictive Value of the Five-miRNA Panel in an American Cohort by Blinded Trial We tested an additional 212 serum samples from an American cohort in a blind trial using the RT-qPCR assay. We aimed to further conÔ¨Årm the diagnostic accuracy of the Ô¨Åve serum miRNA-based biomarker for different ethnicities. First, we analyzed the difference in serum concentrations of miRNAs between the different ethnicities. We observed from RT-qPCR data that the Cq values of serum miRNAs in samples from the American cohort were generally higher than in the Chinese cohort by approximately N 3 cycles. However, the relative concentrations of four miRNAs, including miR-483-5p, miR-193a-3p, miR-25 and miR-7, were comparable between the Chinese and American cohorts 5 except miR-214, since the average Cq value of the internal control let-7d/g/i in the American cohort with a mean (SD) of 24.59 (0.85) was also approximately 3 cycles higher than that in Chinese cohort [22.03(0.92)]. These results suggest that the relative concentrations of the Ô¨Åve selected miRNAs correlate well between different races when let-7d/g/i is used as an internal reference for the RT-qPCR assay. Subsequently, we analyzed the relative levels of the Ô¨Åve miRNAs from the American cohort and classiÔ¨Åed them on the basis of the risk score model of the Ô¨Åve-miRNA panel, with a cutoff value of 2.617. As a result, 103 of 108 NSCLC samples (95%) and 87 of 104 control samples (84%) were classiÔ¨Åed correctly (Table 2). Of the 104 control samples, 48 were from healthy individuals, and the other 56 were from patients who had benign lung nodules detected by CT scan and did not develop lung cancer after a Ô¨Åve-year period of follow-up. The relative concentrations of the Ô¨Åve serum miRNAs were all signiÔ¨Åcantly higher in NSCLC patients than in normal controls and patients with benign nodules (at least P b 0.001) (Fig. 4). In the blind American cohort, the AUCs for the Ô¨Åve miRNAs ranged from 0.696 to 0.964 (Supplemental Table 5). The Ô¨ÅvemiRNA panel was signiÔ¨Åcantly reliable when applied to the testing set, exhibiting a sensitivity of 95%, a speciÔ¨Åcity of 84% and an AUC of 0.952 (95% CI, 0.922‚Äì0.981) (Fig. 3C). Moreover, the risk score was capable of distinguishing NSCLC cases from benign nodules with an AUC of 0.979 (95% CI, 0.959‚Äì1.000) and sensitivity and speciÔ¨Åcity of 95% (Fig. 3D). Taken together, the data suggest that the Ô¨Åve-miRNA panel is also an accurate diagnostic marker for American patients with NSCLC. Fig. 2. The relative contents of the selected Ô¨Åve miRNAs in the sera from patient with NSCLC in the training set and validation set. The asterisks indicate signiÔ¨Åcant differences from normal controls (P b 0.05). ‚ÅéP b 0.05; ‚Åé‚ÅéP b 0.01; ‚Åé‚Åé‚ÅéP b 0.001. Please cite this article as: Wang, C., et al., A Five-miRNA Panel IdentiÔ¨Åed From a Multicentric Case‚Äìcontrol Study Serves as a Novel Diagnostic Tool for Ethnically Diverse No..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.07.034 6 C. Wang et al. / EBioMedicine xxx (2015) xxx‚Äìxxx Table 2 Risk score analysis of NSCLC cases in Chinese cohorts and American cohort. Risk score Chinese cohorts Controls NSCLC Stage I‚ÄìII NSCLC Training set Controls NSCLC Stage I‚ÄìII NSCLC Validation set Controls NSCLC Stage I‚ÄìII NSCLC American cohort Testing (single-blind) set Controlsa NSCLC Stage I‚ÄìII NSCLC a 0‚Äì2.617 N2.617‚Äì13.820 Prediction accuracy (%) 59 8 2 23 74 12 72 90 87 16 0 0 3 19 3 84 100 100 43 8 2 20 55 9 68 87 82 87 5 3 17 103 55 84 95 95 Including normal controls and non-cancer controls (benign nodules). 3.5. Assessment of Risk Score in Early-stage NSCLC We further evaluated the potential of the Ô¨Åve-miRNA panel to diagnose early-stage cancer, for which surgery is most effective. Our training set and validation set were enriched with advanced cancers (stages III and IV), while the testing set contained a relatively even proportion of early-stage patients (stages I and II) (Table 1). The mean risk score of NSCLC cases at early stages was not markedly different from that at later stages (P N 0.05). In Chinese cohorts, 12 of 14 stage I‚ÄìII tumors had a risk score N 2.617 and an accuracy of 86%. In the testing set, using the cutoff value of 2.617 allowed correct prediction of 55 of 58 (95%) stage I‚ÄìII tumors, demonstrating a positive performance for detecting early stages of NSCLCs (Table 2). The results indicate that the Ô¨ÅvemiRNA panel also functions as a strong predictor of early-stage cancer. 3.6. Logistic Regression Analysis of the Five Selected miRNAs and Their Panel Finally, we conducted a forward stepwise binary logistic regression analysis to determine whether the diagnostic usefulness of the Fig. 3. Sensitivity and speciÔ¨Åcity of the Ô¨Åve-miRNA panel in the training, validation and testing sets. (A), ROC curve for the Ô¨Åve-miRNA panel to discern 19 NSCLC cases from 19 normal controls in the training set. (B), ROC curve for the Ô¨Åve-miRNA panel to discern 63 NSCLC cases from 63 normal controls in the validation set. (C), ROC curve for the Ô¨Åve-miRNA panel to discern NSCLC cases from normal controls in the testing set. (D), ROC curve for the Ô¨Åve-miRNA panel to discern 108 NSCLC cases from 56 benign nodules in the testing set. Please cite this article as: Wang, C., et al., A Five-miRNA Panel IdentiÔ¨Åed From a Multicentric Case‚Äìcontrol Study Serves as a Novel Diagnostic Tool for Ethnically Diverse No..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.07.034 C. Wang et al. / EBioMedicine xxx (2015) xxx‚Äìxxx 7 Fig. 4. The relative contents of the selected Ô¨Åve miRNAs in the sera from patient with NSCLC in the testing set. The asterisks indicate signiÔ¨Åcant differences from normal controls (P b 0.05). ‚ÅéP b 0.05; ‚Åé‚ÅéP b 0.01; ‚Åé‚Åé‚ÅéP b 0.001. Ô¨Åve-miRNA panel is affected by clinical features of NSCLC cases, such as TNM stage, tumor subtype, sex, smoking history or age. Sensitivity analyses revealed that the Ô¨Åve-miRNA panel remained a strong predictor of risk of NSCLC regardless of almost all subgroupings of the patients considered in both Chinese and American cohorts (Supplemental Table 6). To further examine whether the Ô¨Åve altered miRNAs and their panel as well as other clinical characteristics are independent potent diagnostic markers for NSCLC, we next performed a forward stepwise univariate logistic regression using the control group as the reference category. Consequently, the odds ratios (ORs) for all of the Ô¨Åve identiÔ¨Åed miRNAs (including miR-483-5p, miR-193a-3p, miR-214, miR-25 and miR-7) were statistically signiÔ¨Åcant in the NSCLC cases in Chinese cohorts, while four of the Ô¨Åve miRNAs (including miR-483-5p, miR-193a-3p, miR-25 and miR-7) were independently correlated with NSCLC patients in American cohort. Moreover, the panel of the Ô¨Åve miRNAs was also signiÔ¨Åcant in the NSCLC patients (P b 0.001) when the cutoff value for the panel was 2.617 in both the Chinese cohorts and American cohort (Supplementary Table 7). These results demonstrated that the 5 miRNAs and their panel are independent potent diagnostic markers for NSCLC. Subsequently, we also performed a multivariate logistic regression analysis to further determine the inÔ¨Çuence of Ô¨Åve miRNAs' levels and clinical features on NSCLC. After adjusting for age, gender and smoking status, we found that miR-25 (OR = 12.3, 95% CI = 1.5‚Äì101.3, P = 0.02) and the Ô¨Åve-miRNA panel (OR = 20.0, 95% CI = 7.9‚Äì50.7, P b 0.001) were still independently associated with NSCLC in Chinese cohorts; Furthermore, miR-25 (OR = 7.1, 95% CI = 2.1‚Äì23.9, P = 0.001), miR-483-5p (OR = 28.6, 95% CI = 8.1‚Äì100.5, P b 0.001) and miR-7 (OR = 91.0, 95% CI = 22.3‚Äì371.1, P b 0.001) were the independent risk factors for NSCLC in American cohort (Supplementary Table 7). Taken together, these results indicated that the 5 identiÔ¨Åed Please cite this article as: Wang, C., et al., A Five-miRNA Panel IdentiÔ¨Åed From a Multicentric Case‚Äìcontrol Study Serves as a Novel Diagnostic Tool for Ethnically Diverse No..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.07.034 8 C. Wang et al. / EBioMedicine xxx (2015) xxx‚Äìxxx miRNAs enable the detection of NSCLC and may potentially serve as independent risk factor for NSCLC patients. 4. Discussion To our knowledge, this study is the Ô¨Årst multiethnic, multicentric, single-blind global analysis of miRNA expression patterns in the sera from patients with NSCLC. We identiÔ¨Åed a new Ô¨Åve-miRNA panel in three independent cohorts from four centers in China and then validated the miRNA panel in a large cohort from America in a blind sensitivity trial. Our results reveal a Ô¨Åve-miRNA panel which can serve as a potential biomarker for diagnosing NSCLC in patients of different races, even in the early stages of the cancer. The panel can also differentiate between malignant lesions and benign nodules that are frequently found by CT scans in high-risk populations. Our Ô¨Åndings expand upon the results of previous studies of circulating-miRNA signatures in patients with NSCLC. Numerous circulating-miRNA signatures have been reported for the detection of NSCLC, but the miRNAs identiÔ¨Åed by different groups vary from one another (Hu et al., 2010; Bianchi et al., 2011; Chen et al., 2012a; Heegaard et al., 2012; Wang et al., 2013). This inconsistency, we suspect, may be attributed to the differences in study design, disease type, sample size, methodology and especially in the method of data normalization. Improper normalization methods may obscure real changes and make artiÔ¨Åcial changes. In these previous studies on NSCLC, several different normalization methods were applied in circulating-miRNA expression analysis, including normalizing miRNA concentrations to serum volume, or using reference genes that include small RNAs, endogenous miRNAs or external non-human miRNAs. In the present study, we utilized a combination of let-7d, let-7g and let-7i as an internal reference for the normalization of serum miRNA concentrations. In a recent study, we comprehensively assessed let-7d/g/i and demonstrated the statistical superiority of this stable reference gene, as it corrects for experimental variations more effectively and achieves more accurate identiÔ¨Åcation than alternative existing methods (Chen et al., 2013). The use of this appropriate reference for normalization of miRNAs in serum improves the reproducibility and sensitivity of the results and guarantees reliable biomarker discovery. To realize true translational relevance and bring circulating miRNAs into routine diagnostics, multiethnic, multicentric, and reproducibly validated studies are indispensable. In the present study, we selected candidate serum miRNAs in three Chinese cohorts Ô¨Årst and then validated the wide applicability of this biomarker in a large American cohort in a blind fashion. We found that a panel of Ô¨Åve serum miRNAs, including miR-483-5p, miR-193a-3p, miR-25, miR-214 and miR-7, were signiÔ¨Åcantly elevated in Chinese patients with NSCLC compared with control groups. The AUCs of ROC curve of this panel were 0.976 and 0.823 for the two Chinese cohorts, respectively. Furthermore, in a single-blind sensitivity trial, using a proper cutoff value of 2.617, the miRNA panel correctly classiÔ¨Åed 103 of 108 (95%) NSCLC cases and 87 of 104 (84%) controls from an American cohort. Our results demonstrate that the effectiveness of the Ô¨Åve-miRNA panel is not limited to Chinese patients; rather, the biomarker also has high sensitivity and speciÔ¨Åcity for the diagnosis of NSCLC in American patients. Our results suggest that this miRNA panel has potential utility as a biomarker for detecting NSCLC in persons of different races. In addition, we observed that the Cq values of all serum miRNAs examined in American cohort (both patients and controls) were generally higher (approximately 3 cycles) than in the corresponding population in Chinese cohort. This result is consistent with a previous report that proÔ¨Åles of circulating miRNAs differed between African Americans and European Americans, suggesting that levels of circulating miRNAs can differ between races (Heegaard et al., 2012). Therefore, it is critical to select a suitable internal reference for quantiÔ¨Åcation of miRNAs in serum. The endogenous internal reference miRNA let-7-d/g/i trio has similar characteristics to other serum miRNAs in that its average Cq value was also approximately 3 cycles higher in American cohort than in Chinese cohorts. Therefore, when ŒîCq was calculated by subtracting the Cq values of let-7-d/g/i from the Cq values of the target miRNAs, the ŒîCq values for the target miRNAs are comparable between the Chinese American cohorts. These results further demonstrate that a set of internal reference miRNA let-7-d/g/i may give highly consistent results across populations and races. One of the weaknesses in our present study is the use of pooled samples for initially global miRNA screening since results generated from pooled samples might yield some inaccurate information and missing useful information due to individual difference, and therefore decreases the speciÔ¨Åcity of the test. Fortunately, by using RT-qPCR validation in a large number of individual serum samples arranged in multiple training and validation sets, we successfully identiÔ¨Åed the Ô¨Åve-miRNA panel and additional bind trial test further conÔ¨Årmed our Ô¨Åndings. Nevertheless, TLDA screening in individual sample should still be preferred and recommended in a similar study in the future. Another notably issue of the present study is that the fold changes of the Ô¨Åve altered miRNAs in TLDA were not matched well with RT-qPCR analysis. The exact reason for this inconsistence, we suspected, is because a 12-cycle preampliÔ¨Åcation was made in TLDA, and the fold-change of miRNA expression between NSCLC patients and control subjects might be ampliÔ¨Åed. Therefore, the TLDA results must be validated using RT-qPCR assay performed at an individual level of serum samples. The diverse, complex molecular events involved in the initiation and development of malignant neoplasm require functional changes not only in the tumor cell growth-related genes and/or tissue-speciÔ¨Åc genes but also in the body's immune response-related genes. Accordingly, there should be multiple miRNAs targeting those genes involved in tumorigenesis. Circulating miRNAs can derive from the tumor itself or from host responses to the tumor (Pritchard et al., 2012; Chen et al., 2012b). Therefore, functionally dysregulated miRNAs in circulation can be classiÔ¨Åed into three groups: (i) tissue-speciÔ¨Åc, (ii) tumor cell growth/cycle-related, and (iii) immune response-related. Of the Ô¨Åve members of our predictive biomarker, miR-7 is a tissue-speciÔ¨Åc miRNA, because it has been found to be directly involved in NSCLC (Chou et al., 2010). miR-25, miR-483-5p and miR-193a-3p are tumor cell growth/cycle-related miRNAs (Zhang et al., 2013; Soon et al., 2009; Yu et al., 2015; Uhlmann et al., 2012). Finally, miR-214 may be categorized as an immune response-related miRNA because it has been found to play a dominant role in endothelial cell function, angiogenesis and exosome-mediated communication between endothelial cells (van Balkom et al., 2013; Ghalali et al., 2014; Schwarzenbach et al., 2012). Therefore, the combination of the identiÔ¨Åed Ô¨Åve serum miRNAs in our study should be more reliable and speciÔ¨Åc than the single miRNA for the detection of NSCLC. In conclusion, we have identiÔ¨Åed a new Ô¨Åve-serum miRNA panel associated with NSCLC. SpeciÔ¨Åcally, we have demonstrated that this group of serum miRNAs can potentially be used as an accurate biomarker for diagnosing NSCLC patients of different races. Further research is necessary to explore whether this panel of miRNAs is also effective for detection of NSCLC patients with other races and how much this panel adds to serum carcinoembryonic antigen. ConÔ¨Çicts of interest The authors have no competing interests to declare. Author Contributions C-Y.Z., C.Z., X.C., M.X. and K.Z. designed the study. C.W., M.D., M.X., N.W., A.VL., R.S-G., X.C., J.W., W.G., X.W. and Y.Z. collected the data. C.W., M.D., A.VL., R.S-G. and S.C. did the data analysis and interpretation. C.Z., C.W. and C-Y.Z. wrote the manuscript. K.Z., M.X. and A.VL. revised the report. S.C., C.W. and X.C. did the statistical analysis. All authors reviewed the report and approved the Ô¨Ånal version. Please cite this article as: Wang, C., et al., A Five-miRNA Panel IdentiÔ¨Åed From a Multicentric Case‚Äìcontrol Study Serves as a Novel Diagnostic Tool for Ethnically Diverse No..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.07.034 C. Wang et al. / EBioMedicine xxx (2015) xxx‚Äìxxx Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ebiom.2015.07.034. References Jemal, A., Bray, F., Center, M.M., Ferlay, J., Ward, E., Forman, D., 2011. Global cancer statistics. CA Cancer J. Clin. 61, 69‚Äì90. D'Addario, G., Fr√ºh, M., Reck, M., Baumann, P., Klepetko, W., Felip, E., et al., 2010. Metastatic non-small-cell lung cancer: ESMO clinical practice guidelines for diagnosis, treatment and follow-up. Ann. Oncol. 21 (Suppl. 5), v116‚Äìv119. Spiro, S.G., Silvestri, G.A., 2005. One hundred years of lung cancer. Am. J. Respir. Crit. Care Med. 172, 523‚Äì529. Siegel, R., Naishadham, D., Jemal, A., 2012. Cancer statistics, 2012. CA Cancer J. Clin. 62, 10‚Äì29. Little, A.G., Rusch, V.W., Bonner, J.A., Gaspar, L.E., Green, M.R., Webb, W.R., et al., 2005. Patterns of surgical care of lung cancer patients. Ann. Thorac. Surg. 80, 2051‚Äì2056. Oken, M.M., Marcus, P.M., Hu, P., Beck, T.M., Hocking, W., Kvale, P.A., et al., 2005. Baseline chest radiograph for lung cancer detection in the randomized prostate, lung, colorectal and ovarian cancer screening trial. J. Natl. Cancer Inst. 97, 1832‚Äì1839. International Early Lung Cancer Action Program Investigators, Henschke, C.I., Yankelevitz, D.F., Libby, D.M., Pasmantier, M.W., Smith, J.P., et al., 2006. Survival of patients with stage I lung cancer detected on CT screening. N. Engl. J. Med. 355, 1763‚Äì1771. Esquela-Kerscher, A., Slack, F.J., 2006. Oncomirs-microRNAs with a role in cancer. Nat. Rev. Cancer 6, 259‚Äì269. Kong, Y.W., Ferland-McCollough, D., Jackson, T.J., Bushell, M., 2012. MicroRNAs in cancer management. Lancet Oncol. 13, e249‚Äìe258. Calin, G.A., Croce, C.M., 2006. MicroRNA signatures in human cancers. Nat. Rev. Cancer 6, 857‚Äì866. Chen, X., Ba, Y., Ma, L., Cai, X., Yin, Y., Wang, K., et al., 2008. Characterization of microRNAs in serum: a novel class of biomarkers for diagnosis of cancer and other diseases. Cell Res. 18, 997‚Äì1006. Mitchell, P.S., Parkin, R.K., Kroh, E.M., Fritz, B.R., Wyman, S.K., Pogosova-Agadjanyan, E.L., et al., 2008. Circulating microRNAs as stable blood-based markers for cancer detection. Proc. Natl. Acad. Sci. U. S. A. 105, 10513‚Äì10518. Eichelser, C., Flesch-Janys, D., Chang-Claude, J., Pantel, K., Schwarzenbach, H., 2013. Deregulated serum concentrations of circulating cell-free microRNAs miR-17, miR34a, miR-155, and miR-373 in human breast cancer development and progression. Clin. Chem. 59, 1489‚Äì1496. Hu, Z., Chen, X., Zhao, Y., et al., 2010. Serum microRNA signatures identiÔ¨Åed in a genomewide serum microRNA expression proÔ¨Åling predict survival of non-small-cell lung cancer. J. Clin. Oncol. 28, 1721‚Äì1726. Bianchi, F., Nicassio, F., Marzi, M., Belloni, E., Dall'olio, V., Bernard, L., et al., 2011. A serum circulating miRNA diagnostic test to identify asymptomatic high-risk individuals with early stage lung cancer. EMBO Mol. Med. 3, 495‚Äì503. 9 Chen, X., Hu, Z., Wang, W., Ba, Y., Ma, L., Zhang, C., et al., 2012a. IdentiÔ¨Åcation of ten serum microRNAs from a genome-wide serum microRNA expression proÔ¨Åle as novel noninvasive biomarkers for nonsmall cell lung cancer diagnosis. Int. J. Cancer 130, 1620‚Äì1628. Heegaard, N.H., Schetter, A.J., Welsh, J.A., Yoneda, M., Bowman, E.D., Harris, C.C., 2012. Circulating micro-RNA expression proÔ¨Åles in early stage nonsmall cell lung cancer. Int. J. Cancer 130, 1378‚Äì1386. Wang, Y., Gu, J., Roth, J.A., Hildebrandt, M.A., Lippman, S.M., Ye, Y., et al., 2013. Pathwaybased serum microRNA proÔ¨Åling and survival in patients with advanced stage non-small cell lung cancer. Cancer Res. 73, 4801‚Äì4809. Luo, Y., Wang, C., Chen, X., Zhong, T., Cai, X., Chen, S., et al., 2013. Increased serum and urinary microRNAs in children with idiopathic nephrotic syndrome. Clin. Chem. 59, 658‚Äì666. Chen, X., Liang, H., Guan, D., Wang, C., Hu, X., Cui, L., et al., 2013. A combination of Let-7d, Let-7g and Let-7i serves as a stable reference for normalization of serum microRNAs. PLoS One 8, e79652. Pritchard, C.C., Kroh, E., Wood, B., Arroyo, J.D., Dougherty, K.J., Miyaji, M.M., et al., 2012. Blood cell origin of circulating microRNAs: a cautionary note for cancer biomarker studies. Cancer Prev. Res. (Phila.) 5, 492‚Äì497. Chen, X., Liang, H., Zhang, J., Zen, K., Zhang, C.Y., 2012b. Secreted microRNAs: a new form of intercellular communication. Trends Cell Biol. 22, 125‚Äì132. Chou, Y.T., Lin, H.H., Lien, Y.C., Wang, Y.H., Hong, C.F., Kao, Y.R., et al., 2010. EGFR promotes lung tumorigenesis by activating miR-7 through a Ras/ERK/Myc pathway that targets the Ets2 transcriptional repressor EFR. Cancer Res. 70, 8822‚Äì8831. Zhang, S., Chen, H., Zhao, X., Cao, J., Tong, J., Lu, J., et al., 2013. REV3L 3‚Ä≤UTR 460 TNC polymorphism in microRNA target sites contributes to lung cancer susceptibility. Oncogene 32, 242‚Äì250. Soon, P.S., Tacon, L.J., Gill, A.J., Bambach, C.P., Sywak, M.S., Campbell, P.R., et al., 2009. miR-195 and miR-483-5p identiÔ¨Åed as predictors of poor prognosis in adrenocortical cancer. Clin. Cancer Res. 15, 7684‚Äì7692. Yu, T., Li, J., Yan, M., Liu, L., Lin, H., Zhao, F., et al., 2015. MicroRNA-193a-3p and -5p suppress the metastasis of human non-small-cell lung cancer by downregulating the ERBB4/PIK3R3/mTOR/S6K2 signaling pathway. Oncogene 34, 413‚Äì423. Uhlmann, S., Mannsperger, H., Zhang, J.D., Horvat, E.√Å., Schmidt, C., K√ºblbeck, M., et al., 2012. Global microRNA level regulation of EGFR-driven cell-cycle protein network in breast cancer. Mol. Syst. Biol. 8, 570. van Balkom, B.W., de Jong, O.G., Smits, M., Brummelman, J., den Ouden, K., de Bree, P.M., et al., 2013. Endothelial cells require miR-214 to secrete exosomes that suppress senescence and induce angiogenesis in human and mouse endothelial cells. Blood 121, 3997‚Äì4006. Ghalali, A., Ye, Z.W., H√∂gberg, J., Stenius, U., 2014. Phosphatase and tensin homolog deleted on chromosome 10 (PTEN) and PH domain and leucine-rich repeat phosphatase cross-talk (PHLPP) in cancer cells and in transforming growth factor Œ≤-activated stem cells. J. Biol. Chem. 289, 11601‚Äì11615. Schwarzenbach, H., Milde-Langosch, K., Steinbach, B., Muller, V., Pantel, K., 2012. Diagnostic potential of PTEN-targeting miR-214 in the blood of breast cancer patients. Breast Cancer Res. Treat. 134, 933‚Äì941. Please cite this article as: Wang, C., et al., A Five-miRNA Panel IdentiÔ¨Åed From a Multicentric Case‚Äìcontrol Study Serves as a Novel Diagnostic Tool for Ethnically Diverse No..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.07.034 CLINICAL RESEARCH STUDY Age and Gender Differences in Quality of Care and Outcomes for Patients with ST-segment Elevation Myocardial Infarction Sripal Bangalore, MD, MHA,a Gregg C. Fonarow, MD,b Eric D. Peterson, MD, MPH,c Anne S. Hellkamp, MS,c Adrian F. Hernandez, MD,c Warren Laskey, MD,d W. Frank Peacock, MD,e Christopher P. Cannon, MD,f Lee H. Schwamm, MD,g Deepak L. Bhatt, MD, MPH,h for the Get with the Guidelines Steering Committee and Investigators a New York University Medical Center, New York; bAhmanson-UCLA Cardiomyopathy Center, Los Angeles, California; cDuke Clinical Research Institute, Durham, NC; dUniversity of New Mexico, Albuquerque; eCleveland Clinic, Cleveland, Ohio; fTIMI Study Group, Boston, Mass; gMassachusetts General Hospital, Boston, Mass; hVA Boston Healthcare System, Brigham and Women‚Äôs Hospital, and Harvard Medical School, Boston, Mass. ABSTRACT BACKGROUND: Young patients (aged ’Ö 45 years) presenting with ST-segment elevation myocardial infarction present unique challenges. The quality of care and in-hospital outcomes may differ from their older counterparts. METHODS: A total of 31,544 patients presenting with ST-segment elevation myocardial infarction and enrolled in the American Heart Association‚Äôs Get With the Guidelines Coronary Artery Disease registry were analyzed. The cohort was divided into those aged 45 years or less and those aged more than 45 years. RESULTS: Young patients accounted for 10.3% of all ST-segment elevation myocardial infarction cases. Compared with older patients, younger patients were less likely to have traditional cardiovascular risk factors and had similar or better quality/performance measures with lower in-hospital mortality (unadjusted rate 1.6 vs 6.5%, P œΩ.0001; adjusted odds ratio [OR], 0.37; 95% confidence interval [CI], 0.29-0.46). Time trend analysis (2002-2008) suggested an increase over time in the ‚Äúall or none‚Äù composite performance measure in both the younger and older patients (68%-97% and 69%-96%, respectively). However, there was significantly lower quality of care and worse outcomes in women (vs men) and in the very young (’Ö35 vs 36-45 years). Significant interaction was seen between age and gender for in-hospital death, such that the gender difference was greater in the younger cohort. Similar interaction was seen for door-tothrombolytic time such that the gender delay was greater in the younger cohort (women:men ratio of means œ≠ 1.73, 95% CI, 1.21-2.45 [younger] vs 1.08, 95% CI, 1.00-1.18 [older]; Pinteraction œ≠ .0031). CONCLUSION: Young patients aged 45 years or less presenting with ST-segment elevation myocardial infarction overall had similar quality of care and in-hospital outcomes as older counterparts. However, quality of care was significantly lower and mortality was higher in young women (vs young men) and the very young (’Ö35 vs 36-45 years). ¬© 2012 Elsevier Inc. All rights reserved. ‚Ä¢ The American Journal of Medicine (2012) 125, 1000-1009 KEYWORDS: Gender; Myocardial infarction; Prognosis; Young. Funding: The Get With The Guidelines-Coronary Artery Disease (GWTG-CAD) program was provided by the American Heart Association. The GWTG-CAD program was supported in part through the American Heart Association Pharmaceutical Roundtable and an unrestricted educational grant from Merck. Conflicts of Interest: See last page of article. 0002-9343/$ -see front matter ¬© 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2011.11.016 Authorship: All authors had access to the data and played a role in writing this manuscript. Requests for reprints should be addressed to Sripal Bangalore, MD, MHA, Director of Research, Cardiac Catheterization Laboratory, The Leon H. Charney Division of Cardiology, New York University School of Medicine, New York, NY 10016. E-mail address: sripal.bangalore@nyumc.org Bangalore et al ST-Segment Elevation Myocardial Infarction in the Young 1001 ST-segment elevation myocardial infarction is a relaSUBJECTS AND METHODS tively uncommon disease in young individuals, accountData Source and Study Population ing for 2% to 10% of ST-segment elevation myocardial Patients enrolled in the Get With the Guidelines-Coronary infarction cases.1-5 Earlier studies demonstrated signifiArtery Disease (GWTG-CAD) registry, a prospective, mulcant differences in the risk factor profile and outcomes in ticenter, national, observational younger patients when compared registry and quality improvement with older patients with ST-seginitiative established by the Amerment elevation myocardial inCLINICAL SIGNIFICANCE ican Heart Association, were chofarction.6-10 It is unknown sen for this analysis.13,14 For the whether the increased rate of risk ‚óè Young patients aged 45 years or less purposes of the present analysis, factors, such as smoking among presenting with ST-segment elevation patients were included if the initial younger patients with ST-segmyocardial infarction overall had simielectrocardiogram showed new ment elevation myocardial inlar quality of care and in-hospital outST-segment elevation or a new farction, seen in the studies from comes as older counterparts. left bundle branch block or a di1970 to 1999, is still applicable agnosis of ST-segment elevation to contemporary cohorts given ‚óè Quality of care was significantly lower myocardial infarction. At the time increased public health measures and mortality was higher in young of the present analysis, the registry to prevent and reduce smoking. women (vs young men) and the very included 47,694 patients with a diIn addition, a subset of ST-segyoung (’Ö35 vs 36-45 years). agnosis of ST-segment elevation ment elevation myocardial in‚óè Substantial differences were observed myocardial infarction enrolled at farction in the very young (’Ö35 380 participating sites between in the use of evidence-based therapies years) is likely related to illicit 2002 and 2008. Patients who were in both younger and older women when drug use8,10 or to nontraditional transferred in or out were exrisk factors, such as oral contracompared with men in the correspondcluded (n œ≠ 16,150), leaving a ficeptive use.9 Very young paing age category. In addition, women nal study population of 31,544 patients with ST-segment elevation (both younger and older) had higher tients from 369 sites. myocardial infarction are likely in-hospital mortality when compared to attribute chest pain to other with men. Quality of Care Measures causes and present late. In addi‚óè Both the door-to-thrombolytic time and tion, physicians taking care of Adherence with evidence-based door-to-balloon time was longer for such patients also are less likely medical therapies was evaluated, to consider cardiac cause for including the acute use (within 24 both younger and older women when such pain in the very young, rehours) of aspirin and ‚ê§-blockers; compared with younger and older men. sulting in withholding of care, the use of evidence-based therapies at discharge (aspirin, ‚ê§inadequate care, or delays in blockers, clopidogrel, angiotencare. The pattern of care and outsin-converting enzyme inhibitors or angiotensin II receptor comes of the very young with ST-segment elevation blocking agents, and statins or lipid-lowering drugs); and myocardial infarction is therefore not well defined. Prior smoking cessation counseling. An ‚Äúall or none‚Äù composite reports have shown that women have a higher mortality performance measure encompassed 6 key quality of care rate in the setting of acute myocardial infarction than measures: patients discharged while taking aspirin, patients men, even in the era of reperfusion therapy.8,11,12 Howdischarged while taking ‚ê§-blockers, patients receiving asever, it is not well defined whether younger women have pirin within 24 hours of presentation, patients with doculower quality of care and worse outcomes. Younger premented left ventricular dysfunction discharged on angiotenmenopausal women are less likely to attribute chest pain sin-converting enzyme inhibitor or angiotensin receptor as due to cardiac cause and present late. Treating physiblockers, smoking cessation counseling, and patients with a cians also are likely biased by traditional teaching, which low-density lipoprotein level greater than 100 mg/dL resuggests that coronary artery disease is a disease of ceiving lipid-lowering drugs. postmenopausal women. Although prior studies8 have explored gender-related differences in patients presenting Outcome Measures with ST-segment elevation myocardial infarction, it is The primary outcome measure was in-hospital all-cause less well known whether there is a gender‚Äìage interaction mortality. Secondary outcome measures included median for both quality of care and in-hospital mortality. door-to-balloon times, median door-to-thrombolytic times, Our objective was to evaluate differences in the risk the proportion of patients meeting the American College of factor profile, quality of in-hospital care, and outcomes of Cardiology/American Heart Association guideline-recomyounger patients when compared with older patients with mended within 90-minute door-to-balloon time, and length ST-segment elevation myocardial infarction. of hospital stay. 1002 The American Journal of Medicine, Vol 125, No 10, October 2012 Statistical Analysis Subjects with ST-segment elevation myocardial infarction were categorized into 2 groups according to their age: younger cohort (’Ö45 years) versus older cohort (œæ45 years). Multivariable models were used to evaluate the association of age groups with quality of care measures and outcomes. For the composite performance measure (number successes/number eligible), an opportunity-level model was used in which each measure contributed an observation. Time trends were assessed within each age group separately, and a time-by-age group interaction test was used to determine whether time trends were different between age groups. All models used a generalized estimating equation approach to take into account within-hospital clustering and were adjusted for gender, race, chronic obstructive pulmonary disease, diabetes, prior heart failure, hypertension, hyperlipidemia, prior myocardial infarction, peripheral artery disease, chronic renal insufficiency, prior cerebrovascular accident/transient ischemic attack, smoking, aspirin or ‚ê§-blocker use within 24 hours, hospital region, bed size, academic status, and in-hospital intra-aortic balloon pump or fibrinolytic therapy. Each risk relationship is presented as an odds ratio (OR) with a 95% confidence interval (CI) for dichotomous outcomes and as the ratio of geometric means, with a 95% CI for continuous outcomes. Two subgroup analyses were performed: or each outcome, gender ORs were estimated within each age cohort using subgroup models, adjusted as above, and an age by gender interaction was tested in the full model; among the younger cohort, patients were divided into those aged 35 years or less versus those aged 36 to 45 years, and these 2 groups were compared for each outcome using a model adjusted as above. All statistical analyses were performed using SAS version 9.1 (SAS Institute Inc, Cary, NC). All P values were 2-tailed, with statistical significance set at .05. RESULTS Baseline Characteristics and In-Hospital Treatment Among the 31,544 patients with ST-segment elevation myocardial infarction, 3257 (10.3%) were categorized as young (aged ’Ö 45 years). When compared with the older cohort, the younger cohort were more likely to be men, black, or Hispanic, and to smoke, but were less likely to have traditional risk factors (Table 1) and had higher body mass index, baseline heart rate, and systolic blood pressure, but were less likely to be taking medications before admission. A high percentage of both groups received aspirin/‚ê§blockers within 24 hours of admission, but this was higher in the younger cohort. Approximately two thirds of the patients received percutaneous coronary intervention with a stent, with a greater proportion of younger patients receiving stents (Table 1). Quality/Performance Measures and In-Hospital Outcomes: Univariate For the overall cohort, the majority of patients received evidence-based therapies, but a greater proportion of young patients were more likely to receive ‚ê§-blockers within 24 hours of admission, angiotensin-converting enzyme inhibitors, and clopidogrel at discharge; to have a recorded blood pressure less than 140/90 mm Hg at discharge and lowdensity lipoprotein levels recorded; and to receive recommendations on rehabilitation, weight management, and smoking cessation counseling when compared with the older cohort (Table 2). The median door-to-balloon time was 85 minutes with 51% of patients achieving a door-toballoon time of 90 minutes or less. There was no difference between the young and the older cohorts for door-to-balloon time or door-to-thrombolytic time (Table 2). In addition, younger patients had a significantly lower in-hospital mortality and length of stay when compared with their older counterparts (Table 2). Quality/Performance Measures and In-Hospital Outcomes: Multivariable In a multivariable model adjusting for baseline covariates, younger patients were more likely to have a recorded blood pressure less than 140/90 mm Hg at discharge, to have low-density lipoprotein levels recorded, and to receive recommendations on rehabilitation, weight management, and smoking cessation (Table 2) when compared with older patients. Younger patients had a 63% lower odds of inhospital mortality that was statistically significant and a shorter length of stay (Table 2). Quality/Performance Measures and In-Hospital Outcomes: Time Trend The ‚Äúall or none‚Äù composite performance measure improved significantly from 68.4% to 97.0% from 2002 to 2008 for younger patients (adjusted OR per year 1.46; 95% CI, 1.35-1.59; P œΩ .0001) and from 68.8% to 96.3% for older patients (adjusted OR per year 1.39; 95% CI, 1.341.44; P œΩ .0001) (Figure 1). During the study period, inhospital all-cause death remained steady for both younger patients (adjusted OR per year 1.13; 95% CI, 0.91-1.41; P œ≠ .26) and older patients (adjusted OR per year 1.03; 95% CI, 0.98-1.08; P œ≠ .21) (Figure 2). Subgroup Analyses Outcomes in Men versus Women. Gender differences in quality/performance measures and outcomes are listed in Table 3. When compared with men in the corresponding age category, both younger and older women were less likely to receive angiotensin-converting enzyme inhibitors/ angiotensin receptor blockers at discharge, to receive lipidlowering therapy, to have a blood pressure less than 140/90 mm Hg at discharge, to have longer door-to-balloon times, and to receive stents, with fewer achieving a door-to-bal- Bangalore et al Table 1 ST-Segment Elevation Myocardial Infarction in the Young 1003 Demographics, Baseline Risk Factors, and In-Hospital Treatment by Age Group Variable Demographics Age (SD), y Men, % Race, % White Black Hispanic Asian Medical history Hypertension, % Diabetes, % Hyperlipidemia, % Smoker (past 12 mo), % Prior MI, % Prior heart failure, % Prior CABG, % Peripheral vascular disease, % Renal insufficiency, % Cerebrovascular accident or transient ischemic attacks, % COPD or asthma, % Vital signs, admission BMI (SD), kg/m2 Heart rate (SD), beats/min Systolic pressure (SD), mm Hg Medications before admission Aspirin, % Aspirin œ© another antiplatelet, % ACEI/ARB, % Beta-blocker, % Statin, % Medications at admission (within 24 h) Aspirin, % Beta-blocker, % Procedures Percutaneous coronary intervention with stent, % Intra-aortic balloon pump, % Left ventricular assist device, % Fibrinolytic therapy, % Ejection fraction (SD), % Hospital characteristics No. of beds (SD), mean Region, % Northeast Midwest South West Academic site, % Overall (N œ≠ 31,544) Young (’Ö45 y) (N œ≠ 3257) Older (œæ45 y) (N œ≠ 28,287) 63.4 (14.4) 65.3 40.2 (4.7) 77.1 66.1 (12.7) 63.9 74.1 7.3 7.9 3.0 66.8 11.5 10.1 2.8 74.9 6.8 7.7 3.1 62.2 9.2 43.1 37.4 17.6 9.4 0.7 5.9 5.9 6.3 47.8 8.1 40.0 65.5 14.6 2.4 0.3 1.8 2.2 1.8 63.7 9.3 43.4 34.5 18.0 10.1 0.8 6.4 6.3 6.8 œΩ.0001 .04 .001 œΩ.0001 œΩ.0001 œΩ.0001 .004 œΩ.0001 œΩ.0001 œΩ.0001 10.6 4.4 11.2 œΩ.0001 P Value œΩ.0001 œΩ.0001 œΩ.0001 28.6 (6.3) 80.7 (20.8) 135.0 (29.9) 30.6 (6.7) 82.6 (18.9) 136.7 (28.2) 28.4 (6.2) 80.5 (21.0) 134.8 (30.1) œΩ.0001 œΩ.0001 .02 36.0 39.3 34.4 10.8 33.4 25.5 28.3 27.7 8.0 25.2 37.1 40.5 35.0 11.1 34.2 œΩ.0001 œΩ.0001 œΩ.0001 .01 œΩ.0001 91.3 78.8 93.7 84.3 91.0 78.2 œΩ.0001 œΩ.0001 64.8 71.5 64.0 œΩ.0001 3.1 0.06 6.8 46.2 (13.5) 2.8 0.06 7.1 48.4 (12.2) 3.1 0.05 6.8 45.9 (13.6) .16 .96 .42 œΩ.0001 412 (255) 448 (271) 407 (253) œΩ.0001 17.0 26.2 32.0 24.9 54.0 17.8 27.6 32.8 21.9 57.1 16.9 26.0 31.9 25.2 53.6 .0005 .0002 Numbers in parentheses represent SD. SD œ≠ standard deviation; MI œ≠ myocardial infarction; CABG œ≠ coronary artery bypass grafting; COPD œ≠ chronic obstructive pulmonary disease; BMI œ≠ body mass index; ACEI œ≠ angiotensin-converting enzyme inhibitor; ARB œ≠ angiotensin receptor blocker. loon time 90 minutes or less and door-to-thrombolytic time 30 minutes or less (Table 3). However, there were no differences in the use of intra-aortic balloon pump (P œ≠ .99) or left ventricular assist device (P œ≠ .36) between younger women and younger men. In addition, both younger and older women had significantly increased in-hospital mortal- 1004 Table 2 The American Journal of Medicine, Vol 125, No 10, October 2012 Quality of Care/Performance Measures and Outcomes by Age Groups Variable Quality measures Beta-blocker within 24 h before/ after arrival, % Door-to-percutaneous coronary intervention time ’Ö90 min, % Door-to-thrombolytic time ’Ö 30 min, % ACEI at discharge, % ACEI/ARB at discharge for acute MI, % Lipid-lowering medications at discharge, % Blood pressure œΩ 140/90 at discharge, % LDL level recorded, % Rehabilitation/activity recommendations, % Clopidogrel at discharge, % Weight management/activity recommendations, % Performance measures Aspirin at discharge, % Beta-blocker at discharge, % Smoking cessation counseling, % LDL œæ 100 mg/dL and lipid-lowering medications, % ACEI/ARB at discharge for left ventricular systolic dysfunction, % Aspirin within 24 h before/after arrival, % Composite performance (No. success/No. eligible) (SD), mean Composite performance (100% compliance), % Time to therapy Door-to-balloon time (min) [median (IQR)] Door-to-thrombolytic time (min) [median (IQR)] Outcomes In-hospital death, % Length of stay (d) [mean (SD)] P Value Multivariable OR (95% CI) Young vs Older 90.0 .02 1.07 (0.95-1.20) .26 52.0 50.8 .45 0.96 (0.87-1.06) .41 34.2 35.8 34.0 .93 0.98 (0.71-1.36) .90 77.6 76.6 79.6 77.2 77.3 76.6 .02 .93 1.06 (0.96-1.17) 0.96 (0.88-1.05) .25 .37 87.5 89.6 87.2 .06 0.97 (0.87-1.09) .65 85.5 89.4 85.1 œΩ.0001 1.23 (1.08-1.41) .002 74.3 85.4 79.8 87.7 73.7 85.1 œΩ.0001 1.16 (1.07-1.25) .003 1.08 (1.01-1.16) .0002 .04 80.4 84.9 84.8 87.1 79.9 84.6 œΩ.0001 1.08 (0.99-1.18) .004 1.10 (1.01-1.20) .10 .02 97.2 96.2 92.2 92.0 97.8 96.5 94.6 93.6 97.1 96.2 91.7 91.7 .08 .49 .0001 .13 1.02 0.96 1.40 0.98 84.9 89.7 84.5 .01 1.17 (0.88-1.55) .29 94.8 95.6 94.7 .06 1.02 (0.88-1.19) .76 94.3 (15.9) 95.4 (13.5) 94.1 (16.1) .02 1.05 (0.97-1.13)* .24 84.3 85.7 84.1 .15 0.94 (0.85-1.04) .22 Overall (N œ≠ 31,544) Young (’Ö45 y) (N œ≠ 3257) Older (œæ45 y) (N œ≠ 28,287) 90.2 91.6 51.0 (0.84-1.23) (0.77-1.19) (1.16-1.69) (0.80-1.20) Multivariable P Value .87 .68 .0004 .85 85 (60-122) 85 (60-119) 85 (60-122) .33 1.00 (0.96-1.04)* .92 40 (25-67) 39 (22-75) 40 (25-67) .48 1.04 (0.90-1.19)* .63 5.98 5.2 (6.3) 1.61 4.2 (5.6) 6.48 5.3 (6.3) œΩ.0001 0.37 (0.29-0.46) œΩ.0001 0.88 (0.86-0.90)* œΩ.0001 œΩ.0001 SD œ≠ standard deviation; ACEI œ≠ angiotensin-converting enzyme inhibitor; ARB œ≠ angiotensin receptor blocker; IQR œ≠ interquartile range; LDL œ≠ lowdensity lipoprotein. *For door-to-balloon and door-to-thrombolytic times and length of stay, ratio of geometric means shown. For Composite Performance, OR for 1 additional performance measure shown. ity and length of stay (trend for younger women). Furthermore, the older women were less likely to receive angiotensin-converting enzyme inhibitors at discharge, to have low-density lipoprotein levels recorded during hospitalization, to be advised on rehabilitation/weight management, to be treated with beta-blockers within 24 hours of presentation, or to be discharged on clopidogrel, aspirin, or beta- blockers. They also had longer door-to-thrombolytic times, and fewer older women met the ‚Äúall or none‚Äù composite performance measure than did older men. In addition, a significant interaction between age and gender (P œ≠ .03) was seen for in-hospital death (Figure 3) such that the gender difference was greater in the younger cohort than the older cohort. A similar interaction was seen for door-to- Bangalore et al ST-Segment Elevation Myocardial Infarction in the Young 1005 Figure 1 Time trend in ‚Äúall or none‚Äù composite performance measures in patients with ST-segment elevation myocardial infarction aged 45 years or less versus those aged more than 45 years. thrombolytic such that the gender delay was greater in the younger cohort when compared with the older cohort (women:men ratio of means 1.73, 95% CI, 1.21-2.45 [younger] vs 1.08, 95% CI, 1.00-1.18 [older], Pinteraction œ≠ .003), with fewer achieving the door-to-thrombolytic goal of 30 minutes or less. Outcomes in the Very Young. Among the 3257 patients aged 45 years or less, 489 (15%) were very young (aged ’Ö35 years). These very young patients had a significantly longer door-to-balloon time, had a lower ‚Äúall or none‚Äù composite performance measure, were less likely to receive a stent, and had higher in-hospital mortality compared with the cohort aged 36 to 45 years (Table 4). In a multivariable model adjusting for baseline covariates, very young patients had a 32% lower odds of the ‚Äúall or none‚Äù composite performance measure, 37% lower odds of receiving a stent, and a nonsignificant trend (P œ≠ .061) toward an increase in the odds of in-hospital death when compared with the patients aged 36 to 45 years (Table 4). DISCUSSION Our analysis of data from approximately 32,000 patients with ST-segment elevation myocardial infarction, from 369 sites, showed that the young patients presenting with STsegment elevation myocardial infarction were more likely to be men, smokers, black, or Hispanic, whose quality of care Figure 2 Time trend in in-hospital all-cause mortality in patients with ST-segment elevation myocardial infarction aged 45 years or less versus those aged more than 45 years. 1006 Table 3 The American Journal of Medicine, Vol 125, No 10, October 2012 Quality of Care/Performance Measures and Outcomes by Age and Gender Subgroups ’Ö45 y Variable Quality measures ACEI at discharge, % ACEI/ARB at discharge for acute MI, % Lipid-lowering medications at discharge, % Blood pressure œΩ 140/ 90 at discharge, % LDL level recorded, % Rehabilitation/activity recommendations, % Beta-blocker within 24 h before/after arrival, % Door-to-percutaneous coronary intervention time ’Ö90 min, % Door-to-thrombolytic time ’Ö30 min, % Clopidogrel at discharge, % Weight management/ activity recommendations, % Performance measures Aspirin at discharge, % Beta-blocker at discharge, % Smoking cessation counseling, % LDL œæ 100 mg/dL and lipid-lowering medications, % ACEI/ARB at discharge for left ventricular systolic dysfunction, % Aspirin within 24 h before/after arrival, % Composite performance (No. success/No. eligible) * [mean (SD)] Composite performance (100% compliance), % Treatment Door-to-balloon time (min)* [median (IQR)] Door-to-thrombolytic time (min)* [median (IQR)] PCI with stent, % Outcomes In-hospital death, % Length of stay (d)* [mean (SD)] Age œ´ Gender œæ45 y Female (N œ≠ 689) Male (N œ≠ 2510) Multivariable OR (95% CI) Male (N œ≠ 18,086) Multivariable OR (95% CI) 76.5 72.1 80.4 78.5 0.84 (0.68-.04) .12 0.73 (0.59-0.89) .002 74.8 74.6 78.5 77.4 0.86 (0.81-0.91) œΩ.0001 0.90 (0.86-0.96) .0005 .99 .07 84.6 90.7 0.63 (0.48-0.83) .001 82.8 89.3 0.70 (0.64-0.75) œΩ.0001 .45 91.4 88.9 1.51 (1.11-2.04) .01 82.1 86.6 0.80 (0.75-0.86) œΩ.0001 œΩ.0001 79.3 87.9 80.1 87.8 1.02 (0.83-1.25) .88 0.98 (0.83-1.17) .87 71.4 82.4 74.8 86.6 0.94 (0.89-0.99) .01 0.87 (0.82-0.92) œΩ.0001 .21 .07 89.5 92.1 0.86 (0.66-1.12) .26 88.4 90.8 0.90 (0.84-0.98) .52 49.1 52.6 0.79 (0.61-1.01) .06 46.6 52.6 0.83 (0.78-0.89) œΩ.0001 .63 16.7 41.1 0.33 (0.13-0.80) .01 30.1 35.7 0.80 (0.64-0.99) .04 .03 83.6 85.2 1.10 (0.80-1.50) .56 75.8 82.2 0.88 (0.82-0.95) .001 .19 87.7 87.0 0.97 (0.77-1.21) .76 82.3 85.6 0.89 (0.83-0.96) .001 .38 96.8 96.5 98.0 96.4 0.77 (0.46-1.30) .33 1.24 (0.66-2.32) .50 96.2 95.0 97.5 96.7 0.78 (0.68-0.90) .0005 0.75 (0.66-0.86) œΩ.0001 .85 .06 94.7 94.4 1.07 (0.67-1.69) .78 91.5 91.7 1.00 (0.88-1.14) .94 .84 90.9 94.1 0.84 (0.47-1.52) .57 88.8 93.0 0.75 (0.63-0.88) .0005 .59 84.6 91.1 0.48 (0.24-0.95) .03 81.5 85.9 0.78 (0.66-0.92) .004 .30 94.0 95.9 0.82 (0.57-1.18) .29 93.8 95.1 0.92 (0.82-1.03) .15 .99 94.2 (15.8) 95.6 (13.0) 0.92 (0.73-1.17) .49 92.9 (18.1) 94.7 (15.0) 0.88 (0.84-0.93) œΩ.0001 .65 83.5 86.0 0.94 (0.71-1.24) .65 81.9 85.1 0.94 (0.88-1.00) .92 Multivariable Female P Value (N œ≠ 9706) Multivariable Interaction P Value P Value .01 .05 88 (63-134) 84 (59-116) 1.18 (1.06-1.31) .003 88 (63-132) 84 (60-119) 1.09 (1.06-1.12) œΩ.0001 .10 67 (38-109) 35 (21-57) 44 (27-73) 39 (24-66) .003 70.2 3.23 4.5 (4.2) 75.6 1.17 4.1 (5.9) 1.73 (1.21-2.45) .002 0.81 (0.67-0.99) .04 2.15 (1.05-4.38) .03 1.05 (0.99-1.11) .09 61.8 8.56 5.8 (5.9) 71.5 5.34 5.1 (6.6) 1.08 (1.00-1.18) .05 0.80 (0.76-0.85) œΩ.0001 .50 1.29 (1.15-1.44) œΩ.0001 1.09 (1.06-1.11) œΩ.0001 .03 .28 SD œ≠ standard deviation; MI œ≠ myocardial infarction; LDL œ≠ low-density lipoprotein; ACEI œ≠ angiotensin-converting enzyme inhibitor; ARB œ≠ angiotensin receptor blocker; IQR œ≠ interquartile range; SD œ≠ standard deviation. *ORs for continuous outcomes: For door to times and length of stay, ratio of geometric means shown. For composite performance, OR for 1 additional performance measure shown. and in-hospital outcomes were similar to or better than that of their older counterparts. In addition with the implementation of the GWTG program, the ‚Äúall or none‚Äù composite performance measure has steadily improved from 2002 to 2008 in both the younger and older patients. There was a significant interaction between age and gender, with young Bangalore et al ST-Segment Elevation Myocardial Infarction in the Young 1007 Figure 3 Age‚Äì gender interaction and in-hospital all-cause mortality. CI œ≠ confidence interval; OR œ≠ odds ratio. women having significantly lower quality of care as assessed by performance measures and worse outcomes than young men. A similar interaction for care and outcomes also was seen among the very young (’Ö35 years) versus those aged 36 to 45 years. ST-segment elevation myocardial infarction seen in the studies published in the last 3 decades continues to be seen in this contemporary cohort.1,2,4,6,7,20,21 In addition, the younger cohort were treated as well or even better than their older counterparts and were more likely to receive evidence-based therapies with lower inhospital mortality and a shorter length of stay. Although prior studies did not consistently report quality of care and performance measures, few studies have shown that the referral rate for revascularization for younger cohort was similar,1-3 with significantly fewer complications during hospitalization,1,3 similar inpatient medical treatment,1,2 lower in-hospital mortality,1-3 and shorter length of stay2 when compared with the older cohort. Others have shown that younger patients are more likely to receive beta-blocker and statin therapy when compared with the older cohort.2 In ST-Segment Elevation Myocardial Infarction in the Young Our data suggest that 10.3% of ST-segment elevation myocardial infarction occurs in patients aged less than 45 years, consistent with previously reported data.15-19 Among the younger cohort, there was a greater proportion of men, black or Hispanics, and smokers, but a lesser proportion of those with traditional cardiovascular risk factors. Of note, the high prevalence of smokers in the younger cohort with Table 4 Quality of Care/Performance Measures and Outcomes by Age Subgroups Variable Performance measures Composite performance (No. success/No. eligible) [mean (SD)] Composite performance (100% compliance), % Treatment Door-to-balloon time (min) [median (IQR)] Door-to-thrombolytic time (min) [median (IQR)] Percutaneous coronary intervention with stent, % Outcomes In-hospital death, % ’Ö35 y (N œ≠ 489) œæ35-45 y (N œ≠ 2768) P Value Multivariable OR (95% CI) ’Ö 35 vs œæ35-45 y 93.3 (16.9) 95.7 (12.8) .002 0.72 (0.57-0.90)* .005 81.2 86.5 .005 0.68 (0.51-0.89) .01 Multivariable P Value 91 (65-138) 84 (59-117) .001 1.11 (0.95-1.29)* .18 45 (25-87) 37 (22-71) .94 1.24 (0.83-1.87)* .29 65.1 76.0 œΩ.0001 0.63 (0.51-0.78) œΩ.0001 2.89 1.39 .01 2.01 (0.97-4.16) .06 OR œ≠ odds ratio; IQR œ≠ interquartile range; SD œ≠ standard deviation. *For door-to-balloon and door-to-thrombolytic times, ratio of geometric means shown. For composite performance, OR for 1 additional performance measure shown. 1008 The American Journal of Medicine, Vol 125, No 10, October 2012 our study, although a greater proportion of younger patients received evidence-based therapies, these differences were no longer present after adjusting for baseline confounders. However, despite multivariable adjustments, younger patients were more likely to have a recorded blood pressure less than 140/90 mm Hg at discharge, to have low-density lipoprotein levels recorded, and to receive recommendations on rehabilitation, weight management, and smoking cessation when compared with older patients. Our time trend analysis showed that with the inception of the GWTG program, there has been a steady increase in the implementation of evidence-based therapies in both the young and the older patients. ST-Segment Elevation Myocardial Infarction: Gender Differences Among the Young Although the younger cohort was treated as aggressively or more aggressively when compared with the older cohort and had better in-hospital outcomes, we observed substantial differences in the use of evidence-based therapies in both younger and older women when compared with men in the corresponding age category. In addition, women (both younger and older) had higher in-hospital mortality when compared with men. We observed a significant interaction such that the younger cohort had greater gender difference in mortality and greater delays in door-to-thrombolytic time compared with the older cohort. Of note, both the door-tothrombolytic time and door-to-balloon time was longer for both younger and older women when compared with younger and older men. Prior analysis from the registry (2001-2006) had shown underuse of evidence-based treatments and higher adjusted mortality rates in women with ST-segment elevation myocardial infarction driven by an increase in early mortality (within 24 hours of admission).8 This analysis extends the prior observation in including patients through 2008 and showing significant interaction with age such that observed difference in mortality and delays in care were greater for the younger cohort. Multiple reports have shown that women have a higher mortality rate in the setting of acute myocardial infarction than men, even in the era of reperfusion therapy.8,11,12 This has been accounted for by a variety of factors: 1) Women present more often with atypical symptoms and present late; 2) in a small proportion, nontraditional factors such as oral contraceptive use9 may be contributory; 3) pregnant women present unique challenges in both the diagnosis and the treatment of ST-segment elevation myocardial infarction because the pathophysiology of ST-segment elevation myocardial infarction in pregnancy is different (likely due to coronary artery dissection), and there is likely withholding/ delaying of therapies such as thrombolytic therapy or aggressive anticoagulation; 4) there was less aggressive treatment of women with evidence-based therapies and invasive treatments8,22-24 either because of selection bias or fears of increased risk of complications; 5) women have more comorbidities than men when presenting with myocardial in- farction,12,25,26 although, in our analysis, we adjusted for baseline covariates. Thus, although ST-segment elevation myocardial infarction in women occurs less often than in men, women experience less favorable acute outcomes. Our data suggest the need for better education of patients and physicians to recognize and treat younger women more aggressively. The ongoing National Heart, Lung, and Blood Institute‚Äìsponsored Variation In Recovery: Role of Gender on Outcomes of Young AMI Patients (VIRGO) trial aims to study this gender differences. ST-Segment Elevation Myocardial Infarction Outcomes in the Very Young Although we observed comparable quality of care, performance measures, and better outcomes in the young compared with the old, the same was not true for the very young (’Ö35 years), in whom there was a 32% lower odds of the ‚Äúall or none‚Äù composite performance measure, 37% lower odds of receiving a stent, and a trend (P œ≠ .06) toward an increase in the odds of in-hospital death when compared with the patients aged 36 to 45 years. Although this is concerning, a number of possible reasons could account for higher in-hospital death: 1) Very young patients are more likely to attribute symptoms to a noncardiac cause and to present late; 2) a subset of these admissions were likely related to illicit drug use,9,10 when physicians attributed chest pain and electrocardiographic changes due to coronary vasospasm rather than plaque rupture; moreover, longstanding illicit drug use has been shown to accelerate the process of atherosclerosis or cause cardiomyopathy with subsequent worse prognosis; and 3) physicians also are less likely to attribute chest pain to a myocardial infarction in the very young (reflected in the longer door-to-balloon time and lower proportion of compliance with evidencebased therapies in the very young). These data strongly urge the need for better education of not only patients but also physicians to recognize and treat very young patients more aggressively. STUDY LIMITATIONS Our data are from a prospective registry. However, this is still the largest series on young patients with ST-segment elevation myocardial infarction. Our data included only hospitalized patients with ST-segment elevation myocardial infarction and did not account for out-of-hospital deaths. In addition, we were able to assess only in-hospital outcomes. In our subgroup analysis, we did not have data on illicit drug use in the very young cohort. In addition, given the lack of angiographic data, we were not able to adjust our analysis for severity and extent of coronary artery disease. Residual measured and unmeasured confounding may account for some or all of these findings. Moreover, disparity in the use of evidence-based therapies could merely be a reflection of physician assessment of increased risk of adverse effect for a particular patient subgroup or a reflection of patient choice, which is not captured in the database. However, Bangalore et al ST-Segment Elevation Myocardial Infarction in the Young disparities in the door-to-balloon time and door-to-thrombolytic times are likely due to treatment biases. In addition, relatively few patients were treated with thrombolytic therapy among participating hospitals, which may limit the generalizability of these findings. CONCLUSIONS In this largest series to date, younger patients with STsegment elevation myocardial infarction were more likely to be men, black or Hispanic, or smokers, whose quality of care and in-hospital outcomes were similar to or better than that of their older counterparts. There was significantly lower quality of care as assessed by performance measures and significantly worse outcomes in young women (vs young men) and the very young (’Ö35 vs 36-45 years), suggesting the need for increasing awareness among these subgroups and the physicians taking care of such patients. References 1. Barbash GI, White HD, Modan M, et al. Acute myocardial infarction in the young‚Äìthe role of smoking. The investigators of the international tissue plasminogen activator/streptokinase mortality trial. Eur Heart J. 1995;16:313-316. 2. Chua SK, Hung HF, Shyu KG, et al. Acute st-elevation myocardial infarction in young patients: 15 years of experience in a single center. Clin Cardiol. 2010;33:140-148. 3. Doughty M, Mehta R, Bruckman D, et al. Acute myocardial infarction in the young‚Äîthe University of Michigan experience. Am Heart J. 2002;143:56-62. 4. Moccetti T, Malacrida R, Pasotti E, et al. Epidemiologic variables and outcome of 1972 young patients with acute myocardial infarction. Data from the gissi-2 database. Investigators of the gruppo italiano per lo studio della sopravvivenza nell‚Äôinfarto miocardico (gissi-2). Arch Intern Med. 1997;157:865-869. 5. McManus DD, Piacentine SM, Lessard D, et al. Thirty-year (1975 to 2005) trends in the incidence rates, clinical features, treatment practices, and short-term outcomes of patients œΩ55 years of age hospitalized with an initial acute myocardial infarction. The Am J Cardiol. 2011;108:477-482. 6. Dolder MA, Oliver MF. Myocardial infarction in young men. Study of risk factors in nine countries. Br Heart J. 1975;37:493-503. 7. Hoit BD, Gilpin EA, Henning H, et al. Myocardial infarction in young patients: An analysis by age subsets. Circulation. 1986;74:712-721. 8. Jneid H, Fonarow GC, Cannon CP, et al. Sex differences in medical care and early death after acute myocardial infarction. Circulation. 2008;118:2803-2810. 9. Kanitz MG, Giovannucci SJ, Jones JS, Mott M. Myocardial infarction in young adults: risk factors and clinical features. J Emerg Med. 1996;14:139-145. 10. Schoenenberger AW, Radovanovic D, Stauffer JC, et al. Acute coronary syndromes in young patients: presentation, treatment and outcome. Int J Cardiol. 2011;148:300-304. 11. Hochman JS, Tamis JE, Thompson TD, et al. Sex, clinical presentation, and outcome in patients with acute coronary syndromes. Global use of strategies to open occluded coronary arteries in acute coronary syndromes iib investigators. N Engl J Med. 1999;341:226-232. 12. Vaccarino V, Krumholz HM, Berkman LF, Horwitz RI. Sex differences in mortality after myocardial infarction. Is there evidence for an increased risk for women? Circulation. 1995;91:1861-1871. 1009 13. LaBresh KA, Ellrodt AG, Gliklich R, Liljestrand J, Peto R. Get with the guidelines for cardiovascular secondary prevention: pilot results. Arch Intern Med. 2004;164:203-209. 14. LaBresh KA, Gliklich R, Liljestrand J, Peto R, Ellrodt AG. Using ‚Äúget with the guidelines‚Äù to improve cardiovascular secondary prevention. Jt Comm J Qual Saf. 2003;29:539-550. 15. Glover MU, Kuber MT, Warren SE, Vieweg WV. Myocardial infarction before age 36: Risk factor and arteriographic analysis. Am J Cardiol. 1982;49:1600-1603. 16. Warren SE, Thompson SI, Vieweg WV. Historic and angiographic features of young adults surviving myocardial infarction. Chest. 1979; 75:667-670. 17. Waters DD, Halphen C, Theroux P, David PR, Mizgala HF. Coronary artery disease in young women: clinical and angiographic features and correlation with risk factors. Am J Cardiol. 1978;42:41-47. 18. Wolfe MW, Vacek JL. Myocardial infarction in the young. Angiographic features and risk factor analysis of patients with myocardial infarction at or before the age of 35 years. Chest. 1988;94:926-930. 19. Zimmerman FH, Cameron A, Fisher LD, Ng G. Myocardial infarction in young adults: angiographic characterization, risk factors and prognosis (coronary artery surgery study registry). J Am Coll Cardiol. 1995;26:654-661. 20. Chouhan L, Hajar HA, Pomposiello JC. Comparison of thrombolytic therapy for acute myocardial infarction in patients aged œΩ 35 and œæ 55 years. Am J Cardiol. 1993;71:157-159. 21. Simonson E, Berman R. Myocardial infarction in young people. Experience in U.S.S.R. Am Heart J. 1972;84:814-822. 22. Barakat K, Wilkinson P, Suliman A, Ranjadayalan K, Timmis A. Acute myocardial infarction in women: contribution of treatment variables to adverse outcome. Am Heart J. 2000;140:740-746. 23. Barron HV, Bowlby LJ, Breen T, et al. Use of reperfusion therapy for acute myocardial infarction in the united states: data from the national registry of myocardial infarction 2. Circulation. 1998;97:1150-1156. 24. Vaccarino V, Rathore SS, Wenger NK, et al. Sex and racial differences in the management of acute myocardial infarction, 1994 through 2002. N Engl J Med. 2005;353:671-682. 25. MacIntyre K, Stewart S, Capewell S, et al. Gender and survival: a population-based study of 201,114 men and women following a first acute myocardial infarction. J Am Coll Cardiol. 2001;38:729-735. 26. Malacrida R, Genoni M, Maggioni AP, et al. A comparison of the early outcome of acute myocardial infarction in women and men. The third international study of infarct survival collaborative group. N Engl J Med. 1998;338:8-14. Conflicts of Interest: Sripal Bangalore: advisory board for Daichii Sankyo. Gregg C. Fonarow: consulting for Novartis Pfizer. Eric Peterson: research funding from BMS/Sanofi, Eli Lilly, Merck, and Ortho McNeil Pharmaceuticals. Anne S. Hellkamp: none. Adrian F. Hernandez: none. Warren Laskey: none. Frank Peacock: scientific advisory board for Abbott, Beckman-Coulter, Biosite, and The Medicines Co; research grants for Abbott, BAS, Biosite, Brahms, Nanosphere, EKR, and The Medicines Co; speakers bureau for Abbott, Biosite, and The Medicines Co; ownership interest in Vital Sensors. Christopher P. Cannon: research grants/support from Accumetrics, AstraZeneca, Glaxo Smith Kline, Merck, and Takeda; advisory board (but funds donated to charity) for Bristol-Myers Squibb/ Sanofi, Novartis, and Alnylam; honorarium for development of independent educational symposia for Pfizer and AstraZeneca; clinical advisor for and equity in Automedics Medical Systems. Lee Schwamm: Chair, GWTG steering committee (unpaid). Deepak L. Bhatt: research grants from Astra Zeneca, Bristol-Myers Squibb, Eisai, Ethicon, Medtronic, Sanofi Aventis, The Medicines Company, and Amarin. Age at First Venous Thromboembolism and Risk of Recurrence A Prospective Cohort Study Lisbeth Eischer, MD, Sabine Eichinger, MD, and Paul A. Kyrle, MD Abstract: Risk of Ô¨Årst venous thromboembolism (VTE) increases with age. We investigated whether age is related to the risk of recurrent VTE. We followed 694 patients for a mean of 40 months after Ô¨Årst unprovoked proximal deep vein thrombosis (DVT) and/or pulmonary embolism (PE) and withdrawal of anticoagulants. We excluded patients with natural inhibitor deÔ¨Åciency, lupus anticoagulant, or cancer; patients who required indeÔ¨Ånite anticoagulation; pregnant women; and women who had VTE related to female hormone use. The endpoint was symptomatic recurrent VTE. VTE recurred in 152 patients (22%). The adjusted hazard ratio (HR) of recurrence for a 10-year increase in age was 0.94 (95% conÔ¨Ådence interval ECI^, 0.82Y1.08; p = 0.4). Compared with patients aged younger than 47 years (1st tercile of patient population) no signiÔ¨Åcant increase in the risk of recurrent VTE was found among patients 47-61 years old (2nd tercile) or patients older than 61 years (3rd tercile) (EHR, 1.25; 95% CI, 0.78Y2.01^ and EHR, 0.93l; 95% CI, 0.56Y1.53^, respectively). Compared to patients older than 80 years, the HR of recurrence among patients younger than 50 years was 1.11 (95% CI, 0.11Y10.3; p = 0.9). After 5 years, probability of recurrence was 32% (95% CI, 24%Y40%) among patients aged less than 47 years; 21% (95% CI, 15%Y28%) among patients 47Y61 years old; and 33% (95% CI, 24%Y42%) among patients older than 61 years ( p = 0.5). Our results show that in patients with Ô¨Årst unprovoked proximal DVT and/or PE, risk of recurrence is not related to age at Ô¨Årst VTE. Regardless of age, these patients have a high risk of recurrence. (Medicine 2009;88: 366Y370) Abbreviations: CI = conÔ¨Ådence interval, CT = computed tomography, DVT = deep vein thrombosis, HR = hazard ratio, OR = odds ratio, PE = pulmonary embolism, VTE = venous thromboembolism. INTRODUCTION T here is compelling evidence that the risk of a Ô¨Årst venous thromboembolism (VTE) exponentially increases with age.1,6,20,22 In a study from the United States, incidence rates of VTE among individuals aged 60Y70 years were approximately 200Y300 per 100,000 per year.1 Likewise, in a pro- spective study of almost 20,000 individuals, the risk of a Ô¨Årst VTE was more than 15-fold increased among individuals aged 85 years or more compared with individuals aged younger than 55 years.21 VTE is a chronic disease that tends to recur. The risk of recurrence depends on the severity and on the number of risk factors in an individual patient. Over the past years, the natural course of VTE has been studied intensively, and several clinical and biochemical risk factors of VTE recurrence have been identiÔ¨Åed.12 These include prior VTE, absence of a temporary risk factor (for instance, surgery, trauma, pregnancy, or estrogencontaining contraceptives), proximal deep vein thrombosis (DVT), symptomatic pulmonary embolism (PE), male sex, cancer, high levels of factor VIII or factor IX, lupus anticoagulant, and hyperhomocysteinemia.12 Data on the effect of age on the risk of recurrent VTE are scarce and conÔ¨Çicting. In a study from Norway, age did not affect the recurrence risk.7 Beyth et al3 followed 124 patients with DVT for 6Y8 years and found a more than 4-fold increased risk of recurrence among patients younger than 65 years. Using a computer registry of more than 36,000 patients, White and colleagues23 found a 6-month incidence of rehospitalization for recurrent VTE of about 6%. Age at index VTE was associated with an odds ratio (OR) of recurrence of 0.85 for each 10-year increase. In contrast, in a study from Sweden more patients aged at least 60 years had a recurrent VTE than the younger patients, with an OR of 1.36.17 In a retrospective population-based study from the United States with a total follow-up of more than 10,000 person-years, the risk of recurrent VTE was increased by 17% per decade of age at incident VTE.8 Likewise, in a study from Italy with a median observation time of 50 months, advanced age was a risk factor of recurrence with an adjusted HR of 1.14 for every 10-year increase in age.15 In a study from Canada, age greater than 65 years conferred a more than 2-fold risk of recurrence among women, but was not associated with an increased recurrence risk among men.16 We followed 694 patients who had had a Ô¨Årst episode of unprovoked proximal DVT and/or PE to evaluate the relationship between age and risk of recurrence. PATIENTS AND METHODS From Department of Medicine I, Medical University of Vienna; and Karl Landsteiner Institute of Thrombosis Research, Vienna, Austria. ¬® sterreichische This work was supported by the Jubila¬®umsfonds of the O National Bank, the Medizinisch-Wissenschaftlicher Fonds des Bu¬®rgermeisters der Bundeshauptstadt Wien, and the Vienna Insurance Group, Vienna, Austria. Received June 1, 2009, and in revised form September 18, 2009. Accepted for publication September 21, 2009. Reprints: Paul Kyrle, MD, Klinik fu¬®r Innere Medizin I, Allgemeines Krankenhaus Wien/Universita¬®tskliniken, Wa¬®hringer Gu¬®rtel 18Y20, 1090-Vienna, Austria (e-mail: paul.kyrle)meduniwien.ac.at). Copyright * 2009 by Lippincott Williams & Wilkins ISSN: 0025-7974 DOI: 10.1097/MD.0b013e3181c29e31 366 www.md-journal.com Patients and Study Design The current study was performed within the framework of the Austrian Study on Recurrent Venous Thromboembolism (AUREC), a large prospective ongoing multicenter cohort study. Patients were included when they were older than 18 years of age; had symptomatic VTE, conÔ¨Årmed by objective investigations; and were treated with oral anticoagulants for 3Y18 months. Exclusion criteria were a previous episode of VTE; VTE related to surgery, trauma, pregnancy, or female hormone intake; DVT distal of the trifurcation of calf veins; upper extremity DVT; deÔ¨Åciency of antithrombin, protein C, or protein S; presence of Medicine & Volume 88, Number 6, November 2009 Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine & Volume 88, Number 6, November 2009 Age and Risk of Recurrent Venous Thromboembolism lupus anticoagulant; cancer at time of enrollment; homozygosity or double heterozygosity for factor V Leiden and/or factor II G20210A, or requirement for long-term antithrombotic treatment. Patients entered the study at the time of discontinuation of oral anticoagulation. At study entry, a detailed history and a physical examination were performed. Three weeks after withdrawal of anticoagulation, patients were screened for biochemical risk factors of VTE including antithrombin, protein C or protein S, lupus anticoagulant, factor V Leiden and prothrombin G20210A. Patients were observed at 6-month intervals for the Ô¨Årst year and once a year thereafter. They were given detailed written information on the symptoms of VTE and were asked to report immediately if such symptoms occurred. A medical history was obtained at each visit. Female patients were strongly discouraged from intake of estrogen-containing oral contraceptives or hormone replacement therapy. Patients received thromboprophylaxis with a low-molecular-weight heparin during high-risk situations such as prolonged immobilization or long-haul air travel. The ethics committee of the Medical University of Vienna approved the study, and all patients gave written informed consent before inclusion in the study. Diagnosis of VTE The diagnosis of VTE was established by a positive Ô¨Ånding on venography or color duplex sonography (in case of proximal DVT). If venography was applied, at least 1 of the following direct or indirect criteria had to be fulÔ¨Ålled: a constant Ô¨Ålling defect seen on 2 views; an abrupt discontinuation of the contrastÔ¨Ålled vessel at a constant level of the vein; and the entire deep vein system without external compression failed to Ô¨Åll, with or without venous Ô¨Çow through collateral veins. Diagnostic criteria for color duplex ultrasonography were the following: visualization of an intraluminal thrombus in a deep vein, incomplete or absent compressibility, and lack of Ô¨Çow spontaneously and after distal manipulation. A diagnosis of PE was considered in case of typical symptoms, that is, chest pain, dyspnea, cough, hemoptysis, and/ or syncope. The diagnosis of PE was conÔ¨Årmed either by ventilation-perfusion lung scanning according to the criteria of the Prospective Investigation of Pulmonary Embolism Diagnosis,14 or by spiral computed tomography (CT) scan dem- onstrating 1 or several low-attenuation areas that partly or completely Ô¨Ålled the lumen of an opaciÔ¨Åed vessel. Study Endpoints The endpoint of the study was recurrent symptomatic DVT conÔ¨Årmed by venography or color duplex sonography, or recurrent symptomatic PE conÔ¨Årmed by ventilation-perfusion lung scanning and/or spiral CT scan according to the aforementioned criteria. Recurrent DVT was diagnosed if the patient had a thrombus in another deep vein in the extremity involved in the previous event, a thrombus in the opposite extremity, or a thrombus in the same venous system with a proximal extension of the thrombus (if the upper limit of the original thrombus had been visible) or the presence of a constant Ô¨Ålling defect surrounded by contrast medium (if the original thrombus had not been visible). Laboratory Analysis Venous blood from fasting patients was collected into 1:10 volume of 0.11 mM trisodium citrate and immediately centrifuged for 20 minutes at 2000 √Ç g. Aliquots of plasma were stored at j80 -C until analysis. Screening for factor V Leiden and prothrombin G20210A and measurement of antithrombin, protein C, and protein S were carried out by standard methods. The diagnosis of lupus anticoagulant was based on criteria of the International Society on Thrombosis and Haemostasis.4 Statistical Analysis Categorical data were compared among groups using contingency-table analyses (the chi-square test). Continuous data (presented as means T standard deviation) were compared using the Mann-Whitney U tests or Kruskal-Wallis-H-tests. All p values were 2-tailed. Survival-time methods were used to analyze the time to recurrence among patients with a subsequent episode of VTE (uncensored observation) or the duration of follow-up among patients without recurrence (censored observations).9 The probability of recurrence was estimated according to the method of Kaplan and Meier.10 Data on patients who left the study or who were lost to follow-up were censored at the time of withdrawal. To test for homogeneity between strata, we applied the log-rank test. Univariate and multivariate Cox proportional hazards models were used to analyze the association between age and the risk of recurrent VTE. Analyses were TABLE 1. Characteristics of 694 Patients With First Unprovoked Proximal Deep Vein Thrombosis and/or Pulmonary Embolism, by Age 1st Tercile (G47 yr) (n = 233) No. (%) 2nd Tercile (47Y61 yr) (n = 224) No. (%) 3rd Tercile (961 yr) (n = 237) No. (%) 37 T 7 54 T 4 69 T 6 166 (71) 67 (29) 164 (73) 60 (27) 139 (59) 98 (41) G0.001 107 (46) 63 (27) 63 (27) 7.6 T 2.8 44 T 40 59 (26) 11 (5) 126 (56) 42 (19) 56 (25) 7.5 T 2.6 45 T 43 68 (31) 9 (4) 117 (49) 55 (23) 65 (28) 7.1 T 2.3 32 T 30 47 (20) 13 (6) 0.2 Age* (yr) Sex Men Women Localization DVT only PE + DVT Isolated PE Anticoagulants* (mo) Observation time* (mo) Factor V Leiden Factor II G20210A p Value 0.1 0.004 0.05 0.7 *Mean T standard deviation. * 2009 Lippincott Williams & Wilkins www.md-journal.com Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 367 Medicine Eischer et al & Volume 88, Number 6, November 2009 TABLE 2. Hazard Ratios of Recurrence, by Age at First Venous Thromboembolism Hazard Ratio (95% CI) Age No. of Patients No. of Recurrences Univariate Multivariate* 233 224 237 55 46 51 1 0.85 (0.58Y1.26) 1.10 (0.75Y1.61) 1 1.25 (0.78Y2.01) 0.93 (0.56Y1.53) 1st tercile (G47 yr)‚Ä† 2nd tercile (47Y61 yr) 3rd tercile (961 yr) *Adjusted for sex, factor V Leiden and factor II G20210A, observation time, and localization of index VTE. ‚Ä†Reference group. adjusted for sex, presence or absence of factor V Leiden or factor II G20210A, observation time, and localization of index VTE. All computations were performed with the use of SPSS software, version 15.0 (SPSS Inc, Chicago, IL). RESULTS Study Population A total of 694 patients with a Ô¨Årst unprovoked proximal DVT and/or PE were followed for a mean of 40 months after discontinuation of oral anticoagulants. Baseline characteristics according to 3 different age-groups are shown in Table 1. The proportion of female patients was lower in the younger agegroups, and the observation period was shorter in the older patients. Of these 694 patients, 165 left the study because of a new diagnosis of cancer (n = 16), pregnancy (n = 5), antithrombotic therapy for reasons other than VTE, including atrial Ô¨Åbrillation and ischemic heart disease (n = 113); or were lost to follow-up (n = 31). Sixteen patients died for reasons other than VTE: cardiac failure (n = 8), cancer (n = 4), cerebral hemorrhage (n = 1), ischemic stroke (n = 1), aortic dissection (n = 1), or suicide (n = 1). Patients were followed until the time of exclusion or death, when data were censored. Recurrence of VTE VTE recurred in 152 (22%) patients. DVT alone occurred in 83 (29 distal and 54 proximal) and PE with or without DVT in 69 patients; 3 of them were fatal. Recurrence was secondary to surgery or trauma in 11 patients (7%). Patients with recurrence were predominantly male (26% vs. 14%, p = 0.001). The proportion of carriers of factor V Leiden (31% vs. 24%) or factor II G20210A (6% vs. 5%) was similar in patients with and without recurrence. Age and Risk of Recurrent VTE Patients with and without recurrent VTE were of similar age at Ô¨Årst VTE (53 yr vs. 54 yr, p = 0.5). When we analyzed age as a continuous variable in a Cox proportional hazard model, we found a hazard ratio (HR) of recurrence of 0.94 (95% conÔ¨Ådence interval ECI^, 0.82Y1.08; p = 0.4) for each 10-year increase in age after adjustment for sex, factor V Leiden and factor II G20210A, observation time, and localization of index VTE. Table 2 shows the HR of recurrence according to 3 different categories (terciles) of patient age. Compared with patients younger than 47 years, no signiÔ¨Åcant increase in the risk of recurrent VTE was found among older patient groups, neither in the univariate analysis, nor after adjustment. When we compared the very old patients, that is, those aged older than 80 years, with patients younger than 50 years, the adjusted risk of recurrence was 1.11 (95% CI, 0.11Y10.3; p = 0.9). After 5 years, the cumulative probability of recurrence was 32% (95% CI, 24%Y40%) among patients aged less than 47 years compared with 21% (95% CI, 15%Y28%) among patients 47Y61 years old (second tercile) and 33% (95% CI, 24%Y42%) among those older than 61 years (third tercile, p = 0.5, Figure 1). FIGURE 1. Kaplan-Meier estimates of the probability of recurrence according to age in terciles. The cumulative probability of recurrence did not differ between the 3 groups (p = 0.5 by the log-rank test). 368 www.md-journal.com * 2009 Lippincott Williams & Wilkins Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine & Volume 88, Number 6, November 2009 We next assessed the effect of age on the risk of recurrent VTE separately for men and women. We found that advancing age was related to recurrent VTE neither in men nor in women. The adjusted HR of recurrence was 0.96 (95% CI, 0.81Y1.13; p = 0.6) in the 469 male patients and 0.80 (95% CI, 0.50Y1.25; p = 0.3) in the 225 female patients for each 10-year increase in age. The overall risk of recurrence was higher in men than in women (HR, 1.8; 95% CI, 1.2Y2.7; p = 0.003). DISCUSSION The principal Ô¨Ånding of the current study is that in patients with a Ô¨Årst episode of unprovoked proximal DVT and/or PE, the risk of recurrence was not related to age at Ô¨Årst VTE. The HR of recurrent VTE was 0.94 (95% CI, 0.82Y1.08) for each 10-year increase in age after adjustment for sex, factor V Leiden and factor II G20210A, observation time, and localization of index VTE. After 5 years, the cumulative probability of recurrence was 32% (95% CI, 24%Y40%) among patients aged less than 47 years compared with 21% (95% CI, 15%Y28%) and 33% (95% CI, 24%Y42%) among patients 47Y61 years old and patients older than 61 years, respectively. Even the very old patients, those over the age of 80 years, had the same risk of recurrent VTE as patients younger than 50 years. Reports on the effect of age on the risk of recurrent VTE are conÔ¨Çicting.3,7,8,15Y17,23 Age was not a risk factor of recurrence in a study from Norway.7 In contrast, some investigators reported a lower risk of recurrent VTE among older patients,3,23 whereas others found an association between advancing age and recurrence.8,15Y17 Discrepancies between the results of these studies and the current study can be explained by differences in patient selection and study design. In 1 study the number of patients was small,3 other studies had a shorter follow-up,16,23 and 2 studies were retrospective.8,23 Several investigators included low-risk patients such as patients with secondary VTE or VTE associated with female hormone intake,3,7,8,15,16,23 distal DVT,3,7,8,17,23 or upper extremity DVT.7 Cancer patients were included in 4 studies.3,7,8,23 Our 694 patients, who were prospectively followed for a mean of 40 months, had an objectively documented Ô¨Årst episode of unprovoked DVT involving the proximal veins of the leg and/or symptomatic PE, and cancer patients were excluded. In routine clinical practice, age at Ô¨Årst VTE is usually taken into consideration when a patient is counseled regarding duration of anticoagulant prophylaxis. Some clinicians are reluctant to treat elderly patients with anticoagulants for a longer period of time because they consider risk of bleeding rather than risk of recurrent VTE as the determinant for duration of anticoagulation. Others, in contrast, are hesitant to recommend extended anticoagulation to younger patients because they assume that these patients may have a low recurrence risk. Many younger individuals dislike long-term anticoagulation simply because of the prospect of a long-time medical treatment. Extended anticoagulation is troublesome because of the necessity of regular laboratory monitoring, dietary restrictions, and inconveniences in professional life. According to our Ô¨Åndings, the risk of recurrence is comparable between younger and older patients. Thus, age at Ô¨Årst VTE should not matter when determining how long thrombosis patients should receive anticoagulation, provided their risk of bleeding is low. Another important observation was that, regardless of age at Ô¨Årst VTE, the risk of recurrence in our cohort was as high as 29% with a lower 95% CI of 24% 5 years after incident VTE and further increased over time. In another series of patients with unprovoked VTE, the recurrence risk was even higher. After 5 years, Prandoni et al15 recorded a likelihood of recurrence of * 2009 Lippincott Williams & Wilkins Age and Risk of Recurrent Venous Thromboembolism 41% (95% CI, 37%Y45%) in patients with idiopathic proximal or distal DVT and/or PE. In a study from England, the recurrence rate among patients with unprecipitated VTE was almost 20% just 2 years after cessation of anticoagulation.2 What are the potential consequences of these Ô¨Åndings for daily clinical practice? Anticoagulant therapy at therapeutic dose effectively protects from recurrent VTE.11,18 Patients will, however, beneÔ¨Åt from long-term anticoagulation only if the risk of major bleeding is outweighed by the likelihood of recurrent VTE. The annual risk of intracranial bleeding during anticoagulant therapy is 0.2%.19 In a large study from Italy, the incidence of fatal bleeding among patients treated with vitamin K antagonists was 0.25 per 100 patient-years.13 On the other hand, the case fatality rate after discontinuing anticoagulant therapy for VTE was 4%Y9%.5 The case-fatality rates of recurrent VTE were even higher in 2 other studies: greater than 10% in a study from Italy and 12% in a study from Norway.7,15 On the basis of these numbers, it is apparent that VTE patients will beneÔ¨Åt from long-term anticoagulation if their recurrence risk exceeds 5% per year, as is indeed the case in patients with a Ô¨Årst unprovoked proximal DVT and/or PE. This concept of prolonged anticoagulation could be particularly important for younger patients, as their risk of bleeding during anticoagulation is usually low, and sequelae of VTEVsuch as worsening of the postthrombotic syndrome if DVT recurs in the same leg, or development of chronic thromboembolic pulmonary hypertensionVcould be prevented. Some strengths and weaknesses of the current study have to be addressed. Our study population was homogenous and consisted of patients with unprovoked proximal VTE and/or PE only. Patients were followed over a long time, and assessment for the presence of molecular thrombophilia was performed in all patients. Our Ô¨Åndings, however, cannot be extrapolated to patients with distal DVT or to patients with DVT related to a removable risk factor. We cannot comment on patients with antithrombin, protein C or protein S deÔ¨Åciency, lupus anticoagulant, combined heterozygosity for factor V Leiden and factor II G20210A, or homozygous factor V Leiden. At our institution, patients with any of these clotting abnormalities are considered to be at high risk of recurrence and receive indeÔ¨Ånite anticoagulation. In conclusion, among patients with a Ô¨Årst unprovoked proximal DVT and/or PE, a relationship between the likelihood of recurrence and age at Ô¨Årst VTE was not detectable. Patients with an unprovoked proximal DVT and/or PE have a high risk of recurrence regardless of age at presentation, and may beneÔ¨Åt from long-term anticoagulation already after their Ô¨Årst VTE episode. REFERENCES 1. Anderson FA Jr, Wheeler HB, Goldberg RJ, Hosmer DW, Patwardhan NA, Jovanovic B, Forcier A, Dalen JE. A population-based perspective of the hospital incidence and case-fatality rates of deep vein thrombosis and pulmonary embolism. The Worcester DVT study. Arch Intern Med. 1991;151:933Y938. 2. Baglin T, Luddington R, Brown K, Baglin C. Incidence of recurrent venous thromboembolism in relation to clinical and thrombophilic risk factors: prospective cohort study. Lancet. 2003;362:523Y526. 3. Beyth RJ, Cohen AM, Landefeld CS. Long-term outcomes of deep-vein thrombosis. Arch Intern Med. 1995;155:1031Y1037. 4. Brandt JT, Triplett DA, Alving B, Scharrer I. Criteria for the diagnosis of lupus anticoagulants: an update. Thromb Haemost. 1995;74:1185Y1190. 5. Douketis JD, Gu CS, Schulman S, Ghirarduzzi A, Pengo V, Prandoni P. The risk for fatal pulmonary embolism after discontinuing www.md-journal.com Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 369 Eischer et al anticoagulant therapy for venous thromboembolism. Ann Intern Med. 2007;147:766Y774. 6. Hansson PO, Welin L, Tibblin G, Eriksson H. Deep vein thrombosis and pulmonary embolism in the general population. The study of men born in 1913. Arch Intern Med. 1997;157:1665Y1670. 7. Hansson PO, Sorbo J, Eriksson H. Recurrent venous thromboembolism after deep vein thrombosis: incidence and risk factors. Arch Intern Med. 2000;160:769Y774. 8. Heit JA, Silverstein MD, Mohr DN, Petterson TM, Lohse CM, O_Fallon WM, Melton LJ III. Predictors of recurrence after deep vein thrombosis and pulmonary embolism: a population-based cohort study. Arch Intern Med. 2000;160:761Y768. 9. Kalbfleisch JD, Prentice RL. The Statistical Analysis of Failure Time Data (Wiley Series in Probability and Statistics). Wiley-Interscience; 1980. 10. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53:457Y481. 11. Kearon C, Gent M, Hirsh J, Weitz J, Kovacs MJ, Anderson DR, Turpie AG, Green D, Ginsberg JS, Wells P, MacKinnon B, Julian JA. A comparison of three months of anticoagulation with extended anticoagulation for a first episode of idiopathic venous thromboembolism. N Engl J Med. 1999;340:901Y907. 12. Kyrle PA, Eichinger S. Deep vein thrombosis. Lancet. 2005;365: 1163Y1174. 13. Palareti G, Leali N, Coccheri S, Poggi M, Manotti C, D_Angelo A, Pengo V, Erba N, Moia M, Ciavarella N, Devoto G, Berrettini M, Musolesi S, for the Italian Study on Complications of Oral Anticoagulant Therapy. Bleeding complications of oral anticoagulant treatment: an inception-cohort, prospective collaborative study (ISCOAT). Lancet. 1996;348:423Y428. 14. The PIOPED Investigators. Value of the ventilation/perfusion scan in acute pulmonary embolism. JAMA. 1990;263:2753Y2759. 15. Prandoni P, Noventa F, Ghirarduzzi A, Pengo V, Bernardi E, Pesavento R, Lotti M, Tormene D, Simioni P, Pagnan A. The risk of 370 www.md-journal.com Medicine & Volume 88, Number 6, November 2009 recurrent venous thromboembolism after discontinuing anticoagulation in patients with acute proximal deep vein thrombosis or pulmonary embolism. A prospective cohort study in 1,626 patients. Haematologica. 2007;92:199Y205. 16. Rodger MA, Kahn SR, Wells PS, Anderson DA, Chagnon I, Le Gal G, Solymoss S, Crowther M, Perrier A, White R, Vickars L, Ramsay T, Betancourt MT, Kovacs MJ. Identifying unprovoked thromboembolism patients at low risk for recurrence who can discontinue anticoagulant therapy. CMAJ. 2008;179:417Y426. 17. Schulman S, Lindmarker P, Holmstrom M, Larfars G, Carlsson A, Nicol P, Svensson E, Ljungberg B, Viering S, Nordlander S, Leijd B, Jahed K, Hjorth M, Linder O, Beckman M. Post-thrombotic syndrome, recurrence, and death 10 years after the first episode of venous thromboembolism treated with warfarin for 6 weeks or 6 months. J Thromb Haemost. 2006;4:734Y742. 18. Schulman S, Granqvist S, Holmstrom M, Carlsson A, Lindmarker P, Nicol P, Eklund SG, Nordlander S, Larfars G, Leijd B, Linder O, Loogna E. The duration of oral anticoagulant therapy after a second episode of venous thromboembolism. The Duration of Anticoagulation Trial Study Group. N Engl J Med. 1997;336:393Y398. 19. Schulman S, Beyth RJ, Kearon C, Levine MN. Hemorrhagic complications of anticoagulant treatment. Chest. 2008;133:257Y298. 20. Silverstein MD, Heit JA, Mohr DN, Petterson TM, O_Fallon WM, Melton LJ III. Trends in the incidence of deep vein thrombosis and pulmonary embolism: a 25-year population-based study. Arch Intern Med. 1998;158:585Y593. 21. Tsai AW, Cushman M, Rosamond WD, Heckbert SR, Polak JF, Folsom AR. Cardiovascular risk factors and venous thromboembolism incidence. Arch Intern Med. 2002;162:1182Y1189. 22. White RH. The epidemiology of venous thromboembolism. Circulation. 2003;107:14Y18. 23. White RH, Zhou H, Romano PS. Length of hospital stay for treatment of deep venous thrombosis and the incidence of recurrent thromboembolism. Arch Intern Med. 1998;158:1005Y1010. * 2009 Lippincott Williams & Wilkins Copyright @ 2009 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. EBIOM-00228; No of Pages 8 EBioMedicine xxx (2015) xxx‚Äìxxx Contents lists available at ScienceDirect EBioMedicine journal homepage: www.ebiomedicine.com Research Paper A Genome-wide Association Study Provides Evidence of Sex-speciÔ¨Åc Involvement of Chr1p35.1 (ZSCAN20-TLR12P) and Chr8p23.1 (HMGB1P46) With Diabetic Neuropathic Pain Weihua Meng a,‚Åé, Harshal A. Deshmukh a, Louise A. Donnelly b, Wellcome Trust Case Control Consortium 2 (WTCCC2), Surrogate markers for Micro- and Macro-vascular hard endpoints for Innovative diabetes Tools (SUMMIT) study group, Nicola Torrance a, Helen M. Colhoun a, Colin N.A. Palmer b, Blair H. Smith a a b Division of Population Health Sciences, Medical Research Institute, Ninewells Hospital and School of Medicine, University of Dundee, Dundee DD2 4BF, UK Centre for Pharmacogenetics and Pharmacogenomics, Medical Research Institute, Ninewells Hospital and School of Medicine, University of Dundee, Dundee DD1 9SY, UK a r t i c l e i n f o Article history: Received 6 July 2015 Received in revised form 30 July 2015 Accepted 1 August 2015 Available online xxxx Keywords: Neuropathic pain GWAS Heritability Sex-speciÔ¨Åc a b s t r a c t Neuropathic pain is deÔ¨Åned as pain arising as a direct consequence of a lesion or a disease affecting the somatosensory system and it affects around 1 in 4 diabetic patients in the UK. The purpose of this genome-wide association study (GWAS) was to identify genetic contributors to this disorder. Cases of neuropathic pain were deÔ¨Åned as diabetic patients with a multiple prescription history of at least one of Ô¨Åve drugs speciÔ¨Åcally indicated for the treatment of neuropathic pain. Controls were diabetic individuals who were not prescribed any of these drugs, nor amitriptyline, carbamazepine, or nortriptyline. Overall, 961 diabetic neuropathic pain cases and 3260 diabetic controls in the Genetics of Diabetes Audit and Research Tayside (GoDARTS) cohort were identiÔ¨Åed. We found a cluster in the Chr1p35.1 (ZSCAN20-TLR12P) with a lowest P value of 2.74 √ó 10‚àí7 at rs71647933 in females and a cluster in the Chr8p23.1, next to HMGB1P46 with a lowest P value of 8.02 √ó 10‚àí7 at rs6986153 in males. SexspeciÔ¨Åc narrow sense heritability was higher in males (30.0%) than in females (14.7%). This GWAS on diabetic neuropathic pain provides evidence for the sex-speciÔ¨Åc involvement of Chr1p35.1 (ZSCAN20-TLR12P) and Chr8p23.1 (HMGB1P46) with the disorder, indicating the need for further research. ¬© 2015 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction Neuropathic pain is deÔ¨Åned by the International Association for the Study of Pain as pain arising as a direct consequence of a lesion or a disease affecting the somatosensory system (Jensen et al., 2011). The prevalence of neuropathic pain is estimated to be around 7% in a general population while in a diabetic population around 1 in 4 patients will suffer from this disorder (van Hecke et al., 2014; Abbott et al., 2011). The current treatment of neuropathic pain is far from satisfactory, with fewer than 30% of patients achieving satisfactory relief of diabetic neuropathic pain (Barrett et al., 2007). Compared to people without pain and patients with non-neuropathic pain, diabetic neuropathic ‚Åé Corresponding author at: Division of Population Health Sciences, School of Medicine, University of Dundee, Dundee DD2 4BF, UK. E-mail addresses: w.meng@dundee.ac.uk (W. Meng), h.deshmukh@dundee.ac.uk (H.A. Deshmukh), l.y.donnelly@dundee.ac.uk (L.A. Donnelly), n.torrance@dundee.ac.uk (N. Torrance), h.colhoun@dundee.ac.uk (H.M. Colhoun), c.n.a.palmer@dundee.ac.uk (C.N.A. Palmer), b.h.smith@dundee.ac.uk (B.H. Smith). pain has a signiÔ¨Åcant negative effect on patients' quality of life (Davies et al., 2006). In addition, the disorder represents a signiÔ¨Åcant economic burden to healthcare systems (Tarride et al., 2006; Dworkin et al., 2010). Cross-sectional epidemiological studies have identiÔ¨Åed multiple risk factors for neuropathic pain. These include older age, female gender, manual occupation, lower educational attainment, and living in a rural area or in poor accommodation (Smith et al., 2007; Torrance et al., 2006). These risk factors are difÔ¨Åcult to modify and are not suitable for clinical intervention, though they are still of academic and political interest. SpeciÔ¨Åcally for diabetic neuropathic pain, modiÔ¨Åable risk factors including smoking, hypertension, obesity, hypercholesterolemia and duration of diabetes have been identiÔ¨Åed (Jensen et al., 2006; Tesfaye et al., 2005). Unfortunately, there are no published clinical trials that suggest a reduction in the incidence or severity of neuropathic pain through addressing these modiÔ¨Åable risk factors. Further effort is required in this area. Studies have found that, although glycaemic control can reduce the incidence of diabetic neuropathy, there is limited impact in decreasing the incidence of accompanying neuropathic pain, even http://dx.doi.org/10.1016/j.ebiom.2015.08.001 2352-3964/¬© 2015 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Please cite this article as: Meng, W., et al., A Genome-wide Association Study Provides Evidence of Sex-speciÔ¨Åc Involvement of Chr1p35.1 (ZSCAN20-TLR12P) and Chr8p23.1 (HMG..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.08.001 2 W. Meng et al. / EBioMedicine xxx (2015) xxx‚Äìxxx with long-term excellent glycaemic control (Callaghan et al., 2012; Marti et al., 2006). Epidemiological studies, such as genetic association studies, can identify independent risk factors which are clinically important, and offer these risk factors as covariates for basic research studies, or as new factors to address clinically. Diabetic neuropathic pain is considered as a complex trait which is affected by both environmental risk factors and genetic risk factors. Unlike well documented environmental risk factors, the understanding of the genetic contributors to neuropathic pain is rather poor, though evidence from animal models and human studies have both conÔ¨Årmed that it is a heritable trait (Devor et al., 2005; Meng et al., 2015). Studies on animal models have proposed candidate genes for neuropathic pain such as P2X7, P2X4, TLR4, and CACNG2 (Chessell et al., 2005; Trang et al., 2009; Nissenbaum et al., 2010; Wang et al., 2013). The Ô¨Årst genomewide association study (GWAS) on diabetic neuropathic pain in humans reported that GFRA2 might be associated with a subgroup of this disorder (Meng et al., 2015). All these candidate genes need further replications to validate their biological roles. We conducted a population-based GWAS of diabetic neuropathic pain in which our case deÔ¨Ånition was matched with previous population-based observational studies of diabetic neuropathic pain (Hall et al., 2013; Dieleman et al., 2008), seeking candidate genes that might not have been identiÔ¨Åed using the previous, more exclusive case deÔ¨Ånition (Meng et al., 2015). A control was deÔ¨Åned as a type 2 diabetic patient who has not been prescribed any of these Ô¨Åve drugs before. Individuals who had a prescription history of amitriptyline, carbamazepine, or nortriptyline were not included as controls because these drugs are often used for the treatment of other medical conditions, as well as neuropathic pain. In other words, diabetic individuals using these drugs could be correctly classiÔ¨Åed as neuropathic pain cases or wrongly classiÔ¨Åed if these drugs were used for treating other disorders such as depression or epilepsy. It is not possible to differentiate these two situations with certainty based on the available clinical information. To avoid the potential for incorrect phenotyping, those individuals were also removed from the control group. We excluded individuals with a history of only one single prescription for any of these Ô¨Åve drugs from both cases and controls. 3.1. Genotyping and Quality Control The quality control steps of the genotype data were applied based on the standard methods that were used for the WTCCC2 studies (GoDARTS & UKPDS Diabetes Pharmacogenetics Study Group et al., 2011), and the SUMMIT studies (Fagerholm et al., 2012). 3.2. Statistical Analysis 2. Methods 2.1. Resources Genetic resources: The Genetics of Diabetes Audit and Research Tayside (GoDARTS) project recruits diabetic patients and non-diabetic matched controls in Tayside, Scotland to identify genetic contributors relating to the susceptibility of diabetes, the complications of diabetes, the response to diabetes treatment and the prognosis of diabetes. (http://diabetesgenetics.dundee.ac.uk/). So far, the project has recruited 9439 diabetic patients who have provided their DNA samples along with written consent to use their clinical data and biological samples for research. Among these 9434 diabetic individuals, 3673 were genotyped by Affymetrix SNP6.0 chips supported by the Wellcome Trust Case Control Consortium 2 (WTCCC2) project (http://www.wtccc.org. uk/ccc2/) and 3254 were genotyped by Illumina OmniExpress chips supported by the Surrogate markers for Micro- and Macro-vascular hard endpoints for Innovative diabetes Tools (SUMMIT) project (http://www.imi-summit.eu/). The GoDARTS study was approved by Tayside Committee on Medical Research Ethics (REC reference 053-04). E-health resources: Since 1993, every person registered with the National Health Service (NHS) in Scotland has been assigned a unique Community Health Index (CHI) number. This number appears in the records of all personal medical activities within the NHS framework which paves the way for anonymous data linkage. The GoDARTS project includes consent from participants for the genetic data to be anonymously linked with datasets sourced from participants' NHS medical histories, including prescribing data, blood test results, radiology examination results, hospital admissions, and outpatient appointments. The current prescription history dataset used in this study covers from Jan, 1993 to Dec, 2013. 3. DeÔ¨Ånitions of Cases and Controls of Neuropathic Pain In this study, we deÔ¨Åned a neuropathic pain case as a type 2 diabetic patient who has a history of multiple usages (minimum twice) of at least one of the following Ô¨Åve medicines which are recommended and effective in diabetic peripheral neuropathy and prescribed uncommonly for other disorders: duloxetine, gabapentin, pregabalin, capsaicin cream (or patch) and lidocaine patch (Attal et al., 2010; National Institute for Health & Care Excellence NICE (UK), 2013; Finnerup et al., 2010). Non-genotyped single nucleotide polymorphisms (SNPs) in the Affymetrix SNP6.0 chips and Illumina OmniExpress chips were imputed by SHAPEIT and IMPUTE2 based on shared reference Ô¨Åles from the 1000 genome phase I datasets (Delaneau et al., 2011; Howie et al., 2009). An r2 score was used to assess the accuracy of an imputed genotype. It is suggested to adopt an r2 N 0.3 to remove imputed SNPs with poor quality. PLINK was the main GWAS software for genetic data manipulation and standardised quality control steps were frequently performed during analyses (For example, SNPs with over 10% genotyping missing were excluded, SNPs with minor allele frequency less than 1% were removed, SNPs which failed Hardy-Weinberg tests (P b 0.000001) were removed, and individuals with more than 10% genotype data missing were not included) (Purcell et al., 2007). SNPs on the sex chromosomes and mitochondrial SNPs were not included in the analyses since we do not have these data. The detection of individuals with different ancestry was done by the multidimensional scaling method implanted in PLINK. A lambda value generated by this method indicates the level of population stratiÔ¨Åcation. The lambda value should be very close to 1 in a homogeneous population with little ancestry mixture. Related samples were identiÔ¨Åed by calculating pi-hat values greater than 0.125 in PLINK. Logistic regression analyses were applied to generate P values for SNP association tests. A P value of less than 10‚àí6 was considered to be a suggestive association, worth further exploration. SNP functional annotation was searched by SNPnexus and Manhattan plots were generated by HaploView (Barrett et al., 2005; Dayem Ullah et al., 2013). Regional visualisation was achieved by LocusZoom (Pruim et al., 2010). The Q‚ÄìQ plot of P values, a tool to assess whether there are confounders and the impact of these potential confounders (different genotyping machines, different genotyping chips, different DNA extraction methods, etc) between cases and controls, was visualised by SNPEVG (Wang et al., 2012). The whole workÔ¨Çow is summarised in Fig. 1. Narrow-sense heritabilities of the overall dataset and sex-speciÔ¨Åc dataset were performed by restricted maximum likelihood analysis using the recognized approach to genome-wide complex trait analysis (GCTA) (Lee et al., 2011). Narrow-sense heritability represents the ratio of total phenotypic variance which is caused by additive genetic effects of individual SNPs (Lee et al., 2011). Comparisons of means of age and BMI between cases and controls were performed using independent t test in SPSS 21 (IBM Corp, New York, USA). The gender difference was evaluated using chi-square (2 √ó 2 tables). Please cite this article as: Meng, W., et al., A Genome-wide Association Study Provides Evidence of Sex-speciÔ¨Åc Involvement of Chr1p35.1 (ZSCAN20-TLR12P) and Chr8p23.1 (HMG..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.08.001 W. Meng et al. / EBioMedicine xxx (2015) xxx‚Äìxxx Fig. 1. WorkÔ¨Çow of the GWAS on neuropathic pain. 4. Results We identiÔ¨Åed 1043 diabetic patients who had a prescription record of minimum twice usage of at least one of the Ô¨Åve relevant neuropathic pain drugs (Duloxetine, Gabapentin, Pregabalin, Capsaicin cream (or patch) and Lidocaine patch, see Methods section for details) among the genotyped diabetic population of the GoDARTS project, representing 15.06% of the cohort. In addition, we found 3759 diabetic individuals who were identiÔ¨Åable as controls, as they had not been prescribed any of these Ô¨Åve drugs, nor other drugs that can be used (nonexclusively) to treat neuropathic pain (amitriptyline, carbamazepine, or nortriptyline). After removing ethnically outlying samples, genetically related samples, type 1 diabetic samples and those who had had a single prescription of neuropathic pain drugs, the Ô¨Ånal cohort for analysis comprised 961 neuropathic pain cases (male = 470, female = 491) and 3260 controls (male = 2021, female = 1239). We then derived data summarising the age and body mass index (BMI) for the overall dataset, male only dataset and female only dataset (Table 1). In the overall dataset, the average age (mean ¬± standard deviation, years) and BMI (mean ¬± standard deviation, kg/m2) in cases were 72.60 ¬± 10.54, and 27.79 ¬± 6.01, respectively. The average age and BMI in controls were 75.51 ¬± 10.79, and 26.91 ¬± 5.51, respectively. There were statistically signiÔ¨Åcant differences in age and BMI between cases and controls as well as in gender (P b 0.01). In the male only dataset, the average age and BMI in cases were 72.71 ¬± 9.96, and 27.06 ¬± 5.01, respectively. The average age and BMI in controls were 74.82 ¬± 10.69, and 26.83 ¬± 4.94, respectively. There was no statistical difference in BMI between cases and controls, but the difference in age was statistically signiÔ¨Åcant (P b 0.01). In the female only dataset, the average age and BMI in cases were 72.48 ¬± 11.08, and 28.49 ¬± 6.56, respectively. The average age and BMI in controls were 76.63 ¬± 10.90, and 27.06 ¬± 6.33, respectively. The differences in age and BMI between cases and controls were statistically signiÔ¨Åcant (P b 0.01). Table 1 Information on covariates between cases and controls. Overall dataset Male only Female only Age BMI Male:Female Age BMI Age BMI Cases Controls P value 72.60 ¬± 10.54 27.79 ¬± 6.01 470:491 72.71 ¬± 9.96 27.06 ¬± 5.01 72.48 ¬± 11.08 28.49 ¬± 6.56 75.51 ¬± 10.79 26.91 ¬± 5.51 2021:1239 74.82 ¬± 10.69 26.83 ¬± 4.94 76.63 ¬± 10.90 27.06 ¬± 6.33 b0.01 b0.01 b0.01 b0.01 N0.05 b0.01 b0.01 Age and BMI (body mass index) are presented as mean ¬± standard deviation. Age is deÔ¨Åned as 2014 ‚Äî birth year. 3 Altogether 6,906,962 genotyped and imputed SNPs survived for analysis, after standardised quality control of genotyping and imputation (r2 N 0.3). Since the lambda value (indicating the level of population stratiÔ¨Åcation) was 1.014 for the cleaned overall dataset, no extra adjustment was adopted based on population stratiÔ¨Åcation. Using logistic regression testing, with age, sex, and BMI as covariates for the overall dataset, there was a peak showing in chromosome 1 on the Manhattan plot (Fig. 2). The associated Q‚ÄìQ plot is shown in Supplementary File 1. Although none of the SNPs reached formal genome-wide signiÔ¨Åcance (5 √ó 10‚àí8), the cluster in Chromosome 1p35.1 (Chr1p35.1), spanning ZSCAN20-TLR12P area, still indicated possible associations. The most signiÔ¨Åcant SNP in this region was rs35260355 in the ZSCAN20 with a lowest P value of 3.84 √ó 10‚àí 7 and an odds ratio (OR) of 1.66 (95% conÔ¨Ådence interval: 1.37‚Äì2.02). Similar logistic regression in the female only dataset found that the peak in the Chr1p35.1 still existed and the top SNP rs71647933 in the ZSCAN20 achieved a lower P value of 2.74 √ó 10‚àí7 with an OR of 2.31 (95% conÔ¨Ådence interval: 1.68‚Äì3.17) (Fig. 3). In the male only dataset, the SNP cluster in the Chr1p35.1 disappeared while a new peak showed in the Chr8p23.1, next to HMGB1P46 and the P value of the top SNP rs6986153 was 8.02 √ó 10‚àí7 with an OR of 1.67 (95% conÔ¨Ådence interval: 1.34‚Äì2.08) (Fig. 4). Table 2 summarises all the signiÔ¨Åcant SNPs found in the regions in the three datasets. Figs. 5 and 6 show the regional plots of the identiÔ¨Åed loci in the female only dataset and the male only dataset, respectively. It was estimated that the narrow-sense heritability of neuropathic pain was 14.7% in the overall dataset, but 30.0% among males, compared with 14.7% among females. 5. Discussion Utilising a genetic dataset and e-health linkage dataset, we performed a GWAS on diabetic neuropathic pain using case and control deÔ¨Ånitions matched with previous population-based epidemiological studies and the results suggested two loci that may be involved with painful diabetic neuropathy. Standard protocols of the assessment of neuropathic pain have been widely agreed for specialist settings and primary care (Haanp√§√§ et al., 2011; Jones & Backonja, 2013). However, there is no common approach or consensus reached by clinicians or researchers to deÔ¨Åne neuropathic pain in population-based settings or in general cohorts. As GoDARTS participants were recruited through community-based clinics and general hospitals, there is no formal record of neuropathic pain status made by specialists. We acknowledge that expert clinical examination would have increased the robustness of the case deÔ¨Ånition in this cohort. However, without clinical examination evidence, it is reasonable to use an alternative, acceptable deÔ¨Ånition to represent neuropathic pain cases. We adopted a pragmatic approach to deÔ¨Åne cases using a multiple prescription history of the Ô¨Åve main drugs used exclusively or mainly to treat neuropathic pain (rather than other disorders) in a diabetic population. A combination of diagnostic codes for type 2 diabetes and prescription of neuropathic pain drugs has been used in previous epidemiological studies to identify patients with painful diabetic neuropathy (Hall et al., 2013; Dieleman et al., 2008). Members of our population-based cohort were already identiÔ¨Åed as having type 2 diabetes, and so our method of identifying neuropathic pain makes this study reasonably consistent with these previous studies. While amitriptyline, carbamazepine, and nortriptyline are also frequently used in neuropathic pain, we considered that these are relatively likely to be used for indications other than neuropathic pain and we did not include individuals who had been prescribed these drugs as either cases or controls. To have a more homogeneous population, we removed individuals with only a single prescription of the Ô¨Åve neuropathic pain drugs from both cases and controls. It has previously been highlighted that patients in primary care with neuropathic pain are often not prescribed any of the speciÔ¨Åc medications for its treatment (Hall et al., 2008; Torrance et al., 2007, 2013). As there is no pain status recorded in the GoDARTS, Please cite this article as: Meng, W., et al., A Genome-wide Association Study Provides Evidence of Sex-speciÔ¨Åc Involvement of Chr1p35.1 (ZSCAN20-TLR12P) and Chr8p23.1 (HMG..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.08.001 4 W. Meng et al. / EBioMedicine xxx (2015) xxx‚Äìxxx Fig. 2. Manhattan plot of the GWAS on neuropathic pain in the overall dataset. X axis represents 22 autosomes. Y axis means the ‚Äìlog10 of P values. The blue line is the cut-off P value of 10‚àí6. Cases and controls included 961 and 3260 samples, respectively. (Only SNPs whose P b 0.01 were used to make the plot). (For interpretation of the references to colour in this Ô¨Ågure legend, the reader is referred to the web version of this article.) no direct assessment of the presence of (neuropathic) pain can be made among cases or controls. Furthermore, we did not assess whether cases or controls had received any other prescriptions for pain, such as opioid medications, and it is possible that some with neuropathic pain were treated with drugs that are not speciÔ¨Åcally indicated for this. Therefore the deÔ¨Ånition in our study is possible to have classiÔ¨Åed some who have neuropathic pain as controls but few controls as cases. The subsequent P values and ORs may be underestimated, though we cannot measure the extent of this. The most signiÔ¨Åcant SNP cluster in the overall dataset was found in Chr1p35.1 with a lowest P value of 3.84 √ó 10‚àí7 at rs35260355, spanning ZSCAN20-TLR12P area. The function of ZSCAN20 (zinc Ô¨Ånger and SCAN domain containing 20) gene is not known yet and it has not been noted to be associated with any disorders. One of the proteins it codes contain typical C2H2 zinc Ô¨Ånger domain, which enables zinc Ô¨Ånger protein to bind other molecules such as RNA and DNA and affect transcription and translation (Krishna et al., 2003). There have been attempts to use zinc Ô¨Ånger proteins to treat neuropathic pain since the receptor speciÔ¨Åc transcription factors of zinc-Ô¨Ånger proteins have been developed to target gene repression in cell line models and in vitro (Tan et al., 2005). It is worth noting that the top SNP from the female only dataset rs71647933 is suggested to be a transcription factor binding site of the Fig. 3. Manhattan plot of the GWAS on neuropathic pain in the female only dataset. X axis represents 22 autosomes. Y axis means the ‚Äìlog10 of P values. The blue line is the cut-off P value of 10‚àí6. Cases and controls included 491 and 1239 individuals, respectively. (Only SNPs whose P b 0.01 were used to make the plot). (For interpretation of the references to colour in this Ô¨Ågure legend, the reader is referred to the web version of this article.) Please cite this article as: Meng, W., et al., A Genome-wide Association Study Provides Evidence of Sex-speciÔ¨Åc Involvement of Chr1p35.1 (ZSCAN20-TLR12P) and Chr8p23.1 (HMG..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.08.001 W. Meng et al. / EBioMedicine xxx (2015) xxx‚Äìxxx 5 Fig. 4. Manhattan plot of the GWAS on neuropathic pain in the male only dataset. X axis represents 22 autosomes. Y axis means the ‚Äìlog10 of P values. The blue line is the cut-off P value of 10‚àí6. Cases and controls included 470 and 2021 individuals, respectively. (onLy SNPs whose P b 0.01 were used to make the plot). (For interpretation of the references to colour in this Ô¨Ågure legend, the reader is referred to the web version of this article.) zinc interaction domain (SNPnexus). Toll-like receptors (TLRs) are a class of proteins which exist in various cell types in the central nervous system, including neuronal and non-neuronal cells (Liu et al., 2012). TLRs share structural and functional similarities. SpeciÔ¨Åcally, the deletion or inhibition of TLR2 and TLR4 in animal models will impair nerve injury-induced neuropathic pain (Kim et al., 2007; Tanga et al., 2005). When using a TLR4 antagonist to treat both wild type mice and TLR4 knockout mice suffering neuropathic pain, pain relief can be achieved in the wild type mice but not in the TLR4 knockout mice (Bettoni et al., 2008). TLR12P is a unitary pseudogene with a transcript but there is no protein product of this gene in the human. The function of its homolog in mice is unclear although it is suggested it may be involved in the immune system against pathogens (Koblansky et al., 2012). There is emerging evidence showing that TLRs are involved in the control of (neuropathic) pain while the mechanisms are still far from being elucidated (Liu et al., 2012). In the females only dataset (1730 individuals), the P value of the SNPs in the cluster were lower than in the overall dataset, indicating that the male samples were not contributing so much to the associations in this cluster, and that the identiÔ¨Åed ZSCAN20-TLR12P locus has a gender speciÔ¨Åc inÔ¨Çuence on diabetic neuropathic pain. This is consistent with the Ô¨Åndings of other TLR genes. Studies have found that variants in TLR genes are genderspeciÔ¨Åcally linked with multiple situations (Roberts et al., 2012). The mechanism of sex-speciÔ¨Åc phenomena is not clear and the evidence for hormone involvement is insufÔ¨Åcient and controversial (Roberts et al., 2012; Bergh√∂fer et al., 2006). We also identiÔ¨Åed a peak in the Chr8p23.1 next to HMGB1P46 when analysing the male only dataset, and the P value of the top SNP rs6986153 was 8.02 √ó 10‚àí7 with an OR of 1.67. HMGB1P46 is a pseudogene of high mobility group box-1 (HMGB1). It is suggested that the induction of high mobility group box-1 in the dorsal root ganglion can contribute to pain hypersensitivity after peripheral nerve injury (Shibasaki et al., 2010). In addition, Feldman et al found that the persistent endogenous release of HMGB1 by sensory neurons contributes to tactile hyperalgesia in a neuropathic pain rat model (Feldman et al., 2012). The synthesis and release of HMGB1 from spinal neurons due to nerve injury facilitates the activity of both microglia and neurons which leads to symptoms of neuropathic pain (Nakamura et al., 2013). It is interesting to know that HMGB1 signalling and TLR pathways, to some extent, are overlapping together (Yu et al., 2006; Velegraki et al., 2012). There is evidence that pseudogenes are involved in the biological process. For example, the low level of high mobility group A1 (HMGA1) was also associated with a high level of HMGA1 pseudogene (HMGA1-p) mRNA (Chiefari et al., 2010). It was observed that knockdown of HMGA1-p RNA in the cells of diabetic patients led to partially restored HMGA1 mRNA levels which suggested a competing relationship between the two types of transcripts. It is therefore hypothesised that a competing relationship might also exist between HMGB1 and its pseudogenes. There were no SNPs found with a P value of less than 5 √ó 10‚àí8 in the overall dataset, male only or female only datasets. Although a P value of 5 √ó 10‚àí 8 is generally adopted as the cut-off P value for GWAS Table 2 SigniÔ¨Åcant SNPs in the overall, female only and male only dataset. Dataset Overall Female Male SNP rs4652898 rs2336244 rs71647933 rs35260355 rs10914731 rs4652898 rs2336244 rs71647933 rs35260355 rs6986153 Chr 1 1 1 1 1 1 1 1 1 8 Position 33940691 33943390 33945601 33945831 33934824 33940691 33943390 33945601 33945831 108072044 Gene ZSCAN20 ZSCAN20 ZSCAN20 ZSCAN20 Intergenic ZSCAN20 ZSCAN20 ZSCAN20 ZSCAN20 Intergenic Minor allele C C G T G C C G T G Minor allele frequency in cases:controls 0.19:0.16 0.18:0.15 0.19:0.16 0.19:0.16 0.21:0.16 0.20:0.16 0.19:0.15 0.20:0.16 0.20:0.16 0.27:0.19 P value OR (95% CI) ‚àí7 7.45 √ó 10 9.07 √ó 10‚àí7 4.88 √ó 10‚àí7 3.84 √ó 10‚àí7 4.25 √ó 10‚àí7 3.70 √ó 10‚àí7 9.00 √ó 10‚àí7 2.74 √ó 10‚àí7 2.81 √ó 10‚àí7 8.02 √ó 10‚àí7 1.63 (1.34‚Äì1.98) 1.67 (1.36‚Äì2.05) 1.65 (1.36‚Äì2.02) 1.66 (1.37‚Äì2.02) 2.25 (1.64‚Äì3.09) 2.29 (1.67‚Äì3.16) 2.39 (1.69‚Äì3.38) 2.31 (1.68‚Äì3.17) 2.30 (1.68‚Äì3.17) 1.67 (1.34‚Äì2.08) Chr, chromosome; SNP, single nucleotide polymorphisms; OR, odds ratio; 95% CI, 95% conÔ¨Ådence interval. P values and ORs were calculated using logistic regression test. Please cite this article as: Meng, W., et al., A Genome-wide Association Study Provides Evidence of Sex-speciÔ¨Åc Involvement of Chr1p35.1 (ZSCAN20-TLR12P) and Chr8p23.1 (HMG..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.08.001 6 W. Meng et al. / EBioMedicine xxx (2015) xxx‚Äìxxx Fig. 5. Regional plot of Chr1p35.1 in females. r2 represents the linkage disequilibrium among SNPs. signiÔ¨Åcance, it has been suggested that this might be too stringent and risks missing important associations (i.e. false negatives) (Do et al., 2014). Using a lower threshold raises the chance of detecting associated SNPs, but also of detecting spurious associations (false positives), and we need to beware of that in interpreting this study. The narrowsense heritability (variance explained by SNPs, excluding genetic variation due to dominance, epistasis, and environment) of diabetic neuropathic pain in the overall dataset was estimated to be 14.7%, which is similar to that found in our previous analysis (Meng et al., 2015). However when calculated by gender, we found males had a higher heritability (30.0%) than females (14.7%). Sex-speciÔ¨Åc heritability has been observed in other traits (Weiss et al., 2005). The reasons behind the different gender-speciÔ¨Åc heritabilities are unknown although it may result from parent-of-origin effects, interaction with sex chromosomes and the sex-speciÔ¨Åc hormonal environment. It is worth considering sex-speciÔ¨Åc genetic effects in future association studies of neuropathic pain. There are some reports indicating that genetic effects are different between genders in determining pain. Experiments in mice found that the Mc1r gene mediates kappa-opioid analgesia in female mice only. Correspondingly in a human study, females with two variant MC1R alleles showed greater analgesic responses from the kappa-opioid, pentazocine, than males and females who did not have Fig. 6. Regional plot of Chr8p23.1 in males. r2 represents the linkage disequilibrium among SNPs. the variant alleles (Mogil et al., 2003). In addition, polymorphisms in the OPRM1 gene have been reported to be associated with pressurerelated pain sensitivity in men but not in women (Fillingim et al., 2005). Sato et al found that there were signiÔ¨Åcant associations between the opioid receptor genes (OPRM1, OPRD1 and opioid OPRK1) and experimental pain sensitivity (Sato et al., 2013). Our results showing genetic differences associated with neuropathic pain between genders are consistent with these Ô¨Åndings, though the biological mechanisms remain unclear and highlight the need for further research in this area. The heritability of neuropathic pain has been calculated as around 30% in rat models (Devor et al., 2005), similar to that measured here among men, though twice that found among women.. Using an additive model integrated in the CaTS, we had 80% power for the overall dataset (961 cases and 3260 controls), assuming a minor disease allele frequency of 0.20, a genotypic relative risk for this variant of 1.31, a prevalence of neuropathic pain in the diabetic population of 0.25, and the signiÔ¨Åcance level is 10‚àí6 (Skol et al., 2006). In our previous analysis, our case deÔ¨Ånition also included evidence of neuropathy, based on recorded results of monoÔ¨Ålament testing (Meng et al., 2015). As we did not consider the results of monoÔ¨Ålament testing in this study, our case deÔ¨Ånition was more inclusive and therefore less speciÔ¨Åc. Although there are power beneÔ¨Åts of including more cases, there is also a possibility that neuropathic pain with and without neuropathy evidence might have separate genetic risk markers, as well as shared genetic mechanisms. No studies have been reported examining whether there is any genetic difference between neuropathic pain with and without neuropathy evidence. The peaks we have identiÔ¨Åed in this paper could reÔ¨Çect some ‚Äògeneral‚Äô genetic mechanisms of neuropathic pain while the different peaks identiÔ¨Åed in our previous GWAS may be speciÔ¨Åcally associated with neuropathic pain with neuropathy evidence (Meng et al., 2015). In other disorders, a phenotype and its subtypes have been shown to have both shared and different genetic risks (Kessler et al., 2013). Similarly, we did not remove those who were prescribed strong opioid drugs from the control group since opioid drugs are neither indicated Ô¨Årst- or second-line treatments, nor commonly used to treat diabetic neuropathic pain (Torrance et al., 2013). A good phenotype, endophenotype and subgroup deÔ¨Ånition should aim to reÔ¨Çect the underlying genetic mechanisms. There are some recent GWAS published in the Ô¨Åeld of pain research. A locus between CCT5 and FAM173B located at Chr5p15.2 has been proposed to be associated with chronic widespread pain (Peters et al., 2013). TAOK3 was suggested to be associated with morphine requirement and postoperative pain in a retrospective paediatric day surgery population (Cook-Sather et al., 2014). Rs11127292 in the MYT1L was found to be associated to Ô¨Åbromyalgia with low comorbidities (Docampo et al., 2014). Another GWAS study suggested rs2952768 in the Chr2q33.3 was involved with analgesic requirements in humans (Nishizawa et al., 2014). These GWAS have shed light on the elucidation of the genetic pathways for pain while further research is needed, including replication studies, functional studies, and agreement on feasible, valid and reproducible phenotype ascertainment. The limitations of our study include that the P values of tops SNPs are only close to GWAS signiÔ¨Åcance but yet reached; no replication study to conÔ¨Årm the results; though the case deÔ¨Ånition is matched with those used epidemiological studies, we might misclassify some cases who have neuropathic pain but not prescribed medications into controls; we might also misclassify an individual into a control group who uses opioid to treat neuropathic pain. We have provided genetic evidence that SNPs in Chr1p35.1 (ZSCAN20-TLR12P) and Chr8p23.1 (HMGB1P46) may be involved with neuropathic pain in diabetes. Sex-speciÔ¨Åc associations are also suggested. Our Ô¨Åndings should be treated with caution and, while we have also presented their consistency with known biological factors, they can only guide the nature of future research, which will be based on the Ô¨Åndings reported in this paper. Any replication of our Ô¨Åndings will help to conÔ¨Årm hypothesised pathways involved in the genetic Please cite this article as: Meng, W., et al., A Genome-wide Association Study Provides Evidence of Sex-speciÔ¨Åc Involvement of Chr1p35.1 (ZSCAN20-TLR12P) and Chr8p23.1 (HMG..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.08.001 W. Meng et al. / EBioMedicine xxx (2015) xxx‚Äìxxx mechanisms of neuropathic pain and provoke research on new potential drug targets for the treatment of pain. Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ebiom.2015.08.001. Competing Interests The authors declare that they have no competing interests. Author Contributions WM analysed the data and prepared the manuscript. HD contributed to the imputation dataset. YL contributed to e-health linkage dataset. HC read the paper and provided suggestions to the discussion. NT and CP reviewed the paper and made a contribution to the discussion. BS contributed to the design of the study, and contributed signiÔ¨Åcantly to the manuscript. Role of the Funding Sources Funding sources did not have any involvement in the study design; the collection, analysis and interpretation of data; writing of the report; or the decision to submit the article for publication. Acknowledgements The authors of this manuscript are grateful to all the participants in GoDARTS. The authors are indebted to the Health Informatics Centre in the School of Medicine, University of Dundee, for their linkage of GoDARTS data to prescribing data. References Abbott, C.A., Malik, R.A., van Ross, E.R., Kulkarni, J., Boulton, A.J., 2011. Prevalence and characteristics of painful diabetic neuropathy in a large community-based diabetic population in the UK. Diabetes Care 34, 2220‚Äì2224. Attal, N., Cruccu, G., Baron, R., Haanp√§√§, M., Hansson, P., Jensen, T.S., Nurmikko, T., European Federation of Neurological Societies, 2010. EFNS guidelines on the pharmacological treatment of neuropathic pain: 2010 revision. Eur J Neurol 17, 1113-e88. Barrett, J.C., Fry, B., Maller, J., Daly, M.J., 2005. Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 21, 263‚Äì265. Barrett, A.M., Lucero, M.A., Le, T., Robinson, R.L., Dworkin, R.H., Chappell, A.S., 2007. Epidemiology, public health burden, and treatment of diabetic peripheral neuropathic pain: a review. Pain Med. 8 (supple2), 50‚Äì62. Bergh√∂fer, B., Frommer, T., Haley, G., Fink, L., Bein, G., Hackstein, H., 2006. TLR7 ligands induce higher IFN-alpha production in females. J. Immunol. 177, 2088‚Äì2096. Bettoni, I., Comelli, F., Rossini, C., Granucci, F., Giagnoni, G., Peri, F., Costa, B., 2008. Glial TLR4 receptor as new target to treat neuropathic pain: efÔ¨Åcacy of a new receptor antagonist in a model of peripheral nerve injury in mice. Glia 56, 1312‚Äì1319. Callaghan, B.C., Little, A.A., Feldman, E.L., Hughes, R.A., 2012. Enhanced glucose control for preventing and treating diabetic neuropathy. Cochrane Database Syst. Rev. 6, CD007543. Chessell, I.P., Hatcher, J.P., Bountra, C., Michel, A.D., Hughes, J.P., Green, P., Egerton, J., MurÔ¨Ån, M., Richardson, J., Peck, W.L., Grahames, C.B., Casula, M.A., Yiangou, Y., Birch, R., Anand, P., Buell, G.N., 2005. Disruption of the P2X7 purinoceptor gene abolishes chronic inÔ¨Çammatory and neuropathic pain. Pain 114, 386‚Äì396. Chiefari, E., Iiritano, S., Paonessa, F., Le Pera, I., Arcidiacono, B., Filocamo, M., Foti, D., Liebhaber, S.A., Brunetti, A., 2010. Pseudogene-mediated posttranscriptional silencing of HMGA1 can result in insulin resistance and type 2 diabetes. Nat. Commun. 1, 40. Cook-Sather, S.D., Li, J., Goebel, T.K., Sussman, E.M., Rehman, M.A., Hakonarson, H., 2014. TAOK3, a novel genome-wide association study locus associated with morphine requirement and postoperative pain in a retrospective pediatric day surgery population. Pain 155, 1383‚Äì1773. Davies, M., Brophy, S., Williams, R., Taylor, A., 2006. The prevalence, severity, and impact of painful diabetic peripheral neuropathy in type 2 diabetes. Diabetes Care 29, 1518‚Äì1522. Dayem Ullah, A.Z., Lemoine, N.R., Chelala, C., 2013. A practical guide for the functional annotation of genetic variations using SNPnexus. Brief. Bioinform. 14, 437‚Äì447. Delaneau, O., Marchini, J., Zagury, J.F., 2011. A linear complexity phasing method for thousands of genomes. Nat. Methods 9, 179‚Äì181. Devor, M., del Canho, S., Raber, P., 2005. Heritability of symptoms in the neuroma model of neuropathic pain: replication and complementation analysis. Pain 116, 294‚Äì301. Dieleman, J.P., Kerklaan, J., Huygen, F.J., Bouma, P.A., Sturkenboom, M.C., 2008. Incidence rates and treatment of neuropathic pain conditions in the general population. Pain 137, 681‚Äì688. 7 Do, D.N., Ostersen, T., Strathe, A.B., Mark, T., Jensen, J., Kadarmideen, H.N., 2014. Genomewide association and systems genetic analyses of residual feed intake, daily feed consumption, backfat and weight gain in pigs. BMC Genet. 15, 27. Docampo, E., Escaram√≠s, G., Gratacos, M., Villatoro, S., Puig, A., Kogevinas, M., Collado, A., Carbonell, J., Rivera, J., Vidal, J., Alegre, J., Estivill, X., Rabionet, R., 2014. Genomewide analysis of single nucleotide polymorphisms and copy number variants in Ô¨Åbromyalgia suggest a role for the central nervous system. Pain 155, 1102‚Äì1109. Dworkin, R.H., Malone, D.C., Panarites, C.J., Armstrong, E.P., Pham, S.V., 2010. Impact of postherpetic neuralgia and painful diabetic peripheral neuropathy on health care costs. J. Pain 11, 360‚Äì368. Fagerholm, E., Ahlqvist, E., Forsblom, C., Sandholm, N., Syreeni, A., Parkkonen, M., McKnight, A.J., Tarnow, L., Maxwell, A.P., Parving, H.H., Groop, L., Groop, P.H., FinnDiane Study Group, 2012. SNP in the genome-wide association study hotspot on chromosome 9p21 confers susceptibility to diabetic nephropathy in type 1 diabetes. Diabetologia 55, 2386‚Äì2393. Feldman, P., Due, M.R., Ripsch, M.S., Khanna, R., White, F.A., 2012. The persistent release of HMGB1 contributes to tactile hyperalgesia in a rodent model of neuropathic pain. J. NeuroinÔ¨Çammation 9, 180. Fillingim, R.B., Kaplan, L., Staud, R., Ness, T.J., Glover, T.L., Campbell, C.M., Mogil, J.S., Wallace, M.R., 2005. The A118G single nucleotide polymorphism of the mu-opioid receptor gene (OPRM1) is associated with pressure pain sensitivity in humans. J. Pain 6, 159‚Äì167. Finnerup, N.B., Sindrup, S.H., Jensen, T.S., 2010. The evidence for pharmacological treatment of neuropathic pain. Pain 150, 573‚Äì581. GoDARTS and UKPDS Diabetes Pharmacogenetics Study Group, Wellcome Trust Case Control Consortium 2, Zhou, K., Bellenguez, C., Spencer, C.C., Bennett, A.J., Coleman, R.L., Tavendale, R., Hawley, S.A., Donnelly, L.A., SchoÔ¨Åeld, C., Groves, C.J., Burch, L., Carr, F., Strange, A., Freeman, C., Blackwell, J.M., Bramon, E., Brown, M.A., Casas, J.P., Corvin, A., Craddock, N., Deloukas, P., Dronov, S., Duncanson, A., Edkins, S., Gray, E., Hunt, S., Jankowski, J., Langford, C., Markus, H.S., Mathew, C.G., Plomin, R., Rautanen, A., Sawcer, S.J., Samani, N.J., Trembath, R., Viswanathan, A.C., Wood, N.W., MAGIC investigators, Harries, L.W., Hattersley, A.T., Doney, A.S., Colhoun, H., Morris, A.D., Sutherland, C., Hardie, D.G., Peltonen, L., McCarthy, M.I., Holman, R.R., Palmer, C.N., Donnelly, P., Pearson, E.R., 2011. Common variants near ATM are associated with glycemic response to metformin in type 2 diabetes. Nat. Genet. 43, 117‚Äì120. Haanp√§√§, M., Attal, N., Backonja, M., Baron, R., Bennett, M., Bouhassira, D., Cruccu, G., Hansson, P., Haythornthwaite, J.A., Iannetti, G.D., Jensen, T.S., Kauppila, T., Nurmikko, T.J., Rice, A.S., Rowbotham, M., Serra, J., Sommer, C., Smith, B.H., Treede, R.D., 2011. NeuPSIG guidelines on neuropathic pain assessment. Pain 152, 14‚Äì27. Hall, G.C., Carroll, D., McQuay, H.J., 2008. Primary care incidence and treatment of four neuropathic pain conditions: a descriptive study, 2002‚Äì2005. BMC Fam. Pract. 9, 26. Hall, G.C., Morant, S.V., Carroll, D., Gabriel, Z.L., McQuay, H.J., 2013. An observational descriptive study of the epidemiology and treatment of neuropathic pain in a UK general population. BMC Fam. Pract. 14, 28. Howie, B.N., Donnelly, P., Marchini, J., 2009. A Ô¨Çexible and accurate genotype imputation method for the next generation of genome-wide association studies. PLoS Genet. 5, e1000529. Jensen, T.S., Backonja, M.M., Hern√°ndez Jim√©nez, S., Tesfaye, S., Valensi, P., Ziegler, D., 2006. New perspectives on the management of diabetic peripheral neuropathic pain. Diab. Vasc. Dis. Res. 3, 108‚Äì119. Jensen, T.S., Baron, R., Haanpaa, M., Kalso, E., Loeser, J.D., Rice, A.S., Treede, R.D., 2011. A new deÔ¨Ånition of neuropathic pain. Pain 152, 2204‚Äì2205. Jones III, R.C., Backonja, M.M., 2013. Review of neuropathic pain screening and assessment tools. Curr. Pain Headache Rep. 17, 363. Kessler, T., Erdmann, J., Schunkert, H., 2013. Genetics of coronary artery disease and myocardial infarction‚Äî2013. Curr. Cardiol. Rep. 15, 368. Kim, D., Kim, M.A., Cho, I.H., Kim, M.S., Lee, S., Jo, E.K., Choi, S.Y., Park, K., Kim, J.S., Akira, S., Na, H.S., Oh, S.B., Lee, S.J., 2007. A critical role of toll-like receptor 2 in nerve injuryinduced spinal cord glial cell activation and pain hypersensitivity. J. Biol. Chem. 282, 14975‚Äì14983. Koblansky, A.A., Jankovic, D., Oh, H., Hieny, S., Sungnak, W., Mathur, R., Hayden, M.S., Akira, S., Sher, A., Ghosh, S., 2012. Recognition of proÔ¨Ålin by toll-like receptor 12 is critical for host resistance to toxoplasma gondii. Immunity 38, 119‚Äì130. Krishna, S.S., Majumdar, I., Grishin, N.V., 2003. Structural classiÔ¨Åcation of zinc Ô¨Ångers: survey and summary. Nucleic Acids Res. 31, 532‚Äì550. Lee, S.H., Wray, N.R., Goddard, M.E., Visscher, P.M., 2011. Estimating missing heritability for disease from genome-wide association studies. Am. J. Hum. Genet. 88, 294‚Äì305. Liu, T., Gao, Y.J., Ji, R.R., 2012. Emerging role of Toll-like receptors in the control of pain and itch. Neurosci. Bull. 28, 131‚Äì144. Marti, C.L., Albers, J., Herman, W.H., Cleary, P., Waberski, B., Greene, D.A., Stevens, M.J., Feldman, E.L., DCCT/EDIC Research Group, 2006. Neuropathy among the diabetes control and complications trial cohort 8 years after trial completion. Diabetes Care 29, 340‚Äì344. Meng, W., Deshmukh, H.A., van Zuydam, N.R., Liu, Y., Donnelly, L.A., Zhou, K., Wellcome Trust Case Control Consortium 2 (WTCCC2), Surrogate Markers for Micro- and Macro-Vascular Hard Endpoints for Innovative Diabetes Tools (SUMMIT) Study Group, Morris, A.D., Colhoun, H.M., Palmer, C.N., Smith, B.H., 2015. A genome-wide association study suggests an association of Chr8p213 (GFRA2) with diabetic neuropathic pain. Eur J Pain 19, 392‚Äì399. Mogil, J.S., Wilson, S.G., Chesler, E.J., Rankin, A.L., Nemmani, K.V., Lariviere, W.R., Groce, M.K., Wallace, M.R., Kaplan, L., Staud, R., Ness, T.J., Glover, T.L., Stankova, M., Mayorov, A., Hruby, V.J., Grisel, J.E., Fillingim, R.B., 2003. The melanocortin-1 receptor gene mediates female-speciÔ¨Åc mechanisms of analgesia in mice and humans. Proc. Natl. Acad. Sci. U. S. A. 100, 4867‚Äì4872. Nakamura, Y., Morioka, N., Abe, H., Zhang, F.F., Hisaoka-Nakashima, K., Liu, K., Nishibori, M., Nakata, Y., 2013. Neuropathic pain in rats with a partial sciatic nerve ligation is Please cite this article as: Meng, W., et al., A Genome-wide Association Study Provides Evidence of Sex-speciÔ¨Åc Involvement of Chr1p35.1 (ZSCAN20-TLR12P) and Chr8p23.1 (HMG..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.08.001 8 W. Meng et al. / EBioMedicine xxx (2015) xxx‚Äìxxx alleviated by intravenous injection of monoclonal antibody to high mobility group box-1. PLoS One 8, e73640. National Institute for Health and Care Excellence NICE (UK), 2013. Clinical Guideline 173 Neuropathic Pain ‚Äî Pharmacological Management: The Pharmacological Management of Neuropathic Pain in Adults in Non-specialist Settings (http:// guidanceniceorguk/CG173 (accessed 30th July 2015)). Nishizawa, D., Fukuda, K., Kasai, S., Hasegawa, J., Aoki, Y., Nishi, A., Saita, N., Koukita, Y., Nagashima, M., Katoh, R., Satoh, Y., Tagami, M., Higuchi, S., Ujike, H., Ozaki, N., Inada, T., Iwata, N., Sora, I., Iyo, M., Kondo, N., Won, M.J., Naruse, N., UeharaAoyama, K., Itokawa, M., Koga, M., Arinami, T., Kaneko, Y., Hayashida, M., Ikeda, K., 2014. Genome-wide association study identiÔ¨Åes a potent locus associated with human opioid sensitivity. Mol. Psychiatry 19, 55‚Äì62. Nissenbaum, J., Devor, M., Seltzer, Z., Gebauer, M., Michaelis, M., Tal, M., Dorfman, R., Abitbul-Yarkoni, M., Lu, Y., Elahipanah, T., delCanho, S., Minert, A., Fried, K., Persson, A.K., Shpigler, H., Shabo, E., Yakir, B., Pisant√©, A., Darvasi, A., 2010. Susceptibility to chronic pain following nerve injury is genetically affected by CACNG2. Genome Res. 20, 1180‚Äì1190. Peters, M.J., Broer, L., Willemen, H.L., Eiriksdottir, G., Hocking, L.J., Holliday, K.L., Horan, M.A., Meulenbelt, I., Neogi, T., Popham, M., Schmidt, C.O., Soni, A., Valdes, A.M., Amin, N., Dennison, E.M., Eijkelkamp, N., Harris, T.B., Hart, D.J., Hofman, A., Huygen, F.J., Jameson, K.A., Jones, G.T., Launer, L.J., Kerkhof, H.J., de Kruijf, M., McBeth, J., Kloppenburg, M., Ollier, W.E., Oostra, B., Payton, A., Rivadeneira, F., Smith, B.H., Smith, A.V., Stolk, L., Teumer, A., Thomson, W., Uitterlinden, A.G., Wang, K., van Wingerden, S.H., Arden, N.K., Cooper, C., Felson, D., Gudnason, V., Macfarlane, G.J., Pendleton, N., Slagboom, P.E., Spector, T.D., V√∂lzke, H., Kavelaars, A., van Duijn, C.M., Williams, F.M., van Meurs, J.B., 2013. Genome-wide association study meta-analysis of chronic widespread pain: evidence for involvement of the 5p15.2 region. Ann. Rheum. Dis. 72, 427‚Äì436. Pruim, R.J., Welch, R.P., Sanna, S., Teslovich, T.M., Chines, P.S., Gliedt, T.P., Boehnke, M., Abecasis, G.R., Willer, C.J., 2010. LocusZoom: regional visualization of genome-wide association scan results. Bioinformatics 26, 2336‚Äì2337. Purcell, S., Neale, B., Todd-Brown, K., Thomas, L., Ferreira, M.A., Bender, D., Maller, J., Sklar, P., de Bakker, P.I., Daly, M.J., Sham, P.C., 2007. PLINK: a tool set for whole-genome association and population-based linkage analyses. Am. J. Hum. Genet. 81, 559‚Äì575. Roberts, B.J., Dragon, J.A., Moussawi, M., Huber, S.A., 2012. Sex-speciÔ¨Åc signaling through Toll-like receptors 2 and 4 contributes to survival outcome of Coxsackievirus B3 infection in C57Bl/6 mice. Biol. Sex Differ. 3, 25. Sato, H., Droney, J., Ross, J., Olesen, A.E., Staahl, C., Andresen, T., Branford, R., Riley, J., Arendt-Nielsen, L., Drewes, A.M., 2013. Gender, variation in opioid receptor genes and sensitivity to experimental pain. Mol. Pain 9, 20. Shibasaki, M., Sasaki, M., Miura, M., Mizukoshi, K., Ueno, H., Hashimoto, S., Tanaka, Y., Amaya, F., 2010. Induction of high mobility group box-1 in dorsal root ganglion contributes to pain hypersensitivity after peripheral nerve injury. Pain 149, 514‚Äì521. Skol, A.D., Scott, L.J., Abecasis, G.R., Boehnke, M., 2006. Joint analysis is more efÔ¨Åcient than replication-based analysis for two-stage genome-wide association studies. Nat. Genet. 38, 209‚Äì213. Smith, B.H., Macfarlane, G.J., Torrance, N., 2007. Epidemiology of chronic pain, from the laboratory to the bus stop: time to add understanding of biological mechanisms to the study of risk factors in population-based research? Pain 127, 5‚Äì10. Tan, S., McNamara, A., Jouvenot, Y., Krisky, D., Wolfe, D., Compos, B., Zhong, X., Glorioso, J., Zhang, H.S., Gregory, P.D., 2005. Zinc Ô¨Ånger protein transcription factors as potential therapeutic agents for the treatment of neuropathic pain [abstract]. Mol. Ther. 11, s250. Tanga, F.Y., Nutile-McMenemy, N., DeLeo, J.A., 2005. The CNS role of Toll-like receptor 4 in innate neuroimmunity and painful neuropathy. Proc. Natl. Acad. Sci. U. S. A. 102, 5856‚Äì5861. Tarride, J.E., Collet, J.P., Choiniere, M., Rousseau, C., Gordon, A., 2006. The economic burden of neuropathic pain in Canada. J. Med. Econ. 9, 55‚Äì68. Tesfaye, S., Chaturvedi, N., Eaton, S.E., Ward, J.D., Manes, C., Ionescu-Tirgoviste, C., Witte, D.R., Fuller, J.H., EURODIAB Prospective Complications Study Group, 2005. Vascular risk factors and diabetic neuropathy. N Engl J Med 352, 341‚Äì350. Torrance, N., Smith, B.H., Bennett, M.I., Lee, A.J., 2006. The epidemiology of chronic pain of predominantly neuropathic origin. Results from a general population survey. J. Pain 7, 281‚Äì289. Torrance, N., Smith, B.H., Watson, M.C., Bennett, M.I., 2007. Medication and treatment use in primary care patients with chronic pain of predominantly neuropathic origin. Fam. Pract. 24, 481‚Äì485. Torrance, N., Ferguson, J.A., Afolabi, E., Bennett, M.I., Serpell, M.G., Dunn, K.M., Smith, B.H., 2013. Neuropathic pain in the community: more under-treated than refractory? Pain 154, 690‚Äì699. Trang, T., Beggs, S., Wan, X., Salter, M.W., 2009. P2X4-receptor-mediated synthesis and release of brain-derived neurotrophic factor in microglia is dependent on calcium and p38-mitogen-activated protein kinase activation. J. Neurosci. 29, 3518‚Äì3528. van Hecke, O., Austin, S.K., Khan, R.A., Smith, B.H., Torrance, N., 2014. Neuropathic pain in the general population: a systematic review of epidemiological studies. Pain 155, 654‚Äì662. Velegraki, M., Koutala, H., Tsatsanis, C., Papadaki, H.A., 2012. Increased levels of the high mobility group box 1 protein sustain the inÔ¨Çammatory bone marrow microenvironment in patients with chronic idiopathic neutropenia via activation of toll-like receptor 4. J. Clin. Immunol. 32, 312‚Äì322. Wang, S., Dvorkin, D., Da, Y., 2012. SNPEVG: a graphical tool for GWAS graphing with mouse clicks. BMC Bioinf. 13, 319. Wang, X., Grace, P.M., Pham, M.N., Cheng, K., Strand, K.A., Smith, C., Li, J., Watkins, L.R., Yin, H., 2013. Rifampin inhibits Toll-like receptor 4 signaling by targeting myeloid differentiation protein 2 and attenuates neuropathic pain. FASEB J. 27, 2713‚Äì2722. Weiss, L.A., Abney, M., Cook Jr., E.H., Ober, C., 2005. Sex-speciÔ¨Åc genetic architecture of whole blood serotonin levels. Am. J. Hum. Genet. 76, 33‚Äì41. Yu, M., Wang, H., Ding, A., Golenbock, D.T., Latz, E., Czura, C.J., Fenton, M.J., Tracey, K.J., Yang, H., 2006. HMGB1 signals through toll-like receptor (TLR) 4 and TLR2. Shock 26, 174‚Äì179. Please cite this article as: Meng, W., et al., A Genome-wide Association Study Provides Evidence of Sex-speciÔ¨Åc Involvement of Chr1p35.1 (ZSCAN20-TLR12P) and Chr8p23.1 (HMG..., EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.08.001 ˇ˛Aggressive non-Hodgkin s lymphoma Lisa Lowry david Linch Abstract the aggressive non-hodgkin s lymphomas are a group of malignancies with rapid symptom onset and a short natural history. presenting features vary depending on the site(s) affected. investigations are aimed at making an accurate diagnosis and classification, assessing stage and prognostic features, and assessing the patient s ability to tolerate aggressive therapy. the Who classification has been widely adopted in routine practice, and has been recently updated. diffuse large B-cell non-hodgkin s lymphoma, Burkitt s lymphoma and t-cell lymphomas will be discussed in this con- tribution, with brief mention of other entities. the mainstay of treatment is multi-agent chemotherapy. the advent of monoclonal antibody therapy has improved outcomes in many B-cell malignancies. however, some aggressive lymphomas, including many t-cell disorders, still carry a poor prognosis, and new treatment strategies are needed. Keywords aggressive; diagnosis; epidemiology; lymphoma; management Aggressive non-Hodgkin s lymphoma (NHL) comprises a group of malignancies characterized by a more rapid onset and shorter natural history than indolent lymphomas. However, many are very chemosensitive and may be curable with appropriate, timely management. Epidemiology and risk factors There are about 10,000 new NHL cases diagnosed each year in the UK,1 representing 4% of all cancers. The age-standardized incidence is 13.4 per 100,000 population, which appears to have doubled in the past 30 years. The risk of developing the disease is increased in congenital and acquired immunodeficiency states, including HIV infection, but also other immune disorders such as rheumatoid arthritis. A variety of chemicals, including pesticides, herbicides and hair dyes, have been implicated but not proven to have a role in lymphoma. In the majority of cases the cause is unknown. The increase in incidence is only partly explained by the increase in immunodeficiency states and better reporting. Lisa Lowry MRCP FRCPath is Lymphoma Clinical Fellow at UCL Cancer Trials Centre, University College London, UK. Competing interests: none declared. David Linch FRCP FRCPath FMedSci is Head of Department of Haematology, at the University College Hospital, Cancer Institute, London UK. Competing interests: none declared. The rate of increase of lymphoma reporting has slowed in recent years and may have reached a plateau. Aggressive NHL accounts for just over half the total cases 2 (Figure 1). Presentation Symptoms and signs arise when masses infiltrate tissues or obstruct organs the possible presenting features are therefore diverse. Most patients have adenopathy at presentation, which may be painful if enlarging rapidly. Extranodal involvement is rel- atively common and may involve any organ or site. B symptoms (weight loss, fevers, night sweats) occur in 30 40% of patients. Spinal cord compression and superior vena cava obstruction, although uncommon, warrant special mention as they repre- sent medical emergencies. Urgent therapy for symptom relief is required; however, diagnostic tissue must be obtained prior to treatment wherever possible to allow an accurate diagnosis and definitive treatment plan. Investigation Investigations are aimed at making an accurate diagnosis and classification, excluding differential diagnoses, and subsequently staging and assessment of other prognostic factors (Table 1). The most important investigation is tissue biopsy and this should be arranged as quickly as possible if high grade lymphoma is sus- pected as any delay in diagnosis and treatment can adversely affect outcome. Biopsies should be reported by a dedicated hae- matopathologist and all new diagnoses should be reviewed at a suitable multidisciplinary team meeting. It is imperative that an adequate sample is obtained. Where possible, an entire lymph node should be excised as an appreciation of nodal architecture is important. A core biopsy may be adequate in some circum- stances but needle aspiration biopsies are unreliable. Steroid therapy should be avoided prior to biopsy, as rapid tumour degeneration may occur, making diagnosis difficult. Prognostic factors Defining groups with differing prognoses can help guide manage- ment decisions (Figure 2). Using the revised International Prog- nostic Index (IPI)3, one point each is assigned to the following five poor prognostic markers: " stage III or IV " age >60 years " elevated lactate dehydrgenase (LDH) " more than one extranodal site involved " WHO performance status e"2. Other poor prognostic features include bulky disease. Classification Many different classification systems have been used in lym- phoma over the past 6 decades. The current internationally rec- ognized system from the World Health Organization (WHO)4 evolved from the revised European-American classification of lymphoid neoplasms (REAL). Both these systems attempt to define real diseases that can be recognised by histopathologists using current standard techniques, and which relate to clinically LymphoproLiferative disorders mediCiNe 37:4 202 © 2008 published by elsevier Ltd. Figure 1 distinct entities. Classification therefore gives important informa- tion regarding natural history, prognosis and guides treatment decisions. Classification systems are also useful to enable all cases of lymphoma to be recognised as such, even unusual and atypical cases. Finally, standardized nomenclature is essential for research and interpretation of clinical trials. There are many recognised subtypes of NHL in the WHO clas- sification (Table 2) and it is beyond the scope of this article to describe each in detail. Low-grade lymphomas are discussed in the previous chapter. Treatment and prognosis Treatment is aimed at cure unless patient co-morbidities (not age) prohibit the use of effective therapies. Supportive care requires a multidisciplinary team approach. Choice of specific treatment and prognosis are highly dependent on the type of lymphoma, and are discussed below. Patient information the diagnosis, treatment and prognosis should be carefully explained, and written information provided LymphoproLiferative disorders Incidence of non-Hodgkin s lymphoma by subtype Diffuse large B cell (33%) T cell (14%) Burkitt s/Burkitt-like (3%) Mantle cell (6%) Follicular (22%) Other (22%) Differential diagnosis and investigation of aggressive non-Hodgkin s lymphoma Differential diagnoses Investigations haematology Biochemistry virology histopathology immunophenotyping Cytogenetics Clonality studies micro-array staging prognostic factors infection (e.g. epstein-Barr, cytomegalovirus, hiv, tuberculosis, toxoplasmosis) Carcinoma Low grade non-hodgkin s lymphoma (NhL)or hodgkin s lymphoma Leukaemia sarcoidosis autoimmune disorders full blood count cytopenias (marrow involvement, hyposplenism, autoimmune cytopenias, anaemia of chronic disease) Blood film circulating lymphoma cells, leucoerythroblastic film associated with marrow infiltration Urea and electrolytes, liver function tests abnormalities may reflect involvement Lactate dehydrogenase and ≤2-microglobulin may be elevated occasionally presence of a paraprotein hypercalcaemia common in adult t-cell leukaemia/lymphoma (atLL) Uric acid may be elevated increased incidence of NhL in hiv positive patients human t-cell lymphotropic virus type 1 (htLv1) associated with atLL see text may be important in differentiating lymphoma from reactive disorders essential for accurate classification some entities have specific associated chromosomal translocations occasionally necessary to confirm presence of a malignant clone. Kappa and lambda staining used for B cell malignancies but gene rearrangement studies may be necessary for t-cell malignancies. Not currently used routinely in diagnosis or to inform management decisions Ct neck/chest/abdomen/pelvis in all patients Bone marrow aspirate + trephine biopsy in all patients Cerebrospinal fluid examination if CNs signs or disease at specific sites paranasal sinus, epidural, testicular or breast. mri superior to Ct at identifying neurological involvement pet scanning at diagnosis desirable see text Table 1 mediCiNe 37:4 203 © 2008 published by elsevier Ltd. Figure 2 where possible. Patients should give informed consent prior to anti-lymphoma therapy. Infections during periods of neutropenia, care must be taken to prevent where possible and aggressively treat infections. Gran- ulocyte colony stimulating factor (GCSF) may be used to hasten neutrophil recovery. Tumour lysis syndrome (TLS) rapid cell death can lead to severe metabolic disturbance (hyperkalaemia, hyperphospha- taemia, hypocalcaemia), renal failure and death. All patients should be adequately hydrated and treated with allopurinol, at least for the first cycle of chemotherapy. Patients at very high risk (Burkitt s lymphoma, large tumour burden, pre-existing renal disease) or in whom there are signs of incipient TLS should receive rasburicase. Fertility issues men should be offered sperm-banking. Female patients may value the opportunity to discuss embryo or egg storage but this is rarely appropriate as treatment delays should be avoided. Diffuse large B cell lymphoma (DLBCL) DLBCL accounts for approximately a third of all NHL and is the most common lymphoma subtype in all parts of the world. It occurs at any age, with a peak in the seventh decade. The bone marrow is involved in approximately 15% of cases. The tumours LymphoproLiferative disorders Event-free survival among 399 patients assigned to chemotherapy with CHOP alone or with CHOP plus rituximab (elderly patients, all IPI groups) 1.0 0.8 0.6 0.4 0.2 0 CHOP plus rituximab 0 CHOP P < 0.001 0.5 1.0 1.5 2.0 Years after randomisation 2.5 3.0 A recent update reported 5-year overall survival of 58% vs 45%. Figure reproduced from Coiffier B, et al. N Engl J Med 2002; 346: 235 42. È¯ 2002 Massachusetts Medical Society. World Health Organization classification of non-Hodgkin s lymphomas B-cell neoplasms B lymphoblastic leukaemia/lymphoma small lymphocytic lymphoma (Chronic lymphocytic leukaemia) Lymphoplasmacytic lymphoma splenic marginal zone lymphoma extranodal marginal zone lymphoma Nodal marginal zone lymphoma follicular lymphoma primary cutaneous follicle centre lymphoma mantle cell lymphoma diffuse large B-cell lymphoma dLBCL associated with chronic inflammation Lymphomatoid granulomatosis primary mediastinal (thymic) large B-cell lymphoma intravascular large B-cell lymphoma aLK positive large B-cell lymphoma plasmablastic lymphoma Large B-cell lymphoma arising in hhv-8 associated multicentric Castleman disease primary effusion lymphoma Burkitt lymphoma B-cell lymphoma, unclassifiable, with features intermediate between diffuse large B-cell lymphoma and Burkitt lymphoma B-cell lymphoma, unclassifiable, with features intermediate between diffuse large B-cell lymphoma and classical hodgkin lymphoma red text: high-grade lymphomas. T-cell and NK-cell neoplasms t lymphoblastic leukaemia/lymphoma systemic eBv positive t-cell lymphoproliferative disease of childhood hydroa vacciniforme-like lymphoma adult t-cell leukaemia/lymphoma extranodal NK/t cell lymphoma, nasal type enteropathy-type t-cell lymphoma hepatosplenic t-cell lymphoma subcutaneous panniculitis-like t-cell lymphoma mycosis fungoides sezary syndrome primary cutaneous Cd30 positive t-cell lymphoproliferative disorders primary cutaneous gamma-delta t-cell lymphoma peripheral t-cell lymphoma, not otherwise specified angioimmunoblastic t-cell lymphoma anaplastic large cell lymphoma, aLK positive Table 2 mediCiNe 37:4 204 © 2008 published by elsevier Ltd. Probability of event-free survival are composed of large cells (nucleus usually at least twice the size of that of a normal lymphocyte) growing in a diffuse (i.e. non-follicular) pattern. By immunophenotyping they type as CD20 positive B cells. DLBC lymphomas are thought to derive from antigen driven B cells, which have completed immunoglobulin rearrangement. Cytogenetic and oncogene abnormalities Many cytogenetic abnormalities have been reported, most com- monly affecting BCL6 and cMYC (encoding transcription factors) and BCL2 (encoding an anti-apoptotic factor). Treatment Immunochemotherapy CHOP chemotherapy (cyclophos- phamide, doxorubicin, vincristine, prednisolone, administered every 21 days for 6 8 cycles) has been the standard of care since the 1970s.5 Recent improvements in response and survival have been seen with the use of rituximab-containing regimens.6 Rituximab is a chimeric anti-CD20 human monoclonal antibody, administered as an intravenous infusion. Standard three-weekly CHOP (CHOP21) has also been bettered by shortening the inter- val between chemotherapy cycles (CHOP14), using growth fac- tor support.7 9 A large randomized trial is currently comparing R-CHOP21 to R-CHOP14 (Figure 3). Central nervous system (CNS) prophylaxis CNS relapse is associated with a very poor prognosis. Overall, the incidence of CNS relapse following chemotherapy is around 5%, with increased risk associated with involvement of certain anatomi- cal sites (testis, breast, paranasal sinuses, epidural space), raised LDH or involvement of more than one extranodal site. Conven- tional chemotherapy does not effectively cross the blood-brain barrier (BBB). In high-risk patients, either additional systemic drugs that do cross the BBB or small doses of intrathecal Ara-C or methotrexate should be given. Radiotherapy these tumours are usually radiosensitive and radiotherapy may be used in combination with chemotherapy or alone to provide local control when treatment is not intended to be curative. Refractory/relapsed disease there is still potential for cure in primary refractory disease or at first relapse, if responsive to sal- vage chemotherapy. Numerous regimens are in use (e.g. ESHAP, IVE, DHAP, ICE, mini-BEAM, with or without rituximab) with no clear evidence to support one over another. Those responding to salvage chemotherapy should undergo stem cell mobilization and collection prior to myeloablative chemotherapy with autol- ogous stem cell rescue. Reduced-intensity allogeneic stem cell transplantation may have a role in patients in whom stem cells could not be harvested or in selected patients relapsing after an autograft (see pages 168 171). Subtypes of DLBCL Mediastinal large B-cell lymphoma typically presents with a bulky mediastinal mass and is more common in younger, female patients. There may be diagnostic confusion with Hodgkin s lym- phoma. Treatment and prognosis is similar to standard DLBCL. Intravascular large B-cell lymphoma is a rare subtype in which lymphoma cells are only found intravascularly. The clinical picture most commonly includes skin rashes, neurological symp- toms, B symptoms and organomegaly without lymphadenopathy. The diagnosis is often delayed due to the unusual presentation, and prognosis is poor, although the response to rituximab may be good. DLBCL occurring as a transformation from low grade lym- phoma carries a poor prognosis. Burkitt s lymphoma (BL) BL is a highly aggressive disease of children and young adults, more common in males. There are three clinical variants, which are particularly interesting in terms of epidemiology and possible aetiology. Endemic BL the original disease described by Denis Burkitt,10 who identified a correlation between endemic regions in equato- rial Africa and climatic factors. This region corresponds to that with endemic malaria. The peak incidence is between the ages of 4 to 7 years, and children typically present with rapidly enlarg- ing tumours of the jaw or other facial bones. Most tumours are Epstein-Barr virus (EBV)-associated (EBV was originally identi- fied from a biopsy of Burkitt s lymphoma from Africa). Sporadic BL uncommon, but occurs throughout the world and accounts for a high proportion of childhood lymphoma. The most common presentation is with an abdominal mass, espe- cially of the ileo-caecal region. As in endemic BL, other extra- nodal involvement is frequent, especially ovaries, kidneys and breasts. HIV-associated BL presentation with nodal disease and marrow involvement are common. All three subtypes are at risk from CNS disease. It is postulated that HIV infection allows for chronic antigenic stimulation from opportunistic infections, and defective T-cell regulation of EBV-infected B cells (although EBV infection is not essential). It is thought that malaria infection plays a similar role in endemic BL. Morphology and immunophenotype Histological sections show a monotonous infiltrate of medium- sized cells, with occasional benign macrophages ingesting apoptotic tumour cells, giving rise to the classical starry sky appearance (Figure 4). On touch imprint, marrow aspirate or cerebrospinal fluid samples, the cells can be seen to have deeply basophilic cytoplasm with vacuolation. Tumour cells express B-cell antigens, membrane IgM with light chain restriction and have a very high proliferation rate (virtually 100% of cells are positive for Ki-67). Genetics The MYC gene is deregulated by translocation to the Ig heavy chain region (t(8;14)), or less commonly light chain region (t(2;8) or t(8;22)). Treatment and prognosis Historically, BL had a poor prognosis with standard therapy. However, with the use of more intensive chemotherapy regimens the prognosis has improved markedly. Lymphoblastic lymphoma Lymphoblastic lymphoma is a neoplasm of immature B or T cells, analogous to acute lymphoblastic leukaemia (ALL), with mediCiNe 37:4 205 © 2008 published by elsevier Ltd. LymphoproLiferative disorders Figure 3 LymphoproLiferative disorders the same management and prognosis. The lymphoma terminol- ogy is used when the abnormal cells are confined to mass lesions with minimal marrow involvement. Primary effusion lymphoma (PEL) PEL is a rare, aggressive tumour of large B-cells, which pres- ents with serous effusions without detectable tumour masses. It is universally associated with human herpes virus 8 (HHV-8), usually in the setting of immunodeficiency. Prognosis is very poor. T cell NHL Generally, T-cell lymphomas have a poorer prognosis than B-cell lymphomas, although there are important differences between subtypes. Cytological, immunophenotypic and molecular fea- tures have so far been less useful at providing clinical and prog- nostic information compared with B cell lymphomas Peripheral T cell lymphoma, unspecified this is the most common subtype, affecting mainly older adults. It is aggres- sive and often presents with advanced stage. The prognosis is poor with 5 year failure free survival (FFS) and overall sur- vival (OS) of 20% and 32% respectively.11 Current regimens aim to intensify therapy compared to CHOP, and strong con- sideration should be given to autografting in first complete remission (CR). Progression-free survival following treatment of diffuse large B cell lymphoma with R-CHOP, by revised IPI group 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 Very good Good Poor p 9 0.001 0123456 Time (years) Very good, 0 adverse factors; good, 1 or 2 adverse factors; poor, 3 or more adverse factors (n=365). This research was originally published in Blood. Sehn LH et al. Revised international prognostic index (R-IPI) is a better predictor of outcome than the standard IPI for patients with diffuse large B-cell lymphoma treated with R-CHOP. Blood 2007; 109: 1857 61. the American Society of Hematology Diffuse large B-cell lymphoma Burkitt s lymphoma T-cell lymphoma a b directed against Cd20 and c Cd3. the h&e stained slide of BL shows several tangible-body macrophages on a background of monomorphic lymphoid cells the so-called starry sky appearance. the immunological stains show positivity with a brownish colour. Cd20 is a pan B-cell marker and Cd3 is a t-cell marker. the malignant cells in dLBCL and BL are Cd20 positive and Cd3 negative (although there are background Cd3 positive cells in the dLBCL section). t-NhL is Cd20 negative and Cd3 positive. histological sections of diffuse large B cell lymphoma (dLBCL), Burkitt s lymphoma (BL) and t cell lymphoma (t-NhL) stained with a haematoxylin and eosin (h&e), b monoclonal antibody c Figure 4 mediCiNe 37:4 206 © 2008 published by elsevier Ltd. Survival (%) Anaplastic large cell lymphoma (ALCL) anaplastic lym- phoma kinase (ALK) positive cases are common in children and young adults and have a relatively good prognosis (70% 5 year OS).11 The t(2; 5) translocation fuses the ALK gene with the nucleophosmin (NPM) gene resulting in a fusion protein with constitutive tyrosine kinase activity. ALK negative ALCL is a sep- arate disease entity, being seen in an older age group and having a worse prognosis with standard therapy (49% 5 year OS). Cutaneous T-cell lymphoma this section encompasses a variety of clinical and pathological entities. Staging is based on the TNM (tumour, nodes, metastases) system. Broadly, disease confined to the skin has a very good long-term prognosis and can often be controlled with phototherapy, topical drug treatment or radiation. Systemic therapies are used in more advanced disease. Adult T-cell leukaemia/lymphoma (ATLL) this is an aggres- sive lymphoma with poor prognosis, associated with infection by human T-cell lymphotrophic virus type 1 (HTLV1). Infection rates, and ATLL, are high in Japan and the Caribbean. There is a long latent period between infection and disease, and even after 30 years, only about 4% of infected people will develop ATLL. Enteropathy-type intestinal T-cell lymphoma this is a rare lymphoma, often but not exclusively developing on a background of overt clinical coeliac disease or histological villous atrophy. It is aggressive, and patients should be considered for intensive therapy. Special circumstances HIV DLBC NHL is over 100 times more common in HIV positive patients than in the general population. Prognosis depends on IPI score and CD4 count, and has improved in the highly active anti- retroviral therapy (HAART) era, although not quite to the level of non-HIV infected patients. Burkitt s lymphoma has an even greater increased incidence in HIV positive patients. Patients should be strongly considered for aggressive treatment, even when presenting very unwell, as predicted outcomes are as good as for non-HIV infected individuals. Primary CNS lymphoma This is a rare condition (2.8 per million per year) but its incidence has been increasing. It accounts for 5% of all brain tumours and approximately 20% of HIV-associated lymphomas. Survival has improved with the development of chemotherapy regimens that cross the blood-brain barrier (e.g. high dose methotrexate), but long-term neurotoxicity remains a problem, particularly in the elderly. Post-transplant lymphoproliferative disease (PTLD) This is a group of EBV driven conditions resulting from iatro- genic immunosuppression. B cell proliferation may be polyclonal or monoclonal and there is clinical heterogeneity, including very aggressive, rapidly growing forms. Some cases will spontaneously regress with reduction of immunosuppression; others require specific treatment. Summary High-grade lymphomas represent a heterogeneous group of con- ditions with variable clinical features. They have an aggressive clinical course, but, if diagnosed promptly and accurately and managed appropriately, many are curable. ˇ˛AIDS and the gastrointestinal tract Brian Gazzard Abstract the pattern of gastroenterological manifestations of hiV infection has changed greatly since the advent of highly active antiretroviral therapy (haart). the previous almost inevitable viral, protozoal and fungal opportunist infections now occur only in those presenting with very low cd4 counts. the prompt institution of haart is probably the most effective treatment for such individuals and with immune restoration the infections are eradicated, although they may need specific treat- ment in the first few months of haart. improvements in understanding the pathogenesis of hiV disease intimately involve the gastrointestinal tract. it is now recognized that shortly after first acquiring the virus there is a profound and possibly irreversible loss of cd4 memory cells in the lamina propria of the gut. translocation of bacterial antigens across the gut appears to be common in hiV infection, and is perhaps related to the loss of cd4 memory cells or a depletion in the cd 17 subset of cells. such translocation is associated with immune activa- tion, which may result in more rapid cd4 cell loss and certainly consid- erably reduced function of the immune system. Keywords aids-related sclerosing cholangitis; Cryptosporidium; cytome- galovirus infection; Microsporidium; oesophageal candidiasis; wasting The short-term morbidity and mortality of the gastrointestinal manifestations of HIV infection have been transformed in the developed world by the introduction of highly active antiretro- viral therapy (HAART). The three most common gastrointestinal manifestations (oesophageal symptoms, wasting and diarrhoea) usually occur in those who are not treated with antiretrovirals because they are unwilling, or because they were unaware of the infection. Oesophageal symptoms1,2 patients. All patients with candidiasis experience pain on swal- lowing (odynophagia) and 70% also have dysphagia. Candidiasis usually responds rapidly to a systemic antifungal agent (flucon- azole, 400 mg single dose). As a result of the immune recovery associated with antiretroviral treatment, relapses of candidiasis and development of resistance to fluconazole are rare. If the latter is suspected, itraconazole solution, 200 mg/day for 2 weeks, is the treatment of choice. Ulceration Cytomegalovirus (CMV) infection may produce discrete ulcers of the lower oesophagus or haemorrhagic oesophagitis. Both respond to intravenous ganciclovir, 5 mg/kg b.d., or foscarnet, 90 mg/kg b.d., for 2 weeks (in those with normal renal func- tion). Maintenance therapy is not needed because relapses are rare within 6 months of beginning HAART. Herpes simplex virus infection less commonly produces dis- crete oesophagitis, sometimes associated with fluid-filled vesi- cles. Treatment comprises aciclovir, 400 800 mg five times daily. In some patients, aciclovir is given intravenously at a dose of 5 mg/kg three times weekly. Ulcers of unknown aetiology large, shallow ulcers (resem- bling aphthous ulcers) also occur. Corticosteroid therapy given orally (prednisolone 40 mg/day) or by intralesional injection (methylprednisolone acetate) may be effective. Wasting Loss of more than 10% of ideal body weight without an obvious cause is an AIDS-defining illness that has become an uncommon presenting feature in the developed world but remains common in resource-poor countries. Continuing weight loss is usually associated with a specific pathogen, most commonly a protozoal Natural history aNd cliNical features What s new? " haart has had a dramatic effect in restoring immune competence in hiV-infected patients, with a profound decrease in the incidence of gastrointestinal manifestations of hiV " haart is now the mainstay of treatment of opportunistic infections affecting the gut " diagnoses in hiV-infected patients with gastrointestinal pathology now more closely parallel those in the immunocompetent Diagnosis of oesophageal symptoms can often be inferred from inspection of the oral cavity (where buccal candidiasis, herpes simplex, oral hairy leucoplakia or aphthous-like ulceration may be seen), but is best confirmed by oesophagoscopy with biopsy. gut infection. Weight loss may be associated with the typical Candidiasis is the most common cause of oesophageal symp- toms and is a diagnostic criterion for AIDS in HIV-positive Brian Gazzard MA MD FRCP is Consultant Physician and Clinical Research Director of the HIV/Genitourinary Medicine Unit at Chelsea and Westminster Hospital, London, UK. Competing interests: none declared. metabolic responses of starvation or cachexia, the latter associ- ated with raised resting energy expenditure and preferential loss of lean body mass rather than fat. Loss of fat from the extremities and face occurring in the fat redistribution syndrome associated with HAART often pres- ents as wasting , although changes in body weight are usually modest. MediciNe 37:7 357 © 2009 Published by elsevier ltd. Management Early intervention with dietary advice and enteral or paren- teral feeding may prevent weight loss, but controlled trials are lacking. Testosterone, 40 mg t.d.s., anabolic steroids or appetite stimulants may increase body weight, although lean body mass increases only in those who are able to exercise. Recombinant growth hormone, 0.1 mg/kg/day for 12 weeks, is expensive, but increases lean body mass in wasted individuals. There is currently no effective treatment for fat redistribution syndrome. Diarrhoea Almost all patients with advanced HIV infection develop diar- rhoea, which is the presenting feature of AIDS in up to 30%. Homosexuals acquire various sexually transmitted enteric infec- tions, collectively termed gay bowel syndrome , including Shigella, Giardia, Campylobacter-like organisms, Entamoeba, Chlamydia, gonorrhoea and syphilis. There is doubt about the pathogenicity of some organisms associated with gay viral syndrome. Bacterial infections Salmonella and Campylobacter: various species acquired from infected poultry and other food sources are the most common cause of bacterial diarrhoea. Salmonella infection, in particular, produces a more marked systemic disturbance in HIV-seropositive patients than in immunocompetent individuals. Septicaemia (20 40% of patients) and a relapsing course (10 20%) are seen, particularly in those with low CD4 counts. Salmonella and Campylobacter infections are usually sensitive to ciprofloxacin, 250 500 mg b.d. In salmonellosis, ciprofloxacin should be given for at least 3 months to reduce the risk of relapse in those not taking HAART. Mycobacterium avium intracellulare: infection with oppor- tunistic Mycobacterium avium intracellulare (MAI) is a common late manifestation of HIV, causing diarrhoea in up to 10% of patients with a CD4 count of less than 50/ºl. Positive stool cul- tures may indicate colonization only, therefore jejunal biopsies (showing MAI-laden macrophages in the lamina propria; see Figure 1) may be necessary to establish MAI as the cause. In MAI infection, azalide antibiotics such as azithromycin, 1 g/day, and clarithromycin, 1 g/day, produce marked improve- ment in systemic manifestations with variable effects on diar- rhoea. The previously poor prognosis of MAI infection has been transformed by HAART, and treatment can usually be discontin- ued within 6 months. Protozoal infections Cryptosporidiosis3 7: human infection with cryptosporidia was first described in 1976 and is the most common opportunistic infection causing diarrhoea in HIV-positive patients. Cryptospo- ridiosis is a zoonosis in rural environments; direct human-to- human transmission is more common in cities. Epidemics of water-borne infections have been described; the infectious dose of cryptosporidia in immunosuppressed patients is likely to be Figure 1 Jejunal biopsy showing macrophages laden with periodic acid-schiff-positive material (Mycobacterium avium intracellulare) in the lamina propria, and villus shortening. below the present detection limits for contamination of metro- politan water supplies (gnotobiotic animals ( germ-free , reared under sterile conditions) are infected with one to five oocysts). Thus, it is sensible to advise all severely immunosuppressed patients to boil their drinking water. This measure, and the intro- duction of HAART, has led to a considerable decrease in the inci- dence of protozoal infections. Cryptosporidium can be identified in stool samples (using a modified Ziehl Neelsen stain) and are occasionally seen on gut biopsy when direct faecal analysis is negative. Although persistent isolation of cryptosporidia for more than 1 month in patients with diarrhoea is diagnostic of AIDS, the infection is subsequently eradicated in about 10 20% of these patients, who recover symptomatically. Cryptosporidiosis usually responds to HAART, with elimination of all detectable cryptosporidia within 6 months, but it may recur as a terminal event when therapy fails. In these individuals, careful attention to electrolyte replace- ment and use of opiates as anti-diarrhoeals prolong life. No medication has been shown to eradicate cryptosporidia from stool, although paromomycin, 500 mg q.d.s., may reduce stool volume by up to 50%. Microsporidia8,9: see Figure 2. Enterocytozoon bieneusi was first recognized as a major cause of diarrhoea in HIV-positive MediciNe 37:7 358 © 2009 Published by elsevier ltd. Natural history aNd cliNical features Figure 2 light microscopy of thin sections of jejunal biopsy material, showing microsporidia (Giemsa stain). patients in 1989. Septata intestinalis has also been associated with diarrhoea. Little is known about the means of acquisition, although contamination of water supplies is probably impor- tant. Most patients with microsporidiosis suffer marked im- munosuppression, wasting and diarrhoea of up to 1 litre/day. Diagnosis is achieved by staining faecal samples with strong trichrome or a fluorescent stain, which demonstrates the spores. Accurate speciation requires electron microscopy of jejunal biopsies. Septata intestinalis can be eradicated by albendazole, 400 mg b.d., leading to resolution of the diarrhoea. HAART is reliably associated with eradication of all types of microsporidia within 6 months of initiation. CMV infection: see Figure 3. In patients with marked immuno- suppression, severe CMV infection of the colon causes abdominal pain, rebound tenderness and diarrhoea, which may be bloody. Toxic dilatation and perforation are complications. Milder infec- tion produces variable diarrhoea with a lower incidence of com- plications. Multiple CMV inclusions with intense inflammation on rectal biopsy confirm the diagnosis. Treatment with gan- ciclovir or foscarnet is effective in the short term, in doses as for oesophageal ulceration. Relapse is uncommon within a few months of beginning HAART. Pathogen-negative diarrhoea is more common in those with little immunosuppression and in low-volume diarrhoea with little weight loss. In most patients, diarrhoea resolves during follow-up, and fewer than 1% have long-term severe symp- toms. Protozoal infection is occasionally found on repeated analysis. Abdominal pain Infection is the cause of abdominal pain in most patients. " Right upper quadrant pain is often associated with changes, suggesting sclerosing cholangitis (diagnosed on cholangiography) or bile duct dilatation (seen on ultrasonography). There may be associated CMV, cryptosporidial or microsporidial infection, but sometimes no cause is found. The pain often improves spontane- ously; if not, sphincterotomy may abolish it. Figure 3 sigmoidoscopy showing inflammation in cytomegalovirus infection. " Right iliac fossa pain is often caused by appendicitis. Appendi- citis may be more common as a consequence of CMV infection. " Low abdominal pain is commonly caused by constipation secondary to opiate use. " Diffuse abdominal pain, often associated with rebound ten- derness, is usually caused by CMV infection of the colon. This is often suggested by plain abdominal radiographs showing colonic mucosal thickening or irregularity. ˇ˛AIDS and the lung Beth sage rob Miller Abstract respiratory problems continue to be a major burden of disease in hiV- infected patients, with more than 50% suffering at least one respira- tory episode. in early hiV infection, respiratory infections are similar to those found in the general population, but more common. Progres- sive hiV-induced immunosuppression increases the risk of opportunistic infections and tumours. highly active antiretroviral therapy (haart) has reduced the incidence of hiV-associated opportunistic infections and tumours, but has had less impact on the incidence of bacterial pneu- monia, tuberculosis and non-hodgkin lymphoma. haart may trigger immune phenomena including over-exuberant and uncontrolled immune response to exogenous antigen most commonly seen in Mycobacterium tuberculosis, where two syndromes have been described: immune reconstitution inflammatory syndrome (iris) and unmasking of asymp- tomatic, latent, indolent or incipient infection (ali3). Keywords aids; hiV infection; pneumonia; Pneumocystis jirovecii; pulmonary infection; tuberculosis The majority of HIV-infected patients suffer at least one signifi- cant respiratory episode during their disease. In early HIV infec- tion, when immune responses are relatively preserved, typical community-acquired infections occur but more frequently than in the general population. With advanced HIV-induced immu- nosuppression, the risk of opportunistic infections and tumours (Table 1) increases.1,2 Use of highly active antiretroviral therapy (HAART) has led to marked reductions in the incidence of HIV-associated oppor- tunistic infections (including Pneumocystis jirovecii pneumonia, cytomegalovirus disease and Mycobacterium avium complex) and tumours (including Kaposi sarcoma). HAART has had com- paratively less impact on the incidence of bacterial pneumonia, tuberculosis (TB) and non-Hodgkin lymphoma. Respiratory problems remain a major disease burden in HIV infection because of limited availability of HAART worldwide (particularly in resource-poor countries), the failure of sustained viral suppression in up to 50% of patients taking HAART, the Beth Sage MBBS MRCP is Specialist Registrar in Respiratory Medicine in the North West Thames region, West Hertfordshire, UK. Competing interests: none declared. Rob Miller FRCP is Reader in Clinical Infection at University College Medical School, London, and Honorary Professor at the London School of Hygiene and Tropical Medicine, UK. Competing interests: none declared. failure of prophylaxis for specific opportunistic infections and late presentation of previously undiagnosed HIV infection. Developing countries In 2007, an estimated 33 million individuals worldwide were infected with HIV. Of these, 67% live in sub-Saharan Africa. TB is the most common lung disease in this population; with up to 70% of sputum smear-positive TB patients co-infected with HIV. Reductions in case rates of TB in specific HIV-infected African communities following the introduction of HAART has supported the hypothesis that HIV makes a significant contribution to the TB epidemic. These observa- tions are repeated in other parts of the world, notably South-East Asia, Eastern Europe and the former USSR. The problem is exacer- bated and perpetuated by the limited availability of HAART. Streptococcus pneumoniae is the most common cause of bacte- rial pneumonia in Africa where increases in incidence, antibiotic resistance, recurrence rates and mortality are almost exclusively associated with HIV infection. Recent reports show an increas- ing incidence of P. jirovecii pneumonia in the developing world, particularly in Africa.3 Bacterial infections Upper respiratory tract infections, acute bronchitis and acute sinusitis occur more commonly in HIV-infected patients than in the general population. Bacterial infections are particularly com- mon in HIV-infected injecting drug users.1 Sinusitis Symptomatic chronic sinus disease affects more than 15% of HIV- infected individuals, particularly those with CD4 counts of less than 200/ºl (normal count is 350 1250/ºl). The wide spectrum of infectious causes and documented cases of lymphoma pre- senting with chronic sinus symptoms highlight the importance of imaging and tissue sampling in the investigation of persistent or recurrent symptoms. Bronchiectasis This is increasingly recognized in patients with advanced HIV dis- ease. It probably arises as a consequence of recurrent P. jirovecii pneumonia or bacterial infection. Pathogens commonly isolated include Strep. pneumoniae and Haemophilus influenzae. Natural history aNd cliNical features What s new? " Pneumocystis jirovecii is the new name for P. carinii, the cause of Pneumocystis pneumonia in humans. the acronym PcP still applies (PneumoCystis Pneumonia) " in populations with access to antiretroviral therapy haart has led to a marked reduction in the incidence of many hiV- associated opportunistic infections " in response to starting haart there may be an over- exuberant and uncontrolled immune response to exogenous antigen. this phenomenon is called immune reconstitution inflammatory syndrome MediciNe 37:7 342 © 2009 elsevier ltd. all rights reserved. Natural history aNd cliNical features a chest radiograph showing diffuse, bilateral infiltrates mimicking Pneumocystis jirovecii pneumonia. the cause was Streptococcus pneumoniae. b chest radiograph showing lobar consolidation. the cause was Staphylococcus aureus. HIV-associated respiratory diseases Infectious diseases Bacterial infections " upper respiratory tract infection " acute bronchitis " acute sinusitis " Bronchiectasis " Bacterial pneumonia " Mycobacterium tuberculosis Fungal infections " Pneumocystis pneumonia " other fungal pneumonia Non-infectious diseases " Kaposi sarcoma " lymphoma " Bronchial carcinoma " Non-specific interstitial pneumonitis " lymphocytic interstitial pneumonitis " hiV drugs as a cause of respiratory disease À% lactic acidosis À% abacavir hypersensitivity À% immune phenomena " chronic obstructive pulmonary disease " Pulmonary arterial hypertension " Pleural disease " Pneumothorax Table 1 Bacterial pneumonia Community-acquired bacterial pneumonia in HIV-infected indi- viduals presents with features similar to those that occur in the HIV-negative population with similar pathogens isolated, includ- ing, most commonly, Strep. pneumoniae and H. influenzae. Infec- tion with Gram-negative organisms and Staphylococcus aureus is seen in advanced HIV disease. Bacteraemia Bacteraemia is common in HIV-infected patients with bacterial pneumonia, irrespective of CD4 count. Chest radiography is often atypical, mimicking P. jirovecii pneumonia in up to 50% of cases (Figure 1). In contrast, radiographic lobar or segmental consoli- dation may be caused by a wide range of bacterial pathogens in addition to Strep. pneumoniae. Empirical ≤-lactam and macrolide regimens are the mainstay of treatment. Complications of bacte- rial pneumonia are common and include intrapulmonary cavita- tion and abscess formation, empyema and death. The relapse rate is high, despite appropriate antibiotic therapy.1 Prevention of bacterial pneumonia The US Centers for Diseases Control and Prevention (CDC), National Institutes of Health (NIH) and the HIV Medicine Association of the Infectious Diseases Society of America (IDSA) recommend that 23-valent polysaccharide pneumococcal vaccine should be given, at the time of diagnosis of HIV infection, to Figure 1 adults and adolescents with CD4 counts of 200/ºl or more, with re-immmunization after 5 years.4 Although humoral responses and clinical efficacy are likely to be impaired in those with CD4 counts of less than 200/ºl, the CDC/NIH/IDSA suggest that immunization should also be offered to this group. Pneumo- coccal immunization with the polysaccharide vaccine is recom- mended in the UK for all stages of HIV infection. Administration of co-trimoxazole in HIV-infected patients is associated with reduced mortality, particularly from respiratory MediciNe 37:7 343 © 2009 elsevier ltd. all rights reserved. disease, and WHO now recommends it as long-term prophylaxis in all HIV-infected patients in Africa, where access to HAART is limited or non-existent. Fungal infections P. jirovecii pneumonia P. jirovecii, formerly known as P. carinii, is the cause of Pneumo- cystis pneumonia (the acronym PCP still applies, PneumoCystis Pneumonia),5 which remains a common AIDS-defining present- ing diagnosis in individuals who are unaware of their HIV serosta- tus. It is also common in HIV-infected patients who are intolerant of or non-compliant with prophylaxis and/or HAART. PCP in immunosuppressed patients was thought to result from reactiva- tion of latent infection acquired in childhood, but is now believed to result from fresh acquisition from an exogenous source. Clinical features and investigations: patients present with non- productive cough and progressive breathlessness over 2 3 weeks or longer, often accompanied by fever. Features of HIV infection, including seborrhoeic dermatitis, extra-genital molluscum and oral Candida infection may be present. The chest is usually clear on auscultation; occasionally, crackles are audible. Cyanosis may be detected in severe (hypoxaemic) disease. Chest radiography may be normal in early PCP. The most common abnormality is bilateral perihilar interstitial infiltrates, which may progress to confluent dif- fuse alveolar shadowing over a period of only a few days. Atypical radiographic appearances occur in up to 20% of patients including upper-zone infiltrates resembling TB, hilar/mediastinal lymphade- nopathy, intrapulmonary nodules and lobar consolidation. Management: treatment is usually started empirically in HIV- infected patients with typical clinical and radiological features and a CD4 count of less than 200/ºl. The diagnosis can be con- firmed by cytology or PCR analysis of bronchoalveolar lavage (BAL) fluid. In practice, bronchoscopy is probably not necessary in those patients who respond rapidly to specific anti-Pneumo- cystis therapy. Several factors present at, or soon after, hospitalization pre- dict poor outcome; these include patient s age, a second or third episode of PCP, low haemoglobin, hypoxaemia, medical co-morbidity, e.g. pregnancy, and pulmonary Kaposi sarcoma.6 Subsequent to admission, development of pneumothorax, admis- sion to the intensive care unit and need for mechanical ventila- tion are all associated with a poor outcome. PCP can be stratified as: " mild (PaO2 (breathing room air >11.0 kPa, SaO2 >96%) " moderate (PaO2 8.0 11.0 kPa, SaO2 91 96%) " severe (PaO2 <8.0 kPa, SaO2 <91%). This is important, as some regimens (e.g. dapsone/trime- thoprim and atovaquone) are ineffective in severe disease and oral therapy may be given to those with mild disease. First-choice treatment for all grades of PCP is high-dose co-trimoxazole (sul- phamethoxazole, 100 mg/kg/day, and trimethoprim, 20 mg/kg/ day) in two to four divided doses orally or intravenously for 21 days. Approximately 64% of patients will successfully complete co-trimoxazole. Treatment-limiting drug toxicity is common and approximately 7% will fail treatment (defined as deterioration after a minimum of 5 days of therapy). Alternative therapy in patients who develop toxicity or who fail co-trimoxazole in mild or moderate disease includes clindamycin (450 600 mg p.o. or i.v. q.d.s) with primaquine (15 mg daily p.o.), dapsone (100 mg daily), plus trimethoprim (20 mg/kg/ day), or atovaquone (750 mg b.d). In severe disease, alternatives include clindamycin with primaquine (doses as above), or i.v. pentamidine (4 mg/kg daily).7 Patients with an admission PaO2 d"9.3 kPa (breathing room air) benefit from adjuvant glucocorticoids such as prednisolone, 40 mg b.d. for 5 days then 40 mg daily on days 6 10 and 20 mg daily on days 11 21. Co-trimoxazole, dapsone and primaquine should be avoided in patients with glucose-6-phosphate dehydrogenase deficiency. Prophylaxis: HIV-infected patients are at increased risk of Pneu- mocystis pneumonia as their CD4 count decreases.1,4 Indications for prophylaxis are given in Table 2. The first-choice agent for primary and secondary prophylaxis is co-trimoxazole, 960 mg daily. Lower doses (960 mg three times weekly or 480 mg daily) may be equally effective and have fewer side effects. Co-trimoxazole also protects against bacterial infec- tions and reactivation of cerebral toxoplasmosis. Co-trimoxazole is associated with a high incidence of adverse reactions, including rash with or without fever in up to 20% of patients. Desensitization may be attempted. Alternatively other, less effective, regimens may be used for prophylaxis. These include the following: " nebulized pentamidine, 300 mg once per month via jet neb- ulizer (once per fortnight if CD4 count <50/ºl) " dapsone, 100 mg daily, with pyrimethamine, 25 mg once weekly (pyrimethamine may also protect against cerebral toxoplasmosis) " atovaquone, 750 mg b.d. " azithromycin, 1250 mg once per week. Withdrawal of prophylaxis: primary prophylaxis may be discon- tinued in patients who, on HAART, exhibit sustained increases in CD4 count (>200/ºl) and suppression of HIV RNA load in blood Natural history aNd cliNical features Indications for prophylaxis of Pneumocystis pneumonia in adult HIV-infected patients Primary prophylaxis " cd4 count <200/ºl " cd4 count <14% of total lymphocyte count " history of hiV-associated constitutional features such as unexplained fever (>3 weeks duration) or oral/pharyngeal candida irrespective of cd4 count " history of another diagnosis that defines aids (e.g. Kaposi sarcoma) " When close monitoring of cd4 count (e.g. at least every 3 months) is not possible, consider starting prophylaxis in patients with cd4 counts of 200 250/ºl Secondary prophylaxis " all patients after an episode of Pneumocystis pneumonia MediciNe 37:7 344 © 2009 elsevier ltd. all rights reserved. Table 2 to below the limit of detection for 3 months or more.1,4 Secondary prophylaxis may be discontinued using similar criteria. If the CD4 count subsequently declines below 200/ºl and/or the HIV RNA load increases despite HAART, prophylaxis should be re-instituted. Other fungal infections Lung infections with Cryptococcus neoformans, Histoplasma cap- sulatum and Coccidiodes immitis are often seen in HIV-infected patients in the USA and Africa, but are uncommon in the UK. Pulmonary manifestations are variable. The diagnosis should be considered in those with a lifetime history of residence in or travel to an endemic area and may be confirmed by examina- tion of BAL fluid or transbronchial biopsy specimens. Treatment is with amphotericin or fluconazole (itraconazole for histoplas- mosis), and maintenance therapy is needed to prevent relapse unless HAART achieves sustained improvements in HIV RNA load and CD4 count (see above).4 Tuberculosis TB may occur at any stage of HIV infection and may result from re-activation of latent TB or re-infection from an exogenous source. TB in an HIV-infected patient is AIDS-defining regardless of CD4 count.1 TB is a notifiable disease in the USA, the UK and most other European countries. Clinical features More than two-thirds of cases of TB in HIV-infected patients present with pulmonary disease. In early HIV disease, when CD4 counts are normal or only slightly reduced, the clinical features are similar to those of adult post-primary disease. Symptoms include cough, dyspnoea, haemoptysis, chest pain, fever with sweats and weight loss. There may be no clinical features sug- gesting underlying HIV infection. Chest radiography often shows upper lobe infiltrates and cavitary changes. The tuberculin test is usually positive, and spontaneously expectorated sputum and BAL fluid (Figure 2) are often smear positive.1 In patients with advanced HIV disease and a low CD4 count TB may be difficult to diagnose. The presentation is often non- specific, with malaise, fatigue, weight loss and fever. In this situation, pulmonary TB often resembles primary infection; radiographic abnormalities include hilar/mediastinal lymphade- nopathy, miliary or diffuse shadowing and pleural effusions. Cavitation is uncommon. Chest radiography may be normal in up to one-third of patients. The tuberculin test is usually nega- tive, and spontaneously expectorated sputum/BAL fluid is often smear negative but culture positive. Extrapulmonary TB is com- mon in HIV-positive patients with CD4 counts of below 150/ºl. Local or disseminated infection may involve lymph nodes, bone marrow (Figure 2), liver and pericardium. Diagnosis Culture and species identification may take up to 6 weeks, so rapid molecular diagnostic tests can help. Mycobacterium tuberculosis infection should be assumed and treatment with conventional therapy instituted if acid-fast and alcohol-fast bacilli are found in a respiratory sample, in blood, in a tissue biopsy or in an aspirate from an HIV-infected patient, regardless of CD4 count, until species identification and drug sensitivities are known.1 Figure 2 Diagnosis using polymerase chain reaction (PCR) analysis is cur- rently limited by poor sensitivity. PCR analysis may be used to distinguish M. tuberculosis from atypical mycobacteria and to identify common mutations in the rpoB gene that are associated with rifampicin resistance. The role of interferon-gamma release assays (IGRAs) in HIV-infected persons is yet to be defined. Management Response to treatment with standard four-drug regimens is equivalent to that in the general population.8 However, compared with non-HIV-infected individuals, there is a higher incidence of adverse reactions to antituberculosis drugs and a greater risk of death. All patients with TB should be encouraged to have an HIV test and HAART should be started in those found to be HIV infected. The optimum time to start HAART is unknown. Current advice suggests that as patients with CD4 counts of <100/ºl are at risk of HIV disease progression or death, HAART should be started as soon as possible, preferably as part of a clinical trial; in those with CD4 counts of 100 200/ºl HAART may be delayed until after 2 months of anti-tuberculosis therapy; and in those with CD4 counts >200/ºl, HAART can be delayed until anti- tuberculosis therapy is completed, providing patients are closely monitored.8 When treatments for the two infections are started at the same time, patients suffer a high pill burden ; ascribing an adverse reaction to a specific drug may be difficult and an Natural history aNd cliNical features a Bronchoalveolar lavage fluid showing numerous acid-fast and alcohol-fast bacilli. (Ziehl Neelsen stain.) b Bone marrow trephine showing numerous granulomata in an hiV-infected patient with extrapulmonary tuberculosis. (h and e stain.) Ziehl Neelsen staining (not shown) showed acid-fast and alcohol-fast bacilli within the granulomata. (By courtesy of dr s lishman.) MediciNe 37:7 345 © 2009 elsevier ltd. all rights reserved. immune reconstitution inflammatory syndromes (IRIS) phe- nomenon may occur (see below).8 HAART and anti-tuberculosis therapies have overlapping toxicities (e.g. peripheral neuropathy occurs in almost one-third of patients given isoniazid and stavu- dine), and there are important drug interactions between rifam- picin and both protease inhibitors and non-nucleoside reverse transcriptase inhibitors. Multi-drug-resistant TB accounts for 1 2% of cases of TB in HIV-infected patients in the UK. Most cases are a consequence of inadequate treatment of TB or poor patient adherence to therapy. Some cases result from exogenous superinfection of a profoundly immunosuppressed HIV-infected patient who is already receiv- ing treatment for drug-sensitive disease. Prophylaxis The WHO recommends that HIV-infected patients co-infected with M. tuberculosis (but without disease) should be given che- moprophylaxis. There are few data to support this. Concerns about toxicity, interactions with HAART, difficulties distinguish- ing disease from infection and single-drug prophylaxis leading to drug resistance highlight the preferred practice in the UK of close clinical monitoring rather than chemoprophylaxis.8 BCG vaccination HIV infection may abrogate the protective response to BCG. Addi- tionally, there are concerns about the safety of this vaccine in HIV-infected patients. The WHO recommends that HIV-infected infants should not be given BCG as the risks of disseminated BCG disease outweigh any potential benefits in this group. Malignant conditions Kaposi sarcoma Kaposi sarcoma (see also pp. 338 341) is the most common HIV-associated malignancy.2 Before the advent of HAART, it affected 15 20% of patients with AIDS and was particularly com- mon in homosexual and bisexual men. It is strongly associated with human herpesvirus 8. Pulmonary Kaposi sarcoma is almost always accompanied by cutaneous or lymphadenopathic Kaposi sarcoma, and palatal disease strongly predicts the presence of pulmonary lesions. Patients usually present with non-specific cough and progressive dyspnoea; haemoptysis is uncommon. Pulmonary Kaposi sarcoma: may affect the airways or lung paren- chyma causing interstitial or nodular infiltrates (Figure 3) and alve- olar consolidation. Hilar/mediastinal lymphadenopathy occurs in about 25% of patients and up to 40% have a pleural effusion. Diagnosis is made at bronchoscopy by the appearance of usually multiple, raised or flat, red or violaceous endotracheal and endo- bronchial lesions (Figure 3). Biopsy is seldom performed since cutaneous Kaposi sarcoma is usually present and diagnostic yield from biopsy is low (<20%). Management: HAART has reduced the incidence of Kaposi sar- coma and may induce remission in established disease. It is often used together with chemotherapy for the treatment of pulmonary Kaposi sarcoma. Kaposi sarcoma-associated pleural effusions are difficult to treat; chemical pleurodesis is seldom successful. Figure 3 Lymphoma High-grade B-cell non-Hodgkin lymphoma, presenting as part of disseminated disease, is the most common HIV-associated lym- phoma in the thorax, although the incidence of Hodgkin disease is also increased in HIV-infected patients.9 The symptoms are non-specific. Typical chest radiographical abnormalities include mediastinal lymphadenopathy and pleural masses or effusions. The prognosis is better if patients treated with chemotherapy also receive HAART.2,9 Bronchial carcinoma Compared with the general population, bronchial carcinoma occurs more often in HIV-infected individuals. Presentation is typically with disseminated disease and the prognosis is therefore poor. Starting HAART does not have any impact on prognosis.2 Natural history aNd cliNical features a chest radiograph showing multiple, nodular infiltrates in a patient with pulmonary Kaposi s sarcoma. b endobronchial lesions of pulmonary Kaposi s sarcoma. MediciNe 37:7 346 © 2009 elsevier ltd. all rights reserved. Non-malignant conditions HIV therapy as a cause of respiratory symptoms " The nucleoside analogue (reverse transcriptase inhibitor) drugs, e.g. didanosine and stavudine, may induce lactic acidosis. Presentation is with progressive dyspnoea; the chest radiograph is normal. The diagnosis is made by interpretation of an arte- rial blood sample. Although the diagnosis is often delayed, with- drawal of HAART is usually associated with recovery. " The nucleoside analogue abacavir may induce a hypersensi- tivity reaction. Clinical features include dyspnoea, cough, phar- yngitis and pulmonary infiltrates. Recovery occurs with drug withdrawal, but in this setting abacavir should not be restarted as there is the risk of a fatal hypersensitivity reaction. " The fusion inhibitor drug enfuvirtide (T20) is associated with an increased risk of bacterial pneumonia. The mechanism is undetermined. " Immune phenomena occurring after starting HAART: in response to starting HAART there may be an over-exuberant and uncontrolled immune response to exogenous antigen (either from an intercurrent or lifetime previous opportunistic infection). This is best characterized for M. tuberculosis, where two syndromes have been described. First are new clinical manifestations, such as development of new peripheral lymphadenopathy, pleural or pericardial effusion or mediastinal/cerebral disease. These phe- nomena, also called IRIS occur a median of 14 days after starting HAART and are more likely if the baseline CD4 count is low and there is a rapid suppression of HIV viral load and recovery in CD4 count. Second is an unmasking of asymptomatic, latent, indolent or incipient infection (ALI3); this tends to occur later (median is 40 days after starting HAART). Non-specific pneumonitis Non-specific pneumonitis presents with symptoms and chest radiographical abnormalities resembling those of PCP. The CD4 count is often normal. Diagnosis requires transbronchial or open- lung biopsy. Most episodes are self-limiting, but prednisolone may be beneficial. Lymphocytic interstitial pneumonitis Lymphocytic interstitial pneumonitis is more common in HIV- infected children than adults. Patients present with dyspnoea and cough.10 Examination may reveal fine, end-inspiratory crackles. Chest radiography shows reticulonodular, diffuse infiltrates. Diag- nosis requires biopsy. Treatment with HAART is often effective. Chronic obstructive pulmonary disease HIV-infected smokers are twice as likely to have chronic obstruc- tive pulmonary disease (COPD) as smokers without HIV infec- tion.2 In the developed world the risk of COPD is associated with high rates of tobacco smoking among HIV-infected persons, injec- tion drug use and recurrent bacterial and opportunistic infections. HIV-infected cigarette smokers should receive targeted smoking cessation advice. Pulmonary arterial hypertension Pulmonary arterial hypertension occurs more commonly among HIV-infected patients than in the general population (incidence = 0.5% vs 0.002%). Patients typically present with progressive exertional dyspnoea.2 Secondary causes of pulmonary arterial hypertension, for example injection drug use or chronic thrombo- embolic disease, should be excluded. Treatment is as for the general population; bosantan and sildenafil have shown efficacy; HAART is associated with improved haemodynamics and survival. Pleural disease Pneumothorax occurs more frequently in HIV-infected patients than in the age-matched general medical population. Cigarette smoking and receipt of nebulized pentamidine are risk factors. PCP should be excluded in any patient presenting with a pneumothorax. Natural history aNd cliNical features Practice points " More than 50% of hiV-infected patients suffer a respiratory episode during the course of their hiV disease " Bacterial infections are more common in hiV-infected patients than in the hiV-negative general population " With widespread use of prophylaxis and haart in North america, europe and australasia, there has been a marked decrease in the incidence of Pneumocystis pneumonia " Worldwide, hiV infection is the strongest risk factor for tB; tB may occur at any stage of hiV infection, is aids-defining and should always be notified. ˇ˛AIDS-related cancers Mark Bower Abstract immunodeficiency, regardless of the underlying cause, is associated with an increased risk of malignancy. in the case of hiV infection, the majority of these cancers are associated with oncogenic virus infection. Whilst the overall risk of all cancers is increased two to three times in people living with hiV infection, there are three aids-defining cancers whose relative risk is dramatically higher in this population. these are Kaposi sarcoma, high-grade B-cell non-hodgkin lymphoma (including primary cerebral lymphoma) and invasive cervical cancer. since the introduction of highly active antiretroviral therapy the incidence of the aids-defining malignan- cies has declined in populations with access to these medications, whilst the effects on the incidence of other cancers has been small. Keywords aids; anal cancer; cervical cancer; haart; hiV; hodgkin lymphoma; Kaposi sarcoma; non-hodgkin lymphoma AIDS-defining malignancies Systemic non-Hodgkin lymphoma Epidemiology and histology: systemic high-grade B-cell non- Hodgkin lymphomas (NHL) occur 60 100 times more com- monly in people with HIV than in the age- and gender-matched general population. The incidence has fallen since the introduc- tion of highly active antiretroviral therapy (HAART), although this fall is more marked for primary cerebral NHL than for sys- temic NHL. Around one-third of these tumours are Burkitt or Burkitt-like lymphomas and two-thirds are diffuse large cell lymphomas. Clinical presentation: half the cases of HIV-associated systemic NHL present with nodal disease, a third with gastrointestinal dis- ease and the remainder with disease at other extranodal sites. Finally, 1% present as effusions without nodal mass of disease; this variant is called primary effusion lymphoma and is associ- ated with Kaposi sarcoma herpesvirus. Mark Bower PhD FRCP FRCPath is a Medical Oncologist specializing in the care of people with HIV-related cancers at the Department of Oncology at Imperial College School of Medicine, The Chelsea and Westminster Hospital, London, UK. Competing interests: Professor Bower has received reimbursement of expenses incurred while attending a symposium and fees for speaking from Gilead, Roche, BMS and Abbott. Clinical management and prognosis: the treatment of systemic AIDS-related NHL involves the administration of combination chemotherapy with intrathecal chemotherapy for those at risk of meningeal relapse. This should be given with concomitant HAART and prophylaxis against Pneumocystis jirovecii, Myco- bacterium avium complex and fungal infections. This combined management approach will result in durable complete remis- sion in 50 60% of patients, who will in effect be cured of their lymphomas. Primary cerebral lymphoma Epidemiology and histology: primary cerebral lymphoma (PCL) is lymphoma that is confined to the cranio-spinal axis without systemic involvement. It is associated with advanced immunosuppression and has a particularly poor prognosis. The incidence of PCL has declined dramatically since the introduc- tion of HAART. Toxoplasmosis and PCL are the most common causes of cerebral mass lesions in HIV and the differential diag- nosis may prove difficult; both occur at low CD4 counts (<50 cell/mm3) and present with headaches and focal neurological deficits. Clinical presentation: clinical features that favour PCL include a more gradual onset over 2 8 weeks and the absence of a fever. CT and MRI scanning usually reveal solitary or multiple ring- enhancing lesions with prominent mass effect and oedema. Although these features occur in both diagnoses, PCL lesions are usually periventricular whilst toxoplasmosis more often affects the basal ganglia. More than 85% of patients with cerebral toxo- plasmosis will respond clinically and radiologically to 2 weeks of antitoxoplasma therapy and this has become the cornerstone of the diagnostic algorithm for cerebral masses in severely immu- nodeficient patients. Immunodeficiency-related PCL are Epstein Barr virus (EBV)-related tumours and detection of EBV DNA in the cerebrospinal fluid (CSF) by polymerase chain reaction has become established as a diagnostic test with a high sensi- tivity and specificity. 18F-Fluorodeoxyglucose-positron emission tomography also helps to differentiate between PCL and cerebral toxoplasmosis (Figure 1). Clinical management and prognosis: the standard treatment modality for PCL in HIV patients is whole-brain irradiation and, in addition, antiretroviral naive patients should start HAART. Some clinicians advocate the use of chemotherapy for patients with PCL and a good performance status. However, the survival remains very poor (median survival 1 3 months). Kaposi sarcoma Epidemiology and histology: Kaposi sarcoma (KS) is caused by infection with a gammaherpesvirus called Kaposi sarcoma her- pesvirus (KSHV) or human herpesvirus 8. The highest levels of KSHV, which is transmitted both horizontally and vertically, are found in saliva and so kissing is thought to be a route of trans- mission. Histologically, KS is characterized by a proliferation of spindle-shaped cells that harbour latent KSHV and are thought to be derived from lymphatic endothelial precursor cells. These spindle cells are accompanied by endothelial cells, fibroblasts and inflammatory cells that form slit-like vascular channels resem- bling neo-angiogenesis. The major clinical differential diagnosis Natural history aNd cliNical features MediciNe 37:7 338 © 2009 elsevier ltd. all rights reserved. Natural history aNd cliNical features Paired magnetic resonance imaging and 18f-fluorodeoxyglucose-positron emission tomography scans of a patient with aids-related primary cerebral lymphoma. Figure 1 is bacillary angiomatosis caused by Bartonella henselae; this is best excluded by biopsy. Clinical presentation: most patients with KS present with skin lesions that are typically multiple, pigmented, raised, painless and do not blanch (see Figure 2). The earliest cutaneous lesions are often asymptomatic innocuous-looking macular-pigmented lesions, which vary in colour from faint pink to vivid purple. Larger plaques occur usually on the trunk as oblong lesions fol- lowing the line of skin creases. Lesions may develop to form large plaques and nodules that can be associated with painful oedema. In addition to a thorough skin examination, inspection of the oral cavity and conjunctivae should be undertaken (see Figure 3). Oral lesions are a frequent accompaniment that may cutaneous Kaposi sarcoma lesions. Figure 2 MediciNe 37:7 339 © 2009 elsevier ltd. all rights reserved. Natural history aNd cliNical features common non-cutaneous Kaposi sarcoma sites. Figure 3 lead to ulceration, dysphagia and secondary infection. A nodu- lar form of KS that is frequently associated with lymphoedema is seen more commonly in people of African descent and often proves particularly difficult to control. The most common sites of visceral KS are the lungs and stom- ach. Pulmonary KS is a life-threatening complication that usually presents with dyspnoea, dry cough and sometimes haemoptyisis, with or without fever. Chest X-ray typically reveals a diffuse reti- culo-nodular infiltrate and pleural effusion (see Figure 4). Gastro- intestinal lesions are usually asymptomatic but may bleed or cause obstruction. The diagnosis is usually confirmed at endoscopy. Clinical management and prognosis: the clinical management of AIDS-associated KS is determined to a large extent by the clini- cal staging, which is shown in Table 1. Patients with T0 disease should commence HAART and in the majority of cases the KS will respond to this alone although it may take 6 12 months. In view of this delay, cosmetically significant localized lesions may be treated with either intralesional vinblastine or localized radio- therapy. Chemotherapy is advocated for advanced cutaneous and visceral KS (stage T1) but is not merited for early disease in view of the potential responses to HAART. Liposomal anthracyclines are considered first-line chemotherapy for advanced KS, and paclitaxel chemotherapy is recommended for anthracycline- resistant disease. KS is no longer associated with an ominous prognosis, although patients with pulmonary involvement still only have a median survival of around 18 months. Cervical cancer Epidemiology and histology: invasive cervical cancer has been included as an AIDS-defining diagnosis since 1993 and the precur- sor lesions, cervical intra-epithelial neoplasia (CIN) or squamous intraepithelial lesion (SIL), also occur more frequently in women with HIV. Human papilloma virus (HPV) has a central role in the pathogenesis and HIV is associated with a high prevalence and persistence of HPV in the cervix, often with multiple HPV genotypes and progression from low-grade SIL to high-grade SIL as well as a high likelihood of relapse of CIN II/III after therapy. The risk of SIL is greatest amongst women with CD4 cell counts less than 200 cells/mm3 and may regress with HAART. These findings mandate close colposcopic surveillance. In many parts of the world, women with HIV are the least likely to have access to cervical screening programmes. Clinical presentation and management: HIV-positive women with cervical cancer generally have more advanced-stage disease Pulmonary Kaposi sarcoma (chest X-ray and ct scan). Figure 4 MediciNe 37:7 340 © 2009 elsevier ltd. all rights reserved. the modified AIDS Clinical trials group staging of KS (1997) Tumour (T) Immune status (I) Ks, Kaposi sarcoma. good risk (all of the following) t0 I0 confined to skin, lymph nodes or minimal oral disease cd4 count > 150/mm3 Poor risk (any of the following) t1 I1 tumour-associated oedema or ulceration extensive oral Ks Gastrointestinal Ks Ks in other non-nodal viscera cd4 < 150/mm3 table 1 at presentation. The management of invasive cervical cancer in HIV-positive women is the same as in the general population. Better uptake of cervical screening and the vaccination of girls against high-risk HPV genotypes prior to acquiring HPV infec- tion may reduce the incidence of invasive cervical cancer in the future. Non AIDS-defining malignancies Overall cancer occurs two to three times more commonly in HIV-positive people than the age-gender matched general popu- lation. Some tumours occur at much higher rates than others, including cancers associated with viral infections such as anal cancer, Hodgkin disease and hepatocellular cancers. However, some cancers with no known viral aetiology, such as seminoma and non-small cell lung cancers, also occur significantly more frequently. Anal cancer Epidemiology and histology: there is an increased incidence of anal carcinoma amongst HIV-positive patients but it is not an AIDS-defining diagnosis. The incidence of anal cancer amongst gay men is estimated to be 35/100,000, which is similar to the incidence of cervical cancer before the introduction of routine cervical screening. Anal cancer is twice as common in HIV- positive gay men as it is in HIV-negative gay men. Anal cancer shares many features with cervical cancer, including association with HPV infection. High-grade squamous intraepithelial lesion (HSIL) or anal intra-epithelial neoplasia (AIN) of the anus is believed to progress to invasive anal cancer in a fashion analogous to the progression from cervical HSIL or cervical CIN to invasive cervical cancer. Clinical presentation and management: patients present with pain, bleeding and a mass. There is no clear correlation between risk of anal cancer and CD4 cell count. Combined modality therapy of chemotherapy with radiation treatment can result in tumour ablation with sphincter preservation. Most tumours can be controlled locally with 5-year survival rates in the range of 65%, although the toxicity appears to be greater than in the HIV-negative population and salvage surgery may be necessary. The future of anal cancer in HIV seropositive people may lie with effective screening of the at-risk population and early inter- vention. Proctoscopy and anoscopic cytology has been found to be an effective method of screening for AIN and if therapeutic interventions could be shown to reduce progression and mor- tality, this would be an attractive strategy. The introduction of effective vaccines against high-risk HPV genotypes could reduce the incidence of this malignancy but only if both boys and girls were vaccinated prior to acquiring HPV infection, which usually occurs shortly after first intercourse. Hodgkin lymphoma Epidemiology and histology: the incidence of Hodgkin lym- phoma is around ten times higher in individuals with HIV, although it is not an AIDS-defining diagnosis. As with B-cell NHL, Epstein Barr herpesvirus (EBV) infection is implicated in the pathogenesis. The histological subtypes in patients with HIV are often less favourable. Clinical presentation and management: patients usually pres- ent with advanced stage disease and B symptoms, bone marrow infiltration and extranodal disease are all common. The optimal chemotherapy schedule has not been determined; however, ABVD, Stanford V and hybrid regimens have been used. Since the era of HAART, the remission rates and overall survival have improved and approach those seen in the general population when matched for stage and other prognostic variables. CLINICAL RESEARCH STUDY Alcohol Quantity and Type on Risk of Recurrent Gout Attacks: An Internet-based Case-crossover Study Tuhina Neogi, MD, PhD,a Clara Chen, MHS,b Jingbo Niu, DSc,a Christine Chaisson, MPH,b David J. Hunter, MD,c Yuqing Zhang, DSca a Clinical Epidemiology Research and Training Unit, Boston University School of Medicine, Boston, Mass; bData Coordinating Center, Boston University School of Public Health, Boston, Mass; cKolling Institute, University of Sydney and Royal North Shore Hospital, New South Wales, Australia. ABSTRACT OBJECTIVES: Although beer and liquor have been associated with risk of incident gout, wine has not. Yet anecdotally, wine is thought to trigger gout attacks. Further, how much alcohol intake is needed to increase the risk of gout attack is not known. We examined the quantity and type of alcohol consumed on risk of recurrent gout attacks. METHODS: We conducted a prospective Internet-based case-crossover study in the US among participants with gout and who had at least one attack during the 1 year of follow-up. We evaluated the association of alcohol intake over the prior 24 hours as well as the type of alcoholic beverage with risk of recurrent gout attack, adjusting for potential time-varying confounders. RESULTS: This study included 724 participants with gout (78% men, mean age 54 years). There was a signiÔ¨Åcant dose-response relationship between amount of alcohol consumption and risk of recurrent gout attacks (P <.001 for trend). The risk of recurrent gout attack was 1.36 (95% conÔ¨Ådence interval [CI], 1.001.88) and 1.51 (95% CI, 1.09-2.09) times higher for >1-2 and >2-4 alcoholic beverages, respectively, compared with no alcohol consumption in the prior 24 hours. Consuming wine, beer, or liquor was each associated with an increased risk of gout attack. CONCLUSIONS: Episodic alcohol consumption, regardless of type of alcoholic beverage, was associated with an increased risk of recurrent gout attacks, including potentially with moderate amounts. Individuals with gout should limit alcohol intake of all types to reduce the risk of recurrent gout attacks. √ì 2014 Elsevier Inc. All rights reserved. The American Journal of Medicine (2014) 127, 311-318 KEYWORDS: Alcohol; Case-crossover; Gout attacks; Internet; Triggers Funding: TN‚Äôs support included National Institute of Arthritis and Musculoskeletal and Skin Diseases (NIAMS) K23 AR055127, Arthritis Foundation Arthritis Investigator Award, American College of Rheumatology Research Education Fund Junior Career Development Award in Geriatric Medicine (T. Franklin Williams Scholars Program); YZ‚Äôs support included NIAMS AR47785, American College of Rheumatology Research Education Foundation Health Professional Investigator Award. ConÔ¨Çict of Interest: None. Authorship: All of the work is original, all authors meet criteria for authorship, including acceptance of responsibility for scientiÔ¨Åc content of the manuscript. All authors had access to the data and a role in writing the manuscript. Boston University Medical Center Institutional Review Board approval protocol number: H-22804. Requests for reprints should be addressed to Tuhina Neogi, MD, PhD, 650 Albany Street, Clinical Epidemiology Unit, Suite X-200, Boston, MA, 02118. E-mail address: tneogi@bu.edu 0002-9343/$ -see front matter √ì 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.amjmed.2013.12.019 Gout, a crystal-induced arthritis associated with hyperuricemia,1 is currently the most common inÔ¨Çammatory arthritis, affecting 8.3 million US adults.2 Recurrent attacks constitute the main clinical burden of gout. Despite available urate-lowering therapies, the risk of recurrent gout attacks remains high, with the risk of having at least one attack in a year being 69%.3 Strategies to prevent not only disease onset but also recurrent attacks are needed, given the rising incidence and prevalence of gout.2,4-6 Alcohol has been recognized anecdotally as a potential risk factor for recurrent gout attacks. However, most studies to date have focused on alcohol consumption in relation to the risk of initial occurrence of gout.7-9 In a large prospective cohort study, total alcohol consumption was strongly associated with an increased risk of incident gout.8 Additionally, the risk of incident gout varied by type of beverage 312 The American Journal of Medicine, Vol 127, No 4, April 2014 consumed, with an increased risk observed for beer and gout-related data (eg, features, duration, medications used, liquor but not wine.8 However, patients often report wine as a number of gout attacks in the prior 12 months), comorbidities, trigger for recurrent gout attacks, and historic depictions of and other medication use. Eligible subjects were those who gout often included wine, although this may have been related reported a gout attack within the previous year, were age 18 in part to lead contamination in the Roman era. Previously, years or older, were residents of the US, provided informed we have reported that overall alcohol consumption increased consent, and agreed to release medical records. We reviewed the risk of recurrent gout attacks; the medical records and checklist however, due to insufÔ¨Åcient cases completed by their physician of CLINICAL SIGNIFICANCE at the time, we were unable to the components of the American evaluate whether moderate intakes College of Rheumatology (ACR) Episodic intake of any type of alcohol, of alcohol and whether speciÔ¨Åc Preliminary ClassiÔ¨Åcation Criteria whether it is beer, wine, or liquor, can types of alcoholic beverage were for Gout.16 Two rheumatologists increase risk of gout attacks. associated with an increased risk of (DJH, TN) reviewed all medical Increasing amounts of alcohol intake of recurrent gout attack.10 Further, records and checklists to determine whether subjects met a diagnosis of gout treatment guidelines vary any type, even at moderate levels, can gout according to the ACR criteria, regarding recommendations about increase risk of gout attacks. using similar methods of conÔ¨Årquantity and type of alcohol Clinicians and patients with gout should mation as used in the Health Prointake.11-13 ClariÔ¨Åcation of the risk therefore consider limiting the consumpfessional Follow-Up Study.8 This for recurrent gout attacks imparted tion of all types of alcohol, not just beer. by speciÔ¨Åc types of alcoholic bevstudy was approved by the instituerages would have practical clinical tional review board of Boston implications for management of University Medical Center. patients with established gout. To address this knowledge gap, we analyzed 724 gout Ascertainment of Gout Attacks subjects that were recruited prospectively from across the For each gout attack that occurred during the 1-year followUS in an Internet-based study. We used a case-crossover up period, we collected the onset date of the attack, study design to quantify the risk of gout attack in relation anatomical location of the attack, clinical symptoms and to amount of alcohol consumption, particularly moderate signs (maximal pain within 24 hours, redness, swelling), intakes, and evaluated whether the effect on recurrent gout medications used to treat the attack (eg, colchicine, nonattacks varied by consumption of speciÔ¨Åc type of alcoholic steroidal anti-inÔ¨Çammatory drugs [NSAIDs], systemic or beverage. intra-articular glucocorticoids), and whether a health care professional was seen for attack management. This method of identifying gout attacks is in keeping with approaches METHODS used in gout trials,17-19 and the provisional deÔ¨Ånition of Ô¨Çare Study Design in patients with established gout that includes only patientThe Boston University online gout study is an Internetreported elements.20 We additionally restricted our gout based case-crossover study conducted over the period of attack deÔ¨Ånitions to those that were treated with at least one 2003-2012 to examine a set of putative risk factors for gout-related medication typically used to treat attacks (listed recurrent gout attacks. The details of the study have been above), those with Ô¨Årst metatarsophalangeal involvement, described previously.10,14,15 In brief, we constructed a Web those with maximal pain within 24 hours, those with redsite (https://dcc2.bumc.bu.edu/GOUT) on an independent ness, and those with a combination of these features (ie, secure server within the Boston University Medical Center those with at least 2, 3, or all 4 features). domain. Recruitment occurred primarily by means of an advertisement on Google linked to the search term ‚Äúgout.‚Äù Ascertainment of Risk Factors Individuals were directed to the study Web site when they Subjects were queried about the frequency and quantity of clicked on this link. The study design and timing of expoa set of putative risk factors (eg, dietary factors, medicasure assessments are illustrated in Figure 1. With this study tion use, physical activity, geography) during the 24 hours design, each subject serves as his or her own control. This before that gout attack (hazard period).15,21 The same self-matching eliminates confounding by factors that are questions also were asked over a 24-hour period when they constant within an individual but differ among study subwere attack-free (control period) at study entry (for those jects (eg, sex, race, socioeconomic status). subjects who entered the study during an intercritical period), and at 3, 6, 9, and 12 (for those subjects who entered the Study Sample study at the time of a gout attack) months of follow-up (Figure 1). The study Web site provided information about the study, and Standardized questions about alcohol intake included for interested potential participants, administered a screening the number of servings of wine, beer (including light questionnaire that collected sociodemographic information, Neogi et al Alcohol and Recurrent Gout Attacks Free of Gout Attack 1-day control period 313 Gout Attack 1-day hazard period Free of Gout Attack 1-day control period 3-month interval 1-year follow-up Study start: 02/2003 Study end: 01/2012 Gout attacks could occur at any time during the 1-year follow-up period in each subject. Hazard period refers to the 1-day period prior to a gout attack. Up to four control periods were selected from the intercritical period every 3 months during the 1 year of follow-up. Exposure to alcohol and other time-varying factors (potential confounders) were compared between hazard and control periods. The current study spanned between February 2003 and January 2012. Figure 1 Case-crossover study design and timing of exposure measurements in relation to gout attacks. beer, ciders, and malt beverages), or liquor (either straight or in a mixed drink) consumed during the prior 24-hour period for control and hazard periods. Explanation and pictorial depiction of standard serving sizes (ie, a 12ounce bottle or can of beer; a 5-ounce glass of wine; and 1-1.5 ounces of liquor)22 were provided with color images. Information on potential confounders, such as diuretic use, food and beverage intake from which purine consumption could be calculated,15 and gout-related medication also were collected during the control and hazard periods. Statistical Analysis The total amount of alcohol intake (grams/day) was estimated based on number of servings reported in a 24-hour period as ([0.57 * # of cocktails (liquor)/day] √æ [0.44 * # of bottles/cans of beer/day] √æ [0.40 * # of glasses of wine/ day]) * 28.35.23 This latter term represents 28.35 grams of alcohol per Ô¨Çuid ounce. One typical drink is approximately 15 grams of alcohol.24 We divided total amount of alcohol consumption in the hazard and control periods into 7 categories: no alcohol consumption, >0-1 drink, >1-2, >2-4, >4-6, >6-8, and more than 8 drinks. Moderate alcohol intake is considered to be no more than 2 drinks per day for men and no more than 1 drink per day for women.22 We grouped the daily consumption of each speciÔ¨Åc alcoholic beverage into the following categories based on their distribution: for wine, no wine consumption, >0-1, >1-2, and >2 servings; for beer and for liquor, no consumption, >0-2, >2-4, >4-6, and >6 servings. We examined the relation of total alcohol intake over 24 hours to the risk of recurrent gout attacks using conditional logistic regression, which takes into account the matching of each subject‚Äôs own hazard and control periods.25 In multivariable regression models, we adjusted for diuretic use, purine intake, gout-related medication use (allopurinol, colchicine, NSAIDs, other urate-lowering therapies), and water intake. To better depict the dose-response relation between alcohol consumption and risk of gout attacks, we used quadratic spline regression to smooth the dose-risk curve.26 We then evaluated the association of alcohol intake with risk of gout attacks according to subgroups deÔ¨Åned by sex, age (<55 vs 55 years), and body mass index (BMI; <30 vs 30). We also evaluated the joint effects of purine intake (<850 mg [median value for 24-hour intake] Table 1 Baseline Characteristics of Participants in the Internetbased Case-crossover Study of Gout, 2003-2012 Participant Characteristic n ¬º 724 Age, y: mean (SD), range BMI, kg/m2: mean (SD), range Male: n (%) Disease duration: mean years (SD), range White: n (%) Completed college: % Household income $50,000: % Mean number of alcoholic beverages per 24-hour period (calculated from 3380 24-hour hazard and control periods) 54.5 (12.5), 21-88 32.1 (6.9), 14.7-69.9 568 (78.5) 8.0 (9.3), 1-55 642 (88.7) 58.1 58.6 1.2 BMI ¬º body mass index. 314 The American Journal of Medicine, Vol 127, No 4, April 2014 regression adjusting for potential confounders listed above as well as consumption of the other types of alcoholic beverages. Odds Ratio RESULTS Number of servings of alcohol in prior 24-hour period Spline regression depicting the relation of total alcohol intake to the risk for recurrent gout attack. Figure 2 Effect of alcohol consumption on risk of recurrent gout attack. vs 850 mg), diuretic use, allopurinol use, colchicine use, and NSAID use with alcohol intake in the prior 24 hours. Finally, we assessed the independent effect of each speciÔ¨Åc type of alcoholic beverage with conditional logistic Table 2 There were 724 participants (mean age 54 years) who completed both hazard and control period questionnaires over a consecutive 12-month period between February 2003 and January 2012. As shown in Table 1, the majority of participants was male (78%), obese (mean BMI 32 kg/m2), and white (89%). Participants were recruited from 49 states and the District of Columbia. Of these participants, 614 (85%) met the ACR Preliminary ClassiÔ¨Åcation Criteria for Gout. Approximately 48% were on uratelowering therapy (allopurinol: 44%; other: 4%); 25% used colchicine for prophylaxis or gout attacks, while 38% used NSAIDs for prophylaxis and gout attacks. During the 1-year follow-up period, there were 1434 gout attacks, primarily occurring in the lower extremity (92%), particularly in the Ô¨Årst metatarsophalangeal joint, and had features of maximal pain within 24 hours, or redness (89%). Eighty-nine percent of these gout attacks were treated with colchicine, NSAIDs, systemic or intra-articular glucocorticoids, or a combination thereof. Approximately 44% of subjects reported any alcohol intake during hazard, control, or both periods. The mean number of standard servings of alcohol was 1.0 during a control period and 1.4 during a hazard period. The risk of recurrent gout attacks increased as the amount of alcohol consumption increased (Figure 2). While having up to one drink in a 24-hour period did not increase the risk of attack signiÔ¨Åcantly (odds ratio [OR] 1.13; 95% conÔ¨Ådence interval [CI], 0.80-1.58), consuming >1-2 drinks in a 24-hour period was associated with 36% higher risk of recurrent attack (OR 1.36; 95% CI, 1.00-1.88), compared with those with no alcohol intake in that period, indicating that a moderate amount of alcohol intake within a 24-hour period may increase the risk of recurrent gout attacks (Table 2). When we limited our analyses to only those subjects who fulÔ¨Ålled the ACR Preliminary ClassiÔ¨Åcation Criteria for Total Alcohol Intake Over the Prior 24-hour Period and the Risk of Recurrent Gout Attacks Number of Servings of Alcohol Over the Prior 24-hour Period Number of Hazard Periods (n ¬º 1434) Number of Control Periods (n ¬º 946) Crude OR Adjusted OR* (95% CI) 0 >0-1 >1-2 >2-4 >4-6 >6-8 >8 P for linear trend 856 93 121 178 94 48 44 1222 145 185 223 105 40 26 1.0 1.12 1.26 1.60 2.13 2.65 3.90 1.0 (referent) 1.13 (0.80-1.58) 1.36 (1.00-1.88) 1.51 (1.09-2.09) 1.87 (1.19-2.93) 2.33 (1.28-4.24) 3.13 (1.63-6.02) <.001 CI ¬º conÔ¨Ådence interval; OR ¬º odds ratio. *Adjusted for purine intake, allopurinol or other urate-lowering therapy, nonsteroidal anti-inÔ¨Çammatory drug, colchicine, diuretic use, and water intake in prior 24-hour period. Neogi et al Alcohol and Recurrent Gout Attacks Gout (n ¬º 614), the results did not change substantially, with the multivariable adjusted ORs (95% CI) being 1.09 (0.76-1.55), 1.36 (0.96-1.93), 1.50 (1.07-2.18), 2.05 (1.263.35), 2.50 (1.33-4.71), and 3.40 (1.63-7.09) for >0-1, >1-2, >2-4, >4-6, >6-8, and >8 servings, respectively, compared with no alcohol intake in the prior 24 hours. When we used more stringent deÔ¨Ånitions of gout attack, the results also were similar. For example, requiring at least 2 of the following features: Ô¨Årst metatarsophalangeal involvement, maximal pain within 24 hours, redness, use of a typical gout attack treatment (n ¬º 687), the corresponding multivariable adjusted ORs were 1.10 (0.78-1.56), 1.38 (0.99-1.92), 1.43 (1.03-1.99), 2.05 (1.30 -3.24), 2.42 (1.32-4.42), and 3.42 (1.76-6.67), respectively. Participants were required to complete their control period questionnaires once every 3 months. It was possible that control periods may have over-represented certain days of the week; for example, when Internet access may have been more accessible, such as in the ofÔ¨Åce. We therefore performed additional analyses according to weekday versus weekend reporting. The effect estimates of alcohol consumption for weekdays were similar to those for the weekend; the adjusted ORs of recurrent gout attacks for >12 drinks in the prior 24 hours were 1.48 for weekdays and 1.30 for weekends. Moderate alcohol consumption (ie, up to 2 drinks/day for men and up to 1 drink/day for women) was associated with a 41% increased risk of recurrent gout attacks for men (adjusted OR 1.41; 95% CI, 1.00-2.01), but not for women (adjusted OR 1.06; 95% CI, 0.49-2.30) compared with those who did not drink any alcohol in the prior 24-hour period, although there were too few women to precisely estimate this effect (P ¬º .4 for interaction by sex). The combined effects of alcohol intake with concurrent intake of purines and use of gout-related medications are shown in Table 3. Increasing numbers of servings of alcohol in combination with either high purine consumption or diuretic use were associated with higher risk of recurrent gout attacks. In contrast, use of allopurinol mitigated the effects of alcohol intake, as did colchicine, although to a lesser extent. NSAID use did not modify the effect of alcohol intake. As shown in Table 4, each type of alcoholic beverage intake was associated with an increased risk of recurrent gout attacks. Consuming >1-2 servings of wine over the prior 24 hours signiÔ¨Åcantly increased the risk of recurrent gout attack (adjusted OR 2.38; 95% CI, 1.57-3.62). For beer, having up to 2 servings and >2-4 servings were associated with a nonsigniÔ¨Åcant 29% and statistically signiÔ¨Åcant 75% higher risk for recurrent gout attack, respectively, compared with no such intake. There also was an increased risk of recurrent gout attacks with increasing amounts of liquor consumption, with those consuming >2-4 servings of such beverages having 1.67 times higher risk of an attack compared with no such intake in the prior 24-hour period. Similar Ô¨Åndings were observed when analyses were limited to those participants who reported only drinking one type of alcoholic beverage during the course of the study. 315 Table 3 Combined Effects of Alcohol Intake and Other Timevarying Risk Factors (Purine Intake, Diuretic Use, Allopurinol Use, Colchicine Use, NSAID use) on Risk of Gout Attack Exposure to Risk Factor in Prior 24 Hours Purine intake <850 mg* <850 mg <850 mg <850 mg 850 mg 850 mg 850 mg 850 mg Diuretic use No No No No Yes Yes Yes Yes Allopurinol use No No No No Yes Yes Yes Yes Colchicine use No No No No Yes Yes Yes Yes NSAID use No No No No Yes Yes Yes Yes Number of Alcohol Servings in Prior 24 Hours Adjusted OR‚Ä† (95% CI) 0 >0-1 >1-2 >2 0 >0-1 >1-2 >2 1.0 0.88 1.50 1.83 2.35 3.16 2.65 4.17 (ref) (0.54-1.45) (1.01-2.23) (1.24-2.68) (1.88-2.93) (2.00-4.99) (1.66-4.24) (2.95-5.89) 0 >0-1 >1-2 >2 0 >0-1 >1-2 >2 1.0 1.26 1.38 1.61 2.40 2.12 3.44 5.82 (ref) (0.85-1.86) (0.96-1.97) (1.17-2.20) (0.59-3.62) (1.02-4.42) (1.70-6.93) (2.94-11.53) 0 >0-1 >1-2 >2 0 >0-1 >1-2 >2 1.0 1.04 1.58 1.74 0.45 0.61 0.43 0.70 (ref) (0.70-1.55) (1.08-2.31) (1.26-2.41) (0.33-0.62) (0.32-1.17) (0.24-0.79) (0.42-1.17) 0 >0-1 >1-2 >2 0 >0-1 >1-2 >2 1.0 1.17 1.37 1.70 0.82 0.59 1.03 1.18 (ref) (0.81-1.68) (0.97-1.93) (1.24-2.32) (0.55-1.20) (0.24-1.47) (0.45-2.40) (0.63-2.19) 0 >0-1 >1-2 >2 0 >0-1 >1-2 >2 1.0 1.06 1.44 1.72 1.36 1.62 1.45 2.03 (ref) (0.72-1.56) (1.00-2.06) (1.25-2.37) (1.03-1.80) (0.84-3.16) (0.80-2.62) (1.29-3.19) CI ¬º conÔ¨Ådence interval; NSAID ¬º nonsteroidal anti-inÔ¨Çammatory drug; OR ¬º odds ratio. *Median value of purine intake in prior 24 hours. ‚Ä†Mutually adjusted for each other as well as other urate-lowering therapies and water intake. Compared with no intake of each speciÔ¨Åc type of alcoholic beverage during the prior 24 hours, the adjusted ORs for a recurrent gout attack were 3.96 (95% CI 1.84-8.52), 3.63 316 Table 4 The American Journal of Medicine, Vol 127, No 4, April 2014 SpeciÔ¨Åc Alcoholic Beverage Intake Over the Prior 24-hour Period and Risk of Recurrent Gout Attacks Number of Servings of SpeciÔ¨Åc Alcoholic Beverages Over the Prior 24-hour Period Wine 0 >0-1 >1-2 >2 P for linear trend Beer 0 >0-2 >2-4 >4-6 >6 P for linear trend Hard liquor 0 >0-2 >2-4 >4-6 >6 P for linear trend Number of Hazard Periods (n ¬º 1434) Number of Control Periods (n ¬º 1946) 1194 102 89 49 Adjusted OR* Adjusted OR‚Ä† (95% CI) 1664 133 80 69 1.0 1.26 2.34 1.35 <.001 1.0 1.25 (0.87-1.80) 2.38 (1.57-3.62) 1.41 (0.86-2.32) <.001 1124 92 99 52 67 1601 129 114 49 53 1.0 1.28 1.73 2.56 2.40 <.001 1.0 1.29 (0.91-1.83) 1.75 (1.19-2.59) 2.60 (1.40-4.81) 2.32 (1.25-4.31) .001 1199 68 60 75 31 1673 113 57 86 17 1.0 0.97 1.66 1.63 2.97 .002 1.0 0.92 (0.62-1.37) 1.67 (1.00-2.78) 1.56 (0.95-2.57) 2.79 (1.26-6.16) .005 *Adjusted for purine intake, allopurinol or other urate-lowering therapy, nonsteroidal anti-inÔ¨Çammatory drug, colchicine, and diuretic use, in prior 24hour period. ‚Ä†Additionally mutually adjusted for other types of alcohol intake. (95% CI, 1.92-6.87), and 4.44 (95% CI, 1.17-16.91) for consumption of up to 2 servings of wine, beer, and liquor, respectively. DISCUSSION Anecdotally, while alcohol has been thought to trigger gout attacks, the results from our study conÔ¨Årm that alcohol intake, potentially even moderate amounts, increases the risk of recurrent gout attacks in a short time following consumption. Further, all types of alcoholic beverages, whether it was wine, beer, or liquor, were associated, to varying degrees, with an increased risk for recurrent gout attacks. These effects were stronger in the presence of high purine intake and diuretic use, while mitigated to varying degrees by allopurinol and colchicine use; NSAIDs did not modify the effects of alcohol intake on risk of recurrent gout attacks. Ethanol ingestion can increase serum urate through both decreased urate excretion and increased urate production. Reduced renal urate excretion can occur because of lactic acidemia associated with acute excessive alcohol intake, as well as the acidemia associated with fasting that is often concomitant with such intake.27,28 Metabolism of ethanol also accelerates adenosine triphosphate degradation into uric acid precursors.28-30 While alcohol deÔ¨Ånitively has been associated with hyperuricemia,31-33 and variably associated with incident gout,7-9 the Ô¨Åndings of our study support the importance of alcohol, regardless of type, as a trigger in established gout. Why might wine not increase the risk for incident gout in an observational cohort, yet appear to increase the risk of recurrent gout attacks? One might expect the effects of ethanol to be similar regardless of the type of alcoholic beverage. Indeed, all types of alcohol can lead to increased urate levels due to a variety of mechanisms, including ethanol content, thereby increasing the risk of gout attacks. However, one may expect a greater effect of beer on hyperuricemia than other types of alcohol because it not only contains ethanol, but also has high levels of guanosine, a purine that is highly absorbable.27,34 On the other hand, individuals who drink wine often have a healthier lifestyle than those who drink beer or spirits. For instance, wine drinkers tend to buy healthier foods and follow healthier diets than beer drinkers.35-38 Thus, the lack of association between wine and incident gout from an observational study may be related to residual confounding from other healthy lifestyle factors. By using a case-crossover study design to assess the triggering effects of alcohol consumption, we minimize such ‚Äúhealthy lifestyle factors‚Äù that vary greatly among individuals but are relatively consistent within an individual. Additionally, risk factors for triggering recurrent gout attacks among individuals with established gout may not be the same as those for incident gout among individuals who are free of gout. Individuals with established gout may have altered renal handling compared with those who do not have gout (ie, at risk for incident gout), and therefore risk factors may affect the 2 groups differently. Further, the short-term Neogi et al Alcohol and Recurrent Gout Attacks effects of a risk factor may differ from its long-term effects. An example of such a paradoxical phenomenon is the wellknown increased Ô¨Çare risk during urate-lowering therapy initiation, whereas over the long term, such therapy reduces the risk of Ô¨Çares. Several characteristics of this study are worth noting. The case-crossover study is an ideal design to assess the acute effect of triggers. Because each participant serves as his/her own control, this study design eliminates the effects of time-invariant confounding factors among individuals.39 Recruitment of a large number of participants from all over the US through the Internet highlights a novel aspect of this study. Finally, the online design enabled participants to enter data in real time, thereby minimizing the potential for recall bias. Our study has some limitations as well. First, although we collected information on major potential time-varying confounders and adjusted for them in the analyses, residual confounding bias may remain. Second, because it is widely assumed that alcohol may trigger gout attacks, recall bias and differential reporting is a possibility. We attempted to minimize these biases by collecting information on a broad range of potential exposures, capturing data in realtime, and ensuring that the study participants were not primed regarding study hypotheses. Third, as with many epidemiologic studies, dietary intake was not independently veriÔ¨Åed. Fourth, allowing some Ô¨Çexibility for participants to choose which day of the week, albeit within a Ô¨Åxed time window, to complete a control period questionnaire can potentially introduce bias. Nevertheless, when we performed additional analyses stratiÔ¨Åed according to weekday versus weekend reporting, results did not vary materially. Finally, like other epidemiologic studies of gout8 and what is common in clinical practice, most of our participants did not have a crystal-proven diagnosis of gout. However, the majority in our study met ACR Preliminary ClassiÔ¨Åcation Criteria for gout or had a physician diagnosis of gout, and the clinical characteristics of participants in our study are similar to what would be expected of gout patients. In summary, the present study supports the role of episodic alcohol intake in triggering gout attacks, even for moderate amounts and regardless of type of alcohol. Thus, in addition to the general medical management of their gout, individuals with established gout should consider limiting all types of alcohol intake as another preventive strategy to reduce their risk for recurrent gout attacks. References 1. Neogi T. Clinical practice. Gout. N Engl J Med. 2011;364:443-452. 2. Zhu Y, Pandya BJ, Choi HK. Prevalence of gout and hyperuricemia in the US general population: the National Health and Nutrition Examination Survey 2007-2008. Arthritis Rheum. 2011;63:3136-3141. 3. Neogi T, Hunter DJ, Chaisson CE, Allensworth-Davies D, Zhang Y. Frequency and predictors of inappropriate management of recurrent gout attacks in a longitudinal study. J Rheumatol. 2006;33:104-109. 4. Arromdee E, Michet CJ, Crowson CS, O‚ÄôFallon WM, Gabriel SE. Epidemiology of gout: is the incidence rising? J Rheumatol. 2002;29:2403-2406. 317 5. Chen SY, Chen CL, Shen ML, Kamatani N. Trends in the manifestations of gout in Taiwan. Rheumatology (Oxford). 2003;42:1529-1533. 6. Lawrence RC, Felson DT, Helmick CG, et al. Estimates of the prevalence of arthritis and other rheumatic conditions in the United States. Part II. Arthritis Rheum. 2008;58:26-35. 7. Campion EW, Glynn RJ, DeLabry LO. Asymptomatic hyperuricemia. Risks and consequences in the Normative Aging Study. Am J Med. 1987;82:421-426. 8. Choi HK, Atkinson K, Karlson EW, Willett W, Curhan G. Alcohol intake and risk of incident gout in men: a prospective study. Lancet. 2004;363:1277-1281. 9. Hochberg MC, Thomas J, Thomas DJ, Mead L, Levine DM, Klag MJ. Racial differences in the incidence of gout. The role of hypertension. Arthritis Rheum. 1995;38:628-632. 10. Zhang Y, Woods R, Chaisson CE, et al. Alcohol consumption as a trigger of recurrent gout attacks. Am J Med. 2006;119:800.e13-e18. 11. Jordan KM, Cameron JS, Snaith M, et al. British Society for Rheumatology and British Health Professionals in Rheumatology guideline for the management of gout. Rheumatology (Oxford). 2007;46: 1372-1374. 12. Zhang W, Doherty M, Bardin T, et al. EULAR evidence based recommendations for gout. Part II: management. Report of a task force of the EULAR Standing Committee for International Clinical Studies Including Therapeutics (ESCISIT). Ann Rheum Dis. 2006;65:1312-1324. 13. Khanna D, Fitzgerald JD, Khanna PP, et al. 2012 American College of Rheumatology guidelines for management of gout. Part 1: systematic nonpharmacologic and pharmacologic therapeutic approaches to hyperuricemia. Arthritis Care Res (Hoboken). 2012;64:1431-1446. 14. Zhang Y, Chaisson CE, McAlindon T, et al. The online case-crossover study is a novel approach to study triggers for recurrent disease Ô¨Çares. J Clin Epidemiol. 2007;60:50-55. 15. Zhang Y, Chen C, Choi H, et al. Purine-rich foods intake and recurrent gout attacks. Ann Rheum Dis. 2012;71:1448-1453. 16. Wallace SL, Robinson H, Masi AT, Decker JL, McCarty DJ, Yu TF. Preliminary criteria for the classiÔ¨Åcation of the acute arthritis of primary gout. Arthritis Rheum. 1977;20:895-900. 17. Terkeltaub RA, Furst DE, Bennett K, Kook KA, Crockett RS, Davis MW. High versus low dosing of oral colchicine for early acute gout Ô¨Çare: twenty-four-hour outcome of the Ô¨Årst multicenter, randomized, double-blind, placebo-controlled, parallel-group, dose-comparison colchicine study. Arthritis Rheum. 2010;62:1060-1068. 18. Sundy JS, Baraf HS, Yood RA, et al. EfÔ¨Åcacy and tolerability of pegloticase for the treatment of chronic gout in patients refractory to conventional treatment: two randomized controlled trials. JAMA. 2011;306:711-720. 19. So A, De Meulemeester M, Pikhlak A, et al. Canakinumab for the treatment of acute Ô¨Çares in difÔ¨Åcult-to-treat gouty arthritis: results of a multicenter, phase II, dose-ranging study. Arthritis Rheum. 2010;62: 3064-3076. 20. Gaffo AL, Schumacher HR, Saag KG, et al. Developing a provisional deÔ¨Ånition of Ô¨Çare in patients with established gout. Arthritis Rheum. 2012;64:1508-1517. 21. Zhang Y, Chen C, Hunter DJ, Chaisson CE, Choi H, Neogi T. Cherry consumption and risk of recurrent gout attacks. Arthritis Rheum. 2012;64:4004-4011. 22. U.S. Department of Agriculture and U.S. Department of Health and Human Services. Dietary Guidelines for Americans 2010. 7th edn. Washington, DC: U.S. Government Printing OfÔ¨Åce; 2010. 23. Zhang Y, Kreger BE, Dorgan JF, Splansky GL, Cupples LA, Ellison RC. Alcohol consumption and risk of breast cancer: the Framingham Study revisited. Am J Epidemiol. 1999;149:93-101. 24. Willett WC, Stampfer MJ, Colditz GA, Rosner BA, Hennekens CH, Speizer FE. Moderate alcohol consumption and the risk of breast cancer. N Engl J Med. 1987;316:1174-1180. 25. Stokes ME, Davis CS, Koch GG. Conditional logistic regression. In: Stokes ME, Davis CS, Koch GG, eds. Categorical Data Analysis Using the SAS System. 2nd edn. Cary, NC: SAS Institute, Inc.; 2000: 271-322. 318 26. Greenland S. Dose-response and trend analysis in epidemiology: alternatives to categorical analysis. Epidemiology. 1995;6:356-365. 27. Eastmond CJ, Garton M, Robins S, Riddoch S. The effects of alcoholic beverages on urate metabolism in gout sufferers. Br J Rheumatol. 1995;34:756-759. 28. Fam AG. Gout, diet, and the insulin resistance syndrome. J Rheumatol. 2002;29:1350-1355. 29. Faller J, Fox IH. Ethanol-induced hyperuricemia: evidence for increased urate production by activation of adenine nucleotide turnover. N Engl J Med. 1982;307:1598-1602. 30. Puig JG, Fox IH. Ethanol-induced activation of adenine nucleotide turnover. Evidence for a role of acetate. J Clin Invest. 1984;74: 936-941. 31. Choi HK, Curhan G. Beer, liquor, and wine consumption and serum uric acid level: the Third National Health and Nutrition Examination Survey. Arthritis Rheum. 2004;51:1023-1029. 32. Gaffo AL, Roseman JM, Jacobs DR Jr, et al. Serum urate and its relationship with alcoholic beverage intake in men and women: Ô¨Åndings from the Coronary Artery Risk Development in Young Adults (CARDIA) cohort. Ann Rheum Dis. 2010;69:1965-1970. The American Journal of Medicine, Vol 127, No 4, April 2014 33. Yu KH, See LC, Huang YC, Yang CH, Sun JH. Dietary factors associated with hyperuricemia in adults. Semin Arthritis Rheum. 2008;37:243-250. 34. Gibson T, Rodgers AV, Simmonds HA, Toseland P. Beer drinking and its effect on uric acid. Br J Rheumatol. 1984;23:203-209. 35. Johansen D, Friis K, Skovenborg E, Gronbaek M. Food buying habits of people who buy wine or beer: cross sectional study. BMJ. 2006;332: 519-522. 36. Tjonneland A, Gronbaek M, Stripp C, Overvad K. Wine intake and diet in a random sample of 48763 Danish men and women. Am J Clin Nutr. 1999;69:49-54. 37. Barefoot JC, Gronbaek M, Feaganes JR, McPherson RS, Williams RB, Siegler IC. Alcoholic beverage preference, diet, and health habits in the UNC Alumni Heart Study. Am J Clin Nutr. 2002;76:466-472. 38. McCann SE, Sempos C, Freudenheim JL, et al. Alcoholic beverage preference and characteristics of drinkers and nondrinkers in western New York (United States). Nutr Metab Cardiovasc Dis. 2003;13:2-11. 39. Maclure M. The case-crossover design: a method for studying transient effects on the risk of acute events. Am J Epidemiol. 1991;133: 144-153. ALCOHOL AND OTHER DRUG DISORDERS Alcohol use disorders What‚Äôs new? James Bell C C Abstract The medical manifestations of alcohol are protean. A drinking history and high index of suspicion for alcohol misuse is part of comprehensive medical assessment. Identification of hazardous and harmful drinking, and provision of brief advice and brief intervention if indicated, should occur in any episode of healthcare. In acute settings, prevention and management of alcohol withdrawal, and of Wernicke‚Äôs encephalopathy, are key priorities. In community settings, long-term management of alcohol dependence through regular monitoring and feedback is an effective way to manage many alcohol-dependent patients. C C C Alcohol and other psychoactive drugs produce reinforcing effects by acting on the ‚Äòreward pathway‚Äô In dependent drinkers, chronic alcohol exposure produces lasting central nervous system changes, placing people at longterm risk of relapse Medical practitioners providing brief advice, monitoring and feedback to patients can produce sustained benefits in a proportion of harmful drinkers Wernicke‚Äôs encephalopathy is a medical emergency requiring supra-physiological doses of parenteral thiamine Hospitals should have local protocols for managing alcohol withdrawal, use screening questionnaires to identify people whose drinking is placing their health at risk, and use withdrawal scales to monitor admitted patients who are at risk of alcohol withdrawal Keywords alcohol dependence; alcohol use disorder; alcohol withdrawal; brief intervention; harmful drinking; hazardous drinking; WernickeeKorsakoff syndrome consumption has increased over the last two decades, and diseases due to the effects of chronic alcohol consumption have also increased. UK survey data suggest that approximately 7.1 million people in England (23% of the population aged 16e64) drink hazardously or harmfully, and 1.1 million people are dependent on alcohol. Within the UK, the prevalence of alcohol-related violence and injuries is related to the number and density of alcohol outlets and the licensing hours.2 Deprived areas suffer higher levels of alcohol-related mortality, hospital admissions, crime, absence from work, school exclusions, teenage pregnancy and road traffic accidents, linked to greater levels of alcohol consumption. Definition Current guidelines recommend that people should not regularly drink more than the daily unit limits: 3e4 units of alcohol for men (equivalent to a pint and a half of 4% v/v beer) and 2e3 units of alcohol for women (equivalent to a 175-mL glass of wine). Drinking above this level constitutes hazardous drinking, as it increases the risk of harmful consequences for the user. Hazardous use refers to patterns of use that are of public health significance despite the absence of any current disorder in the individual user.1 Pathology and pathogenesis Harmful drinking: is defined as a pattern of alcohol consumption causing health problems directly related to alcohol.1 The commonest consequences of harmful drinking result from acute intoxication, which is estimated to contribute to 1 million alcohol-related attendances to emergency departments (ED) in England on Friday and Saturday nights alone. The top 10 presentations are fall, collapse, head injury, assault, accident, generally unwell, gastrointestinal symptoms, cardiac symptoms, psychiatric problems (especially self-harm) and frequent attendance. Alcohol, like other reinforcing drugs, produces effects in the brain ‚Äòreward pathway‚Äô, an array of neural systems involving dopaminergic transmission and endorphin release. Drugs acting on this pathway reduce anxiety, and induce a sense of well-being and confidence. Neurobiological research suggests that the chronic administration of alcohol and other drugs produces enduring changes in brain neurotransmitter systems that leave the user vulnerable to relapse after abstinence has been achieved, on this basis, it is suggested that alcohol dependence should be viewed as a chronic, relapsing brain disease.3 Course of the disease Alcohol dependence: is a behavioural syndrome of impaired control over alcohol use, with drinking becoming habitual and problematic. The most reliable diagnostic feature is the presence of withdrawal symptoms and signs when alcohol is stopped or reduced. The Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) provides criteria for making the diagnosis (Table 1). This article will focus primarily on three issues of importance to all medical practitioners e identification and brief intervention, management of withdrawal, and prevention of WernickeeKorsakoff syndrome (WeKS). However, it is also important to know when to refer to specialist services (e.g. for planned detoxification and relapse prevention). Contrary to prevailing medical pessimism, many dependent drinkers stop drinking or return to controlled, low-risk drinking.4 Whereas a proportion of more severely affected individuals seeking treatment will progress to a relapsing, remitting condition, follow-up studies suggest quite good outcomes for treatment of alcohol dependence, with 50e60% of men and women with alcohol dependence abstaining or showing substantial improvements in functioning during the year after treatment.5 The course of alcohol-related disease depends on whether people cease drinking. Neurological damage due to alcohol is usually considered irreversible, although cessation of drinking can slow or halt disease progression. The 10-year survival of patients presenting with compensated alcohol-related cirrhosis, who remained abstinent or who substantially reduced their intake, was around 60%, compared to around 30% in those who continued to drink. The respective figures in those with decompensated cirrhosis were 50% and less than 10%.5 Epidemiology The incidence of alcohol-related disease in populations rises with increasing per capita alcohol consumption. In the UK, per capita Diagnosis James Bell FRACP FAChAM MD is a Consultant Addiction Physician, Kings Health Partners, London, UK. His research interest is treatment of drug dependence. Conflicts of interest: none declared. MEDICINE 40:12 Medical assessment should include a drinking history, knowledge of the potential adverse health effects of alcohol, and ability to interpret 637 Crown Copyright √ì 2012 Published by Elsevier Ltd. All rights reserved. ALCOHOL AND OTHER DRUG DISORDERS Managing withdrawal DSM Diagnostic Criteria for alcohol dependence.15 Successive episodes of alcohol withdrawal are associated with an increased severity of withdrawal and rate of complications, and with cognitive impairment; the priority is to prevent withdrawal by anticipation, monitoring and early initiation of treatment with long-acting benzodiazepines.11 Hospitals should have readily accessible protocols for monitoring and managing withdrawal. Benzodiazepines can precipitate hepatic encephalopathy, and management of people with decompensated liver disease who are withdrawing from alcohol withdrawal should involve consultation with a healthcare professional experienced in the management of patients with liver disease (see also Drugs for alcohol dependence on pages 686e687 of this issue) (Table 2). Many dependent drinkers can stop drinking without experiencing severe withdrawal. However, the presence of intercurrent illness, such as trauma, fever, or hypoxia, can contribute to severe withdrawal. Alcohol withdrawal begins within approximately 8 h of abstinence and peaks in intensity on the second or third day; symptoms usually diminish by the fourth or fifth day. Seizures tend to occur early in the course of withdrawal. All heavy drinkers admitted to hospital should be monitored for the emergence of symptoms and signs indicating withdrawal, and many hospitals use a structured withdrawal scale, such as the Clinical Institute Withdrawal Scale (CIWA)12 or Alcohol Withdrawal Scale (AWS),13 to monitor withdrawal symptoms. Hospital staff should be familiar with the use and interpretation of an alcohol withdrawal scale.1 Admission to hospital for medically assisted withdrawal should be offered to: people assessed to be at high risk of developing alcohol withdrawal seizures or delirium tremens vulnerable people (e.g. those who are frail, have cognitive impairment or co-morbidities, lack social support, have learning difficulties, or are aged 16 or 17 years).9 After withdrawal, dependent drinkers may benefit from medication e usually either acamprosate or naltrexone e to assist in relapse prevention.1 DSM-IV diagnosis of alcohol dependence requires that a patient must meet at least three out of the following seven criteria during a 12-month period C Tolerance C Withdrawal symptoms or clinically defined Alcohol Withdrawal Syndrome C Use in larger amounts or for longer periods than intended e Persistent desire or unsuccessful efforts to cut down on alcohol use C Time is spent obtaining alcohol or recovering from effects C Social, occupational and recreational pursuits are given up or reduced because of alcohol use C Use is continued despite knowledge of alcohol-related harm (physical or psychological) Table 1 laboratory investigations. Useful and widely available biological markers of alcohol misuse include a raised g-glutamyl transferase (GGT), which has a sensitivity and specificity approaching 60% in men, and 50% in women.6 History taking must be undertaken sensitively, with respect for the patients‚Äô privacy, dignity and confidentiality.1 Patients, particularly those who drink excessively and who are not seeking treatment for it, are often defensive and understate their intake. Alcohol use disorders (AUDs) are common. However, most people with alcohol-related problems present with general complaints such as insomnia, anxiety, sadness, or a range of medical problems.5 For these reasons screening tests for AUDs are recommended. NHS staff should be familiar with the Alcohol Use Disorder Identification Test (AUDIT7), a brief screening questionnaire. Patients scoring 8e14 on AUDIT are at risk of alcohol-related problems, and should receive brief advice and follow-up. Patients scoring more than 14 on AUDIT should receive brief advice and be referred for comprehensive assessment. Those scoring 20 or more should be referred for assessment and management of anticipated withdrawal.1 WernickeeKorsakoff syndrome (WeKS) results from a deficiency in vitamin B1 (thiamine) and is characterized by nystagmus and ophthalmoplegia, mental-status changes, and unsteadiness of stance and gait, symptoms that are difficult to distinguish from intoxication. This classical triad is seen in only 16% of Differential diagnosis The differential diagnosis of alcohol withdrawal is alcohol intoxication, Wernicke‚Äôs encephalopathy, and other causes of delirium e notably, hepatic encephalopathy, post-ictal state, head injury, and effects of other psychoactive drugs. Features of alcohol withdrawal Alcohol withdrawal is characterized by three primary symptom clusters: C Sympathetic nervous system overactivity e raised pulse, blood pressure, sweating, tremor, and temperature C Perceptual disturbances These may be visual, tactile, or, more rarely, auditory. Visual disturbances range from vivid dreams to illusions (misperceiving objects in the environment), to frank hallucinations (seeing objects that are not present). Tactile disturbances (formication, the sensation of things crawling on the skin) are not rare. C Cognitive changes ranging from anxiety to paranoia, and delirium Management Management of dependence/harmful drinking The medical practitioner‚Äôs role is to identify, monitor and give feedback on alcohol use and health consequences in their patients. Such brief intervention has been ranked as the most effective of all treatments for drinking problems.8 In the UK, all health professionals are advised to implement early identification and brief advice (IBA), using presentation for healthcare as a ‚Äòteachable moment‚Äô. Clinicians should link their presentation to the patient‚Äôs alcohol use and offer a 20e40-min consultation with an alcohol specialist nurse.9 Follow-up monitoring by GPs or other services is also important. Screening using GGT to identify heavy drinkers, then monitoring GGT concentration and providing patients with feedback, has been shown to produce short-term differences in drinking and hospitalization, and long-term reduction in deaths due to alcoholrelated causes.10 MEDICINE 40:12 In addition, alcohol withdrawal is characterized by insomnia, and gastrointestinal disturbances (nausea and vomiting). Seizures may occur early in the course of alcohol withdrawal Table 2 638 Crown Copyright √ì 2012 Published by Elsevier Ltd. All rights reserved. ALCOHOL AND OTHER DRUG DISORDERS 2 Babor TF, Caetano R, Casswell S, et al. Alcohol: no ordinary commoditydresearch and public policy. Oxford: Oxford University Press, 2003. 3 Gunzerath L, Hewitt BG, Li TK, Warren KR. Alcohol research: past, present, and future. Ann New York Acad Sci 2011; 1216: 1e23. 4 Dawson DA, Grant BF, Stinson FS, et al. Recovery from DSM-IV alcohol dependence: United States, 2001e2002. Addiction 2005; 100: 281e92. 5 Shuckitt M. Alcohol use disorders. Lancet 2009; 373: 492e501. 6 Conigrave KM, Degenhardt LJ, Whitfi eld JB, et al. CDT, GGT, and AST as markers of alcohol use: the WHO/ISBRA collaborative project. Alcohol Clin Exp Res 2002; 26: 332e9. 7 Saunders JB, Aaasland OG, Babor TF, de la Fuente JR, Grant M. Development of the Alcohol Use Disorders Identification Test (AUDIT): WHO collaborative project on early detection of persons with harmful alcohol consumption. Addiction 1993; 88: 791e804. 8 Miller WR, Wilbourne PL. Mesa Grande: a methodological analysis of clinical trials of treatments for alcohol use disorders. Addiction 2002; 97: 265e77. 9 NICE. Alcohol use disorders: diagnosis and clinical management of alcohol-related physical complications, www.nice.org.uk/guidance/ CG100; 2010. 10 Kristenson H, Sterling A, Nilsson J, Lindg‚Ç¨arde F. Prevention of alcoholrelated deaths in middle-aged heavy drinkers. Alcohol Clin Exp Resl 2002; 26: 478e84. 11 Lingford-Hughes A, Welch S, Nutt D. Evidence-based guidelines for the pharmacological management of substance misuse. Addict Comorbidity: Recommendations Br Assoc Psychopharmacol J Psychopharmacol 2004; 18: 293. 12 Shaw JM. Development of optimal treatment tactics for alcohol withdrawal. I: assessment and effectiveness of supportive care. J Clin Psychopharmacol 1981; 1: 382e7. 13 Saunders JB. Drug treatment in alcoholism. In: Burrow GD, Norman TR, eds. Drugs in Psychiatry, vol. 4. Amsterdam: Elsevier, 1987; 343e369. 14 Sechi G, Serra A. Wernicke‚Äôs encephalopathy: new clinical settings and recent advances in diagnosis and management. Lancet Neurol 2007; 6: 442e55. 15 American Psychiatric Association. Diagnostic and statistical manual of mental disorders, 4th edn. Washington: DC, text revision, 2000. patients. In Western societies, WeKS is usually observed in heavy drinkers, and autopsy studies indicate that the diagnosis was missed in 75 e80% of cases. About 80% of patients with Wernicke‚Äôs encephalopathy who survive develop Korsakoff‚Äôs syndrome, a disabling disorder characterized by severe memory defects. WeKS is a medical emergency, and doses of thiamine between 100 mg and 250 mg per day apparently may not restore central nervous system vitamin status or improve clinical signs.14 Patients who have signs indicative of Wernicke‚Äôs encephalopathy should be treated empirically with high-dose parenteral thiamine given three times daily for 2e3 days. Prognosis and explanation People with alcohol-related disease should be advised that their prognosis depends on ceasing or reducing drinking. Most clinicians recommend long-term abstinence for dependent drinkers, due to the risk of relapse and rapid re-instatement of dependent drinking. However, in patients not willing to consider this, reducing levels of alcohol consumption may also produce significant health benefits. Follow-up Whether with hospital services, local specialist services, or with the GP, follow-up should be offered to all patients with alcohol dependence or alcohol-related disease. People identified as drinking above safe levels should be reviewed by their GP. Prevention Doctors should offer prophylactic parenteral thiamine followed by oral thiamine to harmful or dependent drinkers: if they are malnourished or at risk of malnourishment or they have decompensated liver disease and in addition they attend an emergency department or are admitted to hospital with an acute illness or injury.9 A REFERENCES 1 NICE. NICE clinical guideline 115 alcohol-use disorders: diagnosis, assessment and management of harmful drinking and alcohol dependence, www.nice.org.uk/guidance/CG115; 2011. Practice points Task C Identification Action Alcohol history, diagnosis, screening questions (e.g. AUDIT) Knowledge Alcohol content of drinks, alcohol-related disease, biological markers Skill Sensitive to stigma, preserve privacy and dignity C Prevent complications Prophylactic thiamine, monitor withdrawal manage withdrawal Manage WeKS Signs and symptoms of withdrawal, WernickeeKorsakoff Vigilance regarding risk, use of withdrawal scales (e.g. CIWA) Differential diagnosis C Provide advice and arrange FU Give feedback on health risks, prognosis. Refer for specialist follow up Recommended levels of drinking Local referral agencies Judgemental or critical responses are often counterproductive MEDICINE 40:12 639 Crown Copyright √ì 2012 Published by Elsevier Ltd. All rights reserved. Aligning emergency care with the triple aim: Opportunities and future directions after healthcare reform * Shantanu Agrawal, , * Patrick H. Conway Show more doi:10.1016/j.hjdsi.2014.05.005 Get rights and content Abstract The Triple Aim of better health, better care, and lower costs has become a fundamental framework for understanding the need for broad health care reform and describing health care value. While the framework is not specific to any clinical setting, this article focuses on the alignment between the framework and Emergency Department (ED) care. The paper explores where emergency care is naturally aligned with each Aim, as well as current barriers which must be addressed to meet the full vision of the Triple Aim. We propose a vision of EDs serving as a nexus for care coordination optimally consistent with the Triple Aim and the requirements for such a role. These requirements include: (1) substantial integration in coordinated care models; (2) development of reliable and actionable data on ED quality, population health, and cost outcomes; (3) specific initiatives to control and optimize ED utilization; and (4) payment models which preserve surge and disaster response capacity. Keywords * Emergency care; * Coordinated care; * Payment innovation; * Health policy 1. Introduction: the Triple Aim The Triple Aim has become a fundamental framework for understanding the need for broad health care reform since described in a 2008 article by Berwick and colleagues.1 At its core, the Triple Aim is a “system of linked goals” designed to achieve a high value, equity-based health care system, one that “contribute[s] to the overall health of populations while reducing costs.”2 The first “aim” is Better Care or improving the individual experience of care. This aim is frequently articulated by the six drivers for improvement in the Institute of Medicine (IOM) report Crossing the Quality Chasm – safe, effective, timely, patient-centered, equitable, and efficient care delivery. The second aim, Better Health, captures improvement in the overall health of populations, which includes traditional health care services and disease prevention and health promotion. This aim seeks to integrate numerous aspects of population health, such as socioeconomic, physiological, and behavioral factors to lower disease burden, reduce mortality, and improve health and functional status. The final aim, Lower Costs, encapsulates the full range of expenses in the health care system – public and private payer, consumer out-of-pocket, public health, and indirect expenditures – to understand and lower the true cost of care for populations. The three aims of this framework are in constant tension and can at times be in competition or complementary to one other. A particular delivery system change, such as a new expensive medication, may increase the cost of care while improving care provision or population health. Or provision of services in a low cost environment may improve the timeliness of care delivery but have negative impact on individual or population health outcomes. The framework requires a broad time horizon and viewpoint to be effective; over time, a responsible delivery system could work to correct such imbalances by lowering per capita cost while maintaining quality outcomes and access to care. The Triple Aim does not necessarily require equipoise today, but prioritizes equity above all else and demands a constant drive towards equipoise. The required balance and need to optimize all three aims is what distinguishes this framework from the current market and regulatory approach. Relatively little work has been done to apply the Triple Aim framework to particular clinical environments. This article focuses on the alignment between the Triple Aim and Emergency Department (ED) care, so often depicted as too expensive, uncoordinated, and unintegrated with broader delivery reform. ED care could be viewed as being at odds with the Triple Aim. This paper will explore where emergency care is naturally aligned with each Aim of the framework, as well as current barriers to alignment (Fig. 1). We will also discuss specific innovations in emergency care pushing the field further along toward realization of the Triple Aim. Finally, we will propose a vision of EDs serving as a nexus for care coordination and the requirements for such a role. Fig. 1. Alignment of Emergency Care with the Triple Aim. Figure options 2. Better care: efficient ED diagnosis and care of complex conditions Optimizing care delivery for individuals requires that it be safe, effective, timely, patient-centered, equitable, and efficient. On many of these metrics, ED care shows significant alignment with this Aim. Unlike many specialties, IOM drivers for improvement are explicit in much of ED care. EDs are equipped and designed to quickly and effectively manage acute and life-threatening conditions and are judged routinely on timeliness and efficacy: 90 min for percutaneous intervention of certain heart attacks, three hours for diagnosis and treatment of stroke, four hours to antibiotics in pneumonia, and the “golden hour” of trauma care are some examples. All of this occurs in an unscheduled, unpredictable, and open environment under tremendous volume pressure: the number of ED visits has increased by 34% over the last 15 years and now exceeds 130 million visits annually.3 Improving throughput and efficiency through lean processes is a common approach to enhancing the timeliness and patient-centeredness of ED care.4 EDs are also increasingly being demanded by patients for “first contact care.” Over the last 10 years, treatment location for acute care visits has been shifting from physician offices to EDs – though ED physicians comprise less than 5% of the U.S. physician workforce, they now manage and treat over one-quarter of all acute care encounters.5The highest increase in ED visits by time of day has been during traditional outpatient office hours.6 There are numerous factors leading to this shift, including symptom severity, convenience, the lack of other options, and limited hours or availability of primary care settings. Lack of insurance coverage does not appear to be the primary determinant of this trend, as many studies have documented a correlation between increasing insurance coverage and increasing ED utilization.7, 8, 9 and 10 Outpatient physicians too are demanding ED care for their patients, primarily to expedite diagnostic workups.11 and 12 A major benefit EDs offer is the availability of numerous resources, including a wide range of diagnostics, procedures, services, and access to specialty care. EDs are able to leverage these resources with efficiency gains over other outpatient settings. A patient requiring laboratory and radiologic testing and specialist consultation for the evaluation of a new or changing condition will often be referred to the ED, where such an evaluation can be performed in hours instead of days. The improvement in patient experience can be dramatic. Numerous factors contribute to this trend, including increasingly busy primary care providers, decreasing reimbursement, and the growing burden of complex patients. For both patients and physicians, ED care can offer value unduplicated in other settings. 2.1. Current challenges to care: lack of longitudinal care and crowding Despite their role in improving the individual experience of care, EDs face significant challenges in this aspect of the Triple Aim. One critical limitation is the lack of integration in longitudinal care. EDs typically adopt an episodic approach to care. Emergency care is built to treat symptoms at presentation, not diagnoses over a broad arc of care.13 This tension, for example, produces significant reluctance to perform screening tests in the ED (e.g., HIV screening) or manage chronic conditions without an acute component during the presentation (e.g., hypertension without hypertensive urgency or emergency). Such limits are consistent with an episodic, fee-for-service approach but discordant with a patient-centered, efficient system which would demand reasonable optimization of every patient encounter. Buttressing this attitude is an emerging view of ED care as a preventable, costly departure from a higher, more elegant care plan – essentially a failure of the system.14Policy-makers and health system analysts will, for example, construct episodes or bundles of care which presume that little to no ED treatment is required, despite the numerous co-morbidities, lack of socioeconomic support, or simple disease progression among some patients which all but predict the need for ED care. This view also ignores the current economic model of most EDs, which rely on the rapid disposition of easily managed, low acuity patients for profitability; eliminating such visits could lead to ED closures and harm the social capacity to provide emergency or disaster care without alternative financing. In addition, EDs continue to be poor environments for the experience of care for some patients. Many EDs, particularly those in urban settings, are overrun and forced to have long waiting times or frequent delays in patient care progression. The stress on EDs is evident: between 2001–2008, ED visits increased 15%, average ED length of stay increased 21%, and average occupancy increased 27%.15 This occurs despite numerous systems and processes to expedite patient flow – such as placing stretchers for patient evaluations in hallways and ordering tests and diagnostics at triage. Numerous studies have demonstrated that EDs are getting more crowded, and a 2006 IOM report identified crowding as one of the five most important issues facing emergency care in the United States.16, 17 and 18 ED crowding and boarding – the “holding” of admitted patients in the ED for extended periods – stem from numerous factors typically outside of ED control. Rather, they are the outcome of elective surgical schedules, hospital staffing levels, and broader hospital occupancy and length of stay issues. In essence, ED log-jam is a downstream symptom of hospital-wide forces and decision making. ED patients frequently bear the ethically challenging consequences of competing interests within a hospital to maximize elective procedures and admissions while keeping the ED open for acute and unscheduled care.19 Crowding and boarding can have significant consequences on health care outcomes and the quality of ED care. A comprehensive review of over 360 research studies found ED crowding leads to a statistically significant increase in in-hospital mortality, particularly for intensive care unit (ICU) admissions; left-without-being-seen (LWBS) rates; and timeliness of critical clinical interventions.20 The authors also identified anecdotal evidence linking crowding to adverse events and errors, as other studies have found.21 Finally, the experience of crowding and its outcomes are borne mainly by populations in urban and socioeconomically deprived areas. This speaks directly to its ethical challenges and equitability of care.1 3. Better health: ED care for complex and safety net Patients Berwick and colleagues identified necessary preconditions for implementation of the Triple Aim, among them specifying a “population of concern” for which better health at lower cost can be pursued. EDs are open to all so is it possible to define a specific population of focus for ED care? By virtue of the Emergency Medical Treatment and Labor Act (EMTALA) and the changing role of ED care already discussed, EDs principally serve two specialized and important populations: the disenfranchised and medically complex patients. First among the disenfranchised are uninsured or underinsured patients, the traditional safety net role. About 17% of ED visits are by patients without insurance – though to be clear patients with insurance are equally or more likely to utilize the ED.22 The salient difference is that the un- or under-insured have comparatively fewer options for care.23Uninsured patients have less access to primary and specialty care, often facing appointment delays of up to 21 days. EDs address a higher proportion of their needs, even for conditions which are better handled in other outpatient settings. The result is a higher volume and cost burden on EDs.24 In addition many social ills share the common pathway of ED care. Substance abuse and withdrawal, intimate partner violence, child or elder abuse and neglect are part of the epidemiologic undercurrent of emergency care.25Implementation of the Affordable Care Act (ACA) is expected to significantly reduce the uninsured population through Medicaid and private insurance expansion, but the ED role in serving the disenfranchised will certainly continue. Sick patients with numerous co-morbidities are a second ED “population of concern.” Complex care has increasingly become the norm for emergency physicians. Owing to the growing concentration of acute care visits and the rising use of EDs as diagnostic centers, numerous studies have documented growing medical complexity. Visits among adults older than 65 years of age have increased about 25% over the last 10 years.12 and 26 Nearly every indicator of medical utilization has also increased: use of radiologic studies has increased by about 10%, laboratory testing by over 15%, medication and intravenous fluid administration between 5–15%, procedure rates by about 6%, and lengths of stay by 27%.6 Perhaps most telling is that ICU admissions have risen dramatically: EDs are the single largest source of ICU admissions and the rate of admission increased nearly 50% from 2002–2008.27 This is three times faster than general ED visit rates, implying a huge increase in ED critical care delivery. ED-based intensive care has been credited with significant survival increases in severe sepsis, heart attack, trauma, stroke, and cardiac arrest. Alternative explanations for increased ED intensity of services have been raised, discussed below, but few analysts have argued that ED care has not gotten more complex. 3.1. Current challenges to health: lack of outpatient care options and measuring ED impact As discussed earlier, the lack of significant longitudinal care or adequate integration with such care is the single greatest barrier EDs face with respect to population health. This is especially true for the specific populations of focus in EDs. Despite offering care to the un- or under-insured, EDs do not typically offer preventive services or elective specialty care. Acutely ill, complex patients who are admitted to inpatient settings do not currently require the same level of longitudinal thinking, but this paradigm will shift as delivery and payment reform emphasize home and community based services or reduced hospital admissions. EDs will need to offer or access outpatient solutions for continued management of these patients in order to meet Triple Aim goals. The same holds true for social services addressing substance abuse, violence, or other psychosocial issues. Related to the issue of longitudinal integration is the barrier of measuring ED impacts on the quality and outcomes of population health. As discussed earlier, EDs are not typically included in broader “episodes” of care; characterizing their influence on disease burden, mortality reduction, and improved health and functional status is therefore highly limited. Comprehensive ED-based metrics beyond simple process measures have not been designed, standardized, or implemented. Doing so will require a cultural shift in how emergency care integration is envisioned, as much as technical problem-solving and additional research. 4. Lower costs: a high cost but complex picture Cost is particularly challenging for ED care. A narrative of ED-as-cost-center has emerged as payment and delivery reform take hold, and one often cited solution is to eliminate ED utilization.14 The reasoning is understandable. As detailed earlier, ED care is increasingly resource intensive. While patients are sicker and more complex, part of the utilization can be attributed to a lack of knowledge of patient baseline status, diagnostic duplication, the push to eliminate rare but emergent diagnoses, defensive medicine, easy availability of expensive technology, and perverse reimbursement incentives. Once thought to have low marginal costs due to efficiencies of scale, analyses have demonstrated EDs to have marginal costs $200–400 higher than other outpatient settings.28 Additionally, ED billing charges have risen at a much higher rate than other clinical settings over the last decade; it has emerged as an outlier compared to other billing in Medicare Part B.11 and 29 There are numerous factors underlying this trend – but EDs are clearly more expensive than originally understood. Emergency physicians also routinely make one of the most expensive decisions in medicine – the choice of whether to admit a patient to the hospital. Admissions from the ED have risen about 17% over the last decade, accounting for nearly all of the 4% growth in this metric and about half of all elective and non-elective inpatient admissions.12Admissions from other care settings actually declined over the same period. The financial impact of admissions will likely continue to rise, particularly as readmission penalties are applied. ED cost is far more complicated, however, than these details would suggest. First, EDs comprise a small portion of national healthcare expenditure, just 2–4%.12 They contribute to about 10% of all outpatient spending, which is roughly equivalent to their portion of outpatient visits. Eliminating ED costs would not solve broader healthcare solvency issues. Second, there is evidence that higher intensity ED care may actually lower the rate of preventable hospitalizations. The mean expense of ED visits not resulting in admission increased 77% over the last decade, but costly diagnostic testing such as CT scans or other interventions may obviate far more costly inpatient evaluations.6, 11 and 30 Patients on the “borderline” of admission can be assessed and observed to a sufficient degree for discharge. Third, even if lower ED utilization is the ideal outcome, it remains unclear how to divert patients or shift them to less expensive settings on a sustainable and large-scale basis. There are no universally accepted standards of what defines an urgent condition versus those that could be evaluated in other outpatient settings. Clear admission triggers are also lacking. ED triage itself demonstrates this difficulty as presentation conditions do not always correlate with discharge severity; predicting who will be admitted or what interventions will be necessary is imperfect.31 Illness severity, compromised follow-up conditions, and EMTALA requirements limit how much volume can simply be diverted away from EDs. Also at stake is preserving emergency care capacity such that EDs will be staffed and capable of dealing with a variety of disasters or surge needs, which would not be feasible under current payment models with significant decreases in ED volume.32 and 33 5. Continued innovation: EDs as a nexus for coordination Innovation is at the heart of achieving Triple Aim goals. Several examples are available of new approaches and disruptions to traditional ED care delivery which are helping to meet challenges in care and lower barriers to achieving the Triple Aim. 5.1. Integration in coordinated models Several public and private payment and delivery reforms are underway – through, for example, the Affordable Care Act – which incentivize care coordination. Reforms include the creation of accountable care organizations (ACO), use of patient-centered medical homes, greater emphasis on home and community based services, implementation of value-based purchasing, and application of payment penalties for lack of coordination (e.g., readmission penalties). Integration of ED care is an inherent requirement for success in these programs.14 All of these efforts emphasize improved quality and performance measurement, availability of primary care to appropriately leverage ED care, and a focus on high cost and complex conditions which are among the most commonly evaluated in EDs.32 and 34 5.2. Promotion of high value ED utilization There are numerous examples of interventions designed to limit and focus ED care to expunge elements of waste discussed earlier. Some solve common problems which can lead to excess utilization – for example, the availability of prior diagnostic results to discourage redundant testing.35 Other solutions include placing hospitalists or primary care providers in EDs to offer more longitudinal services and consistent follow-up, which can reduce ED-based services. Kaiser Permanente Mid-Atlantic, for example, provides scheduled ED care in multi-specialty, free-standing centers which increase access to primary care services and provide substantial specialist support to ED physicians. Technology enhancements, such as widely accessible and interoperable electronic health records (EHR) or telemedicine can offer similar benefits to EDs on a macro scale. 5.3. Specialty EDs New models of ED care have emerged with a focus on a variety of patients. Urgent-cares and retail clinics emphasize less acute cases with focused and limited evaluations and quick disposition, which also improve patient experience of care. Such settings challenge the waste and expense of traditional ED care, especially for simpler presentations. Other examples include EDs focused on geriatric care, designed to optimize complex care of the elderly to reduce complications and admissions, and “MidTrack” service areas, which expedite care for patients too complex for fast-tracks but lower priority than the sickest patients.36 Through these divergent approaches, EDs are tailoring solutions in a particular service area for patient groups in greatest need of delivery improvements. 5.4. Efficiency and lean processes To reduce ED crowding, length of stay, and timeliness of acute care processes, several centers have focused on parallel processing models and other principles of lean production. One health care system leader, ThedaCare, has been implementing lean manufacturing approaches for years across numerous clinical services including the ED.37 Among several positive outcomes, improved management of acute MI in the ED has led to door-to-balloon times of 37 min on average, meeting timeliness requirements in 100% of cases. Other approaches include Rapid Entry and Accelerated Care at Triage (REACT) and team triage, which focus on rapid intake at triage and initiation of appropriate evaluations.38 These examples provide important lessons for the continued evolution and improvement of ED care. A more fundamental question is asking what role EDs should play in broader care coordination. How can they serve as an important nexus for coordination? Doing so could significantly alter their role as payment and delivery reforms take hold. Patients and providers are conveying a straightforward message: certain patients are best served by the resources and capabilities of EDs due to the acuity or complexity of their conditions. Attempting to decrease the role of EDs in these cases would require substantial, costly, and likely undesirable alteration of other clinical settings. It may also increase the risk of delay of necessary evaluation or treatment. In addition, despite the expected decline in the uninsured population from implementation of the Affordable Care Act (e.g., expansion of private insurance, coverage of pre-existing conditions, coverage under parental plans), EDs will continue to provide a social safety net for the significant remaining numbers of uninsured individuals and to address a host of psychosocial issues. Serving as a nexus for care coordination would mean embracing these functions to leverage ED care when it is most appropriate and necessary. There are several preconditions for such a change. First, ED care would need substantial integration in different models of care, whether in medical homes, ACOs, capitation, bundled payment, or other delivery system approaches. Some of this integration is structural, such as shared access to patient information – electronic health records, health information exchanges – improved communication, and performance management systems. This integration would extend the boundaries of the ED and create a continuum between outpatient and inpatient services through the nexus of emergency care. Rather than a discordant approach to emergency care, EDs should be considered an extension of outpatient care when patient acuity or complexity demands expedited and high intensity evaluation or treatment. Even expensive evaluations may prove to be cost-effective if focused by longitudinal care providers, informed by ED expertise, and able to leverage easily accessible diagnostics and specialty care. Patients could be offered the right care at the right time and ultimately spend fewer days in the hospital. Cultural integration across providers is also key to allowing shared decision-making, appropriately prioritizing ED patients, and, importantly from an ED standpoint, distributing risk, especially when dealing with complex care or incentivizing outpatient management strategies. ED providers cannot shoulder the burden of medical risk alone as processes and systems are put in place to divert away from inpatient care. Finally, integration is needed across care processes, functions, and roles. Case management and care coordinators need to be more visible members of the ED team, to communicate across the delivery system, reduce duplicative testing, and ensure smooth transitions and hand-offs. While this would address a great many issues, EDs continue to need support in providing safety net care. Public health systems should consider establishing medical homes for vulnerable populations, to coordinate with EDs just as any other integrated system would do. Secondly, EDs must obtain more reliable and actionable data on their quality, population health, and cost outcomes. This means research, development, and validation of measures which incorporate ED care in broader episodes and bundles of care and more specific understanding of the impact of crowding and boarding on the experience and quality of care. Understanding the impact of EDs to a far greater degree is necessary to leverage them appropriately. Ultimately, EDs can function financially under a global payment or capitated framework, but doing so would require far more detailed data on the real costs of appropriately leveraging ED care, as described, rather than interpreting emergency care as capitation failures or non-value driving utilization. Thirdly, initiatives to control and optimize ED resource utilization must be enacted. This includes continuing efforts to make EDs more efficient by stream-lining processes and implementing principles of lean production. ED volume and the stress it places on providers and safety processes is clearly an issue which must be addressed. Capitated systems will need to assess the number of ED beds required to address patient and outpatient provider needs – this may require stabilizing the current decline in beds. EDs need staffing levels and other methods of volume control consistent with EMTALA which better manage current safety and quality risks. Since EDs are evaluating sicker, more complex patients and becoming the primary feeder of ICU cases, staffing should be consistent with standards already in place for higher intensity clinical environments, both in the types of providers and ratio of providers to patients. Specialized EDs or clinical areas dedicated to certain patient types may facilitate this and allow matching of physician extenders to lower acuity cases, while ED physicians and trained intensivists can focus on higher acuity. Delays in necessary care with concomitant adverse outcomes cannot be tolerated in high-quality, integrated systems. This commitment will require making difficult choices which prioritize ED throughput (e.g., bed availability, consultant support) over elective procedures and other financial interests of hospital administration in order to eliminate significant crowding and boarding. Both public and private payers should consider payment reforms to incentivize this strategic need and movement away from boarding as an attractive financial choice. Finally, regardless of the objectives of any single payment or delivery reform, the surge and disaster response capacity of EDs must be preserved as part of the broader national emergency system. No other clinical setting is capable of stepping into this role and offering the same degree of efficiency and impact. Novel payment models must incorporate ED needs and continue investment in this public resource. For this reason, while perhaps the majority of ED revenue can ultimately be capitated with other at-risk models, such as medical homes or ACOs, there may always need to be at least a component of alternative financing. This would incentivize longitudinal care and cost containment while preserving an escape valve for social preparedness. Unlike other clinical environments, emergency care is subject to the most extreme risk by definition and is impacted directly by both natural and man-made disasters. Research can elucidate what staffing and preparedness levels are required to preserve response capacity; paying for this capacity may require broader systems of “emergency capitation” – at the regional or state level for example – or a continued fee-for-service approach for the most unexpected cases. Such mixed financing may be essential for optimal risk management. Emergency Departments connect inpatient and outpatient care and should be major players in health system transformation. As opposed to being viewed as “a setting to be avoided,” delivery systems need to focus on the most appropriate use of emergency care that help achieve better care and better health at lower cost. 6. Disclosure The views expressed in this manuscript represent the authors? views and not necessarily the views or policies of the Centers for Medicare & Medicaid Services. EBioMedicine 2 (2015) 467‚Äì475 Contents lists available at ScienceDirect EBioMedicine journal homepage: www.ebiomedicine.com Original Article Allogeneic Mesenchymal Stem Cells Restore Endothelial Function in Heart Failure by Stimulating Endothelial Progenitor Cells Courtney Premer a,1, Arnon Blum b,1, Michael A. Bellio a, Ivonne Hernandez Schulman a, Barry E. Hurwitz c, Meela Parker c, Christopher R. Dermarkarian a, Darcy L. DiFede a, Wayne Balkan a, Aisha Khan a, Joshua M. Hare a,‚Åé a b c Interdisciplinary Stem Cell Institute, University of Miami Miller School of Medicine, FL, USA Department of Medicine and Cardiology, Baruch Padeh Poria Hospital, Bar Ilan University, Lower Galilee 15208, Israel Department of Psychology, University of Miami Miller School of Medicine, FL, USA a r t i c l e i n f o Article history: Received 25 February 2015 Received in revised form 25 March 2015 Accepted 27 March 2015 Available online 28 March 2015 Keywords: Regenerative medicine Nitric oxide Vascular endothelium-dependent relaxation Vasculogenesis Autografts a b s t r a c t Background: Endothelial dysfunction, characterized by diminished endothelial progenitor cell (EPC) function and Ô¨Çow-mediated vasodilation (FMD), is a clinically signiÔ¨Åcant feature of heart failure (HF). Mesenchymal stem cells (MSCs), which have pro-angiogenic properties, have the potential to restore endothelial function. Accordingly, we tested the hypothesis that MSCs increase EPC function and restore Ô¨Çow-mediated vasodilation (FMD). Methods: Idiopathic dilated and ischemic cardiomyopathy patients were randomly assigned to receive autologous (n = 7) or allogeneic (n = 15) MSCs. We assessed EPC-colony forming units (EPC-CFUs), FMD, and circulating levels of vascular endothelial growth factor (VEGF) in patients before and three months after MSC transendocardial injection (n = 22) and in healthy controls (n = 10). Findings: EPC-colony forming units (CFUs) were markedly reduced in HF compared to healthy controls (4 ¬± 3 vs. 25 ¬± 16 CFUs, P b 0.0001). Similarly, FMD% was impaired in HF (5.6 ¬± 3.2% vs. 9.0 ¬± 3.3%, P = 0.01). Allogeneic, but not autologous, MSCs improved endothelial function three months after treatment (Œî10 ¬± 5 vs. Œî1 ¬± 3 CFUs, P = 0.0067; Œî3.7 ¬± 3% vs. Œî-0.46 ¬± 3% FMD, P = 0.005). Patients who received allogeneic MSCs had a reduction in serum VEGF levels three months after treatment, while patients who received autologous MSCs had an increase (P = 0.0012), and these changes correlated with the change in EPC-CFUs (P b 0.0001). Lastly, human umbilical vein endothelial cells (HUVECs) with impaired vasculogenesis due to pharmacologic nitric oxide synthase inhibition, were rescued by allogeneic MSC conditioned medium (P = 0.006). Interpretation: These Ô¨Åndings reveal a novel mechanism whereby allogeneic, but not autologous, MSC administration results in the proliferation of functional EPCs and improvement in vascular reactivity, which in turn restores endothelial function towards normal in patients with HF. These Ô¨Åndings have signiÔ¨Åcant clinical and biological implications for the use of MSCs in HF and other disorders associated with endothelial dysfunction. ¬© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction Heart failure (HF) remains a leading cause of morbidity and mortality in the United States and worldwide (Go et al., 2014; Bui et al., 2011). The failing circulation is characterized not only by depressed cardiac function, but also by endothelial dysfunction (Blum, 2009). Endothelial dysfunction‚ÄîdeÔ¨Åned by impaired Ô¨Çow-mediated vasodilation (FMD) and endothelial progenitor cell (EPC) function‚Äîproduces increased systemic vascular resistance, which augments stress on the failing heart, ‚Åé Corresponding author at: The Interdisciplinary Stem Institute, University of Miami Miller School of Medicine, Biomedical Research Building, Room 908, 1501 NW 10th Ave., Miami, FL 33136, USA. E-mail address: JHare@med.miami.edu (J.M. Hare). 1 These authors contributed equally to this work. and contributes to HF symptomology (Marti et al., 2012). Endothelial dysfunction is also a crucial component of the pathophysiology of numerous cardiovascular (CV) disorders and manifests in patients with CV risk factors such as atherosclerosis, hypertension, and diabetes mellitus (Schulman et al., 2006). Moreover, EPCs regulate the health of the vasculature by incorporating into the endothelium, replacing injured endothelial cells, and secreting angiogenic factors that activate mature endothelial cells (Zampetaki et al., 2008). Notably, HF patients have decreased circulating EPC levels and bioactivity (Schmidt-Lucke et al., 2005). Mesenchymal stem cells (MSCs), under evaluation as a regenerative therapeutic approach for HF (Hare et al., 2012; Heldman et al., 2014; Karantalis and Hare, in press), have the potential for clinical beneÔ¨Åt in CV disease by virtue of their antiÔ¨Åbrotic, anti-inÔ¨Çammatory, and proangiogenic properties (Williams and Hare, 2011; Cao et al., 2015), and http://dx.doi.org/10.1016/j.ebiom.2015.03.020 2352-3964/¬© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 468 C. Premer et al. / EBioMedicine 2 (2015) 467‚Äì475 their ability to stimulate endogenous progenitor cells (Hatzistergos et al., 2010; Chen et al., 2008). Given this capacity of MSCs and the role of impaired EPCs in human HF (Werner et al., 2005; Shantsila et al., 2007a), we hypothesized that MSCs would stimulate the bioactivity of circulating EPCs and improve endothelial function in the failing circulation. Accordingly, we tested the hypothesis that MSCs stimulate EPC function and augment vascular relaxation in patients with HF due to either idiopathic dilated cardiomyopathy (DCM) or ischemic cardiomyopathy (ICM). Our results show that allogeneic, but not autologous, MSCs improve EPC bioactivity and endothelial function in HF patients, regardless of the etiology. These Ô¨Åndings demonstrate a novel clinical beneÔ¨Åcial effect of allogeneic MSCs transplantation in patients with HF and have implications for all disorders associated with endothelial dysfunction. CD45 by Ô¨Çow cytometry, and no growth of bacteria. On average, autologous MSCs were 92.9 ¬± 3.2% CD105 +/CD45 ‚àí, and the allogeneic MSCs were 96.4 ¬± 0.04% CD105+/CD45‚àí (Supplemental Fig. 1). 2.4. Endothelial Colony Forming Units (EPC-CFUs) Peripheral blood samples were obtained from patients before and three months after MSC injection. EPCs were isolated from samples using Ficoll-Paque and Ô¨Åve million cells were seeded on 6-well Ô¨Åbronectin-coated dishes (BD biosciences) in CFU-Hill medium (stem cell technologies, cat#05900) (Hill et al., 2003; Solomon et al., 2012). The non-adherent cells were collected 48 h later and one million cells were seeded on 24-well Ô¨Åbronectin-coated dishes. On day Ô¨Åve, EPCCFUs were counted in Ô¨Åve wells and the average was obtained. 2. Materials and Methods 2.5. ImmunoÔ¨Çuorescence 2.1. Experimental Design The objective of this study was to analyze the change in endothelial function measured by EPC-CFUs and FMD% after either allogeneic or autologous MSC administration. Patients enrolled in ongoing clinical trials with both ischemic and non-ischemic cardiomyopathy were recruited to this endothelial trial. Power calculations were performed for both our primary endpoint, FMD%, and our secondary endpoint, EPC-CFUs. In order to study the response from autologous versus allogeneic patients for both endpoints, we needed Ô¨Åve autologous and Ô¨Åve allogeneic subjects to be able to reject the null hypothesis. Data was collected at baseline and 3 months post-treatment, and treatment group was blinded. 2.2. Study Population This study entitled ‚ÄúStudying Endothelial Function and Endothelial Progenitor Stem Cells' Colonies Before and After Heart Mesenchymal Stem Cell Transplantation‚Äù is a University of Miami Institutional Review Board approved endothelial trial (#20110543). The HF patients recruited for this study are enrolled in POSEIDON-DCM (NCT01392625), ‚ÄúA Phase I/II, Randomized Pilot Study of the Comparative Safety and EfÔ¨Åcacy of Transendocardial Injection of Autologous Mesenchymal Stem Cells Versus Allogeneic Mesenchymal Stem Cells in Patients with Nonischemic Dilated Cardiomyopathy‚Äù (Mushtaq et al., 2014) and in TRIDENT (NCT02013674), ‚ÄúThe Transendocardial Stem Cell Injection Delivery Effects on Neomyogenesis Study‚Äù. In PODEIDON-DCM, patients were randomized to receive by transendocardial delivery either 100 million autologous or allogeneic MSCs. Autologous MSCs were derived from the patient's bone marrow (iliac crest aspiration) 4‚Äì6 weeks before cardiac catheterization. In the TRIDENT study, ICM patients were randomized to receive either 20 or 100 million allogeneic MSCs transendocardially. Allogeneic MSCs were manufactured by the University of Miami Cell Manufacturing Program (Da Silva and Hare, 2013). Healthy subjects (n = 10) were enrolled ranging in ages from 22‚Äì 58 years and both genders. All subjects provided written informed consent. 2.3. Cell Characterization Both autologous and allogeneic MSCs were manufactured by the Foundation for Accreditation of Cellular Therapy (FACT)-accredited Good Manufacturing Practice (GMP) Cell Production Facility at the Interdisciplinary Stem Cell Institute, University of Miami Miller School of Medicine, as previously described (Hare et al., 2012; Da Silva and Hare, 2013). Cells were released for patient administration after meeting the following criteria: negative for mycoplasma via polymerase chain reaction, ‚â• 70% cell viability, growth assay via colony forming unit-Ô¨Åbroblasts assay, positive for CD105 (N80%) and negative for Colonies were directly Ô¨Åxed on Ô¨Åbronectin-coated dishes using 4% PFA. Cells were blocked in 10% normal donkey serum/0.3% Triton X-100/PBS for 1 h and then incubated in anti-CD31 and anti-VEGFR overnight at 4* (DAKO #235218, Cell Signaling #55B1R). Next, cells were incubated in Alexa Flour 564 anti-mouse and Alexa Flour 488 anti-rabbit for 45 min at room temperature. Lastly, wells were cover slipped with Vectashield plus DAPI. Images were obtained using immunoÔ¨Çuorescent microscopy. 2.6. Flow-Mediated Vasodilation (FMD) Brachial artery diameter measurements and FMD% were performed in the morning, after an overnight fast. The subjects' right arm was immobilized in an extended position, and the brachial artery was scanned via ultrasound 5‚Äì10 cm above the antecubital fossa (Hill et al., 2003; Corretti et al., 2002). A brachial cuff was then inÔ¨Çated to a supra-systolic pressure (40 to 50 mm Hg above systolic pressure) for 5 min. Subsequently, the cuff was deÔ¨Çated and the brachial artery diameter was recorded for 3 min. 2.7. VEGF ELISA Serum vascular endothelium growth factor (VEGF) levels (Invitrogen #KHG0111) were measured in DCM patients at baseline and 6 months after allogeneic (n = 6) or autologous (n = 5) MSC treatment. In ICM patients (n = 4), VEGF was measured at baseline and three months after allogeneic MSC treatment. DCM and ICM patients who received allogeneic MSCs were combined and compared to patients who received autologous. Lastly, VEGF was measured in healthy controls (n = 9), which provided written informed consent. 2.8. Matrigel Assays Human umbilical vein endothelial cells (HUVECs) were grown to passage seven in EGM-2 medium (LONZA). Autologous (n = 7) and allogeneic (n = 5) donor MSCs were grown to 70% conÔ¨Çuence and the conditioned medium was obtained. BrieÔ¨Çy, MSCs were starved in MEM alpha for 24 h at 5%, the supernatant was collected, centrifuged at 2000 g for 10 min, and stored at ‚àí 20¬∞ until use. 50,000 HUVECs were plated on Matrigel (BD Biosciences) in 24-well plates and pretreated with 15 ŒºM L-NAME (Cayman Chemical #80210) dissolved in alpha-MEM (GIBCO) for 45 min. 80% of either MSC conditioned medium (MSC-CM) or plain MEM alpha was added to respective treatment wells, and L-NAME was kept in the medium. After 6 h, six pictures per well were taken and Image J was used to analyze vascular index (tube length √ó tube number). C. Premer et al. / EBioMedicine 2 (2015) 467‚Äì475 469 between groups regarding coronary artery disease (CAD); speciÔ¨Åcally, all patients with ICM had CAD (P = 0.0058). 2.9. Statistical Analysis To assess the difference between autologous and allogeneic groups, an unpaired, two-tailed t-test was used. To measure the difference before and after treatment in each group, both a paired, two-tailed t-test and a one-way ANOVA was utilized. Correlations were measured using Pearson correlation, assuming a Gaussian distribution. Data are presented as mean and standard deviation of the mean. Both D'Agostino-Pearson omnibus normality test and Shapiro‚ÄìWilk normality tests were run to measure within-group variability on all data (only signiÔ¨Åcant differences were reported as D'Agostino-Pearson). Lastly, differences between groups regarding gender, race/ethnicity, history of smoking, and medications were analyzed using a Fisher exact test. 3. Results 3.1. Baseline Characteristics A total of 22 patients were analyzed for this study. Allogeneic (n = 15) and autologous (n = 7) MSCs were administered transendocardially. Baseline characteristics of the study subjects are summarized in Table 1. Patients with DCM were evenly distributed for both age and sex (P = NS, ANOVA). Additionally, there was no difference in age between ICM and DCM patients receiving allogeneic MSCs (P = NS, ANOVA); however patients with ICM were older than patients with DCM receiving autologous MSCs (P b 0.01, ANOVA). There were more White/Hispanic patients with DCM compared to all other treatment groups (P = 0.022, Fisher exact test). Additionally, patients with DCM who received allogeneic MSCs had higher cholesterol than patients with ICM who received allogeneic MSCs (P b 0.05, ANOVA). As expected, there was a signiÔ¨Åcant difference 3.2. EPC-Colony Forming Units (CFUs) and Flow-Mediated Vasodilation (FMD) in Heart Failure Patients and Healthy Subjects Patients with ischemic (n = 6) as well as non-ischemic (n = 16) cardiomyopathy had endothelial dysfunction at baseline, characterized by a reduced ability to form EPC-CFUs and an impaired FMD response. SpeciÔ¨Åcally, patients had decreased EPC-CFU counts compared to healthy controls (4 ¬± 3 vs. 25 ¬± 16, respectively, P b 0.0001, t-test; Fig. 1A) as well as diminished FMD% (5.6 ¬± 3.2 vs. 9.0 ¬± 3.3, respectively, P = 0.01, t-test; Fig. 1B). Patients were further analyzed by disease etiology. There was no difference in EPC-CFUs or FMD% comparing patients with DCM (n = 16) versus ICM (n = 6) at baseline (3 ¬± 3 vs. 6 ¬± 1 CFUs, respectively; 5.9 ¬± 3.6 vs. 5.1 ¬± 2.0%, respectively; Figs. S2A and S3A). Additionally, there was no difference at baseline in EPC-CFUs comparing DCM patients who received allogeneic MSCs (n = 9) to DCM patients who received autologous MSCs (n = 7) (3 ¬± 2 vs. 5 ¬± 4, respectively; Fig. S2B). However, DCM patients receiving autologous MSCs had a higher FMD% at baseline compared to DCM patients receiving allogeneic MSCs (8.3 ¬± 3.5 vs. 4 ¬± 2.4, P = 0.011, t-test; Fig. S3B). Moreover, we did a sub-analysis of the patients that received allogeneic MSCs. At baseline, patients with ICM (n = 6) had more EPC-CFUs than patients with DCM (n = 9) (6 ¬± 1 vs. 3 ¬± 2, P = 0.0011, t-test), and there was no difference in FMD% (5.1 ¬± 2 vs. 2.4 ¬± 5.7, P = 0.3742, t-test). Despite ICM patients having more EPC-CFUs at baseline, the improvement from baseline to 3 months post-injection was the same (11 ¬± 8 vs. 10 ¬± 4 DCM, P = 0.6195, t-test). Comparably, there was no difference in the change in FMD % from baseline to 3 months Table 1 Baseline characteristics of patients (n = 22) and healthy controls (n = 10). Patient data are broken down by etiology and cell treatment: Dilated cardiomyopathy (DCM) patients receiving allogeneic mesenchymal stem cells (MSCs) (n = 9), DCM patients receiving autologous MSCs (n = 7) and ischemic cardiomyopathy (ICM) patients receiving allogeneic MSCs (n = 6). *Asterisks next to P values indicate signiÔ¨Åcant differences between groups. *Asterisks next to values indicate signiÔ¨Åcant difference between noted groups (ANOVA). ‚Ä† indicates that healthy controls were left out of Fisher Exact test. Characteristic DCM allogeneic (n = 9, 27%) DCM autologous (n = 7, 21%) ICM allogeneic (n = 6, 18%) Healthy controls (n = 10, 33%) P-values Age ‚Äî yr. Median Range Male sex ‚Äî no. 60 45‚Äì70 8 (89%) 55* 48‚Äì73 6 (86%) 71* 65‚Äì75 6 (100%) 44 22‚Äì58 7 (70%) *b0.01 Race/ethnicity White White/Hispanic Black 8 (89%) 0 (0%) 1 (11%) 3 (43%) 4 (57%) 0 (0%) 6 (100%) 0 (0%) 0 (0%) 7 (70%) 3 (30%) 0 (0%) 0.114 *0.022 0.451 History of CVD CAD History of smoking‚Ä† 1 (11%) 4 (44%) 0 (0%) 4 (57%) 6 (100%) 4 (67%) 0 (0%) 0 (0%) **0.0058 0.689 Blood pressure-mm Hg (systolic/diastolic) Median 126/77 Range 108‚Äì141/54‚Äì91 108/65 99‚Äì142/59‚Äì84 113/70 85‚Äì134/53‚Äì76 118/67 111‚Äì130/59‚Äì90 0.979 Cholesterol (mg/dL) Median Range 175 112‚Äì207 147* 129‚Äì178 N/A N/A * b 0.05 C-reactive protein (mg/mL) Median 0.3 3 0.15 Range 0.2‚Äì0.4 0.3 N/A 0.2‚Äì0.5 0.923 0.1‚Äì0.6 N/A Medications ASA/NSAIDS ‚Ä† Beta-blockers‚Ä† ACE‚Ä† inhibitors/ARBs Diuretics‚Ä† Statins‚Ä† Antiplatelet‚Ä† Other‚Ä† 4 (57%) 7 (100%) 7 (100%) 6 (86%) 3 (43%) 2 (29%) 7 (100%) 5 (83%) 6 (100%) 6 (100%) 4 (67%) 6 (100%) 3 (50%) 6 (100%) 0 (0%) 0 (0%) 1 (10%) 0 (0%) 1 (10%) 0 (0%) 0 (0%) 202* 129‚Äì282 5 (56%) 7 (78%) 6 (67%) 8 (89%) 3 (33%) 4 (44%) 9 (100%) 0.419 0.5 0.204 0.166 0.0814 *0.03 0.707 NS 470 C. Premer et al. / EBioMedicine 2 (2015) 467‚Äì475 A B Fig. 1. Endothelial function in patients with heart failure, including dilated and ischemic cardiomyopathy. (A) Patients (n = 22) have impaired endothelial progenitor cell-colony forming units (EPC-CFUs) compared to healthy controls (n = 10, *P b 0.0001, t test). (B) Patients (n = 22) have reduced Ô¨Çow-mediated vasodilation (FMD%) compared to healthy controls (n = 10, ‚Ä† P = 0. 01, t-test). post-injection comparing DCM vs. ICM patients that received allogeneic MSCs (3.4 ¬± 2.86 vs. 4.18 vs. 3.16, P = 0.6280, t-test). 3.3. EPC-CFUs in Heart Failure Patients Treated With MSCs Patients were assessed for EPC-CFUs before and 3 months after MSC treatment. Patients who received allogeneic MSCs had a signiÔ¨Åcant improvement in the number of EPC-CFUs post-treatment (Œî10 ¬± 5, P b 0.0001, t-test; Fig. 2A). On the other hand, patients who received autologous MSCs showed no improvement (Œî1 ¬± 3, P = NS, t-test; Fig. 2B). Moreover, we compared allogeneic versus autologous MSC treatment, and allogeneic MSCs were superior in stimulating EPC colony formation (P = 0.0067, t-test). We also compared allogeneic treatment versus autologous treatment in DCM patients. 3 months post-injection, DCM patients who received allogeneic MSCs had signiÔ¨Åcantly higher EPC-CFUs compared to patients who received autologous MSCs (12 ¬± 4 vs. 6 ¬± 5, P = 0.015, t-test; Fig. S2C). Accordingly, DCM patients who received allogeneic MSCs had a signiÔ¨Åcantly greater positive change in EPC-CFUs from baseline to 3 months post-injection compared to DCM patients who received autologous MSCs (10 ¬± 4 vs. 1 ¬± 3, respectively, P b 0.0001, t-test; Fig. S2D). EPC-CFUs were examined for morphology. EPCs from HF patients had disorganized and incomplete colony formation (Fig. 2C and D), resulting in clusters that failed to form functional colonies. Three months after allogeneic MSC treatment, patient colonies were organized and healthy in appearance (Fig. 2E and F). Importantly, we found that these colonies were positive for the endothelial cell markers CD31/PECAM and VEGFR (Fig. 2G). Together, these Ô¨Åndings suggest that transendocardial MSC therapy stimulates EPC bioactivity in patients with HF of both ischemic and non-ischemic etiology. 3.4. FMD in Heart Failure Patients Treated With MSCs All patients were evaluated using brachial artery FMD before and 3 months after MSC injection. Patients who received allogeneic MSCs had a dramatic improvement in FMD% (Œî3.7 ¬± 3%, P = 0.0002, t-test; Fig. 3A). In contrast, patients who received autologous MSCs had no improvement and the majority of patients worsened 3 months posttreatment (Œî-0.46 ¬± 3%, Fig. 3B). We next analyzed the difference between treatment with autologous MSCs and allogeneic MSCs. There was a striking difference between the two cell types (P = 0.005, t-test), suggesting that autologous MSCs do not restore endothelial function in this patient population. We also assessed the correlation between EPC-CFUs and FMD% in all patients and found a highly signiÔ¨Åcant correlation between ŒîFMD% and ŒîEPC-CFUs (P = 0.0004, R = 0.68, Pearson correlation; Fig. 3C). Lastly, we analyzed DCM patients only, comparing autologous vs. allogeneic treatment. While patients who received allogeneic MSCs had similar FMD% compared to patients who received autologous MSCs 3 months after treatment (7.4 ¬± 4.7 vs. 7.9¬±3.1; Fig. S2C), patients who received allogeneic MSCs had a greater change from baseline to 3 months post-injection compared to patients who received autologous MSCs (3.8 ¬± 2.6 vs. 0.7 ¬± 2.9, P = 0.042, t-test; Fig. S3D). Notably, the majority of patients receiving autologous MSCs had no improvement or worsened, while all allogeneic patients improved (P = NS vs. P b 0.0001, t-test). 3.5. Vascular Endothelial Growth Factor (VEGF) in Patients Receiving Autologous and Allogeneic MSCs We measured circulating VEGF levels in patients and healthy control serum. At baseline, patients had profoundly elevated circulating VEGF compared to controls (1130.3 ¬± 803.3 vs. 2.0 ¬± 5.9 pg/mL, P = 0.0009, t-test; within-group, P = 0.21, P b 0.001 respectively, D'Agostino-Pearson omnibus normality test; Fig. 4A). Allogeneic MSCs reduced VEGF levels (‚àí 547.5 ¬± 350.8 pg/mL, P = 0.0015, t-test; within-group P = 0.96, D'Agostino-Pearson omnibus normality test), while patients who received autologous MSCs had an increase in VEGF levels (814.1 ¬± 875.8 pg/mL, Fig. 4B). Furthermore, there was a signiÔ¨Åcant difference between patients who received allogeneic MSCs versus patients who received autologous MSCs (P = 0.0012, t-test; Fig. 4B). VEGF levels correlated with EPC-CFUs (P = 0.026, R = ‚àí 0.421, Pearson correlation; Fig. 4C). Even more striking, there was a signiÔ¨Åcant correlation between the change in VEGF and the change in EPC-CFUs from baseline to 3 months post-treatment (R = 0.863, P b 0.0001, Pearson correlation; Fig. 4D). Notably, high levels of VEGF correlated with low levels of EPCs, evidenced by the autologous group. Conversely, lower levels of VEGF correlated with high levels of EPC-CFUs, illustrated by the allogeneic group. Taken together, these data demonstrate that allogeneic MSCs stimulate EPC mobilization and suppress compensatory elevations in circulating VEGF concentrations. 3.6. Autologous and Allogeneic MSC Paracrine Effect on Endothelial Cells Human umbilical vein endothelial cells (HUVECs), pre-treated with L-NAME to block endogenous nitric oxide (NO) synthesis, were subsequently treated with autologous or allogeneic MSC-conditioned media (CM), and vasculogenesis was examined in Matrigel assays (Fig. 5A‚ÄìD). HUVECs treated with L-NAME exhibited severely impaired vasculogenesis (Fig. 5B), evident by their depressed vascular index (305.2 ¬± 196.8 vs. 1170.9 ¬± 352.6, P = 0.02, t-test; Fig. 5E). Only the addition of allogeneic MSC-CM prevented the L-NAMEinduced impairment in vasculogenesis (vascular index 305.2 ¬± 196.8 L-NAME alone vs. 1113.4 ¬± 296.2 L-NAME + Allogeneic C. Premer et al. / EBioMedicine 2 (2015) 467‚Äì475 A 471 B C D E F G Fig. 2. Endothelial colony forming units in heart failure patients treated with either allogeneic or autologous mesenchymal stem cells (MSCs). (A) Patients treated with allogeneic MSCs had a signiÔ¨Åcant improvement in endothelial progenitor cell-colony forming units (EPC-CFUs) 3 months post-treatment (n = 15, *P b 0. 0001, t-test). (B) Patients treated with autologous MSCs had no change in EPC-CFUs post-treatment (n = 7, P = NS, t-test). (C‚ÄìF) Representative EPC-CFUs plated on Ô¨Åbronectin for 5 days before (C, D) and after (E, F) MSC administration (magniÔ¨Åcation 20√ó). (G) Colonies are positive for the endothelial cell markers CD31 (red) and VEGFR (green) (magniÔ¨Åcation 20√ó). MSC-CM, P b 0.05, ANOVA; Fig. 5E). Overall, these results suggest that allogeneic MSCs are able to restore the vascular potential of endothelial cells. 4. Discussion Patients with cardiomyopathy of ischemic or non-ischemic etiology manifest endothelial dysfunction, characterized by reduced EPC colony formation, impaired FMD, and elevated VEGF levels. The major new Ô¨Ånding of this study is that MSC administration to these patients stimulates EPC bioactivity and restores Ô¨Çow-mediated vasodilation towards normal. Importantly, the ability of allogeneic MSCs greatly exceeded that of autologous MSCs in restoring endothelial function, enhancing EPC colony formation, and suppressing VEGF levels. Together these Ô¨Åndings offer major new clinical insights into the bioactivity of MSCs, suggest a novel therapeutic principle for disorders characterized by 472 C. Premer et al. / EBioMedicine 2 (2015) 467‚Äì475 A B C Fig. 3. Flow-mediated vasodilation (FMD) measurements before and after mesenchymal stem cell (MSC) treatment. (A) Patients treated with allogeneic MSCs had an increase in FMD% 3 months post-injection (n = 15, *P = 0.0002, t-test). (B) Patients treated with autologous MSCs had no signiÔ¨Åcant difference in FMD% 3 months post-injection (n = 7, P = NS, t-test). (C) There is a strong correlation between the absolute change in FMD% and the absolute change in endothelial progenitor cell-colony forming units (EPC-CFUs) from baseline to 3 months post-MSC injection in all patients (*P = 0.0004, R = 0.68, Pearson correlation). endothelial dysfunction, and have implications for the choice of allogeneic vs. autologous MSC cell therapy. There is an increasing awareness of the central role endothelial dysfunction plays in CV disorders. Low numbers of EPC-CFUs strongly correlate with endothelial dysfunction (Werner et al., 2007) and are associated with a high Framingham risk score for adverse CV health outcomes (Hill et al., 2003). Furthermore, circulating EPC levels predict CV events‚ÄîspeciÔ¨Åcally in patients with coronary artery disease, heart failure, and angina (Schmidt-Lucke et al., 2005; Hill et al., 2003; Shantsila et al., 2007b). In addition, elevated levels of circulating VEGF are linked to endothelial dysfunction and HF (Chin et al., 2002; Tsai et al., 2005). In this regard, Eleuteri et al. demonstrated that elevated levels of VEGF correlated with HF disease progression (Eleuteri et al., 2011). Moreover, Wei et al. investigated circulating EPCs and VEGF levels in patients with cerebral aneurysm and found that decreased levels of circulating EPCs and increased levels of plasma VEGF were associated with chronic inÔ¨Çammation in the vascular walls of cerebral arteries and the development of cerebrovascular abnormalities leading to aneurysm formation and rupture (Wei et al., 2011). Thus, endothelial dysfunction is a central feature of CV disease, and may represent a powerful surrogate marker in the development of new treatments for CV disease. MSCs are adult stem cells that are prototypically found in bone marrow and have the capacity to differentiate into multiple cell types (Williams and Hare, 2011). Importantly, they stimulate the proliferation and differentiation of endogenous precursor cells and play a crucial role in maintaining stem cell niches (Williams and Hare, 2011). In addition, MSCs secrete paracrine factors that participate in angiogenesis, cardiomyogenesis, neovascularization, stimulation of other endogenous stem cells, and regulation of the immune system (Gomes et al., 2013; Gnecchi et al., 2008). While MSCs are known to stimulate cardiac precursor cells and cell cycle activity in the heart (Hatzistergos et al., 2010), their role in stimulating other endogenous precursor populations has heretofore been unknown. Here we report that MSCs stimulated endogenous EPC activation, increasing the number and quality of functional EPCs. These Ô¨Åndings suggest that augmentation of EPCs may represent a novel mechanism of action by which MSCs exert favorable biological effects. Over the last decade, there has been an emerging interest in the use of MSCs in CV disorders (Karantalis and Hare, in press; Telukuntla et al., 2013). Clinical trials have demonstrated a major safety proÔ¨Åle for MSC administration, and suggested efÔ¨Åcacy in patients with HF (Hare et al., 2012; Telukuntla et al., 2013); however, underlying mechanism(s) of action continue to be vigorously debated. Our Ô¨Ånding that allogeneic MSC injections in patients with both ischemic and non-ischemic HF results in an improvement in endothelial function, speciÔ¨Åcally by restoring EPC function and FMD and reducing VEGF levels towards normal, offers a major new insight into the mechanisms of action of MSCs. In the study population, increased serum VEGF correlated with diminished EPC-CFUs, consistent with the idea that VEGF plays a compensatory role, a Ô¨Ånding also reported in patients with cerebral aneurysm (Wei et al., 2011). This is also supported by the study of Vasa et al. which showed a diminished response of EPCs to VEGF in patients with CAD (Vasa et al., 2001). Moreover, Alber et al. found that a key beneÔ¨Åcial effect of atorvastatin therapy is reducing the levels of plasma VEGF in patients with CAD (Franz Alber et al., 2002). This coincides with our study using MSCs, rather than a pharmacological intervention, to decrease pathologic VEGF and increase endothelial function. Thus our Ô¨Åndings establish a previously unappreciated therapeutic principle whereby allogeneic MSCs can be employed to stimulate EPC bioactivity, improve arterial physiologic vasodilatory responses, and decrease unfavorable cytokine mobilization in patients with CV disease and other disorders associated with endothelial dysfunction. We found that allogeneic MSCs restored endothelial function in patients to a degree greatly exceeding that of autologous MSCs. One possible explanation for this may be the age of the cells. Recent studies highlight that MSC's therapeutic function declines as a result of aging C. Premer et al. / EBioMedicine 2 (2015) 467‚Äì475 A B C D Autologous Allogeneic 473 Autologous Allogeneic Fig. 4. Serum vascular endothelium growth factor (VEGF) concentration in patients and controls. (A) Patients (n = 14) have a higher level of circulating VEGF compared to controls (n = 9) at baseline (*P = 0.0009, t-test; within-group, P = 0.21, P b 0.001, respectively, D'Agostino-Pearson omnibus normality test). (B) Patients who received allogeneic MSCs (n = 9) had a decrease in VEGF serum levels post-injection (Œî-547.5 ¬± 350.8 pg/mL, ‚Ä†P = 0.0015, t-test; within-group P = 0.96, D'Agostino-Pearson omnibus normality test), while patients who received autologous MSCs (n = 5) had an increase in serum VEGF post-injection (Œî814.1 ¬± 875.8), and there was a difference between the groups (‚Ä†P = 0.0012, t-test). (C) There is a correlation between endothelial progenitor cell-colony forming units (EPC-CFUs) and serum VEGF in patients at both baseline and 3 months post-MSC treatment (R = ‚àí0.421, ‚Ä°P = 0.026, Pearson correlation). (D) The change in EPC-CFUs from baseline to 3 months post-treatment strongly correlated with the change in VEGF (R = ‚àí0.863,*P b 0.0001, Pearson correlation). (EÔ¨Åmenko et al., 2013; Asumda, 2013). EÔ¨Åmenko et al. showed that adipose-derived MSCs from aged patients with coronary artery disease have impaired angiogenic potential (EÔ¨Åmenko et al., 2013). Similarly, Kasper et al. demonstrated that MSC function is altered and diminished with age, speciÔ¨Åcally showing lower actin turnover and therefore decreased motility, decreased antioxidant power, decreased responsiveness to chemical and mechanical signaling, and increased senescence (Kasper et al., 2009). Stolzing et al. also reported a decline in ‚ÄúÔ¨Åtness‚Äù as a result of aging, as evidenced by a decline in colony-forming unitÔ¨Åbroblasts and increase in reactive oxygen species levels and oxidative stress (Stolzing et al., 2008). In our study, all allogeneic stem cell donors were healthy, young donors between the ages of 20 and 35. Patients receiving their own stem cells not only had underlying chronic diseases, but also were older (between the ages of 45 and 75). MSC aging may impair the survival, differentiation, and ability to recruit EPCs to areas of damage, ultimately reducing their therapeutic efÔ¨Åcacy (Asumda, 2013). Additionally, due to underlying patient comorbidities, the autologous MSC microenvironment may be negatively altered due to systematic inÔ¨Çammation. Consistent with this notion, Teraa et al. showed that systemic inÔ¨Çammation affects the bone marrow microenvironment, disturbing EPC function (Teraa et al., 2013). Although more studies are necessary to validate that the advantage evident here is due to the health and age of MSCs, this study supports the encouraging idea of using ‚Äúoff the shelf‚Äù allogeneic MSCs over autologous MSCs. In this study, we report positive systemic effects from local, cardiac transendocardial MSC injections. We have previously shown that MSC engraftment after intramyocardial injection is approximately 10 to 20%, suggesting that these cells migrate and circulate systemically (Quevedo et al., 2009). MSCs are known to secrete anti-inÔ¨Çammatory factors and cytokines (such as IL-2, TGF-Œ≤1, hepatocyte growth factor, NO, prostaglandin 2, and stromal cell derived factor-1), which can modulate the mobilization of EPCs from bone marrow (Williams and Hare, 2011; Iyer and Rojas, 2008). Additionally, MSCs have been shown to secrete paracrine factors that stimulate resident cells (Williams and Hare, 2011). Thus, we propose a potential mechanism whereby allogeneic MSCs injected into cardiac tissue respond to local microenvironment cues, thereby secreting anti-inÔ¨Çammatory and EPC mobilizing factors that ultimately improve endothelial function alleviating cardiac stress. There are several limitations of our study. All ICM patients received allogeneic MSCs, therefore we were unable to study the effect of autologous MSCs in this speciÔ¨Åc HF population. Additionally, patients who received autologous MSCs had higher FMD% at baseline compared to patients who received allogeneic MSCs. Notably, however, all patients receiving autologous MSCs had lower FMD% post-treatment, highlighting the allogeneic advantage. There was also a signiÔ¨Åcantly higher White/Hispanic population in the DCM autologous group compared to other treatment groups. Despite this, there was no difference in baseline or treatment response comparing White/Hispanic patients to White patients (Supplementary Table 1). Furthermore, patients with ICM received different total number of cells (either 20 or 100 million cells). Regardless, all patients had an improvement in endothelial function and there was no intergroup variability. Lastly, there was variability within our control group for circulating VEGF levels. The majority of our controls had too low levels of circulating VEGF to detect, ultimately highlighting the elevated levels of VEGF evident in HF patients. Despite these limitations, we are conÔ¨Ådent our results provide novel insights into the positive endothelial function effect of allogeneic MSCs in patients with HF. In conclusion, this study demonstrates a potent and clinically relevant efÔ¨Åcacy outcome of transendocardial therapy with MSCs in patients with advanced HF. Allogeneic MSCs restore Ô¨Çow mediated brachial artery dilatation, EPC bioactivity, and VEGF levels towards normal. As abnormalities in the vascular function of patients with CV disease is shown to be highly predictive of adverse outcomes and disease 474 C. Premer et al. / EBioMedicine 2 (2015) 467‚Äì475 E * Fig. 5. Effect of autologous and allogeneic mesenchymal stem cell (MSC) treatment on vasculogenesis. (A‚ÄìD) Representative pictures of human umbilical vein endothelial cells (HUVECs) alone (n = 3), HUVECs with the nitric oxide synthase inhibitor, L-NG-Nitroarginine methyl ester (L-NAME) (n = 3), HUVECs with L-NAME and allogeneic MSC-conditioned media (CM) (n = 5), and HUVECs with L-NAME and autologous MSC-CM (n = 7) after 6 h on Matrigel (magniÔ¨Åcation 10√ó). (E) L-NAME greatly reduced vascular index (*P b 0.05, ANOVA), and only allogeneic-CM restored it (*P b 0.05, ANOVA). progression, targeting endothelial function is a signiÔ¨Åcant therapeutic strategy. Together, these Ô¨Åndings offer a new mechanism of action underlying potentially clinically relevant responses to the use of allogeneic MSCs in CV disease. Author Contributions CP: study concept and design, EPC-CFUs, VEGF, and Matrigel experiments, data analysis, drafted manuscript; AB: study concept and design, drafted manuscript; MB: CFU assays, data analysis; IHS: study design, data analysis, manuscript editing; BH: FMD studies, data analysis; MP: FMD studies, data analysis; CD: Matrigel assay, VEGF assay, data analysis; WB: study design; DD: study design; AK: clinical cell manufacturing and characterization; JMH: study design and concept, data analysis, manuscript drafting and editing. Funding Dr. Hare is supported by the National Institutes of Health grants RO1 HL084275, RO1 HL107110, RO1 HL110737, and 5UM HL113460, and the Starr Foundation and the Soffer Family Foundation. Role of Funding Sources No funders played a role in study design, data collection, data analysis, interpretation, or writing of the report. C. Premer et al. / EBioMedicine 2 (2015) 467‚Äì475 ConÔ¨Çict of Interest Dr. Hare has a patent for cardiac cell-based therapy and reports equity interest and board membership in Vestion Inc. Acknowledgments The authors wish to thank Julio Sierra, Cindy Delgado, and Phillip Gonzalez for enrolling and consenting patients, collecting samples, and coordinating patient data. Additionally, the authors wish to thank Vasileios Karantalis for his statistical review. Lastly, the authors would like to thank Irene Margitich for her laboratory help and technical support. Appendix A. Supplementary data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ebiom.2015.03.020. References Asumda, F.Z., 2013. Age-associated changes in the ecological niche: implications for mesenchymal stem cell aging. Stem Cell Res. Ther. 4 (3), 47. Blum, A., 2009. Heart failure‚Äînew insights. Isr. Med. Assoc. J. 11 (2), 105‚Äì111. Bui, A.L., Horwich, T.B., Fonarow, G.C., 2011. Epidemiology and risk proÔ¨Åle of heart failure. Nat. Rev. Cardiol. 8 (1), 30‚Äì41. Cao, Y., Gomes, S., Rangel, E.B., et al., 2015. S-nitrosoglutathione reductase-dependent PPARŒ≥ denitrosylation participates in adipogenesis and osteogenesis. J. Clin. Investig. 125 (4). Chen, L., Tredget, E.E., Wu, P.Y., Wu, Y., 2008. Paracrine factors of mesenchymal stem cells recruit macrophages and endothelial lineage cells and enhance wound healing. PLoS ONE 3 (4), e1886. Chin, B.S.P., Chung, N.A.Y., Gibbs, C.R., Blann, A.D., Lip, G.Y.H., 2002. Vascular endothelial growth factor and soluble P-selectin in acute and chronic congestive heart failure. Am. J. Cardiol. 90 (11), 1258‚Äì1260. Corretti, M.C., Anderson, T.J., Benjamin, E.J., et al., 2002. Guidelines for the ultrasound assessment of endothelial-dependent Ô¨Çow-mediated vasodilation of the brachial artery: a report of the International Brachial Artery Reactivity Task Force. J. Am. Coll. Cardiol. 39 (2), 257‚Äì265. Da Silva, J.S., Hare, J.M., 2013. Cell-based therapies for myocardial repair: emerging role for bone marrow-derived mesenchymal stem cells (MSCs) in the treatment of the chronically injured heart. Methods Mol. Biol. 1037, 145‚Äì163. EÔ¨Åmenko, A., Dzhoyashvili, N., Kalinina, N., et al., 2014. Adipose-derived mesenchymal stromal cells from aged patients with coronary artery disease keep mesenchymal stromal cell properties but exhibit characteristics of aging and have impaired angiogenic potential. Stem Cells Transl. Med. 3 (1), 32‚Äì41. Eleuteri, E., Di Stefano, A., Tarro Genta, F., et al., 2011. Stepwise increase of angiopoietin-2 serum levels is related to haemodynamic and functional impairment in stable chronic heart failure. Eur. J. Cardiovasc. Prev. Rehabil. 18 (4), 607‚Äì614. Franz Alber, H., Dulak, J., Frick, M., et al., 2002. Atorvastatin decreases vascular endothelial growth factor in patients with coronary artery disease. J. Am. Coll. Cardiol. 39 (12), 1951‚Äì1955. Gnecchi, M., Zhang, Z., Ni, A., Dzau, V.J., 2008. Paracrine mechanisms in adult stem cell signaling and therapy. Circ. Res. 103 (11), 1204‚Äì1219. Go, A.S., Mozaffarian, D., Roger, V.L., et al., 2014. Heart disease and stroke statistics‚Äî2014 update: a report from the American Heart Association. Circulation 129 (3), e28‚Äìe292. Gomes, S.A., Rangel, E.B., Premer, C., et al., 2013. S-nitrosoglutathione reductase (GSNOR) enhances vasculogenesis by mesenchymal stem cells. Proc. Natl. Acad. Sci. U. S. A. 110 (8), 2834‚Äì2839. Hare, J.M., Fishman, J.E., Gerstenblith, G., et al., 2012. Comparison of allogeneic vs autologous bone marrow-derived mesenchymal stem cells delivered by transendocardial 475 injection in patients with ischemic cardiomyopathy: The POSEIDON randomized trial. JAMA 308 (22), 2369‚Äì2379. Hatzistergos, K.E., Quevedo, H., Oskouei, B.N., et al., 2010. Bone marrow mesenchymal stem cells stimulate cardiac stem cell proliferation and differentiation. Circ. Res. 107 (7), 913‚Äì922. Heldman, A.W., DiFede, D.L., Fishman, J.E., et al., 2014. Transendocardial mesenchymal stem cells and mononuclear bone marrow cells for ischemic cardiomyopathy: the TAC-HFT randomized trial. JAMA 311 (1), 62‚Äì73. Hill, J.M., Zalos, G., Halcox, J.P.J., et al., 2003. Circulating endothelial progenitor cells, vascular function, and cardiovascular risk. N. Engl. J. Med. 348 (7), 593‚Äì600. Iyer, S.S., Rojas, M., 2008. Anti-inÔ¨Çammatory effects of mesenchymal stem cells: novel concept for future therapies. Expert. Opin. Biol. Ther. 8 (5), 569‚Äì581. Karantalis, V., Hare, J.M., 2015. Use of mesenchymal stem cells for therapy of cardiac disease. Circ. Res. http://dx.doi.org/10.1161/CIRCRESAHA.116.303614 (in Press). Kasper, G., Mao, L., Geissler, S., et al., 2009. Insights into mesenchymal stem cell aging: involvement of antioxidant defense and actin cytoskeleton. Stem Cells 27 (6), 1288‚Äì1297. Marti, C.N., Gheorghiade, M., Kalogeropoulos, A.P., Georgiopoulou, V.V., Quyyumi, A.A., Butler, J., 2012. Endothelial dysfunction, arterial stiffness, and heart failure. J. Am. Coll. Cardiol. 60 (16), 1455‚Äì1469. Mushtaq, M., DiFede, D., Golpanian, S., et al., 2014. Rationale and design of the Percutaneous Stem Cell Injection Delivery Effects on Neomyogenesis in Dilated Cardiomyopathy (The POSEIDON-DCM Study). J. Cardiovasc. Transl. Res. 1‚Äì12. Quevedo, H.C., Hatzistergos, K.E., Oskouei, B.N., et al., 2009. Allogeneic mesenchymal stem cells restore cardiac function in chronic ischemic cardiomyopathy via trilineage differentiating capacity. Proc. Natl. Acad. Sci. U. S. A. 106 (33), 14022‚Äì14027. Schmidt-Lucke, C., R√∂ssig, L., Fichtlscherer, S., et al., 2005. Reduced number of circulating endothelial progenitor cells predicts future cardiovascular events: proof of concept for the clinical importance of endogenous vascular repair. Circulation 111 (22), 2981‚Äì2987. Schulman, I.H., Zhou, M.S., Raij, L., 2006. Interaction between nitric oxide and angiotensin II in the endothelium: role in atherosclerosis and hypertension. J. Hypertens. Suppl. 24 (1), S45‚ÄìS50. Shantsila, E., Watson, T., Lip, G.Y., 2007a. Endothelial progenitor cells in cardiovascular disorders. J. Am. Coll. Cardiol. 49 (7), 741‚Äì752. Shantsila, E., Watson, T., Lip, G.Y.H., 2007b. Endothelial progenitor cells in cardiovascular disorders. J. Am. Coll. Cardiol. 49 (7), 741‚Äì752. Solomon, A., Blum, A., Peleg, A., Lev, E.I., Leshem-Lev, D., Hasin, Y., 2012. Endothelial progenitor cells are suppressed in anemic patients with acute coronary syndrome. Am. J. Med. 125 (6), 604‚Äì611. Stolzing, A., Jones, E., McGonagle, D., Scutt, A., 2008. Age-related changes in human bone marrow-derived mesenchymal stem cells: consequences for cell therapies. Mech. Ageing Dev. 129 (3), 163‚Äì173. Telukuntla, K.S., Suncion, V.Y., Schulman, I.H., Hare, J.M., 2013. The advancing Ô¨Åeld of cell‚Äê based therapy: insights and lessons from clinical trials. J. Am. Heart Assoc. 2 (5). Teraa, M., Sprengers, R.W., Westerweel, P.E., et al., 2013. Bone marrow alterations and lower endothelial progenitor cell numbers in critical limb ischemia patients. PLoS ONE 8 (1), e55592. Tsai, W.C., Li, Y.H., Huang, Y.Y., Lin, C.C., Chao, T.H., Chen, J.H., 2005. Plasma vascular endothelial growth factor as a marker for early vascular damage in hypertension. Clin. Sci. 109 (1), 39‚Äì43. Vasa, M., Fichtlscherer, S., Aicher, A., et al., 2001. Number and migratory activity of circulating endothelial progenitor cells inversely correlate with risk factors for coronary artery disease. Circ. Res. 89 (1), e1‚Äìe7. Wei, H., Mao, Q., Liu, L., et al., 2011. Changes and function of circulating endothelial progenitor cells in patients with cerebral aneurysm. J. Neurosci. Res. 89 (11), 1822‚Äì1828. Werner, N., Kosiol, S., Schiegl, T., et al., 2005. Circulating endothelial progenitor cells and cardiovascular outcomes. N. Engl. J. Med. 353 (10), 999‚Äì1007. Werner, N., Wassmann, S., Ahlers, P., et al., 2007. Endothelial progenitor cells correlate with endothelial function in patients with coronary artery disease. Basic Res. Cardiol. 102 (6), 565‚Äì571. Williams, A.R., Hare, J.M., 2011. Mesenchymal stem cells: biology, pathophysiology, translational Ô¨Åndings, and therapeutic implications for cardiac disease. Circ. Res. 109 (8), 923‚Äì940. Zampetaki, A., Kirton, J.P., Xu, Q., 2008. Vascular repair by endothelial progenitor cells. Cardiovasc. Res. 78 (3), 413‚Äì421. EBioMedicine 2 (2015) 499‚Äì512 Contents lists available at ScienceDirect EBioMedicine journal homepage: www.ebiomedicine.com Original Article Altered Mitochondrial Function, Mitochondrial DNA and Reduced Metabolic Flexibility in Patients With Diabetic Nephropathy Anna Czajka a,1, Saima Ajaz a,1, Luigi Gnudi b, Chandani Kiran Parsade a, Peter Jones a, Fiona Reid c, Afshan N. Malik a,‚Åé a b c Diabetes Research Group, Division of Diabetes and Nutritional Science, Faculty of Life Sciences and Medicine, King's College London, SE1 1UL, UK Cardiovascular Division Department of Primary Care and Public Health Sciences a r t i c l e i n f o Article history: Received 21 January 2015 Received in revised form 27 March 2015 Accepted 3 April 2015 Available online 11 April 2015 Keywords: Diabetic nephropathy Mitochondrial DNA Hyperglycemia Mitochondrial dysfunction Bioenergetic deÔ¨Åcit Bioenergetic health index (BHI) a b s t r a c t The purpose of this study was to determine if mitochondrial dysfunction plays a role in diabetic nephropathy (DN), a kidney disease which affects N 100 million people worldwide and is a leading cause of renal failure despite therapy. A cross-sectional study comparing DN with diabetes patients without kidney disease (DC) and healthy controls (HCs); and renal mesangial cells (HMCs) grown in normal and high glucose, was carried out. Patients with diabetes (DC) had increased circulating mitochondrial DNA (MtDNA), and HMCs increased their MtDNA within 24 h of hyperglycaemia. The increased MtDNA content in DCs and HMCs was not functional as transcription was unaltered/down-regulated, and MtDNA damage was present. MtDNA was increased in DC compared to HC, conversely, patients with DN had lower MtDNA than DC. Hyperglycaemic HMCs had fragmented mitochondria and TLR9 pathway activation, and in diabetic patients, mitophagy was reduced. Despite MtDNA content and integrity changing within 4 days, hyperglycaemic HMCs had a normal bioenergetic proÔ¨Åle until 8 days, after which mitochondrial metabolism was progressively impaired. Peripheral blood mononuclear cells (PBMCs) from DN patients had reduced reserve capacity and maximal respiration, loss of metabolic Ô¨Çexibility and reduced Bioenergetic Health Index (BHI) compared to DC. Our data show that MtDNA changes precede bioenergetic dysfunction and that patients with DN have impaired mitochondrial metabolism compared to DC, leading us to propose that systemic mitochondrial dysfunction initiated by glucose induced MtDNA damage may be involved in the development of DN. Longitudinal studies are needed to deÔ¨Åne a potential cause‚Äìeffect relationship between changes in MtDNA and bioenergetics in DN. ¬© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction Diabetic nephropathy (DN) affects ~one-third of patients with diabetes and develops over a long period of clinical silence (Ritz and Orth, 1999). With the epidemic rise in the incidence of diabetes currently affecting more 350 million people, N100 million people are at risk of DN and of these 30% are likely to progress to end stage renal failure despite therapy (IDF, 2013; Wild et al., 2004). Therefore there is an urgent need to understand the underlying mechanisms of damage in the diabetic kidney in order to design novel therapeutics. Clinical randomised studies such as the Diabetes Control and Complications Trial and the Epidemiology of Diabetes Interventions and Complication have shown that early intensive glycemic control reduces risk, whereas prolonged hyperglycaemia can increase the long term ‚Åé Corresponding author at: Diabetes Research Group, Hodgkin Building, Guy's Campus, King's College London, London Bridge, London SE1 1UL, UK. E-mail address: afshan.malik@kcl.ac.uk (A.N. Malik). 1 Co-Ô¨Årst author. risk of diabetic complications (Kilpatrick et al., 2009). Hyperglycaemia is therefore accepted as a major mediator of renal damage and activates several complex and overlapping biochemical pathways resulting in abnormal signalling in cells (Aronson, 2008; Brownlee, 2001). Diabetes results in increased risk of numerous other complications which affect major organs, including eyes (retinopathy), heart (diabetic cardiomyopathy), blood vessel (peripheral vascular disease) and brain (dementia) (Stratton et al., 2000). The multi-organ impact of diabetes complications resembles mitochondrial genetic disease (Moraes et al., 1991; Wallace, 1999) and suggests a systemic dysfunction in the body. Mitochondria, cellular organelles in the cytosol of eukaryotic cells, harbour their own circular DNA genome which is located outside the nuclear genome, and require both mitochondrial and nuclear genome encoded proteins to function. Mitochondria produce energy in the form of ATP via oxidative phosphorylation (OXPHOS) (Wojtczak and Zab≈Çocki, 2008), therefore the number of mitochondria in a cell is related to the energy requirements of the cell and can vary depending on many factors such as the environment and redox balance of the cell (Michel et al., 2012). Mitochondrial dysfunction can affect key cellular http://dx.doi.org/10.1016/j.ebiom.2015.04.002 2352-3964/¬© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 500 A. Czajka et al. / EBioMedicine 2 (2015) 499‚Äì512 functions, result in a variety of diseases (Wallace, 1999), and altered mitochondrial DNA (MtDNA) levels have been reported in a wide range of human disease (Malik and Czajka, 2013). We and others have reported changes in circulating MtDNA in patients with DN (Lee et al., 2009; Malik et al., 2009). Reduced renal mitochondrial function was postulated in a urinary metobolomics study showing that mitochondrial metabolites were reduced in the urine of DN patients (Sharma et al., 2013), however, it was suggested that the basis for reduced mitochondrial content/function would be difÔ¨Åcult to address in clinical samples. Therefore, despite some evidence that mitochondrial dysfunction may be involved in DN, there is no direct evidence in clinical samples and it is unclear whether measuring mitochondrial function in patients may have any translational potential. Furthermore the underlying molecular mechanisms are not understood. If systemic mitochondrial dysfunction is associated with DN, then we would predict systemic changes in the body which may be the basis of renal microvascular complications in diabetes. Peripheral blood mononuclear cells (PBMCs) from patients have been used widely as surrogates and various parameters associated with DN can be detected in these cells (Hofmann et al., 1999; Ihm et al., 1997; Sourris et al., 2010; Yi et al., 2014); our assumption is that PBMC mitochondria will resemble systemic changes in the body and therefore will act as surrogate cells for kidney tissue. Recently, methodology for measuring bioenergetics in live fractionated blood cells have been developed and it has been suggested that these may be utilised to measure oxygen consumption rates (OCR) and extra cellular Ô¨Çux analysis (ECAR) to indicate potential mitochondrial dysfunction (Chacko et al., 2013) It has been proposed that ECAR and OCR values can be combined using a formula as a single value that deÔ¨Ånes an individual's bioenergetic health index (BHI) which could be used as a prognostic/diagnostic indicator of metabolic stress. In the current paper our major aim was to determine if mitochondrial dysfunction can be detected in patients with DN and to elucidate the potential mechanisms by which hyperglycaemia may affect renal mitochondria. To expand our in-vivo observations we also utilised an experimental in-vitro cell model system to examine the effects of hyperglycaemia on renal mitochondria. We used primary human mesangial cells (HMCs) as transformation can affect cellular bioenergetics. Importantly, glomerular mesangial cells are a target of both metabolic and haemodynamic perturbations in diabetes (Clarkson et al., 2002; Murphy et al., 1999) and have been widely utilised as models of DN. We therefore used PBMCs to examine systemic changes in diabetes and HMCs to examine effect of diabetes on renal mitochondria. 2. Materials and Methods 2.1. Human Subjects Patients were recruited between 2008 and 2014 with written informed consent from Guy's and St Thomas' hospital clinics under ethical approval from the regional Research Ethics Committee (REC; ref number 07/H0806/120). The cross-sectional study adhered to the Ethical Principles for Medical Research Involving Human Subjects, World Medical Association Declaration of Helsinki. A random blood glucose level of ‚â•11.1 or a fasting blood glucose level of ‚â•7 mmol/l was considered to be indicative of diabetes. Type 1 diabetes (T1D) and type 2 Diabetes (T2D) were deÔ¨Åned as follows: T1D: onset before age 35, insulin therapy within 6 months of diagnosis and no breaks in insulin therapy N6 months; T2D: onset after age 35, controlled by diet or established oral hypoglycaemic treatment and/or insulin. Albumin/ creatinine ratio (ACR) was used to assess the level of albuminuria, ACR b 2.5 mg/mmol for men and ACR b3.5 mg/mmol for women, was deÔ¨Åned as cut off for normo to microalbuminuria. Glomerular Ô¨Åltration rate (GFR) was assessed using the ModiÔ¨Åcation of Diet in Renal Disease (MDRD) formula (Stoves et al., 2002). For controls without nephropathy, we used patients with type 1 diabetes and type 2 diabetes with ‚â•20 or ‚â•10 years of diabetes duration, respectively, without a history of albuminuria, with normal renal function and normal blood pressure (‚â§ 130/80 mm Hg) and taking no antihypertensive agents. HC (n = 39) with no history of the disease or current medication were recruited with informed consent, and were age and sex matched with the patient study group. Sample size was calculated from a pilot experiment using Cohens d, which showed that n = 39 for each group was adequate to power the MtDNA study. Table 1 shows the baseline characteristics of subjects used in this study. 2.2. Primary Human Glomerular Mesangial Cells Human kidneys unsuitable for transplantation or normal portion from renal carcinoma were used as a source of primary human mesangial cells (HMCs) which were cultured and characterised as previously described (Gruden et al., 1997; Thomas et al., 2000) and grown in Dulbecco's ModiÔ¨Åed Eagle Medium (DMEM, Sigma Aldrich) supplemented with 10% FBS, ITS and antibiotics. Cells were seeded at an equal density (1 √ó 105/well) in 6-well plates in DMEM containing either 5 mM (normal, NG) or 25 mM (high, HG) glucose for varying time points. 20 mM mannitol in 5 mM glucose DMEM was used as an osmotic control in all experiments. 2.3. Real Time Quantitative PCR Genomic DNA or RNA templates converted to cDNA were used for real time qPCR using SYBR green (Qiagen) as previously described (Shahni et al., 2013). All primer sequences used in the study are listed (Table S3). Absolute quantiÔ¨Åcation was carried out in the presence of dilution standards for each gene. The mRNA copy numbers were calculated by dividing the mean values of target gene relative to the mean values of Œ≤-actin (the most stably expressed gene), MtDNA copy numbers were assessed as mitochondrial to nuclear gene ratio using Beta-2 microglobulin (B2M) as the nuclear control (Malik et al., 2011). 2.4. Extracellular Flux Analysis Extracellular Ô¨Çux analysis is deÔ¨Åned as the measurement of cellular bioenergetics (respiratory activity), based on the measurement of two major energy pathways OXPHOS and glycolysis (Chacko et al., 2013; Ferrick et al., 2008). Table 1 Baseline characteristics of healthy controls and diabetic patients with and without nephropathy with MtDNA content. Variable Healthy controls (HCs) (n = 39) Diabetic controls (DCs) (n = 45) Diabetic nephropathy (DN) (n = 83) Type of diabetes (T1D:T2D) Age (years) Gender (female:male) Diabetes duration (years) BMI (kg/m2) HbA1c (%) ACR (mg/mmol) eGFR (ml min‚àí1 1.73 m‚àí2) Systolic BP (mm Hg) Diastolic BP (mm Hg) Cholesterol (mmol/l) MtDNA content MtDNA (Median) NR 52 ¬± 25 20:19 NR 23 ¬± 4 ND ND ND ND ND ND 34 ¬± 9 32 (35) 27:18 49 ¬± 14 28:17 22 ¬± 10 27 ¬± 4# 10 ¬± 9 0.8 ¬± 0.6 103 ¬± 21 122 ¬± 12 72 ¬± 9 4.4 ¬± 0.8 64 ¬± 75# 38 (469)# 31:52 61 ¬± 14‚Åé‚Åé 37:46 22 ¬± 13 30 ¬± 7‚Åé‚Åé## Data are means ¬± SD, except where otherwise indicated. ‚Åé P b 0.05 compared with DC. ‚Åé‚Åé P b 0.01 compared with DC. ‚Åé‚Åé‚Åé P b 0.001 compared with DC. # P b 0.05 compared with HC. ## P b 0.01 compared with HC. 8¬±1 18 ¬± 50‚Åé‚Åé 66 ¬± 34‚Åé‚Åé 133 ¬± 17‚Åé‚Åé 72 ¬± 9 3.9 ¬± 0.9‚Åé 43 ¬± 52‚Åé‚Åé 26 (298)‚Åé‚Åé‚Åé A. Czajka et al. / EBioMedicine 2 (2015) 499‚Äì512 501 The metabolic proÔ¨Åles of cultured HMCs and freshly isolated human PBMCs s were assessed using the XFe96 Seahorse analyser and XF cell mito stress test kit (Seahorse Biosciences). Oligomycin (ATP synthase blocker) was used to measure ATP turnover and to determine proton leak, the mitochondrial uncoupler ‚Äî FCCP (carbonyl cyanide 4-[triÔ¨Çuoromethoxy] phenylhydrazone) was used to measure maximum respiratory function (maximal OCR). Reserve capacity was calculated as maximal OCR minus the basal respiration. Rotenone (inhibitor of complex I) and Antimycin A (a blocker of complex III), were used to measure non-mitochondrial respiration (Brand and Nicholls, 2011; Dranka et al., 2011). The data generated for OCR and ECAR in PBMCs were used to calculate the Bioenergetic Health Index as described recently by Darley-Usmar group (Chacko et al., 2014) using the BHI = log (reserve capacity) √ó (ATP-linked) / (non-mitochondrial) √ó (proton leak). the help of LG and PJ, KP undertook MtDNA damage experiments, the data analysis was undertaken by AC/SA/AM and evaluated by all authors, the statistical analysis was supported by FR, all authors were involved in the drafting and revision of the manuscript. 2.5. Mitochondrial DNA Damage We Ô¨Årst set out to establish if we could conÔ¨Årm and extend ours and another study showing altered circulating MtDNA in DN patients (Lee et al., 2009; Malik et al., 2009). Using small volumes of peripheral blood, our assays comprised of measurement of MtDNA content, and mRNA content of various mitochondrial-encoded and nuclearencoded mRNAs including genes encoding various mitochondrial subunits, involved in biogenesis, fusion, Ô¨Åssion and mitophagy. In addition we assessed the integrity of MtDNA in these samples. Using a cross-sectional study design, we compared 3 groups of subjects: healthy controls (HCs, n = 39) were volunteers with no history of any disease, diabetes controls (DCs, n = 45) comprised of patients with ‚â•20 years diabetes duration, normal renal function and no history of albuminuria, and diabetic nephropathy patients (DN, n = 83) with a history of or current albuminuria (Table 1). The HC group were age and sex matched with the diabetes patients (Table S1) and had a lower BMI (P b 0.05). The DN patients were older, had higher BMI, albumin/creatinine ratio (ACR), and systolic blood pressure and lower eGFR than DC. We used both T1D and T2D patients in this study and analysed the groups combined as hyperglycaemia plays a role in progression of renal disease independently of the type of diabetes. Analysis of T1D and T2D patients separately showed the same trends (Table S5). MtDNA content was quantiÔ¨Åed as the ratio of mitochondrial genome to nuclear genome (Malik et al., 2011). DC had signiÔ¨Åcantly higher MtDNA content compared to HC (P b 0.05) whereas the DN patients had reduced MtDNA compared to DC (P b 0.001, Fig. 1A). We examined if the decreased MtDNA in DN patients was a consequence of other parameters, as age, BMI, systolic blood pressure, eGFR and cholesterol, were signiÔ¨Åcantly different between the DC and DN groups (Table 1). After adjusting for these variables, stepwise binary regression analysis showed that eGFR, systolic blood pressure and MtDNA levels remained independently associated with DN (P = 0.009) however, age, BMI, and cholesterol were no longer associated with DN. As reduced eGFR and elevated blood pressure are well known risk markers for DN (Alaveras et al., 1997), our data strongly suggest that MtDNA levels represent a new risk marker for DN and the associated increase in cardiovascular morbidity and mortality of these patients. To investigate if the changes in circulating MtDNA were accompanied by altered transcription of mitochondrial mRNAs, we quantiÔ¨Åed mitochondrial genome coded OXPHOS subunits; NADH oxidase subunit VI (ND6), NADH oxidase subunit I (ND1) and cytochrome c oxidase subunit 3 (COX3); and nuclear encoded mitochondrial mRNAs involved in mitochondrial biogenesis (mitochondrial transcription factor A ‚Äî TFAM, Peroxisome proliferator-activated receptor gamma coactivator 1 alpha ‚Äî PGC1A), Ô¨Åssion (Dynamin-1-like protein ‚Äî DRP1), fusion (Mitofusin 1 and 2-MFN1 and MFN2, Optic atrophy 1 ‚Äî OPA1) and mitophagy (PTEN-induced putative kinase 1 ‚Äî PINK1, parkin RBR E3 ubiquitin protein ligase ‚Äî PARK2, Fig. 1B‚ÄìL). As the amount of MtDNA is known to be directly proportional to the amount of mRNA (Hock and Kralli, 2009; Williams, 1986) we expected to see changes in mitochondrial encoded mRNAs. However, despite DNA damage was quantiÔ¨Åed using the elongase method (Furda et al., 2014) by comparing the relative ampliÔ¨Åcation of an 8.9 kb region relative to a 127 bp region in the mitochondrial genome using speciÔ¨Åc primers (Table S3). For the Surveyor Nuclease method (Bannwarth et al., 2006), amplicon A was ampliÔ¨Åed (Table S3) and digested with Surveyor nuclease. Gel electrophoresis was used in both methods to detect MtDNA damage. 2.6. Measurement of ROS Production, Cell Viability and Apoptosis in Human Mesangial Cells Intracellular ROS in was measured using cell-permeant 2‚Ä≤,7‚Ä≤dichlorodihydroÔ¨Çuorescein diacetate (H2DCFDA) oxidation, cell viability was measured using CellTiter-Glo luminescence assay (Promega, UK) in a microplate reader according to the manufacturer's instructions. 2.7. Assessment of Mitochondrial Network Cells were grown in 96-well plates with clear bottoms, Ô¨Åxed with paraformaldehyde and visualized by staining with Mitotracker Red (Invitrogen) according to the manufacturer's instructions. Cells were viewed under Nikon Eclipse Ti-E Inverted Microscope, using CFI S Plan Fluor ELWD 20√ó/0.45NA objective. The stained mitochondrial network was assessed using ImageJ software (version 1.47q: National Institutes of Health; www.rsb.info.nih.gov/ij) using an image-processing algorithm (Koopman et al., 2005). Two different parameters, formfactor F as a measure of degree of branching (calculated as a perimeter2 / 4Œ† ‚àó area) aspect ratio AR (measure of mitochondria length) were used to quantify mitochondrial morphology. 2.8. Statistics Analysis was performed using GraphPad (GraphPad Software, Inc.) and IBM SPSS for Windows software. The distribution of the data was tested using the Kolmogorov‚ÄìSmirnov test (graph pad) and histograms in SPSS and parametric tests were used on raw or log transformed data. For parametric analysis, groups were compared using t-test (2 groups) or one way ANOVA with post-hoc Tukey's multiple comparison test (N2 groups). For non-parametric analysis, groups were compared using Mann‚ÄìWhitney (2 groups) or Kruskal Wallis with Dunn's posthoc test with Bonferroni correction (N2 groups). Data are presented as mean ¬± Standard error of the mean and as median for non-normally distributed data. 2.9. Author Contributions AC and SA generated the cell and human data respectively and are joint Ô¨Årst authors, the study was conceived and designed by AM with 2.10. Funding This work was partly funded by an EFSD/Janssen Programme for the Study of the Role of the Kidney in Diabetes grant; AC and SA are supported by KCL PhD scholarships. 3. Results 3.1. Detection of Changes in Mitochondrial DNA and mRNAs in Blood Samples From DN Patients 502 A. Czajka et al. / EBioMedicine 2 (2015) 499‚Äì512 A HC DC DN B C E F G H J K L N Relative amplification M 150 100 50 0 HC DC DN D I A. Czajka et al. / EBioMedicine 2 (2015) 499‚Äì512 altered MtDNA levels in DC and DN, we did not see any signiÔ¨Åcant alterations in mRNAs levels of mitochondrial encoded subunits (Fig. 1B‚ÄìD) or in most nuclear encoded mitochondrial mRNAs (Fig. 1F‚ÄìJ), suggesting that the altered MtDNA levels are not functional. This idea is further supported by the reduced TFAM levels (Fig. 1E), and a signiÔ¨Åcant reduction in mitophagy mRNAs in diabetes patients (Fig. 1K and L), suggesting that removal of dysfunctional mitochondria is hampered. Despite having higher MtDNA, DC patients have reduced TFAM compared to HC, and despite having lower MtDNA, DN patients have higher TFAM than DC, supporting the idea that the MtDNA may comprise of a mixture of functional and non-functional molecules. One possibility is that MtDNA in these samples comprises of both intact and damaged/mutated MtDNA i.e. there could be a level of heteroplasmy. The integrity of the MtDNA was determined in a subset of patients, using two methods: a PCR based ‚Äúelongase method‚Äù (Furda et al., 2014) and the surveyor nuclease method (Bannwarth et al., 2006). Using the elongase method, we found that relative ampliÔ¨Åcation was lower in DC (84 ¬± 50) and DN (85 ¬± 33) compared to HC (107 ¬± 37), but this difference was not signiÔ¨Åcant (P N 0.05 Fig. 1M). Using surveyor nuclease, we detected mismatches in 5 out of 10 DN patients, but none in 10 HC patients (Fig. 1N). We were unable to conÔ¨Årm presence of mutations by Sanger sequencing. As surveyor nuclease can detect N3% of heteroplasmy whereas with Sanger sequencing detection levels are N 20% (Bannwarth et al., 2006), our data suggest that the DN patients have damaged MtDNA genomes present at between 3% and 20% of the total MtDNA content. 3.2. Reduced Mitochondrial Metabolism in PBMCs From DN Patients The data above showed that DN patients have reduced circulating MtDNA compared to DC patients and an increased level of heteroplasmy. It is possible that this could affect cellular energy production. Therefore we examined whether there is any difference in mitochondrial metabolism by determining the bioenergetic proÔ¨Åle of HC, DC, and DN groups. The baseline characteristics of this subset of patients (Table S2) show similar trends to the whole study group (Table 1). We used freshly puriÔ¨Åed PBMCs which were used for the assays within 2 h of collection from patients. Basal, ATP-linked, maximal oxygen consumption rate (OCR), reserve capacity and extracellular acidiÔ¨Åcation rate (ECAR) were measured using a Seahorse XFe96 analyser. OCR, an indicator of mitochondrial respiration, and ECAR, representative of glycolysis, were normalised to the cell number of PBMCs from HC (n = 10), DC (n = 14) and DN patients (n = 16). Basal respiration (Fig. 2A), ATP-linked respiration (Fig. 2B), basal glycolytic rate (Fig. 2E), proton leak and non-mitochondrial respiration (data not shown) were similar in the 3 groups. However, both maximal respiration and reserve capacity were signiÔ¨Åcantly reduced by ~ 40% (P b 0.05) in DN patients (Fig. 2C and D). These data suggest that whilst energy production capability under normal physiological conditions is likely to be similar in the 3 groups, reduced reserve capacity and reduced maximal respiration in DN patients are suggestive of a compromised response to stress. 503 a trend for slight but non-signiÔ¨Åcant reduction in basal, ATP-linked, maximal respiration and reserve capacity, showing that these cells are able to handle acute stress (Fig. 3A‚ÄìD). In contrast, PBMCs from DN patients were sensitive to acute glucose loading (Fig. 3E and F), as although there was no signiÔ¨Åcant change in basal and ATP-linked respiration, both maximal respiration and reserve capacity were signiÔ¨Åcantly decreased (P b 0.05). Maximal respiration (119.8 ¬± 39.7) decreased 40% (83.7 ¬± 29.1, P = 0.05) and reserve capacity (74.4 ¬± 30) decreased ~ 50% (40.2 ¬± 0.7, P = 0.04) after acute load in DN patients (n = 8). There was no signiÔ¨Åcant difference in basal ECAR between the three groups in high glucose and normal glucose (Fig. 3G). As this data shows that DN patients have a compromised metabolic response, with a reduced maximal respiration and reserve capacity compared to HC and DC, we sought to determine if there is any translational potential in identifying DN patients using the bioenergetics response of PBMCs. Therefore, the bioenergetics data (Fig. 2 A‚ÄìD) were used to calculate the Bioenergetic Health Index (BHI) as described recently by the Darley-Usmar group (Chacko et al., 2014) (Fig. 2F). The mean (+/‚àí SD) BHI value for DN patients (3.0 ¬± 0.4, n = 12, P = 0.01) was signiÔ¨Åcantly lower than DC (3.4 ¬± 0.2, n = 5) (Fig. 2F, Table S4). 3.4. Hyperglycaemia-induced Changes in Mitochondrial DNA and mRNAs in Human Renal Glomerular Mesangial Cells The data presented above suggested that in PBMCs from DN patients there may be a dis-connect between MtDNA levels and mitochondrial mRNA levels. As PBMCs may be reÔ¨Çective of systemic changes in the body (Rudkowska et al., 2011), we wanted to determine if diabetes could affect renal mitochondria. To investigate this we used primary glomerular mesangial cells (HMCs) and examined the effect of hyperglycaemia on their mitochondria. Growth of HMCs in 25 mM glucose (HG) for 4 days resulted in a ~2.5-fold increase in cellular MtDNA content (717 ¬± 157) compared to cells cultured in 5 mM glucose (NG) (280 ¬± 53, n = 4, P b 0.01, Fig. 4A). A time course experiment showed that MtDNA content began to increase within 24 h of incubation in HG, and this increase was statistically signiÔ¨Åcant (P = 0.02) after 72 h (Fig. 4B). To examine whether increased MtDNA content after 4 days of growth in HG was accompanied by increased transcription of mitochondrial mRNAs, we quantiÔ¨Åed mitochondrial genome encoded OXPHOS subunits, and nuclear encoded mitochondrial mRNAs involved in mitochondrial biogenesis, Ô¨Åssion, fusion and mitophagy. Surprisingly, there was a more than 50% reduction in ND6 and COX3 mRNA in HG (Fig. 4C and E, P b 0.01), and no signiÔ¨Åcant change was observed in ND1 mRNA (Fig. 4D, P N 0.05), suggesting that increased MtDNA did not result in increased mitochondrial genome transcription. TFAM mRNA was signiÔ¨Åcantly increased by ~140% (Fig. 4F, P b 0.05), but the levels of the remaining mRNAs did not change in HG (P N 0.05, Fig. 4G‚ÄìM). No changes were observed at the protein level when mitochondrial respiratory complexes I‚ÄìV were measured via Western blot (data not shown). 3.3. Loss of Metabolic Flexibility in Diabetic Nephropathy 3.5. Functional Consequences of Hyperglycaemia in Kidney Cells To investigate whether DN patients have a reduced response to stress, we determined the effect of acute applied stress (Fig. 3A‚ÄìG) in PBMCs. Cell chambers were injected with high glucose (20 mM) and the change in mitochondrial respiration was measured immediately. Acute glucose loading had no impact on PBMCs from HC and DC, with Since nuclear encoded mRNAs involved in the mitochondrial life cycle did not respond to hyperglycaemia in the time frame in which the MtDNA increase was seen; we examined intracellular reactive oxygen species (ROS) (measured using DCF Ô¨Çuorescence which is an Fig. 1. Diabetes associated changes in mitochondrial DNA content and mitochondrial life cycle in circulating cells. DNA and RNA were isolated from 0.1 ml and 1 ml respectively of peripheral blood from HC (n = 39), DC (n = 45) and DN (n = 83) groups. Data are presented as box plots (except for M) with whiskers showing median, quartiles and range which were analysed using non-parametric Kruskal‚ÄìWallis test with Dunn's multiple comparison test. (A) MtDNA content. Real time qPCR was carried out to determine MtDNA content as mitochondrial to nuclear (B2M) ratio. *P b 0.05, ***P b 0.001. (B‚ÄìL) Quantitative analysis of mitochondrial (mito mRNA) and nuclear encoded mRNAs. RNA was converted to cDNA and used as template and real time qPCR was used to quantify mRNA copy numbers relative to 1000 copies of reference gene Œ≤-actin.*P b 0.05, **P b 0.01. (M) MtDNA damage assessed using the elongase method in HC (n = 10), DC (n = 14) and DN (n = 16), data are presented as mean ¬± SEM. (N) Representative agarose gels showing amplicon A (‚àí) and after surveyor nuclease digestion (+). Digested bands in DN patients are illustrated with red arrows. 504 A. Czajka et al. / EBioMedicine 2 (2015) 499‚Äì512 HC DC DN F Fig. 2. Dysfunctional metabolic response in live peripheral blood mononuclear cells (PBMCs) from patients with diabetic nephropathy. PBMCs from HC (n = 10), DC (n = 14) and DN (n = 16) were isolated, seeded at 3 √ó 105 cells/well, and the Seahorse XFe96 extracellular Ô¨Çux analyser was used to measure oxygen consumption rate (OCR) and extracellular acidiÔ¨Åcation rate (ECAR). Data presented as mean ¬± SEM and analysed by one-way ANOVA with Tukey's test. (A) Basal respiration rate. (B) ATP-linked respiration after ATP synthase blocker (oligomycin) injection. (C) Maximal (uncoupled) respiration rate measured after the FCCP injection *P b 0.05. (D) Reserve capacity, measured as a difference between maximal and basal respiration, *P b 0.05. (E) Basal ECAR (glycolysis). (F) BHI calculated for HC (n = 8), DC (n = 5) and DN (n = 12) groups, *P b 0.05. assay of generalised oxidative stress and not mitochondrial speciÔ¨Åc, Kalyanaraman et al., 2012), cell viability, MtDNA damage, and mitochondrial morphology in order to elucidate the underlying mechanisms. There was a signiÔ¨Åcant increase in ROS in HMCs grown in HG (Fig. 5A, P b 0.001). In parallel to the increased ROS, cell viability was signiÔ¨Åcantly reduced in HMCs grown in HG (Fig. 5B). As the increase in ROS was paralleled with the increase in MtDNA (Fig. 4A) and a reduction in MtDNA transcription (Fig. 4C‚ÄìE), we speculated that ROS might damage the MtDNA, and that the damaged MtDNA may then activate inÔ¨Çammation via the TLR9 pathway (Oka et al., 2012). MtDNA damage was signiÔ¨Åcantly increased in cells grown in HG at 4 days (Fig. 5C, P b 0.05), and at the same time both NF-Œ∫B and MYD88 mRNAs were signiÔ¨Åcantly increased (Fig. 5D and E, P b 0.001, P b 0.01), suggesting activation of the TLR9 pathway. MitoTracker Red staining was used to examine the effect of HG on the mitochondrial network in HMCs. Images were acquired at different time points (Fig. 5F). In cells grown in NG, the mitochondrial network was seen as a mixture of elongated and connected mitochondria A. Czajka et al. / EBioMedicine 2 (2015) 499‚Äì512 A B C D E F 505 HC G DC DN Fig. 3. The loss of metabolic Ô¨Çexibility in diabetic nephropathy patients. Live PBMCs from HC (n = 4), DC (n = 4) and DN (n = 8) were isolated, seeded at 3 √ó 105 cells/well and analysed using the Mito Stress Test (Seahorse) in the Seahorse XFe96 analyser. Oxygen consumption rate (OCR) and extracellular acidiÔ¨Åcation rate (ECAR) were determined without (‚àí, red) and with (+, blue) acute glucose load. Data are expressed as mean ¬± SEM, n = 12‚Äì20 assay replicates per sample (A, C, D). Independent Student's t-test. (A) Mean OCR values in HC samples. (B) Representative example from a single HC sample. (C) Mean OCR values DC samples. (D) Representative example from a single DC patient. (E) Mean OCR in DN patients. (F) Representative example from a single DN patient. (G) Mean ECAR from HC, DC and DN. (Fig. 5F top panel), and no signiÔ¨Åcant changes were observed during the duration of the experiment (data not shown). When cells were grown in HG, their mitochondria became more rounded and fragmented (Fig. 5F). Plotting the values for both mitochondrial length/width (aspect ratio AR) versus degree of branching (formfactor F) showed a progressive reduction in both parameters with increased time of HG exposure (Fig. 5F right panel). The aspect ratio was signiÔ¨Åcantly down regulated in cells exposed to HG, and appeared to change within 24 h (Fig. 5G, P b 0.01). This change became highly signiÔ¨Åcant after a longer time of exposure, with the appearance of more mitochondria of a rounded shape (Fig. 5G, P b 0.001). Formfactor F values were also signiÔ¨Åcantly decreased, following 24 h of exposure to HG with the lowest values at day 8 (Fig. 5H, P b 0.001). These results suggest that incubation in HG leads to increased oxidative stress, MtDNA damage, reduced cellular viability, mitochondrial fragmentation and loss of mitochondrial network connection indicating mitochondrial dysfunction. 3.6. Reduced Metabolism in Kidney Cells Grown in High Glucose To assess the effect of continued hyperglycaemia on mitochondrial bioenergetics in renal cells, HMCs were incubated in NG and HG for 4, 8 and 12 days, and basal, ATP-linked, maximal OCR, reserve capacity 506 A. Czajka et al. / EBioMedicine 2 (2015) 499‚Äì512 A B 5 mM glucose (NG) 25 mM glucose (HG) C D F G H I E J K L M Fig. 4. Glucose-induced changes in mitochondrial DNA content and mitochondrial gene transcription in kidney cells. HMCs were cultured in 5 mM glucose (NG) and 25 mM glucose (HG) for 4 days and used to prepare DNA and RNA. Real-time qPCR was carried out to determine MtDNA content using the DNA as template, and expressed as the mitochondrial to the nuclear genome ratio. For mRNA quantiÔ¨Åcation the RNA was converted to cDNA and used as template, mRNA copy numbers for each gene were determined relative to 1000 copies of reference gene Œ≤-actin. Data are presented as mean ¬± SEM, n N 3 independent experiments. Independent Student's t-test. (A) Quantitative measurement of the MtDNA content in HMCs after 4 days, **P b 0.01. (B) MtDNA content in HMCs measured every 24 h, *P b 0.05, **P b 0.01. (C‚ÄìM) Quantitative analysis of mitochondrial (mito mRNA) and nuclear encoded mRNAs in HMCs, *P b 0.05, **P b 0.01. Fig. 5. Functional consequences of hyperglycaemia in kidney cells. HMCs were cultured in 5 mM glucose (NG) and 25 mM glucose (HG) for the times shown. (A‚ÄìB) ROS production and cell viability shown as a % of control. Data are presented as mean ¬± SEM, n = 17‚Äì21 observations. Non-parametric Mann Whitney test, **P b 0.01, ***P b 0.001. (C) Quantitative analysis of MtDNA damage assessed using the elongase method. Data are presented as mean ¬± SEM, n = 2, 5‚Äì6 observations/experiment. Mann‚ÄìWhitney test, *P b 0.05. (D, E) mRNA expression analysis of NF-kB, MYD88 in HMCs. RNA was converted to cDNA and used as template, mRNA copy numbers for each gene were determined relative to 1000 copies of reference gene Œ≤-actin. Data are presented as mean ¬± SEM, n = 3. Independent Student's t-test, **P b 0.01, ***P b 0.001. (F) Mitochondrial morphology in HMCs exposed to high glucose. Mitochondria were labelled with MitoTracker Red CMXRos and images were captured at magniÔ¨Åcation √ó20. Representative images of each condition shown in left panel. Mitochondrial length (aspect ratio ‚Äî AR) was plotted against mitochondrial degree of branching (formfactor F) and shown on the right panel as scatter plots. Scale bar, 50 Œºm. (G‚ÄìH) Quantitative analysis of aspect ratio AR and formfactor F shown as a % of the control. Data are presented as mean ¬± SEM, n = 2, N14 cells/experiment. Non-parametric Kruskal‚ÄìWallis test with Dunn's multiple comparison test, **P b 0.01, ***P b 0.001. A. Czajka et al. / EBioMedicine 2 (2015) 499‚Äì512 507 5 mM glucose (NG) 25 mM glucose (HG) 508 A. Czajka et al. / EBioMedicine 2 (2015) 499‚Äì512 5 mM glucose (NG) 25 mM glucose (HG) and ECAR were measured using a Seahorse XFe96 analyser (Fig. 6A‚ÄìH). OCR and ECAR were both normalised to the protein content as HMCs are growing cells, and presented as pmolesO2/min/Œºg protein. We could detect no differences in the bioenergetic proÔ¨Åle of cells incubated in HG for 4 days, as basal, ATP-linked, maximal respiration and basal glycolytic rate were similar between cells grown in NG and HG A. Czajka et al. / EBioMedicine 2 (2015) 499‚Äì512 (Fig. 6A‚ÄìF). Cells exposed to HG for 8 days had similar basal and ATPlinked respiration and the basal glycolytic rate, but had a reduced maximal respiration (P b 0.05, Fig. 6A‚ÄìE and G). After 12 days of incubation in HG, cells showed signiÔ¨Åcantly reduced basal, ATP-linked, maximal respiration and basal glycolytic rate (P b 0.05, Fig. 6A‚ÄìE and H). No changes were observed in the proton leak and non-mitochondrial respiration for any of the time points (data not shown). The reserve capacity was unchanged at 4 days, but was reduced in cells grown in HG after 8 and 12 days (P b 0.05, Fig. 6D and F‚ÄìH). Mitochondrial respiration in cells incubated in HG was independent of osmolarity or culture time (data not shown). Comparison of basal respiration rate with basal glycolytic rate showed no difference at 4 and 8 days in HG (Fig. 6I and J). However, after 12 days, the cells grown in HG had reduced OCR and ECAR (P b 0.05) which shows that they have shifted to a less metabolically active state (Fig. 6K). 4. Discussion Diabetes increases the risk of multi-organ complications affecting various cell types and is a major cause of blindness, renal failure, and cardiovascular disease (Gordin et al., 2012; He and King, 2004). Our study suggests that systemic mitochondrial dysfunction and glucose induced changes in MtDNA parallel diabetic kidney disease. The role of MtDNA mutations in mitochondrial genetic disease (Hudson et al., 2014; Niaudet, 1998; Tuppen et al., 2010; Wallace and Chalkia, 2013) and of MtDNA as an inÔ¨Çammatory molecule (Collins et al., 2004; Zhang et al., 2010) has been demonstrated. SpeciÔ¨Åc MtDNA mutations can lead to diabetes and kidney disease in patients with mitochondrial genetic disease (D'Aco et al., 2013; Mazzaccara et al., 2012; Seidowsky et al., 2013), suggesting that acquired MtDNA damage may play a role in DN. We present novel evidence showing that glucose induced increase in MtDNA in renal cells precede mitochondrial dysfunction, suggesting a novel cascade of events contributing to mitochondrial dysfunction in patients with DN (Fig. 7). We were interested in comparing DC patients, who had a long duration of diabetes but did not develop kidney disease, with the DN group, who had developed kidney disease, in order to determine if susceptibility for the development of diabetic kidney disease may involve mitochondrial dysfunction. Our data shows that it is possible using PCR based assays to detect differences in MtDNA and mitochondrial mRNAs between diabetes patients who develop kidney disease and those that don't. We used extracellular Ô¨Çux analysis, as previously described (Ferrick et al., 2008), to measure cellular respiration, and showed, using live PBMCs, that DN patients have a diminished bioenergetic response. We also converted the bioenergetics data to the single value ‚ÄúBioenergetic Health Index (BHI)‚Äù which has recently been proposed as a new biomarker for assessing patient health with both prognostic and diagnostic value (Chacko et al., 2014), and we found that DN patients had signiÔ¨Åcantly reduced BHI. To our knowledge this is the Ô¨Årst report demonstrating altered BHI in patients, supporting the hypothesis that mitochondrial dysfunction is involved in DN. In fact, there have been very few studies reporting metabolic function in live PBMCs and none speciÔ¨Åcally in diabetic complications (Hartman et al., 2014; Maynard et al., 2013). We have detected a clear difference in metabolic function between the DC and DN group and our data suggests that inability to cope with energy demand due to MtDNA damage could be a key factor in the progression of DN. 509 To establish whether hyperglycaemia can affect renal mitochondria and contribute to mitochondrial dysfunction, we used primary glomerular mesangial cells, widely used as an experimental model of DN (Clarkson et al., 2002; Thomas et al., 2000). The data derived from the HMC experiments supports the hypothesis that glucose can lead to alterations in MtDNA and mitochondrial function in renal cells and may be considered as representative of an early stage in diabetes. Exposure of HMCs to hyperglycaemia led to changes in mitochondrial morphology and increased MtDNA levels with 24 h, after 4 days there was reduced transcription of the mitochondrial genome, increased ROS and MtDNA damage, and activation of the NF-Œ∫B pathway, and after 8‚Äì12 day HMCs displayed reduced mitochondrial metabolism. These results suggest that the observed similar defect seen in PBMCs occur at a systemic level and could be the cause of the changes we see in mitochondrial parameters in PBMCs from DN patients. Circulating MtDNA levels have been previously reported to be increased in diabetes but there have been contradictory reports in DN (Lee et al., 2009; Malik et al., 2009), possibly due to the fact that earlier studies used mitochondrial primers which may amplify nuclear pseudogenes with high homology to MtDNA (Malik and Czajka, 2013; Malik et al., 2011). In the current study, using unique primers and a protocol designed for accurate MtDNA quantiÔ¨Åcation from blood (Ajaz et al., 2014) we found that DC patients had signiÔ¨Åcantly higher levels of circulating MtDNA compared to HC. Several mitochondrial- and nuclear-encoded mitochondrial mRNAs remained unchanged, apart from reduced TFAM, PINK1 and PARK2, suggesting that the increased MtDNA in DC patients may not be fully functional, and that mitophagy may be impaired. The inhibition of mitophagy and the presence of damaged MtDNA are very likely to lead to a decrease in mitochondrial quality. The DN patients had signiÔ¨Åcantly reduced MtDNA compared to DC. Binary regression analysis showed that eGFR, A/C ratio, and MtDNA levels were independently associated with DN. As eGFR and A/C ratio are well established risk markers of DN, our data suggests that the MtDNA changes could be a risk marker for DN; prospective studies will have to answer this question in the future. A key Ô¨Ånding of our study is the demonstration that exposure to HG leads to a rapid increase in mesangial cell MtDNA, accompanied by unchanging/decreased transcription of mitochondrial encoded mRNAs and little change in nuclear encoded mitochondrial mRNAs in renal cells. The amount of MtDNA is known to be directly proportional to the amount of mitochondrial encoded mRNAs (Hock and Kralli, 2009; Williams, 1986) however in our study increased MtDNA levels in patients (Fig 1A) are not accompanied by increased mitochondrial encoded mRNAs (Fig. 1B‚ÄìD), we see the same trend in cells, where the mitochondrial encoded mRNAs are reduced (Fig. 4C‚ÄìE) whereas the MtDNA in increased (Fig. 4A). As the function of MtDNA is to encode mitochondrial mRNAs, we deduce that the increased MtDNA in not functional. The only exception is the glucose induced up-regulation of TFAM mRNA, which has been previously reported (Choi et al., 2004). Our observation is suggestive of a disconnect between the increase in MtDNA and mitochondrial transcription/translation. A similar disconnect was reported in a study investigating the mechanisms of insulin resistance in muscle, with increased MtDNA but not increased mitochondrial content in myotubes (Aguer et al., 2013). Increased MtDNA in cultured cells was reported in response to hydrogen peroxide induced oxidative stress (Lee and Wei, 2005). HG induced ROS may inhibit mitochondrial biogenesis, as has been shown in other systems Fig. 6. Reduced metabolism in kidney cells grown in high glucose. HMCs were grown in 5 mM glucose (NG, open bars) and 25 mM glucose (HG, solid bars) for 4, 8 and 12 days, and then plated at a density of 3.5 √ó 104 cells/well one day prior to the experiment. The Seahorse XFe96 extracellular Ô¨Çux analyser was used to measure oxygen consumption rate (OCR) and extracellular acidiÔ¨Åcation rate (ECAR), all data were normalised to the protein content and shown as pmoles/min/ug protein. (A) Quantitative measurement of the basal OCR in HMCs, ***P b 0.001. (B) ATP-linked OCR after oligomycin injection. (C) Maximal (uncoupled) OCR after FCCP injection, *P b 0.05, ***P b 0.001. (D) Quantitative measurement of the reserve capacity, measured as a difference between maximal and basal respiration rates, *P b 0.05, ***P b 0.001. (E) Basal ECAR (glycolysis), *P b 0.05. (F‚ÄìH) Representative mito stress test runs showing mitochondrial respiration in HMCs cultured in NG, HG and osmotic control (Man) at 4, 8 and 12 days respectively. (I‚ÄìK) Representative analysis of the basal OCR plotted against basal ECAR in HMCs cultured in NG, HG and osmotic control (Man) at 4, 8 and 12 days respectively. HMCs grown in HG for 12 days had lower OCR/ECAR ratio, metabolic shift is indicated by an arrow. Data are presented as mean ¬± SEM, n = 6‚Äì8 replicates for 2 independent experiments. Independent Student's t-test. 510 A. Czajka et al. / EBioMedicine 2 (2015) 499‚Äì512 Fig. 7. Early glucose induced changes in mitochondria may lead to diabetic nephropathy: a schematic diagram with each box referring to key Ô¨Åndings of this paper given in brackets. The early changes refer to data obtained from cultured renal cells. The late changes refer to data obtained from PBMCs from HC, DC and DN patients. The evidence supporting this schematic in the paper: hyperglycaemia leads to increased MtDNA content (Fig. 4A), altered mitochondrial morphology (Fig. 5, F‚ÄìH), increased intracellular reactive oxygen species (ROS Fig. 5A), mitochondrial DNA damage (Fig. 5C), cellular damage measured as apoptosis/viability (Fig. 5B), and inÔ¨Çammation (Fig. 5, D and E) as early changes in cells. These changes precede and are followed by reduced OCR (basal, maximal, ATP-linked), reserve capacity and ECAR (Fig. 6, A‚ÄìE). These changes may result in damaged mitochondria, blocked mitochondrial biogenesis and an energy deÔ¨Åcit. The long term consequences of such changes could be as shown in the bottom panel. In DC patients these changes are seen as an increase in MtDNA (Fig. 1A) and normal metabolism (Fig. 2, A‚ÄìE) however DN patients have decreased MtDNA (Fig. 1A) accompanied with a dysfunctional metabolic response (Fig. 2A‚ÄìE) resulting in reduced BHI (Fig. 2F). There is evidence of reduced mitophagy in both DC and DN patients (Fig. 1, K and L). (Ballinger et al., 2000) and this view is consistent with reduced/unaltered mitochondrial transcription and translation which we observe. The increase in MtDNA in these conditions could be a compensatory response to the impairment of transcription. We measured cellular respiration in HMCs and found that oxidative metabolism can cope for several days of high stress conditions and continues to provide energy. Cells started to display altered morphology within a few hours of exposure to HG with mitochondrial fragmentation, which may be part of the initiation of an adaptive biogenesis programme to increase MtDNA content. Whilst glucose induced MtDNA increase was evident within 24 h, cells showed no detectable perturbation in their metabolic proÔ¨Åle at this time point. Some changes were seen after 8 days, however it was only after 12 days of culture that signiÔ¨Åcantly reduced ECAR and OCR were observed. To our knowledge this is the Ô¨Årst report of the metabolic response of HMCs, although previously a meeting report on mouse mesangial cells showed similar proÔ¨Åles to ours (Chacko et al., 2010a). Our data agree with a study using rat retinal endothelial cells where growth in HG for 6 days led to reduced basal and maximal respiration, however unlike our study where ECAR was decreased, they found increased ECAR (Trudeau et al., 2010). The reduced ECAR in our cells is suggestive of an energy deÔ¨Åcit and may be a consequence of a ROS induced glycolytic block (Colussi et al., 2000). This idea is supported by our Ô¨Ånding of HG induced ROS in our cells. The importance of mitochondrial ROS in oxidative stress in the kidney is well established, and mitochondrial targeted antioxidants ameliorate certain markers of renal damage (Chacko et al., 2010b). Here we show that the increased ROS is accompanied by increased MtDNA damage and NF-Œ∫B and MYD88 mRNAs. Like bacterial DNA, MtDNA is un-methylated and can initiate immune responses via the intracellular Toll like receptor (TLR) 9 (Higginbotham et al., 2002). The inÔ¨Çammatory properties of MtDNA (Collins et al., 2004; Zhang et al., 2010) and disruption of the normal degradation of MtDNA in the cytosol of cardiomyocytes was shown to cause TLR9-mediated inÔ¨Çammation leading to a heart failure in a mouse model (Oka et al., 2012). It is well established that chronic inÔ¨Çammation is a key mediator in diabetic complications but the underlying mechanisms have not been elucidated (Navarro-Gonzalez et al., 2011). Our data in HMCs of increased MtDNA content, ROS, MtDNA damage, and up-regulation of MYD88 and NF-Œ∫B support the idea of MtDNA induced activation of the inÔ¨Çammatory TLR9 pathway in these cells and are suggestive of a novel mechanism leading to chronic inÔ¨Çammation in DN. MtDNA is usually present in multiple identical copies in the cells, but the location of MtDNA close to the electron transport chain renders it more susceptible to ROS induced damage and mutations. In diabetes patients, MtDNA may be exposed to hyperglycaemia induced oxidative stress (Singh et al., 2011) resulting in random mutations in the mitochondrial genome leading to heteroplasmy, the presence of normal and mutated MtDNA in the same cell. The damaged MtDNA in such cases would only have an impact once it reached a threshold where the normal non-mutated MtDNA is no longer able to compensate for mitochondrial dysfunction (Chinnery, 2002). It has recently been demonstrated that MtDNA heteroplasmic mutations are associated with widely spread A. Czajka et al. / EBioMedicine 2 (2015) 499‚Äì512 chronic diseases, including atherosclerosis and cancer (Sobenin et al., 2014; Ye et al., 2014). We used two methods to detect MtDNA damage, both of which suggest that MtDNA from diabetes samples shows increased damage, with heteroplasmy being detected only in DN samples. Furthermore, HMCs showed a signiÔ¨Åcant increase in MtDNA damage after 4 days of growth in hyperglycaemic conditions, suggesting MtDNA damage in HMCs could play a role in the subsequent bioenergetic deÔ¨Åcit observed in these cells after 8 and 12 days of hyperglycaemia. The combination of the data presented in this paper paint a picture of severely compromised mitochondria in patients with DN. These patients have reduced MtDNA in circulation, with increased MtDNA damage, and decreased mitophagy, the functional impact of these changes is the reduced metabolic Ô¨Çexibility of these cells. Furthermore if we assume that the PBMCs are representative of systemic changes in the body, then it could be predicted that kidney cells in the DN patients will show a similar compromised response, which could play a major role in progression of pathology. Our results support a recent metabolomics study showing that patients with DN have reduced mitochondrial metabolites in urine (Sharma et al., 2013) and we propose that DN could be viewed as a disease of acquired mitochondrial dysfunction. In conclusion, we have shown that metabolic dysfunction can be detected in peripheral blood samples of patients with DN. We have also shown using renal cells in-vitro that hyperglycaemia affects mitochondria, with MtDNA levels changing before other indicators of mitochondrial dysfunction. Our data suggests that the BHI formula can indicate mitochondrial dysfunction in live PBMCs from patients, and therefore could be developed into a non-invasive translational measure of mitochondrial function. It is of importance to determine if the changes we observe in clinical samples precede the onset of DN, and therefore longitudinal studies should be carried out to determine this. The potential of regular and routine monitoring of MtDNA content, integrity and BHI as indicators of DN should be further evaluated. Acknowledgements Special thanks to the patients, and the volunteers without whose samples this work could not have been done, we are indebted to the nursing/clinical staff at the Diabetes clinic at Guy's Hospital, especially Nurse Siew Cohen for her tireless help and constant encouragement. Thanks to Dr Sylvie Bannwarth (Laboratoire de G√©n√©tiqueMol√©culaire, Nice France) for the help with the surveyor nuclease method, Dr John Harris and the Nikon Imaging Centre at KCL, Alex Liversage of Seahorse Biosciences for loaning us a Seahorse XFe96, Professor Kinya Otsu, KCL, Professor Victor Darley-Usmar (University of Alabama, USA) and Dr David Ferrick, Seahorse Biosciences, for the critical evaluation of this manuscript. Appendix A. Supplementary Data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ebiom.2015.04.002. References Aguer, C., Pasqua, M., Thrush, A.B., Moffat, C., McBurney, M., Jardine, K., Zhang, R., Beauchamp, B., Dent, R., McPherson, R., et al., 2013. Increased proton leak and SOD2 expression in myotubes from obese non-diabetic subjects with a family history of type 2 diabetes. Biochim. Biophys. Acta 1832, 1624‚Äì1633. Ajaz, S., Czajka, A., Malik, A.N., 2014. Accurate measurement of circulating mitochondrial DNA content from human blood samples using real-time quantitative PCR. Methods Mol. Biol. 1264. Alaveras, A.E., Thomas, S.M., Sagriotis, A., Viberti, G.C., 1997. Promoters of progression of diabetic nephropathy: the relative roles of blood glucose and blood pressure control. Nephrol. Dial. Transplant. 12 (Suppl. 2), 71‚Äì74. Aronson, D., 2008. Hyperglycemia and the pathobiology of diabetic complications. Adv. Cardiol. 45, 1‚Äì16. Ballinger, S.W., Patterson, C., Yan, C.N., Doan, R., Burow, D.L., Young, C.G., Yakes, F.M., Van Houten, B., Ballinger, C.A., Freeman, B.A., et al., 2000. Hydrogen peroxide- and 511 peroxynitrite-induced mitochondrial DNA damage and dysfunction in vascular endothelial and smooth muscle cells. Circ. Res. 86, 960‚Äì966. Bannwarth, S., Procaccio, V., Paquis-Flucklinger, V., 2006. Rapid identiÔ¨Åcation of unknown heteroplasmic mutations across the entire human mitochondrial genome with mismatch-speciÔ¨Åc Surveyor Nuclease. Nat. Protoc. 1, 2037‚Äì2047. Brand, M.D., Nicholls, D.G., 2011. Assessing mitochondrial dysfunction in cells. Biochem. J. 435, 297‚Äì312. Brownlee, M., 2001. Biochemistry and molecular cell biology of diabetic complications. Nature 414, 813‚Äì820. Chacko, B.K., Reily, C., Srivastava, A., Johnson, M.S., Ye, Y., Ulasova, E., Agarwal, A., Zinn, K.R., Murphy, M.P., Kalyanaraman, B., et al., 2010b. Prevention of diabetic nephropathy in Ins2(+/)‚àí(AkitaJ) mice by the mitochondria-targeted therapy MitoQ. Biochem. J. 432, 9‚Äì19. Chacko, B.K., Reily, C., Benavides, C.A., Johnson, M.S., Darley-Usmar, A.V.M., 2010a. Chronic Hyperglycemia-induced Attenuation of Mitochondrial Reserve Capacity Mediates Mesangial Cell Dysfunction in Diabetes. Chacko, B.K., Kramer, P.A., Ravi, S., Johnson, M.S., Hardy, R.W., Ballinger, S.W., DarleyUsmar, V.M., 2013. Methods for deÔ¨Åning distinct bioenergetic proÔ¨Åles in platelets, lymphocytes, monocytes, and neutrophils, and the oxidative burst from human blood. Lab. Investig. 93, 690‚Äì700. Chacko, B.K., Kramer, P.A., Ravi, S., Benavides, G.A., Mitchell, T., Dranka, B.P., Ferrick, D., Singal, A.K., Ballinger, S.W., Bailey, S.M., et al., 2014. The Bioenergetic Health Index: a new concept in mitochondrial translational research. Clin. Sci. (Lond.) 127, 367‚Äì373. Chinnery, P.F., 2002. Modulating heteroplasmy. Trends Genet. 18, 173‚Äì176. Choi, Y.S., Lee, K.U., Pak, Y.K., 2004. Regulation of mitochondrial transcription factor A expression by high glucose. Ann. N. Y. Acad. Sci. 1011, 69‚Äì77. Clarkson, M.R., Murphy, M., Gupta, S., Lambe, T., Mackenzie, H.S., Godson, C., Martin, F., Brady, H.R., 2002. High glucose-altered gene expression in mesangial cells. Actinregulatory protein gene expression is triggered by oxidative stress and cytoskeletal disassembly. J. Biol. Chem. 277, 9707‚Äì9712. Collins, L.V., Hajizadeh, S., Holme, E., Jonsson, I.M., Tarkowski, A., 2004. Endogenously oxidized mitochondrial DNA induces in vivo and in vitro inÔ¨Çammatory responses. J. Leukoc. Biol. 75, 995‚Äì1000. Colussi, C., Albertini, M.C., Coppola, S., Rovidati, S., Galli, F., Ghibelli, L., 2000. H2O2-induced block of glycolysis as an active ADP-ribosylation reaction protecting cells from apoptosis. D'Aco, K.E., Manno, M., Clarke, C., Ganesh, J., Meyers, K.E., Sondheimer, N., 2013. Mitochondrial tRNA(Phe) mutation as a cause of end-stage renal disease in childhood. Pediatr. Nephrol. 28, 515‚Äì519. Dranka, B.P., Benavides, G.A., Diers, A.R., Giordano, S., Zelickson, B.R., Reily, C., Zou, L., Chatham, J.C., Hill, B.G., Zhang, J., et al., 2011. Assessing bioenergetic function in response to oxidative stress by metabolic proÔ¨Åling. Free Radic. Biol. Med. 51, 1621‚Äì1635. Ferrick, D.A., Neilson, A., Beeson, C., 2008. Advances in measuring cellular bioenergetics using extracellular Ô¨Çux. Drug Discov. Today 13, 268‚Äì274. Furda, A., Santos, J.H., Meyer, J.N., Van Houten, B., 2014. Quantitative PCR-based measurement of nuclear and mitochondrial DNA damage and repair in mammalian cells. Methods Mol. Biol. 1105, 419‚Äì437. Gordin, D., Waden, J., Forsblom, C., Thorn, L.M., Rosengard-Barlund, M., Heikkila, O., Saraheimo, M., Tolonen, N., Hietala, K., Soro-Paavonen, A., et al., 2012. Arterial stiffness and vascular complications in patients with type 1 diabetes: the Finnish Diabetic Nephropathy (FinnDiane) Study. Ann. Med. 44, 196‚Äì204. Gruden, G., Thomas, S., Burt, D., Lane, S., Chusney, G., Sacks, S., Viberti, G., 1997. Mechanical stretch induces vascular permeability factor in human mesangial cells: mechanisms of signal transduction. Proc. Natl. Acad. Sci. U. S. A. 94, 12112‚Äì12116. Hartman, M.L., Shirihai, O.S., Holbrook, M., Xu, G., Kocherla, M., Shah, A., Fetterman, J.L., Kluge, M.A., Frame, A.A., Hamburg, N.M., et al., 2014. Relation of mitochondrial oxygen consumption in peripheral blood mononuclear cells to vascular function in type 2 diabetes mellitus. Vasc. Med. 19, 67‚Äì74. He, Z., King, G.L., 2004. Microvascular complications of diabetes. Endocrinol. Metab. Clin. N. Am. 33, 215‚Äì238 (xi-xii). Higginbotham, J.N., Seth, P., Blaese, R.M., Ramsey, W.J., 2002. The release of inÔ¨Çammatory cytokines from human peripheral blood mononuclear cells in vitro following exposure to adenovirus variants and capsid. Hum. Gene Ther. 13, 129‚Äì141. Hock, M.B., Kralli, A., 2009. Transcriptional control of mitochondrial biogenesis and function. Annu. Rev. Physiol. 71, 177‚Äì203. Hofmann, M.A., Schiekofer, S., Isermann, B., Kanitz, M., Henkels, M., Joswig, M., Treusch, A., Morcos, M., Weiss, T., Borcea, V., et al., 1999. Peripheral blood mononuclear cells isolated from patients with diabetic nephropathy show increased activation of the oxidative-stress sensitive transcription factor NF-kappaB. Diabetologia 42, 222‚Äì232. Hudson, G., Gomez-Duran, A., Wilson, I.J., Chinnery, P.F., 2014. Recent mitochondrial DNA mutations increase the risk of developing common late-onset human diseases. PLoS Genet. 10, e1004369. IDF, 2013. Diabetes Atlas. 6th edition. International Diabetes Federation. Ihm, C.G., Park, J.K., Hong, S.P., Lee, T.W., Cho, B.S., Kim, M.J., 1997. Circulating factors in sera or peripheral blood mononuclear cells in patients with membranous nephropathy or diabetic nephropathy. J. Korean Med. Sci. 12, 539‚Äì544. Kalyanaraman, B., Darley-Usmar, V., Davies, K.J., Dennery, P.A., Forman, H.J., Grisham, M.B., Mann, G.E., Moore, K., Roberts II, L.J., Ischiropoulos, H., 2012. Measuring reactive oxygen and nitrogen species with Ô¨Çuorescent probes: challenges and limitations. Free Radic. Biol. Med. 52, 1‚Äì6. Kilpatrick, E.S., Rigby, A.S., Atkin, S.L., 2009. The Diabetes Control and Complications Trial: the gift that keeps giving. Nat. Rev. Endocrinol. 5, 537‚Äì545. Koopman, W.J., Verkaart, S., Visch, H.J., van der Westhuizen, F.H., Murphy, M.P., van den Heuvel, L.W., Smeitink, J.A., Willems, P.H., 2005. Inhibition of complex I of the electron transport chain causes O‚àí 2 ‚Ä¢-mediated mitochondrial outgrowth. Am. J. Physiol. Cell Physiol. 288, C1440‚ÄìC1450. 512 A. Czajka et al. / EBioMedicine 2 (2015) 499‚Äì512 Lee, H.C., Wei, Y.H., 2005. Mitochondrial biogenesis and mitochondrial DNA maintenance of mammalian cells under oxidative stress. Int. J. Biochem. Cell Biol. 37, 822‚Äì834. Lee, J.E., Park, H., Ju, Y.S., Kwak, M., Kim, J.I., Oh, H.Y., Seo, J.S., 2009. Higher mitochondrial DNA copy number is associated with lower prevalence of microalbuminuria. Exp. Mol. Med. 41, 253‚Äì258. Malik, A.N., Czajka, A., 2013. Is mitochondrial DNA content a potential biomarker of mitochondrial dysfunction? Mitochondrion 13 (5), 481‚Äì492. Malik, A.N., Shahni, R., Iqbal, M.M., 2009. Increased peripheral blood mitochondrial DNA in type 2 diabetic patients with nephropathy. Diabetes Res. Clin. Pract. 86, e22‚Äìe24. Malik, A.N., Shahni, R., Rodriguez-de-Ledesma, A., Laftah, A., Cunningham, P., 2011. Mitochondrial DNA as a non-invasive biomarker: accurate quantiÔ¨Åcation using real time quantitative PCR without co-ampliÔ¨Åcation of pseudogenes and dilution bias. Biochem. Biophys. Res. Commun. 412, 1‚Äì7. Maynard, S., Keijzers, G., Gram, M., Desler, C., Bendix, L., Budtz-Jorgensen, E., Molbo, D., Croteau, D.L., Osler, M., Stevnsner, T., et al., 2013. Relationships between human vitality and mitochondrial respiratory parameters, reactive oxygen species production and dNTP levels in peripheral blood mononuclear cells. Aging (Albany NY) 5, 850‚Äì864. Mazzaccara, C., Iafusco, D., Liguori, R., Ferrigno, M., Galderisi, A., Vitale, D., Simonelli, F., Landolfo, P., Prisco, F., Masullo, M., et al., 2012. Mitochondrial diabetes in children: seek and you will Ô¨Ånd it. PLoS One 7, e34956. Michel, S., Wanet, A., De Pauw, A., Rommelaere, G., Arnould, T., Renard, P., 2012. Crosstalk between mitochondrial (dys)function and mitochondrial abundance. J. Cell. Physiol. 227, 2297‚Äì2310. Moraes, C.T., Shanske, S., Tritschler, H.J., Aprille, J.R., Andreetta, F., Bonilla, E., Schon, E.A., DiMauro, S., 1991. mtDNA depletion with variable tissue expression: a novel genetic abnormality in mitochondrial diseases. Am. J. Hum. Genet. 48, 492‚Äì501. Murphy, M., Godson, C., Cannon, S., Kato, S., Mackenzie, H.S., Martin, F., Brady, H.R., 1999. Suppression subtractive hybridization identiÔ¨Åes high glucose levels as a stimulus for expression of connective tissue growth factor and other genes in human mesangial cells. J. Biol. Chem. 274, 5830‚Äì5834. Navarro-Gonzalez, J.F., Mora-Fernandez, C., Muros de Fuentes, M., Garcia-Perez, J., 2011. InÔ¨Çammatory molecules and pathways in the pathogenesis of diabetic nephropathy. Nat. Rev. Nephrol. 7, 327‚Äì340. Niaudet, P., 1998. Mitochondrial disorders and the kidney. Arch. Dis. Child. 78, 387‚Äì390. Oka, T., Hikoso, S., Yamaguchi, O., Taneike, M., Takeda, T., Tamai, T., Oyabu, J., Murakawa, T., Nakayama, H., Nishida, K., Akira, S., Yamamoto, A., Komuro, I., Otsu, K., 2012. Mitochondrial DNA that escapes from autophagy causes inÔ¨Çammation and heart failure. Nature 485, 251‚Äì255. Ritz, E., Orth, S.R., 1999. Nephropathy in patients with type 2 diabetes mellitus. N. Engl. J. Med. 341, 1127‚Äì1133. Rudkowska, I., Raymond, C., Ponton, A., Jacques, H., Lavigne, C., Holub, B.J., Marette, A., Vohl, M.C., 2011. Validation of the use of peripheral blood mononuclear cells as surrogate model for skeletal muscle tissue in nutrigenomic studies. OMICS 15, 1‚Äì7. Seidowsky, A., Hoffmann, M., Glowacki, F., Dhaenens, C.M., Devaux, J.P., de Sainte Foy, C.L., Provot, F., Gheerbrant, J.D., Hummel, A., Hazzan, M., et al., 2013. Renal involvement in MELAS syndrome ‚Äî a series of 5 cases and review of the literature. Clin. Nephrol. 80, 456‚Äì463. Shahni, R., Czajka, A., Mankoo, B.S., Guvenel, A.K., King, A.J., Malik, A.N., 2013. Nop-7-associated 2 (NSA2), a candidate gene for diabetic nephropathy, is involved in the TGFbeta1 pathway. Int. J. Biochem. Cell Biol. 45, 626‚Äì635. Sharma, K., Karl, B., Mathew, A.V., Gangoiti, J.A., Wassel, C.L., Saito, R., Pu, M., Sharma, S., You, Y.H., Wang, L., et al., 2013. Metabolomics reveals signature of mitochondrial dysfunction in diabetic kidney disease. J. Am. Soc. Nephrol. 24, 1901‚Äì1912. Singh, D.K., Winocour, P., Farrington, K., 2011. Oxidative stress in early diabetic nephropathy: fueling the Ô¨Åre. Nat. Rev. Endocrinol. 7, 176‚Äì184. Sobenin, I.A., Mitrofanov, K.Y., Zhelankin, A.V., Sazonova, M.A., Postnov, A.Y., Revin, V.V., Bobryshev, Y.V., Orekhov, A.N., 2014. Quantitative assessment of heteroplasmy of mitochondrial genome: perspectives in diagnostics and methodological pitfalls. BioMed Res. Int. 2014, 292017. Sourris, K.C., Harcourt, B.E., Penfold, S.A., Yap, F.Y., Morley, A.L., Morgan, P.E., Davies, M.J., Baker, S.T., Jerums, G., Forbes, J.M., 2010. Modulation of the cellular expression of circulating advanced glycation end-product receptors in type 2 diabetic nephropathy. Exp. Diabetes Res. 2010, 974681. Stoves, J., Lindley, E.J., BarnÔ¨Åeld, M.C., Burniston, M.T., Newstead, C.G., 2002. MDRD equation estimates of glomerular Ô¨Åltration rate in potential living kidney donors and renal transplant recipients with impaired graft function. Nephrol. Dial. Transplant. 17, 2036‚Äì2037. Stratton, I.M., Adler, A.I., Neil, H.A., Matthews, D.R., Manley, S.E., Cull, C.A., Hadden, D., Turner, R.C., Holman, R.R., 2000. Association of glycaemia with macrovascular and microvascular complications of type 2 diabetes (UKPDS 35): prospective observational study. BMJ 321, 405‚Äì412. Thomas, S., Vanuystel, J., Gruden, G., Rodriguez, V., Burt, D., Gnudi, L., Hartley, B., Viberti, G., 2000. Vascular endothelial growth factor receptors in human mesangium in vitro and in glomerular disease. J. Am. Soc. Nephrol. 11, 1236‚Äì1243. Trudeau, K., Molina, A.J., Guo, W., Roy, S., 2010. High glucose disrupts mitochondrial morphology in retinal endothelial cells: implications for diabetic retinopathy. Am. J. Pathol. 177, 447‚Äì455. Tuppen, H.A., Blakely, E.L., Turnbull, D.M., Taylor, R.W., 2010. Mitochondrial DNA mutations and human disease. Biochim. Biophys. Acta 1797, 113‚Äì128. Wallace, D.C., 1999. Mitochondrial diseases in man and mouse. Science 283, 1482‚Äì1488. Wallace, D.C., Chalkia, D., 2013. Mitochondrial DNA genetics and the heteroplasmy conundrum in evolution and disease. Cold Spring Harb. Perspect. Med. 3, a021220. Wild, S., Roglic, G., Green, A., Sicree, R., King, H., 2004. Global prevalence of diabetes: estimates for the year 2000 and projections for 2030. Diabetes Care 27, 1047‚Äì1053. Williams, R.S., 1986. Mitochondrial gene expression in mammalian striated muscle. Evidence that variation in gene dosage is the major regulatory event. J. Biol. Chem. 261, 12390‚Äì12394. Wojtczak, L., Zab≈Çocki, K., 2008. Basic mitochondria physiology in cell viability and death. In: Dykens, J.A., Will, Y. (Eds.), Drug-induced Mitochondria Dysfunction. John Wiley & Sons, Inc., New Jersey, pp. 3‚Äì36. Ye, K., Lu, J., Ma, F., Keinan, A., Gu, Z., 2014. Extensive pathogenicity of mitochondrial heteroplasmy in healthy human individuals. Proc. Natl. Acad. Sci. U. S. A. 111, 10654‚Äì10659. Yi, B., Hu, X., Zhang, H., Huang, J., Liu, J., Hu, J., Li, W., Huang, L., 2014. Nuclear NF-kappaB p65 in peripheral blood mononuclear cells correlates with urinary MCP-1, RANTES and the severity of type 2 diabetic nephropathy. PLoS One 9, e99633. Zhang, Y., Lee, A.S., Shameli, A., Geng, X., Finegood, D., Santamaria, P., Dutz, J.P., 2010. TLR9 blockade inhibits activation of diabetogenic CD8+ T cells and delays autoimmune diabetes. J. Immunol. 184, 5645‚Äì5653. Alternative diagnoses at paediatric appendicitis MRI * M.M. Moorea, , , * A.N. Kulaylatb, * J.M. Briana, * A. Khakuc, * M.A. Hulsea, * B.W. Engbrechtb, * S.T. Methrattaa, * D.K.B. Boala Show more doi:10.1016/j.crad.2015.03.001 Get rights and content As the utilization of MRI in the assessment for paediatric appendicitis increases in clinical practice, it is important to recognize alternative diagnoses as the cause of abdominal pain. The purpose of this review is to share our institution's experience using MRI in the evaluation of 510 paediatric patients presenting with suspected appendicitis over a 30 month interval (July 2011 to December 2013). An alternative diagnosis was documented in 98/510 (19.2%) patients; adnexal pathology (6.3%, n = 32), enteritis–colitis (6.3%,n = 32), and mesenteric adenitis (2.2%, n = 11) comprised the majority of cases. These common entities and other less frequent illustrative cases obtained during our overall institutional experience with MRI for suspected appendicitis are reviewed. Introduction MRI for the evaluation of paediatric appendicitis is effective and avoids the potential detrimental effects of ionizing radiation.1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11 and 12 With several articles confirming both the clinical feasibility and diagnostic accuracy of MRI in both paediatric and adult populations, clinical practice is beginning to shift toward the utilization of MRI in lieu of CT.1, 2, 3, 4, 5, 6, 7 and 12 MRI may be employed as either the primary imaging technique or in a staged approach following ultrasound, depending on institutional structure and preference.1, 12, 13 and 14 At our institution, we have been using MRI as the primary imaging technique for the assessment of paediatric appendicitis (using 1.5 and 3 T systems with an 18 channel phased-array flexible body coil). Our protocol consists of four sequences: axial and coronal single-shot turbo spin echo (SS-TSE), without and with fat-saturation. Neither intravenous nor oral contrast media are utilized. Examinations are performed without sedation, with a lower age recommendation of 5 years (although even younger patients may be considered depending on their ability to cooperate in the absence of sedation). As discussion of the optimal MRI protocol or imaging algorithms have been previously described in the literature, they will not be detailed further in this report.1, 2, 3, 4, 5, 6, 7, 12, 15,16 and 17 A representative case of paediatric acute appendicitis is provided in Fig 1. Multiple additional examples and case interpretation pearls have previously been reported by our group.18 Figure 1. Acute appendicitis. A 6-year-old male patient with 1 day of abdominal pain and vomiting. Axial SS-TSE image demonstrates peri-appendiceal inflammation, enlarged appendix, and intraluminal fluid; MRI findings indicating acute appendicitis. Acute appendicitis was confirmed pathologically. Figure options During the diagnostic evaluation of acute abdominal pain, alternative diagnoses may be encountered in the child other than appendicitis.15, 17 and 19 In an institutional review board-approved retrospective analysis of a 30 month interval (July 2011 to December 2013) at our medical centre, 510 paediatric patients underwent MRI for clinically suspected appendicitis with a resultant high degree of diagnostic accuracy (sensitivity 96.8%, specificity 97.4%, positive predictive value 92.4%, and negative predictive value of 98.9%). This analysis also demonstrated favourable clinical and surgical outcomes (71 min was the median time from request to scan; median imaging duration, 11 min; median time from initial assessment to admit order, antibiotic administration, and operating room, 4.1, 4.7, and 9.1 h, respectively).12 Additionally, alternative diagnoses as the aetiology of the clinical presentation were observed in 98 (19.2%) of these patients. Adnexal pathology (6.3%, n = 32), enteritis-colitis (6.3%, n = 32), and mesenteric adenitis (2.2%, n = 11) comprised the majority of cases. A summary of these results is documented in Table 1. Reports of alternative diagnoses encountered following MRI for paediatric appendicitis are very limited. In a recent study by Koning et al., 17which utilized a contrast-enhanced MRI protocol, alternative diagnoses obtained during the evaluation of suspected paediatric appendicitis demonstrated a similar prevalence of the aforementioned diagnoses. Table 1. MRI alternative diagnoses during evaluation of paediatric appendicitis. Diagnosis n (%) (n = 510) Adnexal pathology 32 (6.3%) Enteritis or colitis 32 (6.3%) Mesenteric adenitis 11 (2.2%) Pyelonephritis 5 (1%) Hydronephrosis 5 (1%) Pneumonia 3 (0.6%) Intussusception (2 SB, 1 IC) 3 (0.6%) Small bowel obstruction 1 (0.2%) Midgut volvulus 1 (0.2%) Splenic infarct 1 (0.2%) Cholelithiasis 1 (0.2%) PTLD 1 (0.2%) SBP 1 (0.2%) SMA syndrome 1 (0.2%) SB, small bowel; IC, ileocolic; PTLD, post-transplant lymphoproliferative disorder; SBP, spontaneous bacterial peritonitis; SMA, superior mesenteric artery. Table options With the increasing role of MRI in the evaluation of acute abdominal pain, it is important to review systematically common and unique alternative diagnoses encountered in the evaluation of paediatric acute abdominal pain. There have been no prior pictorial essays focused on alternative diagnoses discovered during paediatric appendicitis MRI. This practical pictorial essay reviews the alternative diagnoses obtained, as well as additional illustrative cases obtained during our overall institutional experience with MRI in the evaluation of paediatric appendicitis. Adnexal pathology (Figure 2, Figure 3 and Figure 4) Adnexal pathology represents a commonly encountered alternative diagnosis (6.3%,n = 32), equal in frequency to enteritis–colitis. Ultrasound is the primary imaging technique of suspected ovarian and adnexal lesions given its diagnostic accuracy, absence of ionizing radiation, and wide availability. 20 and 21 However, given the expanding role of MRI in the evaluation of acute appendicitis, adnexal pathology will be encountered. Consistent with previous reports, commonly encountered ovarian lesions include functional and haemorrhagic cysts. 17, 18 and 21 Less commonly, cystic neoplasms may be encountered, as illustrated by the case of a mucinous cystadenoma. Unlike CT evaluation, which provides very limited morphological evaluation and tissue contrast of the ovarian parenchyma, ovarian pathology is well delineated as part of MRI evaluation.22 Therefore, ultrasound following MRI is often redundant and, at our institution, is reserved for cases where additional delineation is needed. If the ovaries are normal or adnexal pathology is clearly defined on paediatric appendicitis MRI, pelvic ultrasound is not routinely performed. Figure 2. Ovarian cysts: haemorrhagic. A 14-year-old female patient with acute abdominal pain. (a) Axial SS-TSE image with fat-saturation demonstrated a complex cyst (arrow) arising from otherwise normal left ovarian tissue. Subsequent ultrasound (not shown) following MRI confirmed a haemorrhagic cyst. Two years later, now 16 years old, the same patient re-presented with right lower quadrant pain. (b) Axial SS-TSE image demonstrated a complex right ovarian cyst with a reticular internal pattern and dependent debris most consistent a haemorrhagic cyst (arrow). No further emergent imaging was obtained. Figure options Figure 3. Mucinous cystadenoma. A 15-year-old female patient with a 4 day history of worsening right lower quadrant pain. Axial SS-TSE image demonstrated a very large left ovarian cyst (arrow) with several thin internal septa, interpreted as most consistent with a benign neoplasm such as cystadenoma. Intraoperatively, a left ovarian cystectomy was performed with mucinous cystadenoma confirmed at histopathology. Figure options Figure 4. Ovarian torsion. A 10-year-old female patient presented with acute right lower quadrant pain of 12 h duration. Axial SS-TSE images without and with fat saturation demonstrated asymmetric enlargement of the right ovary with peripheralization of follicles and indistinctness of stromal architecture (arrows) with focal adjacent oedema (a–b) interpreted as right ovarian torsion. Laparoscopy was performed without additional imaging. Intraoperatively, right ovarian torsion was confirmed with a 180° degree rotation but no ischaemia. Right oophoropexy was performed. Figure options If adnexal torsion is clinically suspected, ultrasound with Doppler should be performed, rather than MRI. Adnexal torsion is a surgical emergency, characterized by the rotation of the ovary, fallopian tube, or both around a comprised vascular pedicle. However, as the use of MRI expands in the evaluation of acute appendicitis, an unexpected ovarian torsion may also be encountered.17 and 23 The majority of cases are associated with a cyst (most commonly follicular or corpus luteal) or a benign ovarian tumour (most commonly dermoid).21 and 23 Sonographic findings of torsion include asymmetrical enlargement, oedema, peripheralization of follicles, and ipsilateral deviation of the ovary or uterus, which may also be observed with MRI.6, 23 and 24 Enteritis–colitis (Figure 5 and Figure 6) Enteritis–colitis (6.3%, n = 32) is a commonly encountered alternative diagnosis, equal in frequency to adnexal pathology. MRI findings of enteritis and colitis include bowel wall thickening, bowel wall oedema, intraluminal fluid, and adjacent inflammation. Fig 5demonstrates a severe case of infectious colitis that resulted in haemolytic–uraemic syndrome (HUS). HUS is defined by microangiopathic haemolytic anaemia, thrombocytopaenia and renal failure. Shiga toxin from Escherichia coli 0157:H7 is the most common aetiology of HUS. The imaging findings of HUS in paediatric patients have been described on contrast enema, ultrasound, and CT, most often demonstrating colonic wall thickening and irregular luminal narrowing. 25 HUS may also result in colonic stricture. 26 There is only one prior report documenting the subsequent diagnosis of HUS following MRI evaluation of abdominal pain in a 30-year-old pregnant patient. However, in that patient no major abnormality was identified at abdominal MRI. 27 Figure 5. HUS. A previously healthy 11-year-old female patient presented with 2 days of right lower quadrant abdominal pain. (a) Axial and (b) coronal SS-TSE fat-saturation images demonstrate severe wall thickening and oedema of the ascending colon with surrounding inflammation and fluid. Patient was diagnosed with severe colitis, infectious versus inflammatory. Subsequently, stool cultures were positive for Shiga toxin 2 and the patient developed thrombocytopaenia and acute renal failure. Figure options Figure 6. Crohn's disease. A 10-year-old male patient presented with acute right lower quadrant pain with accompanying nausea and anorexia. He was diagnosed with Crohn's disease in the previous year, but without previous exacerbation, and his clinical presentation was concerning for acute appendicitis. Axial SS-TSE image demonstrated bowel wall thickening, irregularity, and hyperintensity of the terminal ileum with surrounding free fluid (arrow); the appendix (not shown) was normal. Patient was subsequently started on infliximab with good response at follow-up with his gastroenterologist. Figure options MRI accurately depicts the inflammatory findings of inflammatory bowel disease (IBD) with additional advantages of superior contrast resolution in comparison to CT and the avoidance of ionizing radiation.28 At MRI, bowel-wall thickening and oedema is usually present in Crohn's disease and is pronounced when inflammation is active. Additional features that may be observed include lymphadenopathy, fistula or abscess formation, and abnormal fold patterns.29 A MRI enterography protocol with oral and intravenous contrast media is preferred; however, an emergent appendicitis protocol may depict many of the hallmark findings of IBD, as illustrated in Fig 6. Mesenteric adenitis (Fig 7) Mesenteric adenitis commonly mimics acute appendicitis in children and is characterized by multiple enlarged lymph nodes in the right lower quadrant and mesentery (most commonly anterior to right psoas muscle) in the absence of other pathology, or in combination with associated enteritis.30 and 31 Although the size criteria for short axis enlargement is often considered 5 mm, a more recent study suggests 8 mm may be more appropriate.32 Mesenteric adenitis is an important alternative diagnosis as it is self-limiting, managed conservatively, and diagnosis helps to facilitate patient discharge. Mesenteric adenitis was present in 2.2% of our patients. Figure 7. Mesenteric adenitis. A 6-year-old girl presented with 1 day of diffuse abdominal pain, fever, and one episode of emesis. Although the abdominal examination was benign, there was a lingering clinical concern for early appendicitis. (a) Coronal SS-TSE image demonstrates multiple prominent lymph nodes situated anterior to the right psoas muscle (arrow). (b) Axial SS-TSE image with fat-saturation demonstrates mild hyperintensity of these nodes (arrow). Resolution of symptoms was confirmed at follow-up with family physician 4 weeks later. Figure options Pyelonephritis (Fig 8) Acute infection of the kidney is a common cause of abdominal pain and hence may mimic the clinical presentation of acute appendicitis. Pyelonephritis can be unifocal, multifocal, mass-like, or produce a striated nephrogram on T2-weighted imaging. Abnormally increased signal intensity within the kidney and adjacent inflammation in the perinephric space have been previously described on MRI.6 MRI can also be useful in identifying perinephric fluid collections. Figure 8. Pyelonephritis. A 7-year-old female patient presented with abdominal pain and fever. Coronal SS-TSE image with fat-saturation demonstrated a thin rim of perinephric fluid, asymmetrically increased signal intensity, and enlargement of the left kidney. Left pyelonephritis was subsequently confirmed by urine culture, which demonstrated Escherichia coli. Figure options Hydronephrosis (Fig 9) Hydronephrosis was present in 1.0% (n = 5) of our patients. Hydronephrosis may be either an asymptomatic finding from pre-existing pathology or the cause of acute abdominal pain. Differentiation between these will rely in part on clinical information, prior imaging studies, and urinalysis. One common cause of acute abdominal pain is a ureteral calculus. Although MRI is insensitive for evaluating renal calculi less than 1 cm, hydronephrosis and perinephric fluid are well demonstrated on MRI and can assist in the diagnosis. 6 Figure 9. Hydronephrosis. A 13-year-old girl with right lower quadrant abdominal pain. MRI was negative for appendicitis; however, axial SS-TSE sequences (a–b) demonstrated moderate right hydroureteronephrosis (arrows). No calculus could be identified at MRI. Clinical symptoms combined with strong family history suggested renal calculus in the setting of hydroureteronephrosis. At subsequent urology outpatient follow-up and ultrasound 3 weeks later, symptoms had resolved, right hydroureteronephrosis had resolved and renal calculi were confirmed (not shown). Figure options Pneumonia (Fig 10) Pathology within the lower lobes of the lungs (n = 3, 0.6%) may present as abdominal pain. The case of lower lobe pneumonia serves as a reminder of the importance of lung base evaluation. As MRI may be the only imaging method employed in the evaluation of acute abdominal pain, the authors believe that the lung bases should be included on at least one sequence, similar to their inclusion on abdominal CT. 1 Figure 10. Pneumonia. A 7-year-old male patient presented with severe mid-abdominal pain, worsening with movement, of 12 h duration. The appendix was normal. Axial SS-TSE image demonstrated left lower-lobe pneumonia (arrow). Figure options Intussusception: ileocolic and small bowel (Figure 11 and Figure 12) Ultrasound has been validated as a first-line diagnostic test for intussusception and is standard practice at our institution.33 If the patient is beyond the expected age range for intussusception or this diagnosis is not clinically suspected (Fig 11), intussusception may be identified at MRI rather than ultrasound. The diagnosis of ileocolic intussusception is well described on radiographs, ultrasound, and CT.34 The presented cases of ileocolic intussusception demonstrate not only the expected “target” sign of intussusceptum within the intussuscipiens, but also high signal intensity fluid and inflammation between these loops of bowel and surrounding the intussuscipiens (Fig 12a). Additionally, despite a reported sensitivity of 97.9%, it is possible, particularly with a smaller intussusception or technically difficult examination, for the diagnosis to be unrecognized sonographically.33 Figure 11. Ileocolic intussusception. A 16-year-old male patient with a 3 day history of right lower quadrant abdominal pain, with associated nausea. Coronal SS-TSE with fat-saturation (a) and axial SS-TSE (b) images show intussusceptum within the transverse colon intussuscipiens (arrow). Air contrast enema was performed with reduction to the level of the ileocaecal valve. As ileocolic intussusception is highly unusual in the adolescent population, diagnostic laparoscopy was performed to complete reduction and evaluate for underlying lead point. No underlying pathology was discovered. Figure options Figure 12. Intussusceptions: ileocolic and small bowel. (a) A 3-year-old male patient with acute abdominal pain, and right upper quadrant tenderness without emesis or diarrhoea. Ultrasound was performed but was limited by patient agitation and motion during examination. Sonographically, no intussusception was identified and the appendix was not visualized. During further clinical observation, the patient developed emesis and the white blood cell count was elevated. MRI was performed for appendicitis with sedation in this rare instance demonstrating a short ileocolic intussusception with a “target” sign on MRI (arrow). (b) Additionally, a 6-year-old male patient with right-sided abdominal pain, fever, emesis, and anorexia. Coronal SS-TSE image with fat-saturation demonstrates a small bowel intussusception within the mid-abdomen (arrow). The patients was observed and given intravenous fluids for gastroenteritis with improvement of symptoms. Figure options Small bowel–small bowel intussusception is another alternative diagnosis that may cause abdominal pain. Unlike ileocolic intussusception, small bowel into small bowel intussusceptions are often short segment, do not demonstrate significant wall oedema, and are usually self-limited (Fig 12b). Jejuno-jejunal intussusception has been previously reported in 13-year-old male patient with Peutz–Jeghers syndrome on MR enterography.28 As demonstrated in this reported case, small bowel intussusceptions occur with increased frequency in certain children (e.g., small bowel polyp syndromes, Henoch–Schönlein purpura). Ataxia telangiectasia (Fig 13) Ataxia telangiectasia is an autosomal recessive disorder characterized by cutaneous telangiectasia, cerebellar ataxia, and immunodeficiency.35 This immunodeficiency often results in pneumonia, sinusitis, and bronchiectasis. These patients also have a substantial likelihood of developing malignancy, particularly lymphoreticular malignancies such as lymphoma and leukaemia. A single gene mutation on chromosome 11 causes an aberration in the DNA repair pathway resulting in greatly heightened sensitivity to the detrimental effects of ionizing radiation.35 and 36 Therefore, it is critical to avoid ionizing radiation in these children. Previous MRI descriptions have focused primarily on the neurological aspects of the disease.36Fig 13 illustrates that MRI in this setting is also applicable to suspected appendicitis. Figure 13. Ataxia telangiectasia. A 15-year-old male patient with ataxia telangiectasia hospitalized with leukaemia. Additional imaging was requested to exclude appendicitis in the setting of increasing abdominal pain. (a) A normal appendix (arrow) within a large amount of ascites was seen on coronal SS-TSE images with fat-saturation. (b) A fluid–fluid level in the pelvis axial SS-TSE images with fat-saturation indicated complicated ascites (arrow). As the patient also had severe autoimmune thrombocytopaenia, the findings were most consistent with haemorrhagic ascites in setting of ataxia telangiectasia. Figure options Midgut volvulus (Fig 14) Midgut volvulus is a well-described paediatric surgical emergency. If imaging evaluation for suspected midgut volvulus is undertaken, upper gastrointestinal (GI) series is the study of choice. If volvulus had been a clinical concern rather than appendicitis, emergent upper GI series rather than MRI would have been performed. In this illustrative case, fortunately, imaging evaluation was prompt with the time of examination request to first MRI sequence of 66 min and the study duration of 9 min. One previous case of malrotation was described in an 11-year-old female patient utilizing MR enterography and a case has also been reported in an adult utilizing MR angiography.28 and 37 MR findings are expected to be the same as CT findings including malposition of bowel, lack of duodenal crossing from right to left inferior to the superior mesenteric artery (SMA), anomalous SMA/superior mesenteric vein (SMV) relationship with a “whirlpool” sign, and proximal gastric/bowel dilatation with decompressed distal small and large bowel.37 and 38 Figure 14. Midgut volvulus. A 4-year-old male patient with 2 days of progressively worsening right lower quadrant abdominal pain followed by vomiting. (a) Coronal SS-TSE image demonstrates dilatation of the fluid-filled proximal duodenum and stomach (arrow). (b) Axial SS-TSE image shows the “whirlpool” sign with a twist of the superior mesenteric vasculature (arrow). The patient was taken emergently to the operating room (without additional imaging) and intraoperatively a midgut volvulus with a 180° rotation (without intestinal compromise) was confirmed, with subsequent Ladd's procedure performed. Figure options Conclusion The purpose of this review was to describe the appearance of various alternative clinical entities encountered during imaging of paediatric appendicitis with MRI. It should be noted that MRI is not the primary imaging technique for many of the alternative pathologies described and that ultrasound often provides comparable and, in particular instances, superior information depending on the aetiology in question.39, 40, 41, 42 and 43The role of MRI in the evaluation of paediatric acute abdominal pain continues to evolve, and its application in relation to ultrasonography and CT has not yet been fully defined. There are particular considerations that must be weighted, including cost-effectiveness, as well as institutional and health system capabilities. Further discussion is beyond the scope of the present manuscript. Although acknowledging these considerations, growing experience continues to validate MRI as a safe and accurate technique for the evaluation of paediatric appendicitis. Utilization of MRI additionally facilitates rapid identification of alternative diagnoses as demonstrated in this review, allowing for appropriate consultation or safe discharge. Given the risks of ionizing radiation, MRI has the potential to supplant CT in the evaluation of paediatric appendicitis and accompanying alternative diagnoses. CLINICAL RESEARCH STUDY Alternative Smoking Cessation Aids: A Meta-analysis of Randomized Controlled Trials Mehdi Tahiri, MD,a,b,c Salvatore Mottillo, MD,a,b,c Lawrence Joseph, PhD,d,e Louise Pilote, MD, MPH, PhD,d,e,f Mark J. Eisenberg, MD, MPHa,c,d a Divisions of Cardiology and Clinical Epidemiology, Jewish General Hospital/McGill University, Montreal, Quebec, Canada; bFaculty of Medicine, University of Montreal, Montreal, Quebec, Canada; cLady Davis Institute for Medical Research, Jewish General Hospital/ McGill University, Montreal, Quebec, Canada; dDepartment of Epidemiology, Biostatistics, and Occupational Health, McGill University, Montreal, Quebec, Canada; eDivision of Clinical Epidemiology, McGill University Health Centre, Montreal, Quebec, Canada; fDivision of Internal Medicine, McGill University Health Centre, Montreal, Quebec, Canada. ABSTRACT BACKGROUND: Acupuncture, hypnotherapy, and aversive smoking are the most frequently studied alternative smoking cessation aids. These aids are often used as alternatives to pharmacotherapies for smoking cessation; however, their efficacy is unclear. METHODS: We carried out a random effect meta-analysis of randomized controlled trials to determine the efficacy of alternative smoking cessation aids. We systematically searched the Cochrane Library, EMBASE, Medline, and PsycINFO databases through December 2010. We only included trials that reported cessation outcomes as point prevalence or continuous abstinence at 6 or 12 months. RESULTS: Fourteen trials were identified; 6 investigated acupuncture (823 patients); 4 investigated hypnotherapy (273 patients); and 4 investigated aversive smoking (99 patients). The estimated mean treatment effects were acupuncture (odds ratio [OR], 3.53; 95% confidence interval [CI], 1.03-12.07), hypnotherapy (OR, 4.55; 95% CI, 0.98-21.01), and aversive smoking (OR, 4.26; 95% CI, 1.26-14.38). CONCLUSION: Our results suggest that acupuncture and hypnotherapy may help smokers quit. Aversive smoking also may help smokers quit; however, there are no recent trials investigating this intervention. More evidence is needed to determine whether alternative interventions are as efficacious as pharmacotherapies. ¬© 2012 Elsevier Inc. All rights reserved. ‚Ä¢ The American Journal of Medicine (2012) 125, 576-584 KEYWORDS: Acupuncture; Alternative smoking cessation aid; Aversive smoking; Hypnotherapy; Meta-analysis; Smoking cessation Funding: MT was supported by a Canadian Cardiovascular Outcomes Research Team summer studentship funded through a Canadian Institutes of Health Research Team Grant in Cardiovascular Outcomes Research. LJ is a Chercheur-National of the Fonds de la Recherche en Sant√© du Qu√©bec (FRSQ). LP is a Chercheur-National of the FRSQ. MJE is a ChercheurNational of the FRSQ. Conflict of Interest: None. Authorship: All authors had access to the data and played a role in writing this manuscript. Reprint requests should be addressed to: Mark J. Eisenberg, MD, MPH, Professor of Medicine, Divisions of Cardiology and Clinical Epidemiology, Jewish General Hospital/McGill University, 3755 C√¥te Ste-Catherine Road, Suite H-421.1, Montreal, Quebec, Canada H3T 1E2. E-mail address: mark.eisenberg@mcgill.ca. 0002-9343/$ -see front matter ¬© 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2011.09.028 Smoking is the most preventable cause of morbidity and mortality in North America and costs the economy $210 billion each year in the United States alone.1-3 Annually, approximately half of the 51 million smokers in North America try to quit for at least 1 day.4-6 Of these smokers attempting cessation, less than half use pharmacologic cessation aids (nicotine replacement therapy, bupropion, or varenicline) because of concerns about potential side effects and limited efficacy.7 Many smokers have turned to alternative smoking cessation aids as a substitute for pharmacotherapies, of which the most studied include acupuncture, hypnotherapy, and aversive smoking. Acupuncture for smoking cessation consists of stimulating specific acupoints on the ear.8 Hypnotherapy consists of inducing an altered state of conscious- Tahiri et al Alternative Smoking Cessation Aids 577 the only difference between arms was the alternative smokness enabling an adherence to smoking cessation.9 Finally, aversive smoking consists of rapidly taking a large number ing cessation aid itself. For randomized controlled trials of puffs in a short period of time.10 with multiple arms per intervention, we reused the control In 2005, 6% of smokers trying to quit underwent acugroup in each comparison. We accounted for this reuse in puncture and 13% underwent hypnotherapy. Furthermore, our analysis, avoiding double-counting of groups from ran30% and 40% of smokers have domized controlled trials with reported an interest in trying acumultiple arms, while using all puncture and hypnotherapy, reavailable data. CLINICAL SIGNIFICANCE spectively.11 These alternative Acupuncture for smoking cesaids are costly, with the total price sation was defined as the stimula‚óè The use of unconventional smoking cesof each therapy ranging from $400 tion of specific acupoints on the sation aids, including acupuncture and to $1000. ear using needles or laser therhypnotherapy, results in substantial inDespite the popularity and high apy.8,12 We included only acucreases in smoking cessation compared cost, the efficacy of alternative puncture studies that controlled with control. smoking cessation aids remains for the intervention using sham unclear. Several randomized conacupuncture (ie, insertion of nee‚óè We recommend that practitioners adtrolled trials have examined their dles or lasers at a location other minister acupuncture and hypnotherapy efficacy; however, widely varying than the ear, where there should be as an alternative to pharmacologic inestimates of their treatment effect no effect). terventions to patients who prefer not have been reported. Our objective Hypnotherapy was defined as to use conventional smoking cessation was to carry out a systematic review the induction of a state of recepaids to quit smoking. and meta-analysis to determine the tive and attentive concentration. ‚óè More studies are needed to evaluate efficacy of alternative smoking cesThis state enables the individual to sation aids. adhere to suggestions and stratewith a more precise estimate the effigies to quit smoking.9 Randomcacy of those alternative aids. ized controlled trials investigating MATERIALS AND METHODS hypnotherapy used control treatments that varied from being Data Sources and Searches placed on a wait list to receiving a booklet. We systematically searched the Cochrane Library, EMBASE, Aversive smoking is a procedure defined as inhaling a Medline, and PsycINFO databases through December 2010 puff of cigarette every 6 seconds for 3 minutes, or until the for randomized controlled trials investigating alternative patient consumes 3 cigarettes, or until the patient is unable smoking cessation aids. Our search strategy combined the to smoke. After a short resting period, this procedure is terms cigarette, nicotine, smoking, and tobacco with the repeated 2 to 3 times per session to deter the patient from following key words: acupuncture, electrotherapy, electrosmoking.10 The control groups for randomized controlled acupuncture, laser therapy, hypnosis, hypnotherapy, avertrials investigating aversive smoking included smoking at a sive smoking, and rapid smoking. We limited our search to regular pace during a session or being placed on a wait list. the English and French languages and to randomized controlled trials carried out in adults. The references of randomized controlled trials, systematic reviews, and metaanalyses were hand-searched for studies not identified in the database search. Study Selection Eligible studies were randomized controlled trials; investigated one of the following alternative smoking cessation aids: acupuncture, hypnotherapy, and aversive smoking; reported long-term smoking cessation at 6 or 12 months; and investigated adults. Randomized controlled trials investigating more than 1 smoking cessation aid also were eligible for inclusion. Specifically, we included 2 types of randomized controlled trials that investigated more than 1 smoking cessation aid: factorial-designed randomized controlled trials and randomized controlled trials with multiple arms per intervention. For factorial-designed randomized controlled trials, we treated these as 2 separate randomized controlled trials and compared intervention arms such that Data Extraction Data from each randomized controlled trial were independently extracted by 2 reviewers. Disagreements were resolved by consensus or, when necessary, a third reviewer. Reviewers extracted information on study design, including the type of intervention, the type of randomization, the type of blinding, the therapists delivering the intervention, and the biochemical validation of data. Extracted baseline participant characteristics included mean number of cigarettes per day, mean Fagerstr√∂m Test for Nicotine Dependence scores, mean age, sex, and race. Treatment characteristics that were extracted included mean number of sessions, mean time of treatment, and type of control intervention. Smoking abstinence was defined as continuous abstinence or point prevalence of abstinence. Continuous abstinence consists of no smoking from the end of treatment to follow-up. Point prevalence of abstinence consists of no smoking for the 7 days preceding follow-up. We interpreted 578 the results of included randomized controlled trials using an intention-to-treat analysis. The variety of outcomes reported in the randomized controlled trials meant that, for any single outcome, we would have to exclude many studies from the metaanalysis. Therefore, reviewers extracted only the most rigorous criterion of smoking cessation reported for each randomized controlled trial.13 The most rigorous criterion uses the most conservative outcome reported in any given randomized controlled trial. Starting from the most rigorous outcome, the criteria are: continuous abstinence at 12 months; continuous abstinence at 6 months; point prevalence abstinence at 12 months; and point prevalence abstinence at 6 months. We calculated abstinence outcomes according to an intention-to-treat analysis. Patients who were randomized but lost to follow-up were considered smokers. Thus, the study can be applicable to real-life smokers who are trying to quit. Quality Assessment We used the Cochrane assessment tool to assess the quality of included randomized controlled trials.14 This assessment tool uses the following criteria to evaluate bias in a randomized controlled trial: sequence generation; allocation concealment; blinding of participants, personnel, and outcome assessors; incomplete outcome data; selective outcome reporting; and other potential threats to validity. Each randomized controlled trial was classified according to the following criteria: high quality, low quality, or unclear. Randomized controlled trials in which smoking cessation was not biochemically validated were automatically considered as unclear for the category of other potential threats to validity. The American Journal of Medicine, Vol 125, No 6, June 2012 were followed less than 6 months (n œ≠ 24), smoking cessation outcomes were not reported (n œ≠ 6), or studies did not use a proper control (n œ≠ 2). We identified an additional randomized controlled trial that met our inclusion criteria from our hand-search of previous studies. Thus, we included a total of 14 randomized controlled trials in our meta-analysis. Among these randomized controlled trials, 6 evaluated acupuncture (823 patients), 4 evaluated hypnotherapy (273 patients), and 4 evaluated aversive smoking (99 patients). Quality Assessment of Included Studies The qualitative risk of bias was relatively low for the 6 randomized controlled trials investigating acupuncture when compared with hypnotherapy and aversive smoking studies (Table 4). In acupuncture randomized controlled trials, the sequence generation, blinding of patients, and outcome data and reporting were generally of high quality. However, only 2 of the 6 acupuncture randomized controlled trials provided biochemical validation of self-reports of smoking cessation. The outcome data in the randomized controlled trial by Kerr et al16 was difficult to interpret. We contacted the authors, who provided the raw data for their study and confirmed the accuracy of their results. In contrast with acupuncture randomized controlled trials, the 4 randomized controlled trials each investigating hypnotherapy and aversive smoking generally provided few or no details of the sequence generation and allocation concealment of patients. Only 1 of the 4 hypnotherapy randomized controlled trials and 1 of the 4 aversive smoking randomized controlled trials reported biochemical validation of smoking status. Baseline and Treatment Characteristics Statistical Analysis We synthesized the results of included studies using the DerSimonian and Laird random effect model, taking into account both within randomized controlled trial and between trial variability. A forest plot was created for each outcome, for which we estimated pooled odds ratios (ORs) with their 95% confidence interval (CI). Heterogeneity was assessed using I2 statistics. In addition, funnel plots were constructed and visually assessed for the possible presence of publication bias. We conducted our analyses using MIX software version 1.3.15 RESULTS Search Results and Study Inclusion A total of 933 potentially relevant abstracts in our initial literature search were identified (Figure 1). Of these abstracts, 111 studies were retrieved and evaluated for eligibility. Thirteen randomized controlled trials met our inclusion criteria and were included in our meta-analysis (Tables 1-3). The 98 remaining studies were excluded because they did not use a randomized controlled trial design (n œ≠ 66), patients Baseline patient characteristics varied among randomized controlled trials (Tables 1-3). The mean age ranged from 30 to 54 years. The mean number of cigarettes per day varied from 16 to 32, and the mean Fagerstr√∂m Test for Nicotine Dependence score varied from 4 to 11. Certified acupuncturists were appointed to provide acupuncture for smoking cessation. For hypnotherapy, providers included psychologists and family physicians with training in hypnotherapy. For aversive smoking, providers included psychologists and clinical researchers with training in aversive smoking. Treatment characteristics also varied among randomized controlled trials. The mean number of sessions ranged from 1 to 20. The mean time of treatment ranged from 20 to 600 minutes for acupuncture, 80 to 480 minutes for hypnotherapy, and 150 to 600 minutes for aversive smoking. Efficacy of Alternative Smoking Cessation Aids A separate meta-analysis was carried out for each alternative smoking cessation aid for which smoking cessation was defined using the most rigorous criterion reported. The point Tahiri et al Alternative Smoking Cessation Aids 579 Records identified through database searching (n = 1226) Additional records identified through other sources (n = 5) Records after duplicates removed (n = 933) Records screened (n = 933) Records excluded (n =821): Reasons: Not relevant (n=695) Books (n=76) Review articles (n=30) No controls (n=20) Full-text articles assessed for eligibility (n = 112) Studies included in qualitative synthesis (n = 14) Full-text articles excluded (n=98): Reasons: No randomization (n=66) Follow up less than 6 months (n=24) Follow up not at 6 and/ or 12 months (n=7) Outcomes not reported (n=1) Studies included in quantitative synthesis (meta-analysis) (n = 14) Figure 1 Flow diagram of studies included in the systematic review and meta-analysis. estimate for acupuncture intervention (OR, 3.53; 95% CI, 1.03, 12.07) suggested that acupuncture substantially increased smoking cessation compared with sham acupunc- Table 1 ture (Figure 2). Hypnotherapy (OR, 4.55, 95% CI, 0.9821.01) might be efficacious at promoting smoking cessation (Figure 3). However, the CI was wide and included 1.0; Randomized Controlled Trials Investigating Acupuncture Treatment Characteristics Most Rigorous Outcome Reported Smoking Abstinence (%) Follow-up (mo) Abstinence Classification Active Study (First Author and Year)* Sample Size Country Male (%) Mean Age (y) Mean CPD Delivered by Kerr 200816 258 UK 56.5 ‚Äî ‚Äî 4 56 6 PP 55 4 Vandevenne 198524 Ta-Peng Wu 200725 Bier 200226 Waite 199827 Gillams 198428 200 France ‚Äî ‚Äî ‚Äî ‚Äî 120 12 CA 40 32 131 Taiwan 84.8 53.7 ‚Äî Trained researcher Trained physician Acupuncturist ‚Äî ‚Äî 6 PP 9 6 103 76 US US 49.6 55.2 46.4 42.5 27.2 ‚Äî Acupuncturist Acupuncturist 20 1 600 60 6 6 PP PP 13 3 2 5 55 US 51 37.5 24 ‚Äî ‚Äî ‚Äî 6 PP 18 15 Sessions (No.) Total Duration of Sessions (min) CA œ≠ continuous abstinence; CPD œ≠ cigarettes per day; MC œ≠ multicenter; PP œ≠ point prevalence; SC œ≠ single-center. *Studies ordered by sample size. Control 580 The American Journal of Medicine, Vol 125, No 6, June 2012 Table 2 Randomized Controlled Trials Investigating Hypnotherapy Treatment Characteristics Study* Lambe 198629 Williams 198830 Pederson 197931 Elkins 200632 Sample Size Country Male (%) Mean Age (y) Mean CPD 180 US 31.5 35.6 26.2 40 US 51.7 ‚Äî 33 Canada 43.1 20 US 30 Delivered by Sessions (No.) Total Duration of Sessions (min) Most Rigorous Outcome Reported Smoking Abstinence (%) Follow-up (mo) Abstinence Classification Active Control 2 80 12 PP 22 20 16 Trained family physician Hypnotherapist 1 150 12 CA 45 0 41.7 28.2 Hypnotherapist 1 ‚Äî 6 CA 53 18 42.7 23.95 Trained psychologist 8 480 6 PP 40 0 CA œ≠ continuous abstinence; CPD œ≠ cigarettes per day; PP œ≠ point prevalence. *Studies ordered by sample size. thus, we cannot draw strong conclusions about the efficacy of hypnotherapy. Aversive smoking (OR, 4.26, 95% CI, 1.26-14.38) substantially increased smoking abstinence compared with control (Figure 4). The wide CIs for the alternative smoking cessation aids prevented their ranking through indirect comparisons (data not shown). DISCUSSION Our study was designed to determine the efficacy of alternative smoking cessation aids (acupuncture, hypnotherapy, and aversive smoking) at promoting long-term smoking cessation. We included only the most rigorous randomized controlled trials, in which alternative smoking cessation aids were compared with an appropriate control and in which smoking cessation outcomes were reported at 6 or 12 months. Acupuncture, hypnotherapy, and aversive smoking were found to increase smoking abstinence by factors of 3.53, 4.26, and 4.55, respectively. On the surface, these results seem promising, especially when compared with Table 3 pharmacotherapies, which are known to increase smoking cessation by a factor of 2 to 2.5.17 However, these results should be interpreted with caution since the CI for each intervention was wide. There is a lack of randomized controlled trials investigating alternative smoking cessation aids, and the patient population included in our meta-analysis was small. Furthermore, in most randomized controlled trials included in our meta-analysis, reports of smoking cessation were not validated by biochemical means. We recommend that physicians encourage the use of alternative smoking cessation aids, particularly in patients hesitant or unable to use pharmacotherapies proven to be efficacious for smoking cessation.17 However, more evidence is needed to determine the precise level of efficacy of alternative smoking cessation aids. Given that up to 40% of individuals trying to quit smoking would consider using an alternative smoking cessation aid, it is surprising to note the lack of research in this area. This lack of research is particularly alarming when one considers that smokers may opt for alternative aids in place of pharmacotherapies. More research is needed to determine Randomized Controlled Trials Investigating Aversive Smoking Most Rigorous Outcome Reported Treatment Characteristics Sample Size Country Male (%) Mean Age (yrs) Mean CPD Delivered by Lando 197533 Barbarin 197834 Lichtenstein 197335 32 US 40 30.9 32 Trained researcher 30 US ‚Äî 40.9 ‚Äî ‚Äî 20 US 42.5 32.3 27 Erickson 198336 17 US 61.5 33 28 Trained psychology graduate students Trained psychology graduate students Study* Sessions (No.) Smoking Abstinence (%) Total Duration of Sessions (min) Months Abstinence Classification Active Control 6 270 12 PP 21 18 10 600 12 PP 40 0 7 177 6 PP 60 30 6 270 6 PP 70 14 CA œ≠ continuous abstinence; CPD œ≠ cigarettes per day; PP œ≠ point prevalence. *Studies ordered by sample size. Tahiri et al Table 4 Alternative Smoking Cessation Aids 581 Quality Assessment of Included Randomized Controlled Trials Acupuncture Quality Assessment* 1. Sequence generation 2. Allocation concealment 3. Blinding (of patients) 4. Incomplete outcome data (smoking cessation) 5. Selective outcome reporting 6. Other sources of bias‚Ä† Hypnotherapy Aversive Smoking Kerr 2008 Vandevenne 1985 Wu 2007 Bier 2002 Waite 1998 Gillams 1984 Lambe 1986 Williams 1988 Pederson 1979 Elkins 2006 Lando 1975 Barbarin 1978 Lichtenstein 1973 Erickson 1983 High Unclear High Unclear High Unclear High Unclear Low Unclear High Unclear High Unclear Unclear Unclear Unclear Unclear Unclear Unclear Unclear Unclear Unclear Unclear Unclear Unclear Unclear Unclear High High High High Unclear High N/A N/A N/A N/A N/A N/A N/A N/A High High High High High High High High High High High High High High High High High High High High High High High High High High High High Low Low High High Low Low Low Low Low High High Low Low Low N/A œ≠ not applicable. *For each criterion, the quality of each randomized controlled trial was classified as high, low, or unclear. ‚Ä†Other sources of bias include the absence of biochemical validation of self-reports of smoking cessation. the efficacy of alternative smoking cessation aids when used individually and when used as adjuncts to conventional pharmacologic interventions. The combination of these aids may be more efficacious than their use in isolation. ulation of acupuncture points is associated with a decrease of adrenocorticotropic hormone, a hormone that usually increases in response to stress. Studies also showed that stimulation of acupoints releases beta-endorphins and betaenkephalin, which are endogenous opiate peptides. By increasing these endogenous opiate peptides, acupuncture may have the capacity to diminish withdrawal symptoms.12 Potential mechanisms for acupuncture as a smoking cessation aid are still under investigation. The latest theory is that one of the points stimulated by needles in the ear corresponds to the closest position of the vagus nerve to the cutaneous surface. Therefore, by stimulating this point, the acupuncturist blocks the outflow of withdrawal symptoms coming from the parasympathetic nervous system through the vagus nerves. In the studies included in our meta-analysis, the procedure was similar from Acupuncture Acupuncture is rooted in traditional Chinese medicine, in which energy (called ‚Äúchi‚Äù) is perceived as the main component of the body and illness is seen as an imbalance of chi. The physical and psychologic symptoms experienced by smokers when they quit are seen as an imbalance of chi.12 Thus, to rebalance the chi, acupuncture points are stimulated by needles or low-level laser, allowing the symptoms experienced by the smoker trying to quit to diminish. Stimulation of acupuncture points may involve the endocrine and neurologic systems. Studies have shown that stim- Study Treatm ent n/N Year Control n/N OR(log scale) Weight(%) OR(95%CI) Kerr 2008 72/130 5/128 19.00% 30.54 (11.71 to 79.65) Vandevenne 1985 43/108 29/92 21.00% 1.44 (0.8 to 2.58) Ta-Peng Wu 2007 6/64 4/67 17.00% 1.63 (0.44 to 6.07) Bier 2002 6/45 3/58 17.00% 2.82 (0.66 to 11.97) Waite 1998 5/40 0/38 10.00% 11.93 (0.64 to 223.58) Gillams 1984 5/28 4/27 17.00% 1.25 (0.3 to 5.26) 137/415 45/410 100% 3.53 (1.03 to 12.07) TOTAL 0.1 Favors Sham 1 10 100 1000 Favors Treatm ent I2 = 85% [69, 93%] Figure 2 Forest plot comparing acupuncture to sham acupuncture. CI œ≠ confidence interval; OR œ≠ odds ratio. 582 The American Journal of Medicine, Vol 125, No 6, June 2012 Study Year Treatm ent n/N Control n/N Lambe 1986 20/90 Williams 1988 Pederson Elkins Weight(%) OR (95%CI) 18/90 39.00% 1.14 (0.56 to 2.34) 9/20 0/20 17.00% 33.87 (1.8 to 636.88) 1979 9/17 3/16 29.00% 4.88 (1.01 to 23.57) 2006 4/10 0/10 16.00% 14.54 (0.67 to 316.69) 42/137 21/136 100% 4.55 (0.98 to 21.01) TOTAL OR (log scale) 0.1 1 10 Favors Control 100 1000 Favors Treatm ent I2 = 67% [3, 89%] Figure 3 Forest plot comparing hypnotherapy to control. CI œ≠ confidence interval; OR œ≠ odds ratio. trial to trial and differed only in the number of sessions (Table 1). cedures were not exactly the same from trial to trial. More randomized controlled trials are needed to evaluate this intervention by following a common procedure led by personnel with training in hypnotherapy. Hypnotherapy Hypnotherapy consists of a technique of induction that brings the patient to a receptive mind and makes the patient more attentive and adherent to the suggestions of the therapist. Hypnotherapy is used in the medical field to treat anxiety and depressive disorders, and also may be efficacious in promoting smoking cessation. When the patient is hypnotized, the therapist suggests that cigarettes are harmful, associated with terrible sensations, and that the patient is able to cope with the withdrawal symptoms.18 Given that the patient is hypnotized, he or she is hypothetically more adherent to those suggestions and thus can better cope with withdrawal symptoms. In our meta-analysis, we noted that the hypnotherapy pro- Study Year Treatm ent n/N Control n/N Lando 1975 3/14 Barbarin 1978 Lichtenstein 1973 Erickson 1983 TOTAL: Aversive Smoking Aversive smoking uses excessive smoke as an aversive stimulus. Aversive smoking exposes the patient to an important amount of smoke in a short period of time. This exposure may make the patient become resistant to the additional intake of cigarettes in the future.19 During the last decade, aversive smoking has been less practiced because of beliefs that it may induce nicotine poisoning and cardiac arrest. However, many studies showed the safety of aversive smoking with close monitoring of heart rate and blood pressure.20 OR (log scale) Weight(%) OR(95%CI) 3/17 34.00% 1.27 (0.21 to 7.58) 6/15 0/15 15.00% 21.21 (1.07 to 420.8) 6/10 3/10 32.00% 3.5 (0.55 to 22.3) 7/10 1/7 20.00% 14 (1.14 to 172.64) 22/49 7/49 100% 4.26 (1.26 to 14.38) 0.1 Favors Control 1 10 100 1000 Favors Treatm ent I2 = 21% [0, 88%] Figure 4 Forest plot comparing aversive smoking to control. CI œ≠ confidence interval; OR œ≠ odds ratio. Tahiri et al Alternative Smoking Cessation Aids Previous Studies Separate reviews on acupuncture, hypnotherapy, and aversive smoking have been conducted by the Cochrane Collaboration.21-23 We carried out a systematic review and meta-analysis including all 3 alternative interventions and attempted to rank them through indirect comparisons. Our meta-analysis was designed to include only the most rigorous randomized controlled trials to obtain the most precise estimates of smoking cessation. Specifically, we only included randomized controlled trials with long-term smoking cessation outcomes, appropriate control groups, and limited bias. In addition, we used a random effect model to account for the heterogeneity between individual randomized controlled trials. We provide a comprehensive review of the most commonly used alternative smoking cessation interventions in one article, thus providing evidence-based guidelines to physicians. Our research suggests that alternative smoking cessation aids may be efficacious at promoting smoking cessation. Unlike previous Cochrane reviews, the current review found a statistically significant effect of acupuncture, albeit with a wide confidence interval. In addition, the effect of hypnotherapy would likely have been significant if more patients had been randomized, thus increasing the power of our meta-analysis. Limitations Our meta-analysis has several limitations. First, because of our strict inclusion/exclusion criteria, the number of studies included in our meta-analysis was limited, as testified by the wide CIs. However, by including only the most rigorous randomized controlled trials, we were able to obtain the most reliable estimates of the efficacy of alternative smoking cessation aids. Second, randomized controlled trials varied in the total duration of each intervention, the mean cigarettes per day, the Fagerstr√∂m Test for Nicotine Dependence, and age. We used a random effect model to account for between-trial heterogeneity. Third, publication bias also is a limitation as it is for virtually any meta-analysis. We did not have enough data to interpret publication bias using our funnel plots (data not shown). Finally, we limited our search to randomized controlled trials published in English and French. However, less than 2% of randomized controlled trials identified in our literature search were published in a language other than English or French. CONCLUSIONS Acupuncture and hypnotherapy are used by a large number of smokers as alternative smoking cessation aids. Our results suggest that these alternative aids may help smokers quit. Thus, we recommend that physicians promote the use of acupuncture and hypnotherapy. Aversive smoking also may help people quit, but because the studies investigating this intervention were old, we believe that new studies are needed to recommend this intervention to physicians. Finally, we believe that more evidence is needed to determine whether alternative 583 interventions are as efficacious or perhaps more efficacious than pharmacotherapies at helping smokers quit. ACKNOWLEDGMENTS The authors thank Jennifer Reoch for help with data extraction. References 1. Chandler MA, Rennard SI. Smoking cessation. Chest. 2010;137:428435. 2. CDC grand rounds: current opportunities in tobacco control. MMWR Morb Mortal Wkly Rep. 2010;59:487-492. 3. Rehm J, Gnam W, Popova S, et al. The costs of alcohol, illegal drugs, and tobacco in Canada, 2002. J Stud Alcohol Drugs. 2007;68:886-895. 4. Centers for Disease Control and Prevention. Tobacco Control State Highlights, 2010. Atlanta: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, National Center for Chronic Disease Prevention and Health Promotion, Office on Smoking and Health, 2010. 5. Health Canada. Canadian Tobacco Use Monitoring Survey, Annual Results. Ottawa: Health Canada, Office of Research, Evaluation and Surveillance, 2009. 6. Cigarette smoking among adults and trends in smoking cessation‚Äî United States, 2008. MMWR Morb Mortal Wkly Rep. 2009;58:12271232. 7. Health Canada. Canadian Tobacco Use Monitoring Survey, Annual Results. Ottawa: Health Canada, Office of Research, Evaluation and Surveillance, 2008. 8. McLellan AT, Grossman DS, Blaine JD, Haverkos HW. Acupuncture treatment for drug abuse: a technical review. J Subst Abuse Treat. 1993;10:569-576. 9. Spiegel D, Frischholz EJ, Fleiss JL, Spiegel H. Predictors of smoking abstinence following a single-session restructuring intervention with self-hypnosis. Am J Psychiatry. 1993;150:1090-1097. 10. Lichtenstein E, Glasgow RE. Rapid smoking: side effects and safeguards. J Consult Clin Psychol. 1977;45:815-821. 11. Sood A, Ebbert J, Sood R, Stevens S. Complementary treatments for tobacco cessation: a survey. Nicotine Tob Res. 2006;8:767-771. 12. Moner S. Acupuncture and addiction treatment. J Addict Dis. 1996; 15:79-100. 13. Silagy C, Lancaster T, Stead L, Mant D, Fowler G. Nicotine replacement therapy for smoking cessation. Cochrane Database Syst Rev. 2004(3):CD000146. 14. Higgins J, Green S, Cochrane Collaboration. Cochrane Handbook for Systematic Reviews of Interventions. Chichester, England; Hoboken, NJ: Wiley-Blackwell; 2008. 15. Bax L, Yu LM, Ikeda N, Tsuruta H, Moons KG. Development and validation of MIX: comprehensive free software for meta-analysis of causal research data. BMC Med Res Methodol. 2006;6:50. 16. Kerr C, Lowe P, Spielholz N. Low level laser for the stimulation of acupoints for smoking cessation: a double blind, placebo controlled randomised trial and semi structured interviews. J Chin Med. 2008; 86:46-51. 17. Eisenberg MJ, Filion KB, Yavin D, et al. Pharmacotherapies for smoking cessation: a meta-analysis of randomized controlled trials. CMAJ. 2008;179:135-144. 18. Covino NA, Bottari M. Hypnosis, behavioral theory, and smoking cessation. J Dent Educ. 2001;65:340-347. 19. Danaher BG. Research on rapid smoking: Interim summary and recommendations. Addict Behav. 1977;2:151-166. 20. Horan JJ, Linberg SE, Hackett G. Nicotine poisoning and rapid smoking. J Consult Clin Psychol. 1977;45:344-347. 21. Hajek P, Stead LF. Aversive smoking for smoking cessation. Cochrane Database Syst Rev. 2004(3):CD000546. 584 22. White AR, Rampes H, Liu JP, Stead LF, Campbell J. Acupuncture and related interventions for smoking cessation. Cochrane Database Syst Rev. 2011;1:CD000009. 23. Barnes J, Dong CY, McRobbie H, Walker N, Mehta M, Stead LF. Hypnotherapy for smoking cessation. Cochrane Database Syst Rev. 2010(10):CD001008. 24. Vandevenne A, Rempp M, Burghard G. Study of the specific contribution of acupuncture to tobacco detoxication. Semaine des Hopitaux. 1985;61:2155-2160. 25. Wu T, Chen F, Liu J, Lin M, Hwang S. A randomized controlled clinical trial of auricular acupuncture in smoking cessation. J Chinese Med Assoc. 2007;70:331-338. 26. Bier I, Wilson J, Studt P, Lin M, Shakleton M. Auricular acupuncture, education, and smoking cessation: A randomized, sham-controlled trial. Am J Public Health. 2002;92:1642-1647. 27. Waite N, Clough J. A single-blind, placebo-controlled trial of a simple acupuncture treatment in the cessation of smoking. Br J Gen Pract. 1998;48:1487-1490. 28. Gillams J, Lewith G, Machin D. Acupuncture and group therapy in stopping smoking. Practitioner. 1989;228:341-344. 29. Lambe R, Osier C, Franks P. A randomized controlled trial of hypnotherapy for smoking cessation. J Fam Pract. 1986;22:61-65. The American Journal of Medicine, Vol 125, No 6, June 2012 30. Williams JM, Hall DW. Use of single session hypnosis for smoking cessation. Addict Behav. 1988;13:205-208. 31. Pederson LL, Scrimgeour WG, Lefcoe NM. Comparison of hypnosis plus counseling, counseling alone, and hypnosis alone in a community service smoking withdrawal program. J Consult Clin Psychol. 1975; 43:920. 32. Elkins G, Marcus J, Bates J, Hasan Rajab M, Cook T. Intensive hypnotherapy for smoking cessation: A prospective study 1. Int J Clin Exp Hypn. 2006;54:303-315. 33. Lando HA. A comparison of excessive and rapid smoking in the modification of chronic smoking behavior. J Consult Clin Psychol. 1975;43:350-355. 34. Barbarin OA. Comparison of symbolic and overt aversion in the self-control of smoking. J Consult Clin Psychol. 1978;46:1569-1571. 35. Lichtenstein E, Harris DE, Birchler GR, Wahl JM, Schmahl DP. Comparison of rapid smoking, warm, smoky air, and attention placebo in the modification of smoking behavior. J Consult Clin Psychol. 1973;40:92-98. 36. Erickson L, Tiffany ST, Martin E, Baker TB. Aversive smoking therapies: A conditioning analysis of therapeutic effectiveness. Behav Res Ther. 1983;21:595-611. ˇ˛a Department of Internal Medicine I, Friedrich Schiller University Jena, Erlanger Allee 101, D-07747 Jena, Germany b c European Journal of Internal Medicine 12 (2001) 366 371 Original article www.elsevier.com / locate / ejim Ambulatory electrogastrography in patients with sclerodermia, delayed gastric emptying, dyspepsia, and irritable bowel syndrome Is there any clinical relevance? Michael Hockea,*, Thomas Seidela, Haiko Sprottc,d, Peter Oelznerb, Klaus Eitnera, a Hans Bosseckert Department of Internal Medicine IV, Friedrich Schiller University Jena, D-07747 Jena, Germany Department of Internal Medicine IV, Friedrich Schiller University Jena, D-07747 Jena, Germany Center of Experimental Rheumatology and WHO Collaborating Center for Molecular Biology and Novel Therapeutic Strategies for Rheumatic d Abstract Diseases, University Hospital Zuerich, CH-8091 Zuerich, Switzerland Received 22 June 2000; received in revised form 7 December 2000; accepted 19 December 2000 Background: Changes in electrogastrographic parameters are described in patients with irritable bowel syndrome, sclerodermia, dyspepsia, and delayed gastric emptying in static measurements. However, no information is available about changes in ambulatory measurements. The objective of this study was to find parameters that discriminate between these diseases using cutaneous 24-h- electrogastrography. Methods: Cutaneous 24-h electrogastrography (EGG) measurements were taken from 20 patients with dyspepsia, 10 patients with systemic sclerosis (sclerodermia, SSc), 7 patients with irritable bowel syndrome (IBS), 7 patients with delayed gastric emptying, and 10 healthy volunteers. Measurements were made using a DIGITRAPPER EGGE (Synectics Medical Inc., Stockholm, Sweden) and the accompanying computerized data analysis package (ElectroGastroGram Version 6.30, Gastrosoft Inc., Synectics Medical Inc., Stockholm, Sweden). Frequency and power were compared pre- and postprandially, as well as during the entire day of measurement. Results: The 24-h measurements in healthy volunteers revealed 45.00%612.12% normal values (2.4 3.7 cpm), 30.10%67.15% bradygastric values (,2.4 cpm), and 24.20%67.76% tachygastric values (.3.7 cpm). There was no significant change in frequency between rest and motion, but there was a significant increase in power (P,0.05). There was significantly more bradygastria in patients with dyspepsia periprandially as well as after 24 h (P,0.01) than in healthy volunteers. The mean power of patients with dyspepsia was significantly higher than that of patients with IBS (P,0.05). Conclusion: Cutaneous 24-h-EGG may be used as an additional means of differentiating between dyspepsia and IBS. Ÿˆ 2001 Elsevier Science B.V. All rights reserved. Keywords: Electrogastrography; Irritable bowel syndrome; Sclerodermia; Dysmotility; Dyspepsia 1. Introduction Electrogastrography (EGG) refers to the measurement of electrical activity of the stomach by placing electrodes on *Corresponding author. Tel.: 149-3641-939-123; fax: 149-3641-939- 404. E-mail addresses: hocke@polkim.med.uni-jena.de (M. Hocke), hocke@polkim.med.uni-jena.de (M. Hocke). the surface of the abdomen [1]. Since its first application in the 1920s [2], several studies have shown an association between electric abnormalities and gastrointestinal motility disorders [3]. Unfortunately, there is a very low signal-to- noise ratio of the EGG [4]. Only the gastric slow waves are available for measurement, but the relationship between gastric slow waves and motility seems to be unclear [5]. Simultaneous recordings of cutaneous and serosal [6] or cutaneous and mucosal [7] gastric myoelectrical activity 0953-6205/01/$ see front matter Ÿˆ 2001 Elsevier Science B.V. All rights reserved. PII: S0953-6205(01)00138-8 have confirmed that the predominant frequency of the cutaneous EGG is the same as that of gastric slow waves. At present, only static measurements of the electrical activity of the stomach are used, due to the great influence of motion artifacts. Yet, in Europe, a commercial, ambulat- ory EGG has been available for clinical practice for a few years [8]. The interpretation of acquired data requires a comprehensive, computerized data analysis system (Elec- troGastroGram Version 6.30, Gastrosoft Inc., Synectics Medical Inc., Stockholm, Sweden). The aim of this study was to determine whether ambulatory EGG provides useful clinical data for the diagnosis of gastric motility diseases. 2. Patients and methods The study was performed on 20 inpatients with dyspep- sia (6 male and 14 female; age 41616 years), 10 inpatients with systemic sclerosis [sclerodermia (SSc); all female; age 46613 years], 7 inpatients with irritable bowel syndrome (IBS; 2 male and 5 female; age 42612 years), 7 inpatients with delayed gastric emptying (4 male and 3 female; age 57.0626.0 years), and 10 healthy volunteers (5 male and 5 female; age 33.069.0 years). The healthy volunteers had no symptoms of gastrointestinal diseases. A radiopaque marker examination was performed in cases of delayed gastric emptying. The patients with SSc had been diagnosed according to previously published criteria [9]. All patients had dyspeptic symptoms. The presence of gastrointestinal symptoms, such as retrosternal pain, epi- gastric pain, epigastric fullness, nausea or vomiting with- out any morphological correlate, were needed for inclusion in the dyspeptic group. Given the small number of patients, no subgroups were made. The diagnosis of IBS from motility type was based on the presence of abdominal pain, abdominal distension, and an abnormal bowel habit with normal laboratory investigations. All patients underwent a gastroscopy and abdominal ultrasound to exclude peptic ulcers, erosions, malignancies, etc. In addition, all patients with IBS underwent a colonoscopy to exclude morphologi- cal causes. None of the subjects had taken any medication with a known effect on gastrointestinal motility in the 2 days before or during the study. The DIGITRAPPER EGGE (Synectics Medical Inc., Stockholm, Sweden) and the accompanying computerized data analysis package (ElectroGastroGram Version 6.30, Gastrosoft Inc., Synec- tics Medical Inc., Stockholm, Sweden) were used to make the measurements. Recording sites on the skin were cleaned with Softasept⁄ˆN (Braun Melsungen AG) alcohol solution. One of the active electrodes was placed in the middle of a line between the sternum and the umbilicus (i.e. the pylorus) and the other was placed 2 cm below the left bottom rib in the medioclavicular line (the antrum). The reference electrode was placed inferiorly, forming an isolateral triangle. The recording time was 24 h. All subjects ate three meals during the course of the study period and rested for 1 h preprandially and for 30 min postprandially. Motions and speaking were not allowed during the rest periods. The eating and sleeping times were marked by subjects by pressing an event button on the DIGITRAPPER and also on an event protocol. The reliability of every measurement was checked with the help of visual analysis on the computer screen. The following parameters were analyzed: (a) the percentages of dominant frequency in the defined normogastric range (2.4 3.7 cpm), in bradygastria (,2.4 cpm), and in tachy- gastria (.3.7 cpm) and (b) the predominant power of measurement. Dominant frequency readings higher than 10 cpm were distinguished from tachygastria because they were as- sumed to arise from outside the stomach. The preprandial, postprandial (after all three meals), and all-day, 24-h parameters were compared. All recordings were made at sampling frequencies of 1 Hz. The data were obtained by running a spectrum analysis. With this technique [using by a fast-Fourier transform (FFT) algorithm], power spectra of overlapping stretches of the electrical signal are com- puted and displayed as a function of time, thus yielding frequency and amplitude information over the course of time [10]. The dominant frequency is calculated as the highest peak of the mean FFT line for a given period. In 12 unselected cases (healthy volunteers), the percentage of dominant frequency and the power at preprandial quiet time (30 min) with preprandial motion time (30 min) were analyzed for detection of motion effects on measurement. Statistical analysis was carried out with Student s t-test or, in the case of significant differences between standard deviations, with an unpaired t-test (Welch). Data are given as mean6S.D. M. Hocke et al. / European Journal of Internal Medicine 12 (2001) 366 371 367 3. Results 3.1. 24 -h EGG in healthy volunteers The measurements in healthy volunteers showed 45.00%612.12% normal values (2.4 3.7 cpm), 30.10%67.15% bradygastric values, and 24.20%67.76% tachygastric values. The all-day predominant power was 1491.506750.59 dB. As shown in Table 1, there was no statistically significant difference between the pre- and postprandial dominant frequencies and the predominant power in healthy volunteers. 3.2. Influence of motion artifacts on 24-h EGG We observed an affect of motion on the predominant frequency in tachygastria (rest 23.16616.91, motion 18.87612.36), in normogastria (rest 46.91626.86, motion 44.45628.98), and in bradygastria (rest 28.91620.72, motion 35.58622.18). However, the differences were not statistically significant (n512). In contrast, the predomi- 368 M. Hocke et al. / European Journal of Internal Medicine 12 (2001) 366 371 Table 1 Preprandial, postprandial, and 24-h values of dominant frequency in bradygastria, normal range, tachygastria, and predominant power in healthy volunteers Preprandial Postprandial 24-h Bradygastria (%) 28.20617.79 30.93618.22 30.1067.15 Normal range (%) 44.97623.40 41.53626.49 45.00612.12 Tachygastria (%) 26.03612.54 26.57617.16 24.2067.76 Power (dB) 2155.4761705.53 2322.7361612.05 1491.506750.59 Fig. 1. Comparison of predominant frequency and predominant power at rest and in motion (n512) in healthy volunteers. nant power was highly significant, as shown in Fig. 1 (rest 1002.836774.24, motion 2760.0862359.27; P,0.05 Welch). 3.3. Findings in patients with delayed gastric emptying and SSc There was only a significant reduction in tachygastria in patients with SSc and delayed gastric emptying in com- parison with healthy volunteers (P,0.05). In patients with SSc, this reduction was even more pronounced in the preprandial phase. No difference in predominant power could be analyzed. There seemed to be a shift from tachygastria to bradygastria in patients with delayed gastric emptying, but a significant result could not be achieved. The results are shown in Tables 2 and 3. Table 2 Preprandial, postprandial, and 24-h values of dominant frequency of bradygastria, normal range, tachygastria, and predominant power in patients with delayed gastric emptying (n57) Preprandial Postprandial 24-h Bradygastria (%) 41.2624.5 38.9623.6 37.0614.2 Normal range (%) 37.8623.2 42.1630.2 42.3622.9 Tachygastria Power (dB) (%) 15.9619.2 2112.061734.6 12.9612.1 2360.962239.1 15.867.7 P,0.05* 1706.761116.8 * Significant reduction in comparison with healthy volunteers. Table 3 Preprandial, postprandial, and 24-h values of dominant frequency in bradygastria, normal range, tachygastria, and predominant power in patients with SSc (n510) Preprandial Postprandial 24-h Bradygastria Normal range (%) Tachygastria (%) Power (dB) (%) 38.3626.3 46.7626.9 14.1612.3 P,0.05* 2444.662234.6 38.3626.0 47.0624.4 13.5611.7 2599.161993.1 32.2615.5 51.2618.1 15.668.6 P,0.05* 1450.56895.7 * Significant reduction in comparison with healthy volunteers. Table 4 Preprandial, postprandial, and 24-h values of dominant frequency dyspepsia (n520) in bradygastria, normal range, tachygastria, and predominant power in patients with M. Hocke et al. / European Journal of Internal Medicine 12 (2001) 366 371 369 Preprandial Postprandial 24-h Bradygastria (%) 48.9626.2 P,0.05* 48.6625.3 43.9614.7 P,0.01* Normal range (%) 32.5625.4 39.8653.6 35.95616.8 Tachygastria (%) 16.7616.2 15.4613.8 16.568.4 P,0.05* Power (dB) 2982.162530.3 2970.162189.8 2155.561112.1 * Significant reduction in comparison with healthy volunteers. Table 5 Preprandial, postprandial, and 24-h values of dominant frequency in bradygastria, normal range, tachygastria, and predominant power in patients with IBS (n57) Bradygastria (%) 35.1627.3 26.1623.5 32.3617.8 Normal range Tachygastria (%) (%) Power (dB) 1913.162232.8 1453.661759.5 1113.36758.9 Preprandial Postprandial 24-h 58.6628.1 65.7623.3 53.9615.2 5.166.5 P,0.01* 7.167.1 P,0.05* 10.766.5 P,0.01* * Significant reduction in comparison with healthy volunteers predominant power in dB. 3.4. Findings in patients with dyspepsia and IBS There was a significant reduction in tachygastria over the 24-h period in both diseases in comparison with normal volunteers. This reduction was very marked in patients with IBS in the 24-h measurement (P,0.01), as well as pre- (P,0.01) and postprandially (P,0.05). A significant increase in bradygastria in patients with dyspepsia all day (P,0.01) and preprandially (P,0.05) was important in comparison with healthy volunteers (Tables 4 and 5). There was a significant shift between 24-h predominant frequencies and power in patients with IBS and dyspepsia. Thus, patients with IBS have a more pronounced shift to normal values than patients suffering from dyspepsia (P,0.05). Additionally, there was a significant reduction in predominant power in patients with IBS (P,0.05). These changes were only significant between these two diseases and not in comparison with healthy volunteers or patients with other diseases. These results are shown in Fig. 2. 4. Discussion Pathologic findings in EGG have been associated with various gastrointestinal diseases. Frequency abnormalities or disturbances of power have been reported in patients with different diseases, such as diabetic gastroparesis [11,12], gastric cancer [13], anorexia nervosa [14], and hyperthyroidism [15]. This non-invasive and well-accepted method has failed to become widely established. The most important reason seems to be the high rate of motion artifacts in ambulatory measurement and their assumed influence on results [16]. Thus far, no data are available for Fig. 2. Comparison of predominant frequency and predominant power in patients with IBS (n57) and in patients with dyspepsia (n520). 370 M. Hocke et al. / European Journal of Internal Medicine 12 (2001) 366 371 the application of ambulatory EGG under clinical con- ditions. Two problems had to be solved in order to find a compromise between reliable results and practicable con- ditions. First, it was necessary to define a motion protocol that included rest times around the meals for a better evaluation of results. This protocol was designed as previously proposed by Chen et al. [8] Second, an analysis of the affects of motion on the frequency and power had to be given. A significant increase in power from predomi- nant frequency was determined in the present study but not a significant shift in predominant frequency. This result was the argument for analysis of all data without elimina- tion of motion artifacts, as recommended by Koch and Stern [17]. In fact, there was a relatively high percentage of bradygastria and tachygastria in healthy volunteers in this study. These results differ from data supplied by Pfaffen- bach et al. [18], who described a normal range of 88%. However, it must be stressed that a sonographic placement of electrodes is useless in this study because of multiple changes in the positions of the subjects. Under these conditions, the previously described rates quoted by Pfaf- fenbach et al. [18] of 17.8% bradygastria, 11.3% tach- ygastria, and 69.4% normogastria could almost be con- firmed in 24-h EGG and was sufficient for the detection of differences between healthy volunteers and patients. It should also be taken into account that the normal range used by Pfaffenbach et al. was 2 4 cpm, whereas this study used a normal range of 2.4 3.7 cpm. Thus, the blind placement of electrodes seems to be adequate for a 24-h EGG, but in a stationary EGG an optimal position of electrodes is necessary to recognize fine differences, as described by Abell et al. [3]. In this study, there was a significant difference in age between healthy volunteers and patients; however, no age- or sex-dependent change in gastric electric frequency has been reported [19]. No significant changes in the predominant frequency or power were detected periprandially in the normal volunteers. However, there was a slight shift in frequency from normal range to bradygastria and an increase in postpran- dial power. These changes are in agreement with those described by other authors who have associated bradygas- tria with antral contractions [20] and a postprandial increase in power with gastric mechanical activity [21] or approach of the gastric antrum to the abdominal surface [22]. A significant decrease in tachygastria was detected in all patients. What these phenomena imply is currently unknown. Data from stationary EGG defines tachygastria as an absence of gastric contractions [23 25]. In this case, an increase in tachygastria would be expected, especially in patients with delayed gastric emptying, as described by other authors [26]. This is in contrast to the present data. However, another phenomenon was detected in patients with SSc. There was an increase in the predominant frequency in the normal range as compared to that in healthy volunteers; however, this was not statistically significant. In addition, there were almost no changes in frequencies or power periprandially. These data confirm the findings of Pfaffenbach et al. [27]. The stable behavior of frequency and power can be interpreted as an expression of neural alteration that may have diminished modulating influences on motor activity in this case [28]. The most important finding in this study was a significant increase in the rate of bradygastria in patients with dyspepsia. To- gether with the previously described increased antral motility in bradygastria [20], this could be an indication for motility disturbances in patients with dyspepsia. In this regard, findings by Troncon et al. [29] concerning the abnormal intragastral distribution of food, with an em- phasis on the antral area in dyspeptics, acquire a new importance. The assumed hypothesis concerning antral motility disturbances is also supported by the sonographic findings of Bolondi et al. [30], who described an increase in postprandial stretching of the antrum, and findings by Holtmann et al. [31], who described a sensory defect in intragastric balloon distention. Surprisingly, patients with IBS had almost normal frequency and power in com- parison with healthy volunteers. This result, however, is especially important with regard to dyspeptic patients. It can be demonstrated that there is a significant reduction in predominant power and an increase in normal range frequencies in patients with IBS. In light of the postulated significance of the distance of electrodes from the gastric surface [32], there may be an association with the often flatulent colonic condition of patients with IBS. However, the differences between IBS and dyspeptic patients are important. It is difficult to discriminate between these two diseases entities [33] in clinical practice. The present results support the introduction of 24-h EGG in clinical practice. The commercial, ambulatory EGG was able to detect differences between healthy volunteers and patients, providing a new, useful tool for the discrimination of functional diseases. This conclusion contrasts with that of Bortolotti [34] who, at the moment, is unable to find a clinical role for cutaneous EGG. Further investigations are necessary to confirm these preliminary data. 5. Notation EGG, Electrogastrography; cpm, counts per minute; Hz, Herz; FFT, fast-Fourier transformation; IBS, irritable bowel syndrome; S.D., standard deviation CLINICAL RESEARCH STUDY SAMe-TT2R2 Score, Time in Therapeutic Range, and Outcomes in Anticoagulated Patients with Atrial Fibrillation Pilar Gallego, MD, PhD,a,b Vanessa Rold√°n, MD, PhD,b Francisco Marin, MD, PhD,c Jos√© G√°lvez, (Student),b Mariano Vald√©s, MD, PhD,c Vicente Vicente, MD, PhD,b Gregory Y.H. Lip, MDa a University of Birmingham Centre for Cardiovascular Sciences, City Hospital, Birmingham, UK; bDepartment of Hematology and Clinical Oncology, Hospital Universitario Morales Meseguer, University of Murcia, Spain; cDepartment of Cardiology, Hospital Universitario Virgen de la Arrixaca, University of Murcia, Spain. ABSTRACT BACKGROUND: Oral anticoagulation is highly effective in preventing stroke and mortality in nonvalvular atrial Ô¨Åbrillation patients. However, the efÔ¨Åcacy and safety of vitamin K antagonists (the main oral anticoagulation drug used) strongly depends upon the quantity of anticoagulation control, as reÔ¨Çected by the average percentage of the time in therapeutic range of international normalized ratio 2.0-3.0. An easy, simple prediction of which atrial Ô¨Åbrillation patients are likely to do well on vitamin K antagonists (with good average time in therapeutic range) could guide decision-making between using vitamin K antagonists (eg, warfarin) and non-vitamin K antagonist oral anticoagulants. METHODS AND RESULTS: In a consecutive cohort of nonvalvular atrial Ô¨Åbrillation patients attending our anticoagulation clinic, we tested the hypothesis that the new Sex, Race, Medical history, Tobacco use, Race (SAMe-TT2R2) score was a predictor for good average time in therapeutic range, and second, this would translate into adverse events in a ‚Äúreal world‚Äù cohort of patients with nonvalvular atrial Ô¨Åbrillation. The incidence of bleeding, adverse cardiovascular events (including stroke/thromboembolism), and mortality during the follow-up was higher with increasing SAMe-TT2R2 score. The SAMe-TT2R2 score was predictive for the composite of all adverse events (hazard ratio 1.32 [95% ConÔ¨Ådence Interval 1.17-1.50]; P <.001), adverse cardiovascular events (1.52 [1.28-1.83]; P <.001), and all-cause mortality (1.41 [1.161.67]; P ¬º .001). A trend was also observed for major bleeding events (1.23 [0.99-1.53]; P ¬º .059). CONCLUSION: In a ‚Äúreal world‚Äù cohort of consecutive patients with nonvalvular atrial Ô¨Åbrillation, a high SAMe-TT2R2 score (reÔ¨Çecting poor anticoagulation control with poor time in therapeutic range) was associated with more bleeding, adverse cardiovascular events, and mortality during follow-up. √ì 2014 Elsevier Inc. All rights reserved. The American Journal of Medicine (2014) 127, 1083-1088 KEYWORDS: Anticoagulation; Atrial Ô¨Åbrillation; Bleeding; Mortality; SAMe-TT2R2 score; Stroke Oral anticoagulation is highly effective in preventing stroke and mortality in nonvalvular atrial Ô¨Åbrillation patients.1 However, the efÔ¨Åcacy and safety of vitamin K antagonists depends upon the quality of anticoagulant control, as reÔ¨Çected by the average percentage of the time in therapeutic range of international normalized ratio 2.0-3.0. Various Funding: This work was supported by PI11/1256 from ISCIII and FEDER. ConÔ¨Çicts of Interest: All authors report none speciÔ¨Åcally relevant to this manuscript. PG holds a grant from the Spanish Foundation Alfonso Mart√≠n Escudero. VR has received funding for consultancy and lecturing from Bayer, Bristol-Myers Squibb and Boehringer Ingelheim. FM has received funding for research, consultancy, and lecturing from Boston ScientiÔ¨Åcs, Bayer, AstraZeneca, Daiichi-Sankyo, and Boehringer Ingelheim. GYHL has received funding for research, consultancy, and lecturing from different manufacturers of drugs used for the treatment of atrial Ô¨Åbrillation, including AstraZeneca, Bayer, Boehringer Ingelheim, Astellas, SanoÔ¨Å-Aventis and Daiichi-Sankyo. Authorship: All authors had a role in drafting and writing the manuscript. Requests for reprints should be addressed to Gregory Y. H. Lip, MD, University of Birmingham Centre for Cardiovascular Sciences, City Hospital, Birmingham B18 7QH, UK. E-mail address: g.y.h.lip@bham.ac.uk 0002-9343/$ -see front matter √ì 2014 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.amjmed.2014.05.023 1084 The American Journal of Medicine, Vol 127, No 11, November 2014 studies have shown how a high time in therapeutic range This simple score (SAMe-TT2R2) could help decisiontranslates into a lower risk of stroke and bleeding while making by identifying those atrial Ô¨Åbrillation patients that on oral anticoagulation.2-4 A recent European consensus would do well on vitamin K antagonists (with a high average time in therapeutic range, with SAMe-TT2R2 document recommends that an average individual time in therapeutic range should be >70% for optimal efÔ¨Åcacy and score ¬º 0-1), or conversely, those likely to have antisafety outcomes whilst on a vitamin K antagonist, and this is coagulation control (ie, poor average time in therapeutic also recommended in international range, with SAMe-TT2R2 score guidelines.5 The vitamin K antago!2) who require additional inCLINICAL SIGNIFICANCE terventions to achieve acceptable nists were the main oral anticoaguanticoagulation control or be canlant drug used until the introduction In a ‚Äúreal world‚Äù cohort of consecutive didates for NOACs. This score of the non-vitamin K antagonist patients with nonvalvular atrial Ô¨Åbrillawas derived from the Atrial Fioral anticoagulants (NOACs, pretion, a high SAMe-TT2R2 score (reÔ¨Çecting brillation Follow-up Investigation viously referred to as novel or new poor anticoagulation control with poor of Rhythm Management (AFFIRM) oral anticoagulants). time-in-therapeutic range [TTR]) was trial population and externally valiThe difÔ¨Åculties in achieving a associated with more bleeding, adverse dated in a small ‚Äúreal world‚Äù cohort high time in therapeutic range, as cardiovascular events, and mortality of anticoagulated nonvalvular well as the inconvenience of regduring follow-up. atrial Ô¨Åbrillation patients.7 Further ular anticoagulation monitoring and the various food/drug revalidations in large ‚Äúreal world‚Äù The SAMe-TT2R2 score, an easy, simple strictions associated with the cohorts are necessary, also to assess prediction of which atrial Ô¨Åbrillation vitamin K antagonists, have led to whether this score (which reÔ¨Çects patients are likely to do well on vitamin K the recent introduction of the time in therapeutic range) translates antagonists (VKA) (with good average NOACs, which offer improved into adverse outcomes in antiTTR), could guide decision-making beefÔ¨Åcacy, safety, and convenience coagulated nonvalvular atrial Ô¨Åbriltween using VKAs and non-VKA oral compared with the vitamin K anlation patients. anticoagulants. tagonists.5 Even in clinical trials, In a consecutive cohort of nonvalvular atrial Ô¨Åbrillation papatients in centers with high tients attending our anticoagulation average time in therapeutic range clinic, we tested the hypothesis that new SAMe-TT2R2 have less marked difference in efÔ¨Åcacy outcomes between NOACs and warfarin.6 score is a predictor for average time in therapeutic range. Second, we investigated whether the latter translated into An easy, simple prediction of which atrial Ô¨Åbrillation adverse events (major bleeding, adverse cardiovascular events, patients are likely to do well on vitamin K antagonists (with mortality) in our ‚Äúreal world‚Äù cohort of patients with nongood average time in therapeutic range) could guide valvular atrial Ô¨Åbrillation. decision-making between using vitamin K antagonists and NOACs. Recently, Apostolakis et al7 proposed and validated the SAMe-TT2R2 score (Sex, Age [<60 years], Medical history [at least 2 of the following: hypertension, METHODS diabetes, coronary artery disease/myocardial infarction, peripheral arterial disease, congestive heart failure, previous We recruited consecutive patients with permanent or stroke, pulmonary disease, hepatic or renal disease], and paroxysmal atrial Ô¨Åbrillation on a vitamin K antagonist from Treatment [interacting drugs, eg, amiodarone for rhythm our outpatient anticoagulation clinic. In order to homogecontrol] [all 1 point]; as well as current Tobacco use [2 nize the baseline cohort of patients, only patients who had points] and Race [non-Caucasian; 2 points]) (see Table 1). an international normalized ratio between 2.0 and 3.0 during the previous 6 months were included. Patients with prosthetic heart valves or valvular atrial Ô¨Åbrillation were excluded from the study, as well as those presenting with acute Table 1 The SAME-TT2R2 Score coronary syndrome, stroke (ischemic or embolic), any heS Sex (female) 1 modynamic instability, or hospital admission or surgical A Age (<60 years) 1 intervention in the preceding 6 months. A history of maM Medical history* 1 e lignancy was allowed if the patient‚Äôs expected survival T Treatment (rhythm control strategy) 1 duration was more than 6 months and they were not reT Tobacco use (within 2 years) 2 ceiving chemotherapy or radiotherapy at study entry. A R Race (non-Caucasian) 2 complete medical history was recorded at inclusion. Followup was performed through visits to the anticoagulation *DeÔ¨Åned as more than 2 of the following: hypertension, diabetes, coronary artery disease/myocardial infarction, peripheral arterial disclinic, the hospital electronic medical records system, or, ease, congestive heart failure, previous stroke, pulmonary disease, hewhen unavailable or there were persisting doubts, by telepatic or renal disease. phone interview. Gallego et al SAMe-TT2R2 Score for Anticoagulation Control Baseline stroke risk was assessed using the CHA2DS2VASc (Cardiac failure or dysfunction, Hypertension, Age over 75 years [Doubled], Diabetes, Stroke [Doubled] e Vascular disease, Age between 65 and 74 years, and Sex category [Female]) score, as described in recent guidelines.5,8 The HAS-BLED bleeding risk score was calculated as a measure of baseline bleeding risk, as the result of adding one point to Hypertension, Abnormal renal/liver function (one point each), Stroke, Bleeding history or predisposition, Labile international normalized ratio, Elderly (age over 65 years), and Drugs/alcohol concomitantly (one point for each one).9 Adverse cardiovascular endpoints (mainly thromboembolic) were deÔ¨Åned as stroke/transient ischemic attack, peripheral embolism, acute coronary syndrome, acute heart failure, and cardiac death. Bleeding events were assessed by the 2005 International Society on Thrombosis and Haemostasis criteria, including fatal bleeding or symptomatic bleeding in a critical area or organ, such as intracranial, intraspinal, intraocular, retroperitoneal, intra-articular or pericardial, or intramuscular with compartment syndrome or bleeding causing a fall in hemoglobin level of 20 g/L (1.24 mmol/L) or more, or leading to transfusion of 2 or more units of whole blood or red cells.10 Finally, we recorded all-cause mortality, and whether the cause of death was secondary to a cardiovascular event (stroke/transient ischemic attack, peripheral embolism, acute coronary syndrome, acute heart failure, or cardiac death) or a hemorrhagic one. At 6 months after inclusion, we calculated the percentage of international normalized ratio measures within the therapeutic range during the previous 6 months, as an estimation of the time in therapeutic range. The protocol study was approved by the Ethical Committee from University Hospital Morales Meseguer, and patients gave informed consent to participation in the study. Statistical Analysis Continuous variables were tested for normality by the Kolmogorov-Smirnov test. Continuous variables are presented as a mean √Ü SD or median (interquartile range [IQR]), as appropriate, and categorical variables as a percentage. Cox models were used to determine the association between clinical risk factors and bleeding, as well as cardiovascular events and mortality. The independent effect of clinical variables on prognosis was calculated using a Cox proportional hazard regression model, and analysis of variance models were used to compare means between different groups. The c-statistic was calculated using receiver operating characteristic curves. A P <.05 was accepted as statistically signiÔ¨Åcant. Statistical analyses were performed using SPSS 15.0 for Windows (SPSS, Inc., Chicago, Ill). RESULTS We studied 972 patients (49% male, median age 76 years, IQR 70-80) (Table 2). The median CHA2DS2-VASc score 1085 Table 2 Baseline Clinical Characteristics Baseline Characteristics n ¬º 972 Male sex Age, median (IQR) Age <60 years Hypertension Diabetes mellitus Heart failure History of stroke or TIA Hepatic impairment Renal impairment Coronary artery disease Hypercholesterolemia Current smoking habit Current alcoholic consumption Previous bleeding episode Concomitant malignant disease Concomitant treatment Antiplatelet therapy Angiotensin-converting enzyme inhibitors Angiotensin-renin blockers Calcium antagonist Beta-blockers Statins Digoxin Diuretics CHA2DS2-VASc score, median (IQR) HAS-BLED score, median (IQR) SAME-TT2R2 score, median (IQR) 478 76 66 796 249 350 182 11 94 182 303 136 24 79 58 (49%) (70-80) (7%) (82%) (26%) (36%) (19%) (1%) (10%) (19%) (31%) (14%) (2.5%) (8%) (6%) 156 244 214 208 293 202 179 401 4 2 2 (16%) (25%) (22%) (21%) (30%) (21%) (18%) (41%) (3-5) (2-3) (1-2) CHA2DS2-VASc ¬º Cardiac failure or dysfunction, Hypertension, Age !75 [Doubled], Diabetes, Stroke [Doubled] √Ä Vascular disease, Age 6574 and Sex category [Female]; HAS-BLED ¬º Hypertension, Abnormal Renal/Liver Function, Stroke, Bleeding History or Predisposition, Labile international normalized ratio, Elderly, Drugs/Alcohol Concomitantly; IQR ¬º interquartile range; TIA ¬º transient ischemic attack. was 4 (IQR 3-5), and 93% had a CHA2DS2-VASc score !2. The median HAS-BLED score was 2 (IQR 2-3). Median follow-up was 952 (784-1078) days, and during this period, 107 patients had an adverse cardiovascular event (4.22%/year): of these, 35 were strokes (1.38%/year), 42 were acute coronary syndrome events (1.65%/year), and 31 acute heart failure events (1.22%/year). Of the cohort, 77 patients presented with major bleeding (3.04%/year) and 91 patients died (3.59%/year). In the following 6 months after inclusion, mean time in therapeutic range was 78.0% √Ü 19.98%. Predictive Value for Anticoagulation Control In our anticoagulated patients, increasing baseline SAMeTT2R2 score was associated with signiÔ¨Åcantly lower mean time in therapeutic range at 6 months after inclusion. This ranged from 79.7% for patients with SAMe-TT2R2 score of 0 points to 72.3% for patients with 5 points (P ¬º .043) (Table 3). 1086 Table 3 The American Journal of Medicine, Vol 127, No 11, November 2014 Baseline SAMe-TT2R2 SAMe-TT2R2 Score Score TTR at 6-Month Follow-up Mean (√Ü SD) n ¬º 972 Low Borderline High 0-1 2 !3 79.67(√Ü 19.46) 78.40 (√Ü 20.28) 74.25 (√Ü 20.24) 431 (44%) 332 (34%) 208 (22%) TTR ¬º time in therapeutic range. Predictive Value of the SAMe-TT2R2 Score for Bleeding, Cardiovascular Events, and Death The SAMe-TT2R2 score had a c-statistic for adverse cardiovascular events of 0.62 (95% conÔ¨Ådence interval [CI], 0.57-0.68; P <.001); for bleeding, the c-statistic was 0.55 (95% CI, 0.49-0.62; P ¬º .117). For all-cause mortality, the c-statistic was 0.62 (95% CI, 0.55-0.68; P <.001). Major Adverse Events The incidence of both adverse cardiovascular events (including stroke/thromboembolism) and bleeding and mortality during the follow-up was higher with increasing SAMe-TT2R2 score at baseline (Figure). On unadjusted (crude) analyses, the SAMe-TT2R2 score was predictive for the composite of all adverse events (hazard ratio 1.32 [95% conÔ¨Ådence interval 1.17-1.50]; P <.001), adverse cardiovascular events (hazard ratio 1.52 [95% conÔ¨Ådence interval 1.28-1.83]; P <.001], and for allcause mortality (1.41 [95% conÔ¨Ådence interval 1.16-1.71]; P ¬º .001). Only a nonsigniÔ¨Åcant trend was seen regarding major bleeding events (1.23 [0.99-1.53]; P ¬º .059). DISCUSSION In a ‚Äúreal world‚Äù cohort of consecutive patients with nonvalvular atrial Ô¨Åbrillation, we have shown that a low SAMe-TT2R2 score was a signiÔ¨Åcant predictor for good anticoagulation control (as reÔ¨Çected by time in therapeutic range). A high SAMe-TT2R2 score (reÔ¨Çecting poor anticoagulation control with poor time in therapeutic range) translated into more bleeding, adverse cardiovascular events (including stroke/thromboembolism) and mortality during follow-up. To the best of our knowledge, we have also validated the SAMe-TT2R2 score for the Ô¨Årst time in consecutive atrial Ô¨Åbrillation patients taking acenocoumarol treatment as their oral anticoagulant (rather than warfarin). Nonvalvular atrial Ô¨Åbrillation patients are at high risk for cardiovascular events and mortality. Our population was an elderly cohort with nonvalvular atrial Ô¨Åbrillation where over 50% were aged >75 years. Moreover, our population had an intermediate-high thrombotic risk and multiple prevalent comorbidities. Even in patients with good drug compliance, various acute or clinical decompensate events may lead to unstable international normalized ratio control. Therefore, good prior international normalized ratio control does not necessarily imply good control in the future. Also, the largest randomized trial of genotype-guided (pharmacogenetically based) dosing of warfarin did not result in improved anticoagulation control.11 Thus, the ability to have a simple, practical way to predict who is likely to do well on vitamin K antagonists, or those less likely to have good anticoagulation control where NOACs would beneÔ¨Åt, could help decision-making in everyday clinical practice.12,13 International normalized ratio control and quality of anticoagulation is most often assessed by the time in therapeutic range. The average individual time in therapeutic range is known to increase over time, as more international normalized ratio Ô¨Çuctuations occur during the Ô¨Årst 3 months following initiation of vitamin K antagonists.14 Good adherence to the therapy, as well as minimizing drug or diet 20 18 16 14 Adverse cardiovascular events 12 10 Major bleeding 8 Death 6 4 2 0 0 1 Figure 2 3 4 5 Adverse events according to SAMe-TT2R2 score at baseline. Gallego et al SAMe-TT2R2 Score for Anticoagulation Control interactions, inÔ¨Çuences time in therapeutic range outcomes. In the present study, we selected experienced anticoagulated patients with excellent anticoagulation control at baseline to ensure some degree of patient homogeneity, and thus, adherence to the therapy was expected to be high and not affect subsequent measurements. However, we found that the SAMe-TT2R2 score was still able to predict which patients were prone to subsequent instability and poor international normalized ratio control, even among previously anticoagulated patients, with good time in therapeutic range at inclusion. Slight differences were observed, among the different range of SAMe-TT2R2 scores, due to our inclusion criteria, given that all patients had shown previously excellent international normalized ratio control. The SAMe-TT2R2 score was found to be predictive for both adverse cardiovascular events and mortality, but a trend was observed only regarding major bleeding events. It should be borne in mind that the anticoagulation control in our cohort was optimal, and therefore, only few major bleeding events occurred during the follow-up (with a rate of 3%/year). Although the SAMe-TT2R2 score at baseline is related to outcomes (mainly adverse cardiovascular events and mortality, but also to major bleeding), the low c-statistics show a moderate discriminatory ability of the survival model. This might be related to low outcome event rates (mainly few major bleeding events during the follow-up, with an annual rate of 3%/year), due to the good control of anticoagulation therapy in our (selected) cohort at baseline. Given the modest discriminatory ability (as reÔ¨Çected by c-statistics) of the SAMe-TT2R2 score in our well-controlled anticoagulated clinic population, it is unclear how the score would perform in a less meticulously cared-for group of patients. Of note, we studied a white population, without any prevalence of other ethnic races, which is one of the risk factors for poor time in therapeutic range, based on the SAMe-TT2R2 score.7 Thus, none of the patients could reach a SAMe-TT2R2 score of >6 points. Moreover, high scores were poorly represented in our cohort, as only 6 patients had a score of 5 and none had a score of 6, perhaps reÔ¨Çecting our selection criteria, as stable international normalized ratio was required at baseline. Indeed, patients with the highest SAMe-TT2R2 scores did not achieve good anticoagulation control and therefore were probably excluded from the present cohort. Limitations The possibility of some selection bias cannot be excluded as our patients were clinically stable at study entry, so unstable patients who are more prone to have adverse events were excluded, unlike other clinical studies where patients recruited would present with a time in therapeutic range ranging from 60% to 75%. Similarly, we selected for our cohort ‚Äúanticoagulation experienced‚Äù patients with proven adherence and good anticoagulation control during the 1087 6 months before inclusion. Thus, patients with erratic anticoagulation (who would probably have higher SAMeTT2R2 score values) were excluded from the analysis. This speciÔ¨Åc selection of patients (stable on acenocoumarol with high initial time in therapeutic range at entry) would perhaps limit the generalizability of our Ô¨Åndings to less wellcontrolled populations. Furthermore, we used acenocoumarol (the most widely used vitamin K antagonists in Spain), which differs from warfarin, with a shorter half-life, which is perceived to lead to more unstable anticoagulation, although no results comparing the different vitamin K antagonists have been published. We also presumed that patients with high SAMe-TT2R2 score would be ‚Äúsuitable‚Äù to be treated with NOAC, but we have not studied such a population. Finally, we studied a white population, without any other races being studied, and thus, further studies in ethnically diverse populations would be required. Ongoing prospective studies will address all these aspects. In conclusion, a low SAMe-TT2R2 score was a signiÔ¨Åcant predictor for good anticoagulation control (as reÔ¨Çected by time in therapeutic range), whilst a high SAMe-TT2R2 score (reÔ¨Çecting poor anticoagulation control) was associated with more bleeding, adverse cardiovascular events, and mortality during follow-up. Thus, the simple SAMe-TT2R2 score may aid decision-making by identifying those atrial Ô¨Åbrillation patients likely to do well on vitamin K antagonist (score 0-2) or those likely to have poor(er) anticoagulation control (score >2) who could be better off being started on NOACs as initial therapy, or be ‚ÄúÔ¨Çagged up‚Äù for more aggressive efforts to improve anticoagulation control. References 1. Hart RG, Pearce LA, Aguilar MI. Meta-analysis: antithrombotic therapy to prevent stroke in patients who have nonvalvular atrial Ô¨Åbrillation. Ann Intern Med. 2007;146(12):857-867. 2. Gallagher AM, Setakis E, Plumb JM, Clemens A, van Staa TP. Risks of stroke and mortality associated with suboptimal anticoagulation in atrial Ô¨Åbrillation patients. Thromb Haemost. 2011;106(5):968-977. 3. Morgan CL, McEwan P, Tukiendorf A, Robinson PA, Clemens A, Plumb JM. Warfarin treatment in patients with atrial Ô¨Åbrillation: observing outcomes associated with varying levels of INR control. Thromb Res. 2009;124(1):37-41. 4. Wan Y, Heneghan C, Perera R, et al. Anticoagulation control and prediction of adverse events in patients with atrial Ô¨Åbrillation: a systematic review. Circ Cardiovasc Qual Outcomes. 2008;1(2):84-91. 5. Camm AJ, Lip GY, De Caterina R, et al. 2012 focused update of the ESC Guidelines for the management of atrial Ô¨Åbrillation: an update of the 2010 ESC Guidelines for the management of atrial Ô¨Åbrillation‚Äî developed with the special contribution of the European Heart Rhythm Association. Europace. 2012;14(10):1385-1413. 6. Wallentin L, Yusuf S, Ezekowitz MD, et al. EfÔ¨Åcacy and safety of dabigatran compared with warfarin at different levels of international normalised ratio control for stroke prevention in atrial Ô¨Åbrillation: an analysis of the RE-LY trial. Lancet. 2010;376(9745):975-983. 7. Apostolakis S, Sullivan RM, Olshansky B, Lip GY. Factors affecting quality of anticoagulation control amongst atrial Ô¨Åbrillation patients on warfarin: the same-TT2R2 score. Chest. 2013;144(5):1555-1563. 8. Lip GY, Nieuwlaat R, Pisters R, Lane DA, Crijns HJ. ReÔ¨Åning clinical risk stratiÔ¨Åcation for predicting stroke and thromboembolism in atrial Ô¨Åbrillation using a novel risk factor-based approach: the euro heart survey on atrial Ô¨Åbrillation. Chest. 2010;137(2):263-272. 1088 The American Journal of Medicine, Vol 127, No 11, November 2014 9. Pisters R, Lane DA, Nieuwlaat R, de Vos CB, Crijns HJ, Lip GY. A novel user-friendly score (HAS-BLED) to assess 1-year risk of major bleeding in patients with atrial Ô¨Åbrillation: the Euro Heart Survey. Chest. 2010;138(5):1093-1100. 10. Lip GY, Andreotti F, Fauchier L, et al. Bleeding risk assessment and management in atrial Ô¨Åbrillation patients. Executive Summary of a Position Document from the European Heart Rhythm Association [EHRA], endorsed by the European Society of Cardiology [ESC] Working Group on Thrombosis. Thromb Haemost. 2011;106(6): 997-1011. 11. Kimmel SE, French B, Kasner SE, et al. A pharmacogenetic versus a clinical algorithm for warfarin dosing. N Engl J Med. 2013;369(24): 2283-2293. 12. De Caterina R, Husted S, Wallentin L, et al. Vitamin K antagonists in heart disease: current status and perspectives (Section III). Position Paper of the ESC Working Group on Thrombosis‚ÄîTask Force on Anticoagulants in Heart Disease. Thromb Haemost. 2013;110(6):1087-1107. 13. De Caterina R, Husted S, Wallentin L, et al; European Society of Cardiology Working Group on Thrombosis Task Force on Anticoagulants in Heart Disease. General mechanisms of coagulation and targets of anticoagulants (Section I). Position Paper of the ESC Working Group on Thrombosis‚ÄîTask Force on Anticoagulants in Heart Disease. Thromb Haemost. 2013;109(4):569-579. 14. Garcia DA, Lopes RD, Hylek EM. New-onset atrial Ô¨Åbrillation and warfarin initiation: high risk periods and implications for new antithrombotic drugs. Thromb Haemost. 2010;104(6):1099-1105. A mobile health infrastructure to support underserved patients with chronic disease * Susan L. Moorea, * Henry H. Fischera, b, , , * Andrew W. Steelea, b, * M. Joshua Durfeea, * David Ginosara, b, * Cecilia Rice-Petersona, * Jeffrey D. Berschlinga, * Arthur J. Davidsona, b Show more doi:10.1016/j.hjdsi.2013.12.016 Get rights and content Abstract Background Chronic diseases are the global leading cause of death, but the US health system is poorly designed to support patients with chronic disease. Underserved patients report high rates of cell phone use and interest in using mobile technology for health care. A mobile health infrastructure may help transform health care delivery for underserved patients with chronic disease. Problem This study assessed the feasibility of integrating mobile health infrastructure with clinical information systems and the electronic medical record (EMR) to support patients with chronic disease through automated, bidirectional text messaging. Goals Three priority areas of chronic disease management were targeted. Existing self-management support was expanded, and new support for laboratory test scheduling and medication management was created. Strategy Adult patients (n=135) with diabetes selected preferred content and scheduling for self-management message prompts. Outreach messages were sent to patients overdue for laboratory tests and medications. Manual review of pharmacy and laboratory outreach data was conducted for quality assurance. Focus groups were held to solicit patient perspectives. Results Patients sent over 6500 response messages with response rates of 53.7% (blood sugar), 48.8% (step counts), and 31.9% (blood pressure). Laboratory data integration was achieved, but pharmacy data gaps required ongoing manual review. Focus group participants reported improved self-management and information awareness. Implications HIT was used to address dependency on visit-bound disease management in a novel, low-cost way. Use of a mobile health infrastructure was feasible. Text messaging solutions may mitigate barriers to access and enhance support for patients with chronic disease. Keywords * Diabetes; * Text message; * SMS; * Chronic disease; * Mobile health; * Safety net 1. Background Chronic diseases are worldwide killers.1 and 2 In the US, 70% of deaths and 78% of health care expenditures are due to chronic disease.3 and 4 Nearly half of Americans have one chronic disease and a quarter have two or more, but the design of the US health system is poorly suited for chronic disease management.5, 6 and 7 Competing demands limit providers? ability to manage chronic disease during the standard 20-min clinic visit.7,8 and 9 Racial and ethnic minorities and low-income and uninsured adults experience significantly higher chronic disease prevalence and disparities in receiving recommended treatment when compared to white, higher-income, and insured adults5,10, 11, 12 and 13; these same groups also encounter significantly higher barriers to health care access, rendering visit-focused chronic disease management insufficient for their needs.8, 9, 12, 14 and 15 Health information technology (HIT) can reduce chronic disease costs and improve outcomes.16, 17, 18, 19 and 20 Consumer-focused health informatics solutions have been associated with better management of chronic disease outside the clinic, including improvements in blood pressure, glycemic, and asthma control.20, 21, 22, 23, 24, 25, 26 and 27Cell phone technology is widely available and frequently used for health purposes: 85% of American adults own cell phones, of whom 31% use their phones to obtain health information – a near 100% increase in 2 years.28 Of US cell phone owners, 80% have used text messaging.28 Text message-based diabetes management has been associated with improved glycemic control29 and with patient reports of increased health information awareness and social support.30 Cell phone ownership and use for health information are significantly higher among Latinos and Blacks than among whites,28,31 and 32 indicating the potential of this technology to help overcome disparities in chronic disease management. 2. Organizational context These issues were examined in a pilot study designed to demonstrate the feasibility of using mobile health infrastructure in an urban safety net setting to improve chronic disease management for adult patients with diabetes. The US health care safety net is the source of care for patients with little or no health insurance. It encompasses health delivery systems and providers who share a mission of caring for patients regardless of ability to pay,33 and includes organizations such as health departments, public hospitals, and federally qualified health centers (FQHCs). Safety net patients predominantly include racial and ethnic minorities, children and the elderly, people with disabilities, and other groups with low income or socioeconomic status.34, 35 and 36 Denver Health (DH), an integrated health care delivery system, is the primary safety net for patients in Denver, Colorado. The system includes a 525-bed acute care hospital with a Level I trauma center, Denver Public Health, and 15 school-based clinics and 8 FQHCs distributed throughout Denver neighborhoods. Study participants were recruited from 2 FQHCs: Westside Community Health Center (Westside) and Webb Center for Primary Care (Webb). Over 115,000 patients receive primary care at DH, with over 350,000 clinic visits annually. Half of DH patients are uninsured, and 65% live below 185% of the federal poverty level. Sixty percent represent racial and ethnic minorities, and 25% speak a native language other than English. DH developed its technology infrastructure over almost two decades and has extensive experience with HIT-supported patient care. Integration between the electronic medical record (EMR), clinical systems such as those for pharmacy and laboratory services, and multiple disease registries assists providers with improving patient care at the point of service. A recent infrastructure addition is the Patient Relationship Management (PRM) software platform. The PRM system was originally designed to support patients with diabetes through automated text messaging for appointment reminders and to prompt for and collect patient-reported blood sugar measurements. A previous pilot study demonstrated the potential of this approach.30 3. Personal context The project team expanded the PRM infrastructure with funding through the Agency for Healthcare Research and Quality?s Multiple Chronic Conditions Research Network (Grant R24 HS 019453-01). The team was led by a primary care physician who also serves as the leader for DH's diabetes collaborative and who had previous experience with integrating DH's diabetes registry and the EMR. Other team members contributed multidisciplinary expertise in health informatics, nursing, health services research, and public health (Table 1). Together, these individuals formed a committed, experienced project team with both the necessary knowledge and the drive to pursue continued improvement in chronic disease management through mobile technology. Table 1. Project team composition and expertise. Project role/title Background and multidisciplinary expertise Principal investigator Primary care physician with expertise in general internal medicine and chronic disease management; leader of DH?s diabetes collaborative; health informatics experience through project-based integration of a diabetes registry with the EMR Director, medical informatics Primary care physician who oversees clinical information technology system selection, design, and implementation activities at DH Director, public health informatics Primary care physician who provides technical and clinical leadership for Colorado?s health information exchange and regional health information organization projects Nurse case manager Registered nurse with experience in telephone-based diabetes management IT project manager Certified project management professional with public health expertise; oversaw coordination of project activities with DH?s eHealth Services (eHS) department processes and personnel. Physician champion Primary care physician champion and provider educator with strong interest in chronic disease management Chief quality officer Primary care physician with chronic disease management and quality improvement research experience Director, quality improvement research Primary care physician with chronic disease management and quality improvement research experience Director, community outreach and patient navigation Research and program implementation experience in creating and testing national best-practice models of care Health services researchers Two health services researchers, both with previous IT industry experience and current expertise in public health, quantitative and qualitative methods and analyses, and research project coordination EMC software development team Team of health IT industry consultants who contributed practical system design, software development, implementation and integration skills. Table options 4. Problem The team?s challenge was to expand text message-based support for chronic disease management and to integrate the PRM platform with clinical systems as data sources and with the EMR to deliver patient-reported measurements automatically to providers at the point of care. The team discussed measures of success and identified clinical domains for inclusion in program design and infrastructure development. Additional input came from DH leaders and key stakeholders about parallel and competing organizational priorities and what information systems and architecture were best suited for integration. Three areas of chronic disease management were targeted as priorities. Existing self-management support was expanded, and new support for laboratory test scheduling and medication management was created and implemented. Patient response rate to text message prompts for home measurements was the primary outcome; a secondary outcome was the percent of responses correctly formatted by patients and thus automatically handled by the PRM system. Process outcomes included PRM capacity to successfully prompt for multiple self-management measurements on varying schedules, to identify patients overdue for laboratory tests and statin cholesterol medications and to automatically transfer patient self-reported measurements to the EMR. The Colorado Multiple Institutional Review Board approved the study prior to patient enrollment. 5. Strategy A 9-month pilot study ending April 30, 2012 explored the expanded infrastructure?s feasibility to engage adult patients with diabetes through text message prompts. Eligible patients were on the diabetes registry, over 18 years old, spoke either English or Spanish as their primary language, and had access to a text message-capable cell phone. The study group (n=135) was demographically similar to clinic profiles ( Table 2) and was predominantly female (65%) and Latino (65%), with substantial participation among whites (25%) and African Americans (8%). Table 2. Comparison of demographics for study group vs. clinic populations. Study group (N=135) % Westside (N=1695) % Webb (N=1827) % Age 20–29 3.5 2.4 2.0 30–39 5.6 7.2 6.8 40–49 21.7 18.0 17.9 50–59 40.6 27.6 32.3 60–69 21.0 27.1 23.8 >70 7.7 17.7 17.3 Gender Male 35.0 38.3 45.5 Female 65.0 61.7 54.5 Race/Ethnicity Black 8.9 3.6 11.2 Hispanic/Latin@ 68.1 80.7 52.1 White 22.2 13.3 32.3 Other/unknown 0.7 2.5 4.4 Table options Patients were prompted by text message to submit self-management measurements and also received text message outreach if overdue for laboratory tests or medication refills. Patients were given a glucometer, blood pressure cuff, and/or pedometer according to their preferences. Study personnel recorded patients? preferred primary language and self-measurement prompt schedules, selected from a list of available options. Patients were told not to use text messages for urgent health issues, and that PRM was staffed only during regular business hours. Patients were also advised that DH could not guarantee privacy over commercial carrier networks, and that messages might be viewed by others if their phones were shared, lost or stolen. Previous self-management support was expanded from prompting for blood sugar measurements thrice weekly to prompting for blood sugar, blood pressure, and/or step count measurements with a variety of scheduling options. Integration between PRM and the EMR was established to transfer patient self-reported measurements to the EMR through automated, structured reports available to clinicians at the point of care (Fig. 1). The system automatically acknowledged all communications from patients through return text confirmation of message receipt. Patient-submitted measurements were automatically evaluated against established clinical ranges, and automated activity lists (“work queues”) were created for personal follow-up by the nurse case manager (CM) when patients? measurements were out of range. PRM also identified and flagged incorrectly formatted inbound messages from patients for CM follow-up. Fig. 1. Automated EMR self-report results form. Figure options Laboratory test prompts were generated by PRM for patients overdue for cholesterol or hemoglobin A1c (HbA1c) testing based on automated data queries from clinical systems. Patients overdue for testing were offered scheduling assistance through text prompts weekly for 3 weeks or until a lab test record was detected, whichever was first. When a patient responded confirming interest in test scheduling, PRM created a work queue activity for the CM to prepare a lab order. Subsequent lab test results were automatically conveyed to the patient?s primary care provider and transferred to the EMR. Manual review of EMR data was conducted for quality assurance purposes to evaluate PRM success at identifying overdue patients; however, false negatives, or the failure to identify overdue patients, were not evaluated. Overdue medication outreach was offered to patients who were late refilling statin medications prescribed for cholesterol control, based on automated query and analysis of pharmacy data. Patients were classified as late if over 7 days had elapsed past their expected refill date. Refill dates were calculated from multiple pharmacy data elements: monthly doses dispensed; daily doses prescribed; “fill date,” the date a pharmacist filled a prescription; “date sold,” the date the patient retrieved the medication; patient allergies; and “medication reconciliation,” a field manually updated by health care providers during a clinic visit to confirm patients? current medications. Late-refilling patients received four consecutive text messages asking about barriers encountered, and a work queue item was created for the CM to problem-solve with patients. Manual EMR record review was conducted to confirm system identification of overdue patients, but did not evaluate for false negatives. Patients? perceptions of the text messaging program were solicited through focus groups in English and Spanish, conducted among a subset of study participants with group composition based on frequency of text message response. Focus groups were led by a trained facilitator, who worked with a medical interpreter to ensure clear communication with Spanish-speaking participants. 6. Challenges Challenges were encountered throughout development and implementation. At the individual level, patients exhibited a learning curve in text message response formatting. Formatting errors included typographical errors, errors resulting from misperception of prompts, and system errors generated by the presence of personalized “signatures” appended to patients? messages. Incorrectly-formatted messages which could not be interpreted by the system were automatically flagged for CM follow-up. The CM problem-solved with patients, improving response formatting and patient comprehension, and PRM message processing logic was refined to exclude signature content. At the intervention level, medication outreach challenges occurred with pharmacy data quality. Delays in updating the “date sold”? field at some pharmacy sites resulted in fragmentation between data repositories and erroneous refill date calculation; manual review was required to prevent inappropriate outreach. Certain pharmacy data could not be reliably accessed, including allergy and medication reconciliation fields. Consequently, the system was unable to identify patients who had recently stopped taking a statin medication due to adverse reaction or provider recommendation. A medication cessation question was added to the barrier assessment text message series to identify these patients. Extensive problem-solving around pharmacy data limited outreach to patients overdue for medications (n=5) and precluded useful quantitative analysis. At the system level, transmission over external commercial networks led to identification of several technical challenges. Due to high volume, outbound text messages were sent through a commercial gateway (MessageMedia, San Francisco, CA) for distribution to telephone service providers? networks and subsequent delivery to patients? phones. When delivery failed due to patients? phones being outside service areas or temporarily disconnected, telephone service providers? message handling methods varied. Some carriers stored messages for subsequent attempts, while others rejected messages after a single failure. Upgrades made to the gateway during the project resulted in patients reporting non-receipt of expected messages. The PRM communications interface with the gateway was adapted, and message transmission successfully resumed. Additional post-upgrade issues were observed due to the vendor?s transition from “long-code,” 10-digit telephone numbers to “short-code,” 5-digit identifiers for text messages. Short-code numbers are classified as commercial messages and given priority transmission over telephone providers? networks to ensure higher rates of delivery; however, several participants had opted not to receive short-code commercial messages from their cell phone carriers, which led to PRM messages attached to short-code identifiers being inadvertently rejected. These messages were returned as “failed” (6% of outbound texts). 6.1. Results – self-management support PRM sent 11,429 self-management prompts according to patient-selected options and schedules. Patients replied with 6850 text messages, of which 6512 (95.1%) were correctly-formatted. Response rates among engaged patients (60%; >10 responses) to blood sugar, step count, and blood pressure prompts were 0.71, 0.66, and 0.52 respectively compared to negligible response rates among less-engaged patients. Response rate details are shown in Table 3. Less-engaged patients were demographically similar to engaged patients in age, with slightly more women (72% vs. 65%) and Latinos (72% vs. 68%). The CM logged 547 completed phone calls in response to out-of-bounds measurements. Low readings yielded 481 calls among 54 patients, and high readings resulted in 66 calls among 14 patients. Table 3. Response frequencies to home measurement prompts. Number of unique patients Number of patients receivinga Total number of prompts Prompt rate per patient Number of patients with ≥1 response Total number of responses Mean number of responses Response rate (#responses/#prompts) Engaged (≥10 responses) 82 Blood glucose 66 5121 77.6 66 3619 54.8 0.71 Step count 28 2903 103.7 28 1911 68.3 0.66 Blood pressure 22 425 19.3 9 222 24.7 0.52 Less engaged (1-9 responses) 19 Blood glucose 11 560 50.9 10 14 1.4 0.03 Step count 9 617 68.6 7 23 3.3 0.04 Blood pressure 2 2 1.0 0 0 0.00 Non-responders (0 responses) 34 Blood glucose 26 1087 41.8 47.8 Step count 5 444 88.8 55.3 Blood pressure 7 270 38.6 All patients Blood glucose 103 6768 65.7 76 3633 47.8 0.54 Step Count 42 3964 94.4 35 1934 55.3 0.49 Blood pressure 31 697 22.5 9 222 24.7 0.32 Total 135 176 11,429 64.9 101 5789 57.3 0.51 a Patients may be in more than one category. Table options 6.2. Results – laboratory outreach Integration between PRM and laboratory systems successfully enabled automated outreach to patients overdue for laboratory tests. Prompts were sent to 73 (54%) patients, of whom 18 (24.6%) presented for testing within 4 weeks. This number was insufficient to draw conclusions about impact on clinical or process outcomes. 6.3. Results – medication outreach Although a workflow was created to identify patients overdue for medication refill, pharmacy data quality issues required ongoing manual review to ensure accuracy, preventing full automation. 6.4. Results – focus groups Three patient focus groups were held, two in English (n=8) and one in Spanish (n=6). Active and engaged discussion occurred in all sessions. Spanish-language group participation was enhanced in that two participants brought relatives as part of a family-focused approach to text message-based diabetes management, and two other participants brought friends from among the wider DH patient community to encourage them to sign up for the program. This engagement of patients? social networks in discussion was naturally-occurring and not prompted in the focus group invitation. Participants expressed positive reactions to the overall program, believing it helpful and important, and reported improved health status awareness and adherence to self-management activities. Messaging schedule familiarity was reported to the extent that patients were aware of service interruptions resulting in missed messages. Non-response to message prompts was associated with factors other than lack of engagement, including temporary lack of access to a shared cell phone and not answering prompts when no results were available, e.g., if a self-monitoring test had not been performed. Participants suggested improvements, including support for sending independent “status update” or “missed test” messages and specifying fasting or non-fasting blood sugar measurements, and were interested in additional communication methods such as social media, interactive voice response calls, email, and web-based portals. All participants had unlimited texting plans and expressed no concern about message cost. 7. Unresolved questions and lessons for the field Patients successfully interacted with a mobile health infrastructure through automated, bidirectional text message communication. The integration of text message-based diabetes support with clinical data sources to trigger automated outreach and home monitoring data transmission to the point of care is unique based on our literature review. The feasibility study demonstrated that this infrastructure can be used to collect and integrate patient-provided measurements with an EMR and to promote self-management within a safety net health system. Consumer health informatics applications which link patients with their Patient Centered Medical Home (PCMH) and their health information through configurable, patient-centered channels can transform health service delivery. Mobile health solutions which facilitate engagement with health systems may mitigate barriers to access and support reinforcement of healthy behaviors. From the health system perspective, this infrastructure addresses an unsustainable dependency on visit-bound care in a novel, low-cost way. Expanding this work to engage more patients over a longer period would allow evaluation of this care model?s impact on clinical and health utilization outcomes. While mobile technology has a burgeoning role in a transformed health care system, there are additional considerations involved in scaling services for larger populations. Federal regulation of electronic and automated telephonic communications (including text messages) for business purposes under the Telephone Consumer Protection Act and CAN-SPAM Act requires ongoing revision to institutional consent processes and support for patient-based contact and preference configuration. Ensuring protection of patients? health information under the HIPAA Privacy and Security Rule requires additional review and risk assessment as content becomes potentially more identifiable and penalties more severe.37 That said, patients found this model of text message-mediated care highly acceptable. The use of cell phone text messaging made the intervention accessible to patients of all ages. A configurable, patient-centered approach allowed participants to individualize outreach content, format, language, and frequency. Future health communication tools should enhance support for chronic disease management in ways that are responsive to patient needs. Fig. 2 illustrates one potential expansion of the approach used in this study. Increasing patient exposure to the content of the health ecosystem brings opportunities to improve between-visit communication and health outcomes.38 and 39Achieving a transformed health care system will require well-tested, usable communication tools that support patients, reduce health care costs and improve health outcomes while creating value for both patients and providers. Fig. 2. Conceptual model of text message-mediated care. EBIOM-00184; No of Pages 10 EBioMedicine xxx (2015) xxx‚Äìxxx Contents lists available at ScienceDirect EBioMedicine journal homepage: www.ebiomedicine.com A Multi-Dimensional and Integrative Approach to Examining the High-Risk and Ultra-High-Risk Stages of Bipolar Disorder Kangguang Lin a,b,c,d, Guiyun Xu b,c,‚Åé, Nichol M.L. Wong a,d, Huawang Wu e, Ting Li b, Weicong Lu b, Kun Chen b,c, Xiaodong Chen b, Bingyin Lai b,c, Liuxia Zhong b,c, Kwok-fai So f,g,h, Tatia M.C. Lee a,d,f,i,‚Åé‚Åé a Laboratory of Neuropsychology, The University of Hong Kong, Hong Kong Department of Affective Disorders, Guangzhou Brain Hospital, AfÔ¨Åliated Hospital of Guangzhou Medical University, Guangzhou, China Laboratory of Emotion and Cognition, Guangzhou Brain Hospital, AfÔ¨Åliated Hospital of Guangzhou Medical University, Guangzhou, China d Laboratory of Cognitive Affective Neuroscience, The University of Hong Kong, Hong Kong e Department of Radiology, Guangzhou Brain Hospital, AfÔ¨Åliated Hospital of Guangzhou Medical University, China f State Key Laboratory of Brain and Cognitive Sciences, The University of Hong Kong, Hong Kong g GMH Institute of CNS Regeneration, Jinan University, Guangzhou, China h Department of Ophthalmology, The University of Hong Kong, Hong Kong i Institute of Clinical Neuropsychology, The University of Hong Kong, Hong Kong b c a r t i c l e i n f o Article history: Received 13 April 2015 Received in revised form 25 June 2015 Accepted 27 June 2015 Available online xxxx Keywords: Bipolar disorder Affective disorder Network analysis High-risk design Neuroimaging Cognition Ultra-high-risk a b s t r a c t Background: Validating the high-risk (HR) and ultra-high-risk (UHR) stages of bipolar disorder (BP) may help enable early intervention strategies. Methods: We followed up with 44 offspring of parents with BP, subdividing into the HR and UHR categories. The offspring were aged 8‚Äì28 years and were free of any current DSM-IV diagnoses. Our multilevel, integrative approach encompassed gray matter (GM) volumes, brain network connectivity, neuropsychological performance, and clinical outcomes. Findings: Compared with the healthy controls (HCs) (n = 33), the HR offspring (n = 26) showed GM volume reductions in the right orbitofrontal cortex. Compared with the HR offspring, the UHR offspring (n = 18) exhibited increased GM volumes in four regions. Both the HR and UHR offspring displayed abnormalities in the inferior occipital cortex regarding the measures of degree and centrality, reÔ¨Çecting the connections and roles of the region, respectively. In the UHR versus the HR offspring, the UHR offspring exhibited upwards-shifted small world topologies that reÔ¨Çect high clustering and efÔ¨Åciency in the brain networks. Compared with the HCs, the UHR offspring had signiÔ¨Åcantly lower assortativity, which was suggestive of vulnerability. Finally, processing speed, visual‚Äì spatial, and general function were impaired in the UHR offspring but not in the HR offspring. Interpretation: The abnormalities observed in the HR offspring appear to be inherited, whereas those associated with the UHR offspring represent stage-speciÔ¨Åc changes predisposing them to developing the disorder. ¬© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction Bipolar disorder (BP) is a major psychiatric disorder that is disabling and recurrent. It is highly heritable (more than 70%) (Tsuang and Faraone, 2000) and is characterized by hypo/manic episodes. Pioneering work on the trajectory of BP has identiÔ¨Åed early-risk syndromes that precede the ofÔ¨Åcial onset (Akiskal et al., 1985; Duffy et al., 2014; Mesman et al., 2013) and are proposed to represent early stages in the development of BP (Duffy et al., 2014). Identifying the early stages of ‚Åé Correspondence to: G. Xu, Department of Psychiatry, Guangzhou Brain Hospital, AfÔ¨Åliated Hospital of Guangzhou Medical University, 36 MingxinLoad, Guangzhou, Guangdong Province 510370, China. ‚Åé‚Åé Correspondence to: T.M.C. Lee, Laboratory of Neuropsychology, Rm 656, The Jockey Club Tower, The University of Hong Kong, Pokfulam Road, Hong Kong. E-mail addresses: xuguiyun2908@hotmail.com (G. Xu), tmclee@hku.hk (T.M.C. Lee). BP is important for prevention and early intervention strategies, which have been shown to be successful in early psychosis and can avert or delay the transition from clinically ultra-high-risk conditions to a full psychotic disorder (McGorry et al., 2009). The early symptoms (and syndromes) that precede full-blown BP are usually non-speciÔ¨Åc during childhood (e.g., anxiety, sleep disturbance, and attention deÔ¨Åcit hyperactivity disorder (ADHD) symptoms/signs) and then manifest as adjustment disorder during early adolescence and later as subthreshold depression and/or hypomania that falls short of the ofÔ¨Åcial criteria (Akiskal et al., 1985; Correll et al., 2014; Duffy et al., 2014; Egeland et al., 2000; Mesman et al., 2013). In a Canadian follow-up study (up to 16 years) of the offspring of parents with BP, the accumulated incidence of major mood disorders (depressive spectrum and bipolar disorders) was unfortunately high ‚Äî 83.3% (Duffy et al., 2014). Another Dutch study of genetically high-risk offspring (12 years of follow-up) http://dx.doi.org/10.1016/j.ebiom.2015.06.027 2352-3964/¬© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Please cite this article as: Lin, K., et al., A Multi-Dimensional and Integrative Approach to Examining the High-Risk and Ultra-High-Risk Stages of Bipolar Disorder, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.027 2 K. Lin et al. / EBioMedicine xxx (2015) xxx‚Äìxxx showed that 72% of the offspring developed a lifetime DSM-IV (Diagnostic and Statistical Manual of Mental Disorders, 4th Edition) axis I disorder (Mesman et al., 2013). Furthermore, in modeling the developmental stages of BP ‚Äì from non-speciÔ¨Åc symptoms, followed by minor mood disorder, major depressive episodes, and Ô¨Ånally hypo/manic episodes ‚Äì once entering the model, these high-risk offspring progressed linearly through the stages without skipping any stage (Duffy et al., 2014), which suggests a progressive process and an urgent need for intervention. Recently, clinical staging models for BP that cut across the dichotomous category of the present classiÔ¨Åcation systems have been proposed with attempts at better understanding the underlying pathophysiology and providing potential targets for early intervention (Frank et al., 2014; Scott et al., 2013). The clinical staging models hypothesize that there are stages (0, 1a and 1b) that precede full-blown BP. Stage 0 and stage 1a may together represent a phase of biological vulnerability (i.e., genetic risk) with no or mild, non-speciÔ¨Åc symptoms that can be subsumed into the HR stage (Scott et al., 2013), whereas stage 1b can be described as ultra-high-risk (UHR), manifesting subthreshold syndromes, alterations in cortical (and subcortical) volumes, and deÔ¨Åcits in cognitive function (Frank et al., 2014). Interrogating the underlying pathophysiology prior to the fullblown manifestations can help clarify some confounding effects. For example, extant neuroimaging research on established BP has suggested that the volume of gray matter (GM) decreases in the regions underpinning emotion processing and regulation and cognitive processes, such as the ventrolateral prefrontal cortex (vlPFC), orbitofrontal cortex (OFC) and amygdala (Phillips and Swartz, 2014). Nevertheless, these changes more likely reÔ¨Çect the net effect of inherited neuropathological vulnerability and a number of contributing factors including illness progression, symptoms (e.g., psychotic symptoms), comorbid conditions, and medications such as lithium that can normalize or increase GM volumes (Kalmar et al., 2009; Moore et al., 2000; Moorhead et al., 2007; Nugent et al., 2006; Strasser et al., 2005). Brain network analysis has been increasingly adopted in neuroscience research. This method provides useful information about brain organization in terms of how spatially segregated brain regions are integrated globally via connecting Ô¨Åber tracts (i.e., an anatomical network) to form an integrated system (for a comprehensive review, see (Bullmore and Sporns, 2009)). Accumulating evidence suggests that inter-regional integration is crucial for cognitive performance, particularly for effortful psychological tasks, such as working memory (Kitzbichler et al., 2011). Moreover, brain network analysis is particularly helpful for research on those psychiatric disorders that are conceptualized as dysconnectivity syndromes, such as schizophrenia and BP, disorders that may be caused by the failure of integrating spatially distributed brain regions to form a large-scale network (Catani and ffytche, 2005). Among measures of brain network topology, smallworld properties are useful for describing anatomical connectivity networks that reÔ¨Çect a high clustering of functionally associated regions with short path length (high efÔ¨Åciency) (Bassett and Bullmore, 2006; Bullmore and Sporns, 2012). These properties were reported to be heritable in twin studies (Smit et al., 2008), related to cognitive performance (Micheloyannis et al., 2009), and signiÔ¨Åcantly altered in schizophrenia (Bassett et al., 2008; Liu et al., 2008) (a disorder sharing many overlapping features with BP, including genetics). In contrast, assortativity, an index that can be used to measure the robustness to assaults (i.e., structurally abnormal regions) of a brain network (Newman, 2002), may potentially assist in capturing the vulnerability of the UHR stage of BP. At the macroscopic level, cognitive deÔ¨Åcits have been demonstrated to be an important aspect of full-blown BP that adversely affects the quality of life in people with the disorder (Wingo et al., 2009). A signiÔ¨Åcant research gap is whether cognitive deÔ¨Åcits are the consequences of the course of illness and its related factors, such as medications (e.g., valproate) and recurrent subthreshold syndromes (Martinez-Aran et al., 2004; Rosa et al., 2014; Xu et al., 2012), or are inherited, e.g., the deÔ¨Åcits in visual‚Äìspatial memory (Ferrier et al., 2004) and working memory observed in the ‚Äúunaffected‚Äù relatives of patients with BP (Kulkarni et al., 2010). Indeed, our previous study showed that patients with BP, following clinical remission from depression after six-week treatments, did not recover their processing speed and visual‚Äìspatial memory functioning (Xu et al., 2012). To Ô¨Åll the research gap, it is necessary to investigate whether cognitive deÔ¨Åcits exist in genetically high-risk individuals at the very early stages before the onset of full-blown syndromes. A multi-dimensional approach can inform complementary and mutually informative connections between different levels of descriptions across stages, thus yielding patterns of abnormalities (Phillips and Kupfer, 2013). Such an approach may assist in understanding how neural substrates affect cognitive function and behavioral phenotypes across stages. Given the above considerations, this study applied a multidimensional approach to investigating neural correlates and cognitive function at the HR and UHR stages of BP. We began by proposing operational UHR criteria for BP and delineating the clinical characteristics and general functions for both the HR and UHR stages. We Ô¨Årst searched for structural abnormalities in the GM that form the neural basis of the functional system by comparing HR offspring (stage 0 and stage 1a) with healthy controls ‚Äì neuroanatomical endophenotypes ‚Äì and further tested the validity of UHR (stage 1b) by comparing UHR with HR offspring as an attempt to dissect stage-speciÔ¨Åc (i.e., adaptive) from inherited alterations. Then, in a regional-level network connectivity analysis, we compared the degree and centrality of the prior identiÔ¨Åed abnormal regions (reÔ¨Çecting the connections and roles of a region, respectively, see Section 2), which could help elucidate whether volumetric changes in GM (i.e., GM reduction) were a consequence of the lack of connecting Ô¨Åber bundles. We also quantitatively examined whether the signiÔ¨Åcance (centrality) of the structurally abnormal regions would be affected within the brain networks. Next, a whole-brain network topology analysis was applied to encapsulate all the individual abnormalities that reÔ¨Çect the whole-brain network properties, including small-world and assortativity properties. Finally, at the cognitive function level, we delineated whether cognitive deÔ¨Åcits predate the onset, and if so, at what stages; we then searched for their neural correlates. We hypothesized that HR offspring versus healthy controls had decreased GM volumes in the regions underlying emotion processing and regulation. Compared with HR offspring, UHR offspring might exhibit decreased GM volumes in the regions (e.g., the OFC) related to emotion processing and regulation and in the regions (e.g., the parietal and occipital cortex) involved in the cognitive deÔ¨Åcits of bipolar disorder (e.g., processing speed). For the brain network measures, HR and UHR offspring may display impairments in the degree and centrality of those abnormal regions identiÔ¨Åed by prior GM volume analyses when compared to healthy controls. At the whole-brain level, smallworld properties in UHR offspring might be lower (worse) than those in HR offspring and healthy controls. Compared to healthy controls, UHR offspring might display lower values in assortativity, which is suggestive of vulnerability. Finally, we hypothesized that the deÔ¨Åcits in processing speed and visual‚Äìspatial memory existed prior to the ofÔ¨Åcial onset of bipolar disorder (in the HR and/or UHR stages); there may be a correlation between the cognitive deÔ¨Åcits and GM volumes in those structurally abnormal regions. 2. Methods 2.1. Participants This study derived and extended from the project ‚ÄúThe Recognition and Early Intervention of Prodromal Bipolar Disorders (REI-PBD)‚Äù, whose aims were to identify bipolar prodromal syndromes and to evaluate the effectiveness of exercise interventions for UHR offspring (ClinicalTrials.gov IdentiÔ¨Åer: NCT01863628). The participants were Please cite this article as: Lin, K., et al., A Multi-Dimensional and Integrative Approach to Examining the High-Risk and Ultra-High-Risk Stages of Bipolar Disorder, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.027 K. Lin et al. / EBioMedicine xxx (2015) xxx‚Äìxxx offspring of parents with BP (bipolar I or bipolar II disorder), and the parents referred their offspring to the project; they were recruited from March 2013 to January 2015. Parents with BP were initially identiÔ¨Åed from the Guangzhou Brain hospital (a tertiary national hospital) ‚Äì the earliest psychiatric hospital in China ‚Äì where they received psychiatric services and from population-based registers. The study was approved by the Institutional Review Board of the Guangzhou Brain Hospital. Written informed consent was obtained from all participants and their parents when appropriate (i.e., age b 18 years). 2.2. Procedures Parents with BP with offspring aged 8‚Äì28 years lacking a diagnosis of any psychiatric disorder were interviewed; the diagnosis of proband BP was based on the Structured Clinical Interview for DSM-IV-TR Axis I Disorders, Patient Edition (SCID-I/P) and all available medical information, including blood and brain-imaging tests and treatment history. The selected age range for the offspring is important for testing our hypotheses, given that i) the typical onset of BP is in adolescence or early adulthood, ii) some individuals with bipolar disorder can manifest symptoms as early as 8 years old, long before onset (Akiskal et al., 1985; Duffy et al., 2014; Mesman et al., 2013), and iii) the period is developmentally crucial for structural maturation (Paus et al., 2008). The offspring underwent in-depth and systematic assessments and were prospectively followed up at several time points (scheduled at weeks 1, 2, 4, 8, and 12 for the UHR offspring) during the Ô¨Årst 3 months and then yearly (up to 2 years) or anytime there were changes in symptoms. In this process, we Ô¨Årst applied a self-made, 74-item symptom checklist (which consisted of a constellation of symptoms/signs encompassing the commonly reported prodromal symptoms/signs and the deÔ¨Åning symptoms/signs by the inclusion criteria (below)) to assess current symptoms and a modiÔ¨Åed retrospective instrument based on the Bipolar Prodrome Symptom Scale-Retrospective: Patient Version (BPSS-RPt) (Correll et al., 2007) to assess past symptoms. Second, the offspring aged b18 years were assessed using the Schedule for Affective Disorders and Schizophrenia for School-Aged Children: Present and Lifetime Version (K-SADS-PL) or the SCID-I/P for offspring N 18 years to exclude any psychiatric disorders. Third, the Hamilton Depression Rating Scale (HAM-D), the Young Mania Rating Scale (YMRS), and the Brief Psychiatric Rating Scale (BPRS) were implemented to assess the extent of depression, mania, and psychotic symptoms, respectively. Finally, the Global Assessment of Functioning Scale (GAF) was used to assess general function, and a family history of psychiatric disorders was conÔ¨Årmed using the Family Interview for Genetic Studies (FIGS). Of the 747 offspring with parents with BP screened by research psychiatrists, 554 were not eligible according to the inclusion and exclusion criteria (e.g., having a psychiatric diagnosis, in treatment, or too young/old), 130 declined further participation, and 19 offspring did not complete the MRI data collections. The resultant 44 offspring were included in this study, with 18 offspring deÔ¨Åned as UHR by the operational criteria we tentatively proposed (below). Given that non-speciÔ¨Åc symptoms are not uncommonly endorsed by high-risk offspring and that they are usually transient and can be contingent, we subsumed stage 0 and stage 1a into the HR stage, which is in agreement with the clinical staging model for psychotic and severe mood disorders (Scott et al., 2013). Thirty-three age-matched healthy controls who were free of any psychiatric disorders and had no family history of psychiatric disorders were recruited by advertisement and word of mouth. They were systematically assessed in a similar fashion as the high-risk offspring. 2.3. Inclusion and Exclusion Criteria The criteria for the UHR offspring were i) having at least one biological parent with bipolar disorder (bipolar I and bipolar II) and ii) manifesting at least one of the following deÔ¨Åning syndromes: 1) two (or more) hypomania symptoms lasting at least 4 days but not meeting 3 DSM-IV hypomanic episode criteria; 2) two (or more) major depressive symptoms lasting at least 1 week but falling short of a major depressive episode; 3) one (or more) of the following attenuated psychotic symptoms lasting at least 10 min for each manifestation and 2‚Äì7 manifestations per week for at least 3 months ‚Äî ideas of reference, odd ideas, odd beliefs, unusual perceptual experiences, bizarre thoughts or speech, grandiosity, suspicious ideas, paranoid ideas, odd mannerisms, hallucinations, disorganized/catatonic behaviors; and 4) two (or more) of the hyperactivity and impulsivity symptoms/signs deÔ¨Åned by the DSM-IV for attention deÔ¨Åcit hyperactivity disorder (ADHD) that were observable by teachers, peers, and/or parents. The construction of this UHR criteria referred to the clinical staging model for psychotic disorder and severe mood disorder recently proposed by Scott and colleagues (Scott et al., 2013), which is based on intervention data in early psychosis (McGorry et al., 2006) and on the studies of prodromal symptoms of bipolar disorder (Bechdolf et al., 2012). According to the clinical staging model (Scott et al., 2013), in the UHR stage, high-risk individuals manifest subthreshold syndromes that fall short of criteria for a psychiatric disorder. Thus, this study considered any established psychiatric disorders as exclusion criteria (below). Moreover, because commonly reported ‚Äúprodromal‚Äù symptoms consist of manic and depressive symptoms and ADHD hyperactivity and impulsivity symptoms/signs (Bechdolf et al., 2012), these symptoms were included in the UHR criteria. Though attenuated psychotic symptoms were reported prior to bipolar disorder, they were less frequently presented compared to manic and depressive symptoms (Correll et al., 2014). As such, the UHR criteria required the participants to have at least one biological parent with bipolar disorder (Duffy, 2014b) to increase the speciÔ¨Åcity of the sub-syndromes in terms of progression into bipolar disorder. This requirement may be particularly helpful for the ADHD symptoms and attenuated psychotic symptoms that may lack speciÔ¨Åcity for predicting bipolar disorder. The following conditions were excluded: DSM-IV-deÔ¨Åned disorders; serious general medical illness; mental retardation; the prescription of psychoactive drugs or thyroxine; the inability to complete neuropsychological tests because of physical conditions; and drug or alcohol use. 2.4. Neuroimaging Data T1-weighted and diffusion tensor images were acquired using a 3.0 Tesla scanner (Philips, Best, Netherlands). The parameters of the T1weighted images and the processing of these images are reported in supplementary Appendix A. The parameters used to obtain the diffusion tensor images and the methods of constructing the anatomical network connectivity matrix are fully reported in Appendix B. Figure S1 describes a schematic of the brain network construction. Graph metrics were computed in MATLAB with the Brain Connectivity Toolbox (www.brain-connectivity-toolbox.net). Two node-level network measurements were applied to investigate our research questions: i) degree of a region (node), indicating the number of tracts (edges) linking it to the rest of the network; and ii) the betweenness centrality of a region, measuring the fraction of the shortest paths between any pair of regions that pass through it, which is indicative of its structural or functional importance. The small-world property is thought to reÔ¨Çect an economical trade-off between maximizing adaptive values and minimizing wiring costs (i.e., high clustering and high efÔ¨Åciency) (Bassett and Bullmore, 2006; Bullmore and Sporns, 2012). Assortativity indicates that nodes of high degrees tend to connect with each other and can reÔ¨Çect the robustness to assaults to networks that suffer from the structural integrity of the network. The mathematic definitions of all the network properties are reported in Appendix C. 2.5. Neuropsychological Performance We applied the MATRICS Consensus Cognitive Battery (MCCB), suggested by the International Society for Bipolar Disorder-Battery for Please cite this article as: Lin, K., et al., A Multi-Dimensional and Integrative Approach to Examining the High-Risk and Ultra-High-Risk Stages of Bipolar Disorder, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.027 4 K. Lin et al. / EBioMedicine xxx (2015) xxx‚Äìxxx Table 1 Demographic and clinical data in the offspring groups and healthy controls. Age at assessment, years Gender, female/male Right handedness, % Years of education HAM-D YMRS BPRS Global Assessment Scale TONI-3 HR offspring N = 26 UHR offspring N = 18 HC N = 33 F or X2 p Post hoca 17.7 (5.4) 15/11 96.2 10.1 (3.9) 0.8 (1.2) 0 (0) 18.2 (0.5) 94.9 (2.8) 26.1 (8.7) 16.2 (7.0) 9/9 94.4 7.9 (4.5) 8.0 (11.8) 4.4 (3.4) 23.9 (8.4) 79.6 (14.7) 26.3 (9.1) 15.9 (4.4) 18/15 100 10.1 (3.8) 0.3 (1.0) 0.2 (0.9) 18.4 (1.1) 93.5 (4.4) 29.9 (8.0) 0.841 0.254 1.663 2.06 1 11.721 22.724 12.725 23.654 1.818 0.435 0.881 0.435 0.135 b0.001 b0.001 b0.001 b0.001 0.170 NA NA NA NA B N A; B N C B N A; B N C B N A; B N C A N B; C N B NA Abbreviations: HC, healthy controls; HR, high-risk; UHR, ultra-high-risk; NA, not applicable; HAM-D, Hamilton Depression Rating Scale; YMRS, Young Mania Rating Scale; BPRS, Brief Psychiatric Rating Scale; and TONI-3, Test of Nonverbal Intelligence, Third Edition. Note: A = high-risk offspring; B = ultra-high-risk offspring. a The threshold for signiÔ¨Åcance was p b 0.05 with the Bonferroni correction. visual cortex (occipital region) in cognitive functioning (Cooke et al., 2015; Fears et al., 2015). Thus, these two regions were added to the mask. Age, gender, handedness, and total intracranial volume (the brain volumes had been demodulated during the processing of the T1weighted images, Supplementary Appendix A) were included as covariates. To reduce multiple statistical testing, multivariate analysis of variance (MANOVA) was Ô¨Årst applied to compare the degree (and centrality) of the regions that were identiÔ¨Åed by prior comparisons of gray matter volumes between groups across the offspring and healthy control groups. Permutation tests were further applied to test the signiÔ¨Åcantly different measures identiÔ¨Åed by the MANOVA analysis, which could be considered an attempt to control for type I error (Camargo et al., 2008). MANOVA was applied to compare neuropsychological performance among groups, with age, gender, and years of education as covariates. Mixed-effect regression models were applied by modeling membership (HR and UHR offspring) as a Ô¨Åxed factor and GM volumes as covariates to examine the relationships between the GM volumes and the cognitive deÔ¨Åcits (beta, Wald X2, 95% conÔ¨Ådence interval (CI) and p value were reported). Assessment of Neurocognition (ISBD-BANC), to assess neuropsychological performance. We emphasized processing speed and visual‚Äìspatial memory because our previous Ô¨Åndings suggested that these functions would not recover after clinical remission (Xu et al., 2012). 2.6. Statistical Analysis Analysis of variance (ANOVA) and Chi-square tests were applied for the demographic and clinical data. As in the T1-weighted image analysis, two planned comparisons ‚Äì HR offspring versus healthy controls and UHR versus HR offspring ‚Äì were conducted in the Statistical Parametric Mapping 8 software Package (SPM8; Wellcome Department of Cognitive Neurology, London, UK). A preliminary whole-brain analysis was performed using a statistical threshold of p b 0.001 and an extent threshold of k N 50. Then, regions of an a priori mask that met the threshold were analyzed with FWE-correction at the individual voxel level for the mask. The signiÔ¨Åcance threshold was set at p b 0.05. The mask was drawn using the Wake Forest University Pick Atlas (Maldjian et al., 2003), which was based on previous research on bipolar disorder (Nugent et al., 2006) and UHR psychosis (Pantelis et al., 2003), including the superior prefrontal, orbital, cingulate, superior temporal cortex and cerebellum. Moreover, different types of neuroimaging data have suggested the involvement of the parietal and occipital cortex in bipolar disorder (Cerullo et al., 2014; Fears et al., 2015; James et al., 2011; Yuksel et al., 2015). Our previous study found that patients with bipolar disorder did not recover their visual‚Äìspatial memory and processing speed after clinical remission from depression, implying that the deÔ¨Åcits in these two domains may be serving as potential neurocognitive endophenotypes (Xu et al., 2012). Furthermore, there is some evidence suggesting an association between GM volumes and processing speed in bipolar patients (Fears et al., 2015) and the role of the 3. Results 3.1. Clinical Characteristics and General Function The demographic and clinical characteristics are shown in Table 1. Of the 18 UHR offspring, 12 (66.7%) manifested manic sub-syndrome, nine (50.0%) manifested depressive sub-syndrome, Ô¨Åve (27.8%) manifested hyperactivity and impulsivity symptoms/signs, and four manifested attenuated psychotic symptoms (22.2%). The age of the Ô¨Årst symptom and the duration were on average 11.7 (SD = 5.4) years old and 4.4 (SD = Table 2 Differences in gray matter volumes for high-risk offspring versus healthy controls and ultra-high-risk offspring. Regions BA Cluster sizea MNI coordinatesb X HR offspring versus HC Right OFC Right cerebellum lobe 11 NA 90 935 UHR versus HR offspring Right SFG Right PCC Left middle and inferior OG Left SPC 10 31 31/30 7 118 118 187 59 T tests PFWE-correctedc Y Z 6 8 62 ‚àí54 ‚àí20 ‚àí6 4.10 4.41 0.023 0.029 15 15 ‚àí48 ‚àí15 59 ‚àí70 ‚àí79 ‚àí73 3 15 0 39 3.89 4.24 4.19 3.78 0.023 0.017 0.033 0.047 Abbreviations: HR, high-risk; HC, healthy controls; BA, Brodmann's areas; OFC, orbitofrontal cortex; SFG, superior frontal gyrus; PCC, posterior cingulate cortex; OG, occipital gyrus; and SPC, superior parietal cortex. a Statistics at voxel-level set to a minimum uncorrected threshold of p b 0.001, k N 50 voxels. b Montreal Neurological institute coordinates in millimeters. c Family-wise error correction for multiple comparisons. Please cite this article as: Lin, K., et al., A Multi-Dimensional and Integrative Approach to Examining the High-Risk and Ultra-High-Risk Stages of Bipolar Disorder, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.027 K. Lin et al. / EBioMedicine xxx (2015) xxx‚Äìxxx 3.6) years, respectively. Three individuals developed full-blown BP identiÔ¨Åed after follow-up. Apart from the depressive, manic, and psychotic symptoms, general function measured by the GAF was signiÔ¨Åcantly worse in the UHR offspring (n = 26) than in the HR offspring and healthy controls (n = 33) (p b 0.05, Bonferroni correction). 3.2. Voxel-Based Morphometry (VBM) Analysis Compared with the healthy controls, the HR offspring showed signiÔ¨Åcantly lower GM volumes in the right OFC and the right cerebellum. Compared to the HR offspring, the UHR offspring had signiÔ¨Åcantly greater GM volumes in four regions, including the right superior frontal gyrus, posterior cingulate cortex, left parietal cortex, and the right middle and inferior occipital gyrus (Table 2 and Fig. 1, FWE-corrected p b 0.05). 3.3. Regional-Level and Whole-Brain Network Analysis For the abnormal regions that were identiÔ¨Åed in the prior VBM analysis, we examined the degree and centrality of these regions, which reÔ¨Çects the connection (of a region to the rest) and the importance of a region in the brain network, respectively. Among the Ô¨Åve regions, the MANOVA analysis revealed that the degree of the left inferior occipital gyrus was signiÔ¨Åcantly different across the three groups (F = 5.541, df = 2,68, p = 0.006). The centrality of this region was also signiÔ¨Åcantly different across the three group (F = 3.22, df = 2,68, p = 0.047) (Figure S2). The permutation tests (the permutation distribution is shown in the supplementary materials) further revealed that both the 5 HR and UHR offspring had signiÔ¨Åcantly fewer degrees of this region than the healthy controls (p = 0.003 and p = 0.03, respectively), thus indicating fewer connections from this region to other regions. The centrality of the region was signiÔ¨Åcantly lower in the HR offspring than in the healthy controls (p = 0.021), whereas the UHR offspring showed a similar trend, although it did not reach signiÔ¨Åcance (p = 0.07). A secondary analysis was conducted to examine the centrality of the anterior cingulate cortex (ACC) in the offspring, as abnormalities in this region have been reported in individuals who met the criteria of ‚Äúat-risk mental state‚Äù (ARMS) (Lord et al., 2011). As shown in Table S1, both the HR and UHR offspring displayed signiÔ¨Åcantly lower centrality in comparison to healthy controls (permutation tests, p b 0.05). The centrality of the ACC was signiÔ¨Åcantly lower in the UHR offspring than that in the HR offspring (permutation test, p = 0.006). In the global-level network analysis (Table 3), the UHR offspring displayed a signiÔ¨Åcantly higher small-world property than the HR offspring (p = 0.030). As for assortativity, the UHR offspring displayed signiÔ¨Åcantly lower values than the healthy controls (p = 0.020), which is suggestive of vulnerability, whereas the HR offspring did not display signiÔ¨Åcant differences from the healthy controls (p = 0.189). 3.4. Cognitive Function and Correlation Analysis As shown in Table 3, compared to the healthy controls, the MANOVA revealed that the UHR offspring displayed deÔ¨Åcits in both processing speed and visual-working memory (F = 6.109, df = 1,50, p = 0.017; F = 4.669, df = 1,50, p = 0.036, respectively). The HR offspring Fig. 1. (a) Turquoise and pink represent the regions showing signiÔ¨Åcant changes in gray matter volumes in the high-risk offspring versus healthy controls and in the high-risk versus ultrahigh-risk offspring, respectively. (b) The regions in which the high-risk offspring had signiÔ¨Åcantly lower gray matter volumes than the healthy controls. (c) and (d) The regions in which the ultra-high-risk offspring displayed signiÔ¨Åcantly higher gray matter volumes than the high-risk offspring. Please cite this article as: Lin, K., et al., A Multi-Dimensional and Integrative Approach to Examining the High-Risk and Ultra-High-Risk Stages of Bipolar Disorder, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.027 6 K. Lin et al. / EBioMedicine xxx (2015) xxx‚Äìxxx Table 3 Network properties of neuropsychological performance in the high-risk offspring, ultra-high-risk offspring, and healthy controls. Groups Contrasts HR UHR HC Small-world property (SD) 3.11 (0.30) 3.38 (0.50) 3.22 (0.44) Assortativityb (median) 0.03 (0.03) 0.01 (0.02) 0.05 (0.04) Permutation testsa p value HR versus HC UHR versus HC HR versus UHR HR versus HC UHR versus HC HR versus UHR 0.836 0.875 0.030 0.190 0.020 0.828 MANOVA Processing speedc 60.7 (12.3) 51.7 (10.5) 61.7 (8.7) Visual‚Äìspatial memoryd 25.0 (5.8) 23.2 (5.6) 27.0 (5.1) HR versus HC UHR versus HC HR versus HC UHR versus HC F(1,50) p value 0.261 6.109 0.087 4.669 0.612 0.017 0.769 0.036 Abbreviations: HR, high-risk; UHR, ultra-high-risk; HC, healthy controls; SD, standard deviation; and MANOVA, Multivariate Analysis of Variance. Notes: a Number of resamples = 5000; the distribution of mean differences is shown in the supplementary materials. b Shown as the mean (median); other data: mean (SD). c Measured by the Brief Assessment of Cognition in Schizophrenia: Symbol Coding. d Measured by the Brief Visuospatial Memory Test. were not signiÔ¨Åcantly impaired in either processing speed or visualworking memory (F = 0.261, df = 1,59, p = 0.612; F = 0.087, p = 0.769, respectively). The mixed-effect regression models found that processing speed was signiÔ¨Åcantly related to the GM volumes of the three regions, including the right posterior cingulate cortex (p = 0.005), right superior frontal gyrus (p = 0.011), and left superior parietal cortex (p = 0.027). There were no signiÔ¨Åcant relationships between visual‚Äìspatial memory and the GM volumes of the regions (p N 0.05) (Table S2). 4. Discussion This study provides preliminary evidence that some pathophysiological changes (i.e., GM volumes and brain network measures) may become apparent in certain early stages (i.e., the HR and UHR stages) preceding the ofÔ¨Åcial onset of bipolar disorder. Some results were unexpected. For example, instead of decreased GM volumes, the UHR versus HR offspring displayed increased GM volumes in some regions. UHR versus HR offspring showed an upward-shifted small-world property that reÔ¨Çects high clustering and short path lengths (discussed below). At the macroscopic level, our data suggest that deÔ¨Åcits in processing speed and visual‚Äìspatial memory may exist in the UHR stage but not in the HR stage. Moreover, the deÔ¨Åcits in processing speed may be related to the GM volumes in the right superior frontal gyrus, posterior cingulate cortex, and the left superior parietal cortex. 4.1. Changes in GM Volumes In terms of structural abnormalities, the HR offspring showed a volumetric reduction in the right OFC and the right cerebellum. The OFC, the most inferior and ventral part of the prefrontal cortex, has been found to be involved in set shifting and reversal learning (Rygula et al., 2010), encoding reward values in reward processing (Grabenhorst and Rolls, 2011), decision-making processes (Bechara et al., 2000), and emotion processing (Liu et al., 2012), which may underlie certain characteristics of BP, such as mood lability, mood dysregulation, and reward sensitivity (Phillips and Swartz, 2014). Decreased OFC volumes have been reported in pediatric and adult BP patients (James et al., 2011; StanÔ¨Åeld et al., 2009), but a reduction in this region has also been reported to relate to depressive symptoms (Nery et al., 2009). By minimizing the effects of symptoms, this study indicates that the decreased volumes in the two regions may represent genetic susceptibility. In addition, the connections of the OFC with other regions might not be impaired in the HR offspring in terms of the degree of this region within the brain networks. Orbitofrontal reduction has also been observed in those individuals with prodromal symptoms of psychosis who were subsequently predisposed to psychosis (mainly schizophrenia and BP with psychotic features) (Pantelis et al., 2003). In the UHR versus HR offspring, there were increased GM volumes in several regions (most of which were located in the parietal and occipital cortex), suggestive of stage-speciÔ¨Åc changes that may predispose subjects to the full development of BP. This Ô¨Ånding is in agreement with research on UHR psychosis, which shows that volumetric changes in prefrontal regions are related to genetic susceptibility, whereas volumetric changes in the parietal and temporal regions are related to the transition stage into psychotic disorder (the stage in which attenuated psychotic symptoms are manifested) (Lawrie et al., 1999; Pantelis et al., 2003). This Ô¨Ånding also coincides with morphometric studies reporting that the structural abnormalities in unaffected relatives of BP proband were relatively restricted to the prefrontal cortex, whereas the widespread volumetric changes identiÔ¨Åed in pediatric BP extended to the temporal, occipital and parietal cortexes and the amygdala, some of which were associated with symptoms (Chang et al., 2005; Frazier et al., 2005; Hajek et al., 2013; Matsuo et al., 2012; Nery et al., 2009). The increased GM volumes observed in the UHR offspring are unlikely to be protective factors because the selected age range for the offspring represents the most at-risk time for the onset of full-blown BP. Moreover, the UHR offspring manifested varying subsyndromes, and some of them received a diagnosis of full-blown BP during follow-up. Furthermore, we found that there was an inverse relationship between the GM volumes from these regions and cognitive function. During normative brain maturation from childhood to post-adolescence, gray matter reductions are found in these regions, including in the superior frontal, parietal and occipital regions. Such gray matter reductions are associated with increased brain growth (Sowell et al., 2001). As such, the observation of increased GM volumes in the UHR offspring might indicate disruptions in the process of brain maturation that may predispose an individual to developing BP. To test this speculation, we conducted a complementary analysis to examine differences in the relationship of age to GM volumes in the four regions between the HR and UHR offspring. Our speculation is supported by the results of this analysis, which showed that there were differential relationships between age and GM volumes across the HR and UHR offspring groups (supplementary Figure S7). In addition, the increased volumes in UHR offspring may relate to the phenomenon of ‚Äúallostasis,‚Äù which refers to the body's reactions to repeated adverse physical or psychosocial conditions that cause stress (McEwen, 2000). Allostasis is fundamentally crucial for adaptation, the Please cite this article as: Lin, K., et al., A Multi-Dimensional and Integrative Approach to Examining the High-Risk and Ultra-High-Risk Stages of Bipolar Disorder, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.027 K. Lin et al. / EBioMedicine xxx (2015) xxx‚Äìxxx maintenance of homeostasis and survival in the short term, which might manifest as increased GM changes. However, over longer intervals, prolonged reactions/over-compensations may lead to neurotoxicity effects; such neurotoxicity effects could manifest as brain structure atrophy (Drevets et al., 1997; Kasai et al., 2003; Moorhead et al., 2007), as reported in individuals at UHR for psychosis over a period of 12 months or longer (Pantelis et al., 2003). 4.2. Brain Network Properties At the Ô¨Åner-grain level, we observed that both the HR and UHR offspring displayed signiÔ¨Åcantly fewer degrees of the inferior occipital region. This region is not critical for emotion processing and regulation, dysfunction of which may underlie the clinical characteristics of BP (Phillips and Swartz, 2014), e.g., mood instability and manic symptoms. Rather, the occipital cortex plays a key role in sensory function and visual‚Äìspatial memory in particular (Cooke et al., 2015; Fears et al., 2015). We performed an exploratory analysis (Pearson correlation) showing that the degree of the inferior occipital cortex was positively associated with processing speed and visual‚Äìspatial memory in the healthy controls (p b 0.05). The Ô¨Ånding of fewer degrees of the inferior occipital regions is consistent with the observation of deÔ¨Åcits in processing speed and visual‚Äìspatial memory in the UHR offspring. This observation seems to support the notion that abnormalities of the occipital region may be related to the impairments of cognitive function in BP (Fears et al., 2015; Lyoo et al., 2006). Additionally, it is possible that the involvement of the inferior occipital region reÔ¨Çects the high vulnerability to developing severe forms of bipolar disorder because the abnormalities of this site are relatively common in psychotic bipolar disorder, schizoaffective disorder, (Ivleva et al., 2013), and poor-outcome BP (Bonne et al., 1996). One study on schizophrenia found signiÔ¨Åcantly changed degrees in some regions, including the inferior occipital cortex (Bassett et al., 2008). Moreover, the ARMS criteria were deÔ¨Åned by Yung et al. using the UHR criteria for the onset of psychotic disorders (Yung et al., 2005). Lord et al. found that the centrality of this site was associated with the severity of psychotic symptoms in ARMS individuals (Lord et al., 2011). Using the present UHR criteria (which were more speciÔ¨Åc to BP), this study found decreased centrality in the ACC in UHR versus HR offspring, implying that the centrality of the ACC may be related to the severity of symptoms that are not limited to psychotic symptoms. Small-world properties, which are characterized by high clustering and short path lengths (high efÔ¨Åciency), support rapid synchronization and information transfer across spatially separated regions (Bullmore and Sporns, 2012). The development of a small-world topology is a dynamic process, evolving from early brain development (b 2 years) (Fan et al., 2011) and demonstrating more long-distance connections during late childhood and adolescence (Chen et al., 2013). Longdistance connections facilitate information processing across regions (high efÔ¨Åciency) and lead to shorter path lengths between spatially separated regions (i.e., the path length is shorter compared to when those regions were connected via a number of shorter path-connecting nodes in between). Research has shown that brain networks increase efÔ¨Åciency by conÔ¨Åguring more long-distance connections while performing effortful cognitive tasks (Kitzbichler et al., 2011). Thus, the increased incidence of small-world topology in UHR offspring may be tentatively interpreted as a compensation for impaired cognitive function by conÔ¨Åguring more long-distance connections in the brain network. However, these long-distances connections, which likely integrate ‚Äúhub‚Äù regions, are expensive in terms of the wiring and metabolic costs. They are subject to selective attack in the networks during pathophysiological processes (Bullmore and Sporns, 2012). In what may be a tradeoff between adaptive value and wiring cost, psychiatric disorders such as schizophrenia have been reported to be associated with abnormal shifts in small-world properties (Bassett et al., 2008; Bullmore and Sporns, 2012; Liu et al., 2008). Alternatively, synaptic pruning (which 7 results in the reduction of Ô¨Åber connections and gray matter density) is known to occur during the normative development of the brain from childhood through adolescence (Penzes et al., 2011); thus, the upwards-shifted small world topology in the UHR offspring may be due to disruptions in synaptic pruning that can lead to brain networks of higher clustering and efÔ¨Åciency. Whole-brain network properties help to uncover dysfunction in an entire system comprised of interconnected regions, providing more comprehensive information than just a focus on individual regions (Bullmore and Sporns, 2009). Assortativity was able to capture the clinically deÔ¨Åned UHR stage, which is a critical stage prior to ofÔ¨Åcial onset that includes varying sub-syndromes. 4.3. Neuropsychological Performance and the Neural Correlates At the cognitive level, our Ô¨Ånding is that processing speed and visual‚Äìspatial memory may not be impaired at the HR stage; however, as the disease progresses further into the UHR stage, deÔ¨Åcits in these domains can become apparent. Although emerging before ofÔ¨Åcial onset, genetically high-risk individuals who manifest no or mild, nonspeciÔ¨Åc symptoms may have intact cognitive functioning in these domains. Our Ô¨Ånding is at odds with previous Ô¨Åndings that suggested cognitive deÔ¨Åcits were inherited abnormalities in bipolar disorder. These studies examined neuropsychological performance in the ‚Äúunaffected‚Äù relatives of individuals with BP (Drysdale et al., 2013; Kulkarni et al., 2010). Thus, the discrepancy probably lies in the deÔ¨Ånition of ‚Äúunaffected‚Äù. By applying the clinical staging model, this study allowed the further subdivision of the unaffected status into the HR and UHR stages. In the HR stage, the impact of subthreshold symptoms on neuropsychological performance could be minimized. Indeed, previous research has shown that cognitive deÔ¨Åcits are not characteristic of genetically high-risk, unaffected offspring (i.e., those with intact visual information processing) (Duffy et al., 2009) or of individuals later diagnosed with BP (both high and low academic performance was associated with risk for bipolar disorder) (MacCabe et al., 2010) but were apparent in those high-risk individuals who later developed psychotic affective disorder and schizophrenia (Cannon et al., 2008). Moreover, we observed a negative correlation between processing speed performance and the GM volumes of the regions in which the UHR offspring displayed abnormalities, including the right posterior cingulate cortex, superior frontal gyrus, and left superior parietal cortex. This result resonates with the observation that UHR offspring exhibited increased GM volumes and deÔ¨Åcits in processing speed. The GM volumes in the parietal cortex have been reported to be negatively correlated with executive function in bipolar patients (Haldane et al., 2008). However, a recent multi-generational family study of bipolar disorder by Fears et al. reported that the GM volume of the posterior cingulate cortex was positively correlated with processing speed in bipolar patients, although the correlation did not survive the FDR-correction (Fears et al., 2015). They also found a positive correlation between the GM volumes in both the post cingulate and superior parietal cortex and visual‚Äìspatial memory. This observation is not supported by our study. Although the etiology of the GM change is unknown, factors such as age and affected status may contribute to the discrepancy. 4.4. Limitations The following limitations must be noted for the interpretation of these results. First, the follow-up in this study was relatively short (up to 2.5 years), and only 17% of the UHR offspring have been conÔ¨Årmed to develop bipolar disorder by follow-up. Additionally, the sample size was relatively small. These Ô¨Åndings need to be replicated in large, independent samples. Moreover, given that certain psychiatric disorders (e.g., substance use disorder and major depression) can precede the onset of bipolar disorder (Duffy et al., 2014) and this study Please cite this article as: Lin, K., et al., A Multi-Dimensional and Integrative Approach to Examining the High-Risk and Ultra-High-Risk Stages of Bipolar Disorder, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.027 8 K. Lin et al. / EBioMedicine xxx (2015) xxx‚Äìxxx excluded DSM-IV diagnosable disorders, the Ô¨Åndings of this study may not be generalizable to other populations in which psychiatric disorders are comorbid with or precede the sub-syndromes speciÔ¨Åed in the UHR criteria (e.g., subthreshold depression). The UHR criteria for bipolar disorder need to be reÔ¨Åned to better represent the full spectrum of symptoms. Furthermore, the manic sub-syndrome duration requirement (at least 4 days) may be conservative. Offspring with manic subsyndromes may already be in a relatively ‚Äúlate‚Äù prodromal stage of bipolar disorder. The deÔ¨Ånition of bipolar disorder not otherwise speciÔ¨Åed (bipolar NOS) by DSM-IV is vague. Applying a duration of at least 4 days, the Course and Outcome of Bipolar Youth (COBY) study found that youths with bipolar NOS had a high rate of conversion into bipolar I or bipolar II disorder (29% in two years and 38% in four years) (Birmaher et al., 2006, 2009). Finally, the etiology of bipolar disorder may be heterogeneous (Duffy, 2014b). The current clinical staging model may be biased towards severe cases of bipolar disorder, in which a proportion of the UHR offspring displayed attenuated psychotic symptoms (the symptoms were partly overlapping with the UHR criteria for psychosis despite genetic susceptibility for BP). Thus, the Ô¨Åndings of this study may not be generalizable to other high-risk populations. Declaration of Interests We declare no competing interests. Acknowledgments The project was funded by the National Natural Science Foundation of China (NSFC, 81471375); the Key Medical discipline of Guangzhou, China (GBH2014-ZD04); the Science and Technology Planning Project of Guangdong Province, China (2011B031800154); and a Research Grant Council Humanities and Social Sciences Prestigious Fellowship (HKU703-HSS-13). Role of the Funding Source The funders of the study had no role in study design, data collection, data analysis, data interpretation, or the writing of the report. All authors had access to the data, and all authors agreed to submit the paper for publication. Appendix A. Supplementary Data 4.5. Conclusions Our data suggest that the underlying abnormalities of BP may become apparent long before the ofÔ¨Åcial onset of BP. SpeciÔ¨Åcally, the decreased GM volumes in the OFC observed in the HR offspring may serve as neuroanatomical endophenotypes. In this stage, genetically high-risk offspring may not yet have cognitive deÔ¨Åcits in processing speed and visual memory. In the next stage, UHR offspring may display increased GM volumes in the right posterior cingulate cortex, superior frontal gyrus, and left superior parietal cortex, which may partly explain the cognitive deÔ¨Åcits found in UHR offspring. Although some network properties, such as small-world property, were signiÔ¨Åcantly different between the HR and UHR offspring, the nature of these changes are currently hard to interpret, and these Ô¨Åndings need to be replicated. We tentatively speculate that some changes may be related to compensatory reactions, which may be worthy of future investigation. Clinical staging models are scientiÔ¨Åcally and clinically important, and they emphasize progression from a very early stage prior to the subsyndromes. At this stage, common confounding effects (including symptoms, illness duration and medications) are minimized, and the possibility of investigating the underlying pathophysiology is maximized. More importantly, identifying abnormalities prior to full-blown onset opens up the possibility of putting into practice early intervention (i.e., when cognitive deÔ¨Åcits are observed in the UHR stage) or even primary intervention strategies (i.e., when intact cognitive function but disrupted brain connections are observed in the HR stage), of particular importance because BP is currently considered irremediable and becomes progressively worse. A range of early interventions that show preliminary evidence of effectiveness, including sleep hygiene, nutraceuticals (e.g., omega-3 fatty acids), mindfulness, emotion regulation strategies, and anti-inÔ¨Çammatory agents (Duffy, 2014a), might be justiÔ¨Åed for offspring at familiar risk and those in the UHR stage in particular. Contributors TMCL and GX conceptualized the project. KL and GX prepared the original protocol with later input from KFS and TMCL. TMCL and KFS supervised the project's overall undertaking. KL, HW, TL, WL, KC, XC, BL, and LZ implemented the project. KL and NMW managed the data and performed the data analysis. KL, NMW, WL, and XC constructed the Ô¨Ågures. KL drafted the manuscript. GX, KFS, and TMCL provided critical comments on the Ô¨Ånal version of the manuscript. Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ebiom.2015.06.027. References Akiskal, H.S., Downs, J., Jordan, P., Watson, S., Daugherty, D., Pruitt, D.B., 1985. Affective disorders in referred children and younger siblings of manic-depressives. Mode of onset and prospective course. Arch. Gen. Psychiatry 42, 996‚Äì1003. Bassett, D.S., Bullmore, E., 2006. Small-world brain networks. Neuroscientist 12, 512‚Äì523. Bassett, D.S., Bullmore, E., Verchinski, B.A., Mattay, V.S., Weinberger, D.R., MeyerLindenberg, A., 2008. Hierarchical organization of human cortical networks in health and schizophrenia. J. Neurosci. 28, 9239‚Äì9248. Bechara, A., Damasio, H., Damasio, A.R., 2000. Emotion, decision making and the orbitofrontal cortex. Cereb. Cortex 10, 295‚Äì307. Bechdolf, A., Ratheesh, A., Wood, S.J., Tecic, T., Conus, P., Nelson, B., Cotton, S.M., Chanen, A.M., Amminger, G.P., Ruhrmann, S., Schultze-Lutter, F., Klosterkotter, J., Fusar Poli, P., Yung, A.R., Berk, M., McGorry, P.D., 2012. Rationale and Ô¨Årst results of developing atrisk (prodromal) criteria for bipolar disorder. Curr. Pharm. Des. 18, 358‚Äì375. Birmaher, B., Axelson, D., Strober, M., Gill, M.K., Valeri, S., Chiappetta, L., Ryan, N., Leonard, H., Hunt, J., Iyengar, S., Keller, M., 2006. Clinical course of children and adolescents with bipolar spectrum disorders. Arch. Gen. Psychiatry 63, 175‚Äì183. Birmaher, B., Axelson, D., Goldstein, B., Strober, M., Gill, M.K., Hunt, J., Houck, P., Ha, W., Iyengar, S., Kim, E., Yen, S., Hower, H., Esposito-Smythers, C., Goldstein, T., Ryan, N., Keller, M., 2009. Four-year longitudinal course of children and adolescents with bipolar spectrum disorders: the Course and Outcome of Bipolar Youth (COBY) study. Am. J. Psychiatry 166, 795‚Äì804. Bonne, O., Krausz, Y., GorÔ¨Åne, M., Karger, H., GelÔ¨Ån, Y., Shapira, B., Chisin, R., Lerer, B., 1996. Cerebral hypoperfusion in medication resistant, depressed patients assessed by Tc99m HMPAO SPECT. J. Affect. Disord. 41, 163‚Äì171. Bullmore, E., Sporns, O., 2009. Complex brain networks: graph theoretical analysis of structural and functional systems. Nat. Rev. Neurosci. 10, 186‚Äì198. Bullmore, E., Sporns, O., 2012. The economy of brain network organization. Nat. Rev. Neurosci. 13, 336‚Äì349. Camargo, A., Azuaje, F., Wang, H., Zheng, H., 2008. Permutation ‚Äî based statistical tests for multiple hypotheses. Source Code Biol. Med. 3, 15. Cannon, T.D., Cadenhead, K., Cornblatt, B., Woods, S.W., Addington, J., Walker, E., Seidman, L.J., Perkins, D., Tsuang, M., McGlashan, T., Heinssen, R., 2008. Prediction of psychosis in youth at high clinical risk: a multisite longitudinal study in North America. Arch. Gen. Psychiatry 65, 28‚Äì37. Catani, M., Ffytche, D.H., 2005. The rises and falls of disconnection syndromes. Brain 128, 2224‚Äì2239. Cerullo, M.A., Eliassen, J.C., Smith, C.T., Fleck, D.E., Nelson, E.B., Strawn, J.R., Lamy, M., DelBello, M.P., Adler, C.M., Strakowski, S.M., 2014. Bipolar I disorder and major depressive disorder show similar brain activation during depression. Bipolar Disord. 16, 703‚Äì712. Chang, K., Karchemskiy, A., Barnea-Goraly, N., Garrett, A., Simeonova, D.I., Reiss, A., 2005. Reduced amygdalar gray matter volume in familial pediatric bipolar disorder. J. Am. Acad. Child Adolesc. Psychiatry 44, 565‚Äì573. Chen, Z., Liu, M., Gross, D.W., Beaulieu, C., 2013. Graph theoretical analysis of developmental patterns of the white matter network. Front. Hum. Neurosci. 7, 716. Cooke, S.F., Komorowski, R.W., Kaplan, E.S., Gavornik, J.P., Bear, M.F., 2015. Visual recognition memory, manifested as long-term habituation, requires synaptic plasticity in V1. Nat. Neurosci. 18, 262‚Äì271. Correll, C.U., Penzner, J.B., Frederickson, A.M., Richter, J.J., Auther, A.M., Smith, C.W., Kane, J.M., Cornblatt, B.A., 2007. Differentiation in the preonset phases of schizophrenia and Please cite this article as: Lin, K., et al., A Multi-Dimensional and Integrative Approach to Examining the High-Risk and Ultra-High-Risk Stages of Bipolar Disorder, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.027 K. Lin et al. / EBioMedicine xxx (2015) xxx‚Äìxxx mood disorders: evidence in support of a bipolar mania prodrome. Schizophr. Bull. 33, 703‚Äì714. Correll, C.U., Hauser, M., Penzner, J.B., Auther, A.M., Kafantaris, V., Saito, E., Olvet, D., Carrion, R.E., Birmaher, B., Chang, K.D., DelBello, M.P., Singh, M.K., Pavuluri, M., Cornblatt, B.A., 2014. Type and duration of subsyndromal symptoms in youth with bipolar I disorder prior to their Ô¨Årst manic episode. Bipolar Disord. 16, 478‚Äì492. Drevets, W.C., Price, J.L., Simpson Jr., J.R., Todd, R.D., Reich, T., Vannier, M., Raichle, M.E., 1997. Subgenual prefrontal cortex abnormalities in mood disorders. Nature 386, 824‚Äì827. Drysdale, E., Knight, H.M., McIntosh, A.M., Blackwood, D.H., 2013. Cognitive endophenotypes in a family with bipolar disorder with a risk locus on chromosome 4. Bipolar Disord. 15, 215‚Äì222. Duffy, A., 2014a. Interventions for youth at risk of bipolar disorder. Curr. Treat. Options Psychiatry 37‚Äì47. Duffy, A., 2014b. Toward a comprehensive clinical staging model for bipolar disorder: integrating the evidence. Can. J. Psychiatry 59, 659‚Äì666. Duffy, A., Hajek, T., Alda, M., Grof, P., Milin, R., MacQueen, G., 2009. Neurocognitive functioning in the early stages of bipolar disorder: visual backward masking performance in high risk subjects. Eur. Arch. Psychiatry Clin. Neurosci. 259, 263‚Äì269. Duffy, A., Horrocks, J., Doucette, S., Keown-Stoneman, C., McCloskey, S., Grof, P., 2014. The developmental trajectory of bipolar disorder. Br. J. Psychiatry 204, 122‚Äì128. Egeland, J.A., Hostetter, A.M., Pauls, D.L., Sussex, J.N., 2000. Prodromal symptoms before onset of manic-depressive disorder suggested by Ô¨Årst hospital admission histories. J. Am. Acad. Child Adolesc. Psychiatry 39, 1245‚Äì1252. Fan, Y., Shi, F., Smith, J.K., Lin, W., Gilmore, J.H., Shen, D., 2011. Brain anatomical networks in early human brain development. Neuroimage 54, 1862‚Äì1871. Fears, S.C., Schur, R., Sjouwerman, R., Service, S.K., Araya, C., Araya, X., Bejarano, J., Knowles, E., Gomez-Makhinson, J., Lopez, M.C., Aldana, I., Teshiba, T.M., Abaryan, Z., Al-Sharif, N.B., Navarro, L., Tishler, T.A., Altshuler, L., Bartzokis, G., Escobar, J.I., Glahn, D.C., Thompson, P.M., Lopez-Jaramillo, C., Macaya, G., Molina, J., Reus, V.I., Sabatti, C., Cantor, R.M., Freimer, N.B., Bearden, C.E., 2015. Brain structure‚Äìfunction associations in multi-generational families genetically enriched for bipolar disorder. Brain 138, 2087‚Äì2102. Ferrier, I.N., Chowdhury, R., Thompson, J.M., Watson, S., Young, A.H., 2004. Neurocognitive function in unaffected Ô¨Årst-degree relatives of patients with bipolar disorder: a preliminary report. Bipolar Disord. 6, 319‚Äì322. Frank, E., Nimgaonkar, V.L., Phillips, M.L., Kupfer, D.J., 2014. All the world's a (clinical) stage: rethinking bipolar disorder from a longitudinal perspective. Mol. Psychiatry 20, 23‚Äì31. Frazier, J.A., Breeze, J.L., Makris, N., Giuliano, A.S., Herbert, M.R., Seidman, L., Biederman, J., Hodge, S.M., Dieterich, M.E., Gerstein, E.D., Kennedy, D.N., Rauch, S.L., Cohen, B.M., Caviness, V.S., 2005. Cortical gray matter differences identiÔ¨Åed by structural magnetic resonance imaging in pediatric bipolar disorder. Bipolar Disord. 7, 555‚Äì569. Grabenhorst, F., Rolls, E.T., 2011. Value, pleasure and choice in the ventral prefrontal cortex. Trends Cogn. Sci. 15, 56‚Äì67. Hajek, T., Cullis, J., Novak, T., Kopecek, M., Blagdon, R., Propper, L., Stopkova, P., Duffy, A., Hoschl, C., Uher, R., Paus, T., Young, L.T., Alda, M., 2013. Brain structural signature of familial predisposition for bipolar disorder: replicable evidence for involvement of the right inferior frontal gyrus. Biol. Psychiatry 73, 144‚Äì152. Haldane, M., Cunningham, G., Androutsos, C., Frangou, S., 2008. Structural brain correlates of response inhibition in Bipolar Disorder I. J. Psychopharmacol. 22, 138‚Äì143. Ivleva, E.I., Bidesi, A.S., Keshavan, M.S., Pearlson, G.D., Meda, S.A., Dodig, D., Moates, A.F., Lu, H., Francis, A.N., Tandon, N., Schretlen, D.J., Sweeney, J.A., Clementz, B.A., Tamminga, C.A., 2013. Gray matter volume as an intermediate phenotype for psychosis: Bipolar-Schizophrenia Network on Intermediate Phenotypes (B-SNIP). Am. J. Psychiatry 170, 1285‚Äì1296. James, A., Hough, M., James, S., Burge, L., Winmill, L., Nijhawan, S., Matthews, P.M., Zarei, M., 2011. Structural brain and neuropsychometric changes associated with pediatric bipolar disorder with psychosis. Bipolar Disord. 13, 16‚Äì27. Kalmar, J.H., Wang, F., Spencer, L., Edmiston, E., Lacadie, C.M., Martin, A., Constable, R.T., Duncan, J.S., Staib, L.H., Papademetris, X., Blumberg, H.P., 2009. Preliminary evidence for progressive prefrontal abnormalities in adolescents and young adults with bipolar disorder. J. Int. Neuropsychol. Soc. 15, 476‚Äì481. Kasai, K., Shenton, M.E., Salisbury, D.F., Hirayasu, Y., Lee, C.U., Ciszewski, A.A., YurgelunTodd, D., Kikinis, R., Jolesz, F.A., McCarley, R.W., 2003. Progressive decrease of left superior temporal gyrus gray matter volume in patients with Ô¨Årst-episode schizophrenia. Am. J. Psychiatry 160, 156‚Äì164. Kitzbichler, M.G., Henson, R.N., Smith, M.L., Nathan, P.J., Bullmore, E.T., 2011. Cognitive effort drives workspace conÔ¨Åguration of human brain functional networks. J. Neurosci. 31, 8259‚Äì8270. Kulkarni, S., Jain, S., Janardhan Reddy, Y.C., Kumar, K.J., Kandavel, T., 2010. Impairment of verbal learning and memory and executive function in unaffected siblings of probands with bipolar disorder. Bipolar Disord. 12, 647‚Äì656. Lawrie, S.M., Whalley, H., Kestelman, J.N., Abukmeil, S.S., Byrne, M., Hodges, A., Rimmington, J.E., Best, J.J., Owens, D.G., Johnstone, E.C., 1999. Magnetic resonance imaging of brain in people at high risk of developing schizophrenia. Lancet 353, 30‚Äì33. Liu, Y., Liang, M., Zhou, Y., He, Y., Hao, Y., Song, M., Yu, C., Liu, H., Liu, Z., Jiang, T., 2008. Disrupted small-world networks in schizophrenia. Brain 131, 945‚Äì961. Liu, J., Blond, B.N., van Dyck, L.I., Spencer, L., Wang, F., Blumberg, H.P., 2012. Trait and state corticostriatal dysfunction in bipolar disorder during emotional face processing. Bipolar Disord. 14, 432‚Äì441. Lord, L.D., Allen, P., Expert, P., Howes, O., Lambiotte, R., McGuire, P., Bose, S.K., Hyde, S., Turkheimer, F.E., 2011. Characterization of the anterior cingulate's role in the atrisk mental state using graph theory. Neuroimage 56, 1531‚Äì1539. Lyoo, I.K., Sung, Y.H., Dager, S.R., Friedman, S.D., Lee, J.Y., Kim, S.J., Kim, N., Dunner, D.L., Renshaw, P.F., 2006. Regional cerebral cortical thinning in bipolar disorder. Bipolar Disord. 8, 65‚Äì74. 9 MacCabe, J.H., Lambe, M.P., Cnattingius, S., Sham, P.C., David, A.S., Reichenberg, A., Murray, R.M., Hultman, C.M., 2010. Excellent school performance at age 16 and risk of adult bipolar disorder: national cohort study. Br. J. Psychiatry 196, 109‚Äì115. Maldjian, J.A., Laurienti, P.J., Kraft, R.A., Burdette, J.H., 2003. An automated method for neuroanatomic and cytoarchitectonic atlas-based interrogation of fMRI data sets. Neuroimage 19, 1233‚Äì1239. Martinez-Aran, A., Vieta, E., Reinares, M., Colom, F., Torrent, C., Sanchez-Moreno, J., Benabarre, A., Goikolea, J.M., Comes, M., Salamero, M., 2004. Cognitive function across manic or hypomanic, depressed, and euthymic states in bipolar disorder. Am. J. Psychiatry 161, 262‚Äì270. Matsuo, K., Kopecek, M., Nicoletti, M.A., Hatch, J.P., Watanabe, Y., Nery, F.G., Zunta-Soares, G., Soares, J.C., 2012. New structural brain imaging endophenotype in bipolar disorder. Mol. Psychiatry 17, 412‚Äì420. McEwen, B.S., 2000. Allostasis and allostatic load: implications for neuropsychopharmacology. Neuropsychopharmacology 22, 108‚Äì124. McGorry, P.D., Hickie, I.B., Yung, A.R., Pantelis, C., Jackson, H.J., 2006. Clinical staging of psychiatric disorders: a heuristic framework for choosing earlier, safer and more effective interventions. Aust. N. Z. J. Psychiatry 40, 616‚Äì622. McGorry, P.D., Nelson, B., Amminger, G.P., Bechdolf, A., Francey, S.M., Berger, G., RiecherRossler, A., Klosterkotter, J., Ruhrmann, S., Schultze-Lutter, F., Nordentoft, M., Hickie, I., McGuire, P., Berk, M., Chen, E.Y., Keshavan, M.S., Yung, A.R., 2009. Intervention in individuals at ultra-high risk for psychosis: a review and future directions. J. Clin. Psychiatry 70, 1206‚Äì1212. Mesman, E., Nolen, W.A., Reichart, C.G., Wals, M., Hillegers, M.H., 2013. The Dutch bipolar offspring study: 12-year follow-up. Am. J. Psychiatry 170, 542‚Äì549. Micheloyannis, S., Vourkas, M., Tsirka, V., Karakonstantaki, E., Kanatsouli, K., Stam, C.J., 2009. The inÔ¨Çuence of ageing on complex brain networks: a graph theoretical analysis. Hum. Brain Mapp. 30, 200‚Äì208. Moore, G.J., Bebchuk, J.M., Wilds, I.B., Chen, G., Manji, H.K., 2000. Lithium-induced increase in human brain grey matter. Lancet 356, 1241‚Äì1242. Moorhead, T.W., McKirdy, J., Sussmann, J.E., Hall, J., Lawrie, S.M., Johnstone, E.C., McIntosh, A.M., 2007. Progressive gray matter loss in patients with bipolar disorder. Biol. Psychiatry 62, 894‚Äì900. Nery, F.G., Chen, H.H., Hatch, J.P., Nicoletti, M.A., Brambilla, P., Sassi, R.B., Mallinger, A.G., Keshavan, M.S., Soares, J.C., 2009. Orbitofrontal cortex gray matter volumes in bipolar disorder patients: a region-of-interest MRI study. Bipolar Disord. 11, 145‚Äì153. Newman, M.E., 2002. Assortative mixing in networks. Phys. Rev. Lett. 89, 208701. Nugent, A.C., Milham, M.P., Bain, E.E., Mah, L., Cannon, D.M., Marrett, S., Zarate, C.A., Pine, D.S., Price, J.L., Drevets, W.C., 2006. Cortical abnormalities in bipolar disorder investigated with MRI and voxel-based morphometry. Neuroimage 30, 485‚Äì497. Pantelis, C., Velakoulis, D., McGorry, P.D., Wood, S.J., Suckling, J., Phillips, L.J., Yung, A.R., Bullmore, E.T., Brewer, W., Soulsby, B., Desmond, P., McGuire, P.K., 2003. Neuroanatomical abnormalities before and after onset of psychosis: a cross-sectional and longitudinal MRI comparison. Lancet 361, 281‚Äì288. Paus, T., Keshavan, M., Giedd, J.N., 2008. Why do many psychiatric disorders emerge during adolescence? Nat. Rev. Neurosci. 9, 947‚Äì957. Penzes, P., Cahill, M.E., Jones, K.A., VanLeeuwen, J.E., Woolfrey, K.M., 2011. Dendritic spine pathology in neuropsychiatric disorders. Nat. Neurosci. 14, 285‚Äì293. Phillips, M.L., Kupfer, D.J., 2013. Bipolar disorder diagnosis: challenges and future directions. Lancet 381, 1663‚Äì1671. Phillips, M.L., Swartz, H.A., 2014. A critical appraisal of neuroimaging studies of bipolar disorder: toward a new conceptualization of underlying neural circuitry and a road map for future research. Am. J. Psychiatry 171, 829‚Äì843. Rosa, A.R., Magalhaes, P.V., Czepielewski, L., Sulzbach, M.V., Goi, P.D., Vieta, E., Gama, C.S., Kapczinski, F., 2014. Clinical staging in bipolar disorder: focus on cognition and functioning. J. Clin. Psychiatry 75, e450‚Äìe456. Rygula, R., Walker, S.C., Clarke, H.F., Robbins, T.W., Roberts, A.C., 2010. Differential contributions of the primate ventrolateral prefrontal and orbitofrontal cortex to serial reversal learning. J. Neurosci. 30, 14552‚Äì14559. Scott, J., Leboyer, M., Hickie, I., Berk, M., Kapczinski, F., Frank, E., Kupfer, D., McGorry, P., 2013. Clinical staging in psychiatry: a cross-cutting model of diagnosis with heuristic and practical value. Br. J. Psychiatry 202, 243‚Äì245. Smit, D.J., Stam, C.J., Posthuma, D., Boomsma, D.I., de Geus, E.J., 2008. Heritability of ‚Äúsmallworld‚Äù networks in the brain: a graph theoretical analysis of resting-state EEG functional connectivity. Hum. Brain Mapp. 29, 1368‚Äì1378. Sowell, E.R., Thompson, P.M., Tessner, K.D., Toga, A.W., 2001. Mapping continued brain growth and gray matter density reduction in dorsal frontal cortex: inverse relationships during postadolescent brain maturation. J. Neurosci. 21, 8819‚Äì8829. StanÔ¨Åeld, A.C., Moorhead, T.W., Job, D.E., McKirdy, J., Sussmann, J.E., Hall, J., Giles, S., Johnstone, E.C., Lawrie, S.M., McIntosh, A.M., 2009. Structural abnormalities of ventrolateral and orbitofrontal cortex in patients with familial bipolar disorder. Bipolar Disord. 11, 135‚Äì144. Strasser, H.C., Lilyestrom, J., Ashby, E.R., Honeycutt, N.A., Schretlen, D.J., Pulver, A.E., Hopkins, R.O., Depaulo, J.R., Potash, J.B., Schweizer, B., Yates, K.O., Kurian, E., Barta, P.E., Pearlson, G.D., 2005. Hippocampal and ventricular volumes in psychotic and nonpsychotic bipolar patients compared with schizophrenia patients and community control subjects: a pilot study. Biol. Psychiatry 57, 633‚Äì639. Tsuang, M.T., Faraone, S.V., 2000. The genetic epidemiology of bipolar disorder. In: Marneros, A., Angst, J. (Eds.), Bipolar Disorders: 100 Years After Manic-Depressive Insanity. Kluwer Academic, Zurich, Switzerland. Wingo, A.P., Harvey, P.D., Baldessarini, R.J., 2009. Neurocognitive impairment in bipolar disorder patients: functional implications. Bipolar Disord. 11, 113‚Äì125. Xu, G., Lin, K., Rao, D., Dang, Y., Ouyang, H., Guo, Y., Ma, J., Chen, J., 2012. Neuropsychological performance in bipolar I, bipolar II and unipolar depression patients: a longitudinal, naturalistic study. J. Affect. Disord. 136, 328‚Äì339. Please cite this article as: Lin, K., et al., A Multi-Dimensional and Integrative Approach to Examining the High-Risk and Ultra-High-Risk Stages of Bipolar Disorder, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.027 10 K. Lin et al. / EBioMedicine xxx (2015) xxx‚Äìxxx Yuksel, C., Du, F., Ravichandran, C., Goldbach, J.R., Thida, T., Lin, P., Dora, B., Gelda, J., O'Connor, L., Sehovic, S., Gruber, S., Ongur, D., Cohen, B.M., 2015. Abnormal highenergy phosphate molecule metabolism during regional brain activation in patients with bipolar disorder. Mol. Psychiatry http://dx.doi.org/10.1038/mp.2015.13 (Epub ahead of print). Yung, A.R., Yuen, H.P., McGorry, P.D., Phillips, L.J., Kelly, D., Dell'Olio, M., Francey, S.M., Cosgrave, E.M., Killackey, E., Stanford, C., Godfrey, K., Buckby, J., 2005. Mapping the onset of psychosis: the comprehensive assessment of at-risk mental states. Aust. N. Z. J. Psychiatry 39, 964‚Äì971. Please cite this article as: Lin, K., et al., A Multi-Dimensional and Integrative Approach to Examining the High-Risk and Ultra-High-Risk Stages of Bipolar Disorder, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.027 EBioMedicine 2 (2015) 882‚Äì888 Contents lists available at ScienceDirect EBioMedicine journal homepage: www.ebiomedicine.com Original Article A Mycobacterium tuberculosis Dormancy Antigen Differentiates Latently Infected Bacillus Calmette‚ÄìGu√©rin-vaccinated Individuals DelÔ¨Åna Pe√±a a,b, Ana I. Rovetta a,b, Rodrigo E. Hern√°ndez Del Pino b, Nicol√°s O. Amiano a,b, Virginia Pasquinelli b, Joaqu√≠n M. Pellegrini a,b, Nancy L. Tateosian a,b, Agust√≠n Rolandelli a,b, Marisa Gutierrez c, Rosa M. Musella d, Domingo J. Palmero d, Mar√≠a M. Gherardi e, Juan Iovanna f, H. Eduardo Chuluyan g, Ver√≥nica E. Garc√≠a a,b,‚Åé a Instituto de Qu√≠mica Biol√≥gica, Facultad de Ciencias Exactas y Naturales (IQUIBICEN), UBA (Universidad de Buenos Aires)‚ÄìCONICET, Intendente G√ºiraldes 2160, Pabell√≥n II, 4¬∞piso, Ciudad Universitaria, C1428EGA Buenos Aires, Argentina b Departamento de Qu√≠mica Biol√≥gica, Facultad de Ciencias Exactas y Naturales, UBA, Intendente G√ºiraldes 2160, Pabell√≥n II, 4¬∞piso, Ciudad Universitaria, C1428EGA Buenos Aires, Argentina c Secci√≥n Bacteriolog√≠a de la Tuberculosis, Hospital General de Agudos ‚ÄúDr. E. Tornu‚Äù, Combatientes de Malvinas 3002, 1427 Buenos Aires, Argentina d Divisi√≥n Tisioneumonolog√≠a Hospital F.J. Mu√±iz, Uspallata 2272, C1282AEN Buenos Aires, Argentina e INBIRS, Facultad de Medicina, UBA, Paraguay 2155, C1121ABG Buenos Aires, Argentina f Centre de Recherche en Canc√©rologie de Marseille (CRCM), INSERM U1068, CNRS UMR 7258, Aix-Marseille Universit√© Institut Paoli-Calmettes, Parc ScientiÔ¨Åque et Technologique de Luminy, Marseille, France g Centro de Estudios Farmacol√≥gicos y Bot√°nicos (CEFYBO), Facultad de Medicina, UBA, Paraguay 2155, C1121ABG Buenos Aires, Argentina a r t i c l e i n f o Article history: Received 1 April 2015 Received in revised form 22 May 2015 Accepted 26 May 2015 Available online 30 May 2015 Keywords: Latent tuberculosis Active tuberculosis Diagnosis IFN-gamma Antigens a b s t r a c t IFN-Œ≥ release assays (IGRAs) are better indicators of Mycobacterium tuberculosis infection than the tuberculin skin test (TST) in Bacillus Calmette‚ÄìGu√©rin (BCG)-vaccinated populations. However, IGRAs do not discriminate active and latent infections (LTBI) and no gold standard for LTBI diagnosis is available. Thus, since improved tests to diagnose M. tuberculosis infection are required, we assessed the efÔ¨Åcacy of several M. tuberculosis latency antigens. BCG-vaccinated healthy donors (HD) and tuberculosis (TB) patients were recruited. QuantiFERON-TB Gold In-Tube, TST and clinical data were used to differentiate LTBI. IFN-Œ≥ production against CFP-10, ESAT-6, Rv2624c, Rv2626c and Rv2628 antigens was tested in peripheral blood mononuclear cells. LTBI subjects secreted signiÔ¨Åcantly higher IFN-Œ≥ levels against Rv2626c than HD. Additionally, Rv2626c peptide pools to which only LTBI responded were identiÔ¨Åed, and their cumulative IFN-Œ≥ response improved LTBI discrimination. Interestingly, whole blood stimulation with Rv2626c allowed the discrimination between active and latent infections, since TB patients did not secrete IFN-Œ≥ against Rv2626c, in contrast to CFP-10 + ESAT-6 stimulation that induced IFN-Œ≥ response from both LTBI and TB patients. ROC analysis conÔ¨Årmed that Rv2626c discriminated LTBI from HD and TB patients. Therefore, since only LTBI recognizes speciÔ¨Åc epitopes from Rv2626c, this antigen could improve LTBI diagnosis, even in BCG-vaccinated people. ¬© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction Tuberculosis (TB) continues to be one of the most prevalent infectious diseases worldwide, with 9 million new cases reported and 1.5 million deaths in 2013 (W.H.O., 2014). Furthermore, an estimated one third of the world population is latently infected with Mycobacterium tuberculosis (LTBI) and at risk of disease reactivation (Dye et al., 1999). The existence of such a huge reservoir of the bacteria denotes the ‚Åé Corresponding author at: Departamento de Qu√≠mica Biol√≥gica ‚Äî IQUIBICEN, Facultad de Ciencias Exactas y Naturales, Universidad de Buenos Aires, Intendente G√ºiraldes 2160, Pabell√≥n II, 4¬∞piso, Lab QB23, Ciudad Universitaria, C1428EGA Buenos Aires, Argentina. E-mail address: vgarcia@qb.fcen.uba.ar (V.E. Garc√≠a). need of a rapid diagnosis of TB infection for its early detection and control. The tuberculin skin test (TST) has been long the most used tool for the diagnosis of M. tuberculosis infection (Vukmanovic-Stejic et al., 2006). However, TST speciÔ¨Åcity is limited since it employs antigens shared with environmental mycobacteria and Mycobacterium bovis Bacillus Calmette‚ÄìGu√©rin (BCG) (Shingadia and Novelli, 2008). Furthermore, TST cannot differentiate active disease from LTBI, requiring additional tests to establish the diagnosis. The discovery of the diagnostic potential of the M. tuberculosis antigens ESAT-6 and CFP-10 led to the development of IFN-Œ≥ release assays (IGRAs) (Diel et al., 2011). These antigens are encoded in the region of deletion-1 locus present in M. tuberculosis, but absent in BCG and most environmental mycobacteria, making IGRAs more speciÔ¨Åc than TST (Diel et al., 2011; http://dx.doi.org/10.1016/j.ebiom.2015.05.026 2352-3964/¬© 2015 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). D. Pe√±a et al. / EBioMedicine 2 (2015) 882‚Äì888 Ganguly et al., 2008). However, a deÔ¨Åciency of current IGRAs is that, like TST, they do not discriminate active from latent infection, a key problem in cases with uncertain clinical symptoms or smear negative sputum samples, like extra-pulmonary TB cases (Chegou et al., 2011). Though this is not the main objective of IGRAs, such discrimination would be desired, since it would bring greater speciÔ¨Åcity to the assay, providing an extra tool for fast diagnosis of active infection. In addition, there is as yet no gold standard for the diagnosis of LTBI. Besides, several vaccine studies are being currently performed employing antigens present in existing IGRAs, with many having demonstrated the potential of ESAT-6 to confer protection against M. tuberculosis (Davila et al., 2012; Cervantes-Villagrana et al., 2013; Reither et al., 2014; van Dissel et al., 2011). Therefore, if these vaccines prove to be successful and become commonly used, it might signify the end of ESAT-6 as a diagnostic marker, since immune response to this antigen would reÔ¨Çect both M. tuberculosis infection and vaccination, losing its present speciÔ¨Åcity. Finally, the performance of available IGRAs exhibits great variability among different groups of immunocompromised patients, being dependent upon the nature and extent of immunodeÔ¨Åciency (Sauzullo et al., 2015). Thus, the reasons described above support the search for better assays and/or new immunological biomarkers to diagnose M. tuberculosis infection, which might complement or improve current assays to diagnose both latent and active infections. LTBI is thought to be associated with a dormancy state of the pathogen and several antigens up-regulated during these conditions have been analyzed as potential diagnostic markers (Schuck et al., 2009). We studied the response of healthy subjects, LTBI individuals and TB patients to several M. tuberculosis antigens. DosR regulon-encoded antigens Rv2624c, Rv2626c and Rv2628 were selected based on previous reports that suggested they were recognized by household contacts or TST+ individuals (Leyten et al., 2006; Chegou et al., 2012). Our results demonstrated that LTBI individuals recognized speciÔ¨Åc epitopes from Rv2626c that induced the secretion of signiÔ¨Åcant amounts of IFN-Œ≥, in sharp contrast to non-infected individuals. Moreover, our Ô¨Åndings indicate that Rv2626c would be a useful antigen to discriminate LTBI individuals from both active TB patients and non-infected healthy subjects in a BCG-vaccinated population. Additional studies will validate the potential of Rv2626c to discriminate LTBI from TB patients and healthy donors in non-BCG-vaccinated populations. 2. Methods 2.1. Study Subjects BCG-vaccinated healthy adults lacking a history of TB (household contacts and healthcare workers) were recruited at Argentinean Referral Hospitals between January 2012 and August 2014. Among this group of individuals, diagnosis of LTBI was established using QuantiFERON-TB Gold In-Tube (QFT-GIT; Qiagen, USA; according to the manufacturer's directions) and TST (see below) tests. LTBI diagnosis was assigned to any subject with a positive QFT-GIT/TST and no clinical or radiological evidence of active TB. In the event of discordant QFT-GIT/TST results, individuals were assigned to the corresponding group on the basis of the QFT-GIT result. The group of healthy donors (HD) was comprised by adult individuals without TB disease (tested by chest X-rays and analysis of acid-fast bacilli in sputum) and with negative QFT-GIT/TST. HIV-uninfected patients with active TB were evaluated at Dr. F. Mu√±iz or Dr. E. Torn√∫ Hospitals (Buenos Aires, Argentina). Diagnosis of TB disease was established based on clinical and radiological data together with culture-conÔ¨Årmation and the identiÔ¨Åcation of acid-fast bacilli in sputum. Patients included in this study had received less than one week of anti-TB therapy. Information regarding demographic data and prior TB exposure was obtained at the time of recruitment. A total of 56 LTBI individuals, 56 TB patients and 60 HD participated in this study. All participants provided written informed consent for sample collection and subsequent analysis. The protocols conducted were 883 approved by the Ethical Committee of the Dr. F. Mu√±iz and the Dr. E. Torn√∫ Hospitals. 2.2. Study Inclusion and Exclusion Criteria for Individuals Participating in the Study Inclusion criteria: a) adult (over 18 years old) men and women with active pulmonary TB and b) healthy volunteers with high level of exposure to M. tuberculosis (household contacts of TB patients and healthcare workers of National Referral Hospitals for TB). All recruited subjects were BCG-vaccinated. QFT-GIT and TST assays were used to determine the presence of LTBI among individuals without clinical or microbiological diagnosis of active TB. All TB patients included in the study had a positive culture for M. tuberculosis. Exclusion criteria: a) HIV positive or positive serology to other viral or bacterial infections; b) patients with diabetes, cancer, autoimmune diseases or other conditions that may affect the immune system of the individual; c) pregnant women and d) children. Among the population of active TB patients we excluded: a) patients with multidrug-resistant tuberculosis (MDR-TB) infection and b) patients with more than seven consecutive days of anti-TB treatment. Individuals with indeterminate QFT-GIT results were also excluded from the study. 2.3. Tuberculin Skin Test (TST) TST was administered by medical staff according to the Mantoux method (American Thoracic Society, CDC, 2000). BrieÔ¨Çy, 2 tuberculin units (0.1 ml) of puriÔ¨Åed protein derivative (Instituto Malbr√°n, Buenos Aires, Argentina) were intradermally injected into the inner surface of the forearm. The skin induration was measured (in mm) after 48‚Äì72 h by trained personnel at one of the medical centers. A positive TST was deÔ¨Åned as an induration size ‚â•10 mm, according to National Guidelines (Zerbini, 2013). 2.4. Antigens Recombinant antigens (Rv2624c, Rv2626c, Rv2628, CFP-10, ESAT-6) were produced in Escherichia coli BL21 (DE3) pLysS and puriÔ¨Åed using a nickel nitrilotriacetic system (Ni-NTA Agarose, Qiagen), following the manufacturers' directions. After purity control, proteins were concentrated using Amicon (Millipore, USA) commercial columns and then Detoxi-Gel Endotoxin Removing Resin (Pierce, USA) was used to remove possible endotoxin traces. CFP-10 and ESAT-6-induced IFN-Œ≥ responses were compared to those elicited with the respective protein reference standard obtained from BEI Resources (NIAID, NIH; CFP-10 Recombinant Protein Reference Standard, NR-14869 and ESAT-6 Recombinant Protein Reference Standard, NR-14868). Cell lysate from the virulent M. tuberculosis H37Rv strain (Mtb-Ag) was prepared by probe sonication (BEI Resources, NIAID, NIH: M. tuberculosis, Strain H37Rv, Whole Cell Lysate, NR-14822). 2.5. Peptides Overlapping synthetic peptides (13‚Äì15-mers, overlapped by 11 amino acids; 36 in total) spanning the sequence of Rv2626c were synthesized by Biomatik Corp. (Canada) using Fmoc chemistry. Peptide purity was N80%, as assayed by HPLC, and the composition was veriÔ¨Åed by mass spectrometry. Lyophilized peptides were dissolved in dimethyl sulfoxide, aliquoted and stored at ‚àí70 ¬∞C. For in vitro stimulation, peptides were arranged in pools composed of six peptides, obtaining six consecutive pools (A‚ÄìE) covering the complete Rv2626c sequence. 2.6. Cell preparation and reagents Peripheral blood mononuclear cells (PBMCs) were isolated by centrifugation over Ficoll-Hypaque (GE Healthcare, USA) and cultured (1 √ó 106 cells/ml) with/without the different M. tuberculosis antigens (Rv2624c, Rv2626c, Rv2628, CFP-10, ESAT-6; 2.5 Œºg/ml) or Rv2626c peptide pools (each peptide at 5 Œºg/ml) with RPMI 1640 (Gibco, USA) supplemented with L-glutamine, penicillin/streptomycin and 10% human serum (Sigma-Aldrich, USA), during different times. BrieÔ¨Çy, PBMCs were stimulated for Ô¨Åve days and then cell free supernatants were collected to determine the levels of IFN-Œ≥ by ELISA (Human IFN-Œ≥ ELISA MAX‚Ñ¢ Standard Kit, BioLegend, USA). In different experiments, cells were incubated with the antigens for four days and then harvested to determine IFN-Œ≥ expression by Ô¨Çow cytometry (see below). 2.7. Whole blood stimulation BrieÔ¨Çy, 0.5 ml of heparinized blood was treated with Rv2626c or CFP-10 + ESAT-6 (2.5 Œºg/ml) for 24 h at 37 ¬∞C. Afterwards, plates were centrifuged and plasma was recovered to determine IFN-Œ≥ production by ELISA. 2.8. Flow cytometry After four days of PBMCs stimulation, GolgiStop¬Æ (BD Biosciences, USA) reagent was added for the last 5 h of culture. Cells were then harvested and stained for intracellular IFN-Œ≥ (eBioscience, USA) and surface CD4 (BioLegend) expression using CytoÔ¨Åx/Cytoperm‚Ñ¢ Ô¨Åxation/ permeabilization buffers following the manufacturer's instructions (BD Biosciences) (Supplementary appendix). Negative control samples were incubated with irrelevant isotype-matched antibodies in parallel with experimental samples. Samples were analyzed on a FACS ARIA II Ô¨Çow cytometer (BD Biosciences). IFN-Œ≥Œ≥ (pg/ml) <50 50-200 200-400 400-600 >600 Non TB QFT- GIT D. Pe√±a et al. / EBioMedicine 2 (2015) 882‚Äì888 Mtb-Ag Rv2626c CFP-10 ESAT-6 884 + + + + + + + Fig. 1. Heat map for IFN-Œ≥ responses to Mtb-Ag, Rv2626c, CFP-10 and ESAT-6 M. tuberculosis antigens in subjects without active tuberculosis. Heat map showing the IFN-Œ≥ responses to Mtb-Ag, Rv2626c, CFP-10 or ESAT-6 antigens as measured in cell free supernatants of peripheral blood mononuclear cells (PBMCs) from subjects without active tuberculosis (non-TB; n = 20). Colors represent pg/ml of IFN-Œ≥ produced in each individual, where high values are shown in red and low values in light yellow, as shown in the Ô¨Ågure scale. Right column displays the result of QuantiFERON-TB Gold In-tube (QFT-GIT) test. 2.9. Statistics Individuals with a similar pattern of IFN-Œ≥ cytokine secretion were grouped together as clusters and presented as a heat map; with heat distances computed using Euclidean distances as similarity measure, employing NetWalker 1.0 software. The Mann‚ÄìWhitney test was used to analyze differences between unpaired samples. Fisher's exact test was used for categorical variables. Receiver operating characteristic (ROC) curve analysis was conducted to analyze the predictive value of IFN-Œ≥ response to Rv2626c or CFP10 + ESAT-6, calculating the area under the curve (AUC) and the 95% conÔ¨Ådence interval (CI). Analyses were performed using GraphPad Prism 5 software. p b 0.05 was considered statistically signiÔ¨Åcant. 3. Results 3.1. Evaluation of IFN-Œ≥ Responses to M. tuberculosis Antigens in BCG-Vaccinated People To assess the frequency of LTBI individuals among healthy BCGvaccinated people highly exposed to M. tuberculosis, we analyzed the IFN-Œ≥ response to several dormancy (Rv2626c, Rv2628, Rv2624c) and speciÔ¨Åc early secretory (ESAT-6, CFP-10) antigens from the pathogen. Initially, IFN-Œ≥ production against the mentioned antigens or Mtb-Ag was evaluated in PBMCs from 20 healthy individuals. As shown by the heat map in Fig. 1, all subjects produced elevated amounts of IFN-Œ≥ against Mtb-Ag (N600 pg/ml), as expected in BCG-vaccinated individuals. Interestingly, 7 out of 20 (35%) of the persons tested also responded simultaneously with high intensity to three antigens: Rv2626c, ESAT-6 and CFP-10. In fact, the three antigens induced N600 pg/ml of IFN-Œ≥ in 5 out of the 7 subjects. The two exceptions were one individual in which Rv2626c and ESAT-6 induced 400‚Äì600 pg/ml and CFP-10 induced N600 pg/ml of IFN-Œ≥, and another person in which Rv2626c induced 200‚Äì400 pg/ml and CFP-10 and ESAT-6 induced N600 pg/ml. In contrast to our data resulting from stimulation with Rv2626c, ESAT-6 and CFP-10 showing that 7 individuals simultaneously secreted high levels of IFN-Œ≥ to those antigens (all above 200 pg/ml), none of those subjects secreted IFN-Œ≥ against Rv2624c, and only 2 produced levels of IFN-Œ≥ above 200 pg/ml in response to Rv2628 stimulation (data not shown). Therefore, Fig. 1 represents the results obtained with antigens that displayed noticeable differences among individuals: Rv2626c, ESAT-6 and CFP-10. We then characterized the healthy population that secreted high IFN-Œ≥ levels to the mentioned three antigens with the QFT-GIT test. Fig. 1 shows that all the subjects that highly responded to Rv2626c, ESAT-6 and CFP-10 simultaneously were QFT-GIT+, which would be indicative of LTBI; while all those individuals that did not respond were QFT-GIT‚àí. In view of the Ô¨Åndings described above, we next investigated the response to M. tuberculosis antigens in a larger population of QFT-GIT‚àí HD and QFT-GIT+ LTBI individuals (Table 1). We also stimulated all subjects' PBMCs simultaneously with the combination of antigens included in the QFT-GIT assay, i.e.,: CFP-10 + ESAT-6. As shown in Table 2, LTBI produced signiÔ¨Åcantly higher IFN-Œ≥ levels against CFP-10, ESAT-6 and CFP-10 + ESAT-6 than HD. Interestingly, LTBI subjects also secreted signiÔ¨Åcantly more elevated IFN-Œ≥ levels upon Rv2626c stimulation than HD (Table 2), suggesting that stimulation with Rv2626c might also be useful to discriminate between HD and LTBI. To conÔ¨Årm these Ô¨Åndings D. Pe√±a et al. / EBioMedicine 2 (2015) 882‚Äì888 Table 1 Characteristics of Individuals without active TB infection. Characteristics Individuals without active M. tuberculosis infection (N = 116) Age (years) Sex Male Female Country of birth Argentina Other Latin American countries TST Positive Negative p value QFT-GIT (‚àí) 52% (N = 60) 40.5 ¬± 11.9 QFT-GIT (+) 48% (N = 56) 42.0 ¬± 11.6 38% (N = 23) 62% (N = 37) 40% (N = 22) 60% (N = 34) 0.85 (b) 88% (N = 53) 12% (N = 7) 80% (N = 45) 20% (N = 11) 0.31 (b) 14% (N = 8) 86% (N = 52) 78% (N = 44) 22% (N = 12) b0.001 (b) 0.49 (a) Abbreviations: QuantiFERON-TB Gold In-tube, QFT-GIT; tuberculin skin test, TST. (a) Age values are displayed as Mean ¬± SEM. p values were calculated by the Mann‚Äì Whitney U test for continuous variables; (b) p values were calculated by Fisher's exact test for categorical variables. we used Ô¨Çow cytometry, a more sensitive technique than ELISA (Supplementary Fig. 1). In agreement to the results obtained by ELISA, PBMCs from LTBI displayed signiÔ¨Åcantly higher numbers of IFN-Œ≥ secreting CD4+ cells in response to Rv2626c as compared to HD (Supplementary Table 1). Regarding the other antigens tested, in both HD and LTBI, Rv2624c induced a very weak IFN-Œ≥ response (Table 2). Moreover, although LTBI subjects secreted low IFN-Œ≥ levels in response to Rv2628, they were not signiÔ¨Åcantly different compared to those produced by HD (Table 2). In order to evaluate the potential use of Rv2626c in discriminating LTBI vs. HD, we next performed a ROC analysis for the IFN-Œ≥ responses to this antigen (shown in Table 2), obtaining signiÔ¨Åcant results for AUC analysis (AUC, 0.86; p b 0.001; 95% CI: 0.78‚Äì0.93). Together, these Ô¨Åndings demonstrated that Rv2626c might allow discriminating LTBI individuals from HD. 3.2. IdentiÔ¨Åcation of Immunodominant Peptide Pools Derived from Rv2626c Antigen Since a low percentage of HD showed a weak IFN-Œ≥ production against Rv2626c (Table 2), and an ideal diagnostic test requires the highest discrimination among groups under study, we next searched for the immunodominant regions of Rv2626c to which only LTBI responded. For this, we employed overlapping peptides arranged in pools composed of six peptides each, covering the entire Rv2626c protein. As can be observed in Table 3, pools B, D and F induced high IFN-Œ≥ levels, which signiÔ¨Åcantly discriminated between LTBI and HD, while pools C and E also distinguished between both groups but with a weaker signiÔ¨Åcance (Table 3). In contrast, although stimulation with pool A did induce the production of IFN-Œ≥ in LTBI subjects, no signiÔ¨Åcant differences with HD were detected (Table 3). Therefore, pool A would belong to a less speciÔ¨Åc region of Rv2626c than pools B, D and F, the 885 most speciÔ¨Åc and immunogenic ones, which could strongly differentiate the two groups of individuals. We next analyzed the difference in the levels of IFN-Œ≥ secreted by LTBI and HD in response to Rv2626c antigen or each of the most speciÔ¨Åc peptide pools. We observed that upon stimulation with Rv2626c, LTBI subjects showed an eight times greater response than HD (Table 2), whereas with the pools, the response was 70 (pool B), 23 (pool D) or 15 (pool F) times greater in LTBI than in HD (Table 3), clearly improving the discrimination between HD and LTBI individuals. Since each LTBI subject recognized the individual peptide pools with different intensity, we then calculated the cumulative IFN-Œ≥ response to the most speciÔ¨Åc and immunogenic pools (B, D and F) to capture the overall immune response to the tested peptide pools for each person. As shown in Table 3, our results showed a signiÔ¨Åcantly higher cumulative IFN-Œ≥ response to these pools in the group of LTBI individuals than in HD. Thus, our Ô¨Åndings indicate that LTBI individuals recognized epitopes from Rv2626c antigen and showed a strong cumulative IFN-Œ≥ response, in clear contrast to HD that display weak or no response to those peptides. 3.3. IFN-Œ≥ Responses to M. tuberculosis SpeciÔ¨Åc Antigens in Healthy Donors, LTBI Individuals and TB Patients Since we classiÔ¨Åed our population using QFT-GIT assay, and considering that this test uses whole blood for more practical and faster testing, we also analyzed the response to Rv2626c stimulating whole blood. Given that both current IGRAs discriminate M. tuberculosis infection (active or latent) from non-infected individuals, we also examined the response of active TB patients (Table 4). We Ô¨Årst investigated the response of PBMCs from TB patients, LTBI and HD to Rv2626c and CFP-10 + ESAT-6 (Fig. 2A). We found that, as expected, both active and latently infected individuals (TB patients and LTBI) secreted significantly higher IFN-Œ≥ amounts in response to CFP-10 + ESAT-6 as compared to HD. Interestingly, we observed that stimulation with Rv2626c allowed the discrimination between active and latent infections (Fig. 2A) since TB patients did not secrete IFN-Œ≥ against this antigen (Fig. 2A), in sharp contrast to the results obtained with CFP10 + ESAT-6. When we performed our experiments in whole blood, we obtained similar results to those observed in PBMCs: LTBI subjects secreted signiÔ¨Åcantly higher amounts of IFN-Œ≥ in response to Rv2626c than HD. In addition, like in PBMCs (Fig. 2A), we observed that TB patients did not produce IFN-Œ≥ against Rv2626c (Fig. 2B), in sharp contrast to the results obtained with CFP-10 + ESAT-6 (Fig. 2B). Furthermore, the ROC analysis for IFN-Œ≥ responses to Rv2626c in whole blood reinforced the potential of Rv2626c antigen for discriminating LTBI individuals from non-LTBI individuals (patients with active TB or HD; Fig. 2C; AUC, 0.86; p b 0.001; 95% CI: 0.78‚Äì0.93). Additionally, by using the data obtained in whole blood and the Youden index (Youden, 1950; Perkins and Schisterman, 2006), we were able to establish an optimal cut off point of N63.35 pg/ml of IFN-Œ≥, with 78.26% sensitivity and 89.36% speciÔ¨Åcity. For comparison, Fig. 2D also shows the ROC curve obtained after analyzing the IFN-Œ≥ response to CFP-10 + ESAT-6 for the discrimination of M. tuberculosis infected individuals (LTBI + TB patients) versus HD. In conclusion, our Ô¨Åndings show that the response Table 2 IFN-Œ≥ production against M. tuberculosis antigens by PBMCs from non-TB individuals. pg/ml of IFN-Œ≥, Mean ¬± SEM (range, pg/ml) N CFP-10 ESAT-6 CFP-10 + ESAT-6 Rv2626c Rv2624c Rv2628 HD 45 LTBI 56 250 ¬± 52.8 (1‚Äì1102) 1452 ¬± 202.0 (1‚Äì4127) b0.001 196 ¬± 43.3 (2‚Äì895) 973 ¬± 156.2 (5‚Äì3958) b0.001 248 ¬± 76.3 (8‚Äì578) 1517 ¬± 210.0 (7‚Äì3888) b0.001 69 ¬± 24.5 (1‚Äì476) 549 ¬± 106.6 (14‚Äì2892) b0.001 36 ¬± 13.9 (1‚Äì104) 39 ¬± 24.9 (1‚Äì169) 0.84 33 ¬± 14.4 (1‚Äì150) 135 ¬± 60.2 (1‚Äì461) 0.15 p value Peripheral blood mononuclear cells (PBMCs) from healthy donors (HD) and latently M. tuberculosis infected individuals (LTBI) were cultured with M. tuberculosis antigens CFP-10, ESAT-6, CFP-10 + ESAT-6, Rv2624c, Rv2626c or Rv2628 (2.5 Œºg/ml) for Ô¨Åve days. Then, cell free supernatants were recovered and the production of IFN-Œ≥ was evaluated by ELISA. p values were calculated using the Mann‚ÄìWhitney test for unpaired samples. 886 D. Pe√±a et al. / EBioMedicine 2 (2015) 882‚Äì888 Table 3 IFN-Œ≥ production against overlapping peptides of Rv2626c by PBMCs from non-TB individuals. pg/ml of IFN-Œ≥, Mean ¬± SEM HD LTBI p value N Pool A Pool B Pool C Pool D Pool E Pool F Cumulative response to pools B, D and F 33 40 7.6 ¬± 4.09 216.6 ¬± 82.7 0.09 3.9 ¬± 1.6 275.9 ¬± 93.1 b0.001 4.1 ¬± 6.0 119.2 ¬± 35.7 0.03 18.4 ¬± 18.4 416.2 ¬± 120.3 b0.001 4.9 ¬± 3.4 48.7 ¬± 20.7 0.02 34.4 ¬± 20.0 517.1 ¬± 119.3 b0.001 55.3 ¬± 20.92 1177.0 ¬± 253.1 b0.001 Peripheral blood mononuclear cells (PBMCs) from healthy donors (HD) and latently M. tuberculosis infected individuals (LTBI) were cultured with six peptide pools (A‚ÄìF) derived from Rv2626c (5 Œºg/ml of each peptide) for Ô¨Åve days and then IFN-Œ≥ production was evaluated in cell free supernatants by ELISA. The cumulative response to pools B, D and F was calculated for each individual as the sum of the secretion of IFN-Œ≥ in response to each of the three pools. p values were obtained though the Mann‚ÄìWhitney test for unpaired samples. to Rv2626c might be useful to differentiate between HD and LTBI, similar to ESAT-6 and CFP-10, antigens employed in IGRAs like QFT-GIT. However, in contrast to QFT-GIT assay, stimulation with Rv2626c would also allow the discrimination between active and latent infections. Therefore, our results indicate that Rv2626c might be useful not only to diagnose LTBI, but also, to differentiate latent from active M. tuberculosis infection. 4. Discussion A main health obstacle to control TB is that latent bacilli may remain viable for decades and reactivate later to cause active TB. Thus, there is a need to identify LTBI individuals and treat them to eliminate the risk of developing TB. Since there is no diagnostic gold standard for LTBI and several reports have demonstrated the potential use of antigens employed in current IGRAs as protective vaccines, it urges to Ô¨Ånd new M. tuberculosis antigens to detect latently infected individuals (Davila et al., 2012; Cervantes-Villagrana et al., 2013; Reither et al., 2014; van Dissel et al., 2011; Sester et al., 2011). Currently, both TST and IGRAs are used for LTBI diagnosis worldwide. Still, in BCG-vaccinated populations IGRAs may be preferable to TST since they are more speciÔ¨Åc for M. tuberculosis infection (Pai et al., 2014). In fact, in our study 13% of the individuals assessed had discordant results between TST and IGRA tests (Supplementary Fig. 2). To develop T cell-based assays that discriminate active TB and LTBI, improving the current available tests, it is necessary to identify new host markers expressed in response to M. tuberculosis antigens that are uniquely recognized by LTBI individuals (Chegou et al., 2012). Table 4 Characteristics of patients with active TB infection. Characteristics Tuberculosis patients (TB) (N = 56) Age (years) (a) Sex (b) Male Female Country of birth Argentina Other Latin American countries TST Positive Negative AFB in sputum smear Positive Negative Radiological lesions (c) Severe Moderate Mild 34.5 ¬± 13.4 86% (N = 48) 14% (N = 8) 53% (N = 30) 47% (N = 26) 98% (N = 55) 2% (N = 1) 84% (N = 47) 16% (N = 9) 80% (N = 45) 20% (N = 11) 0% (N = 0) (a) Age values are displayed as Mean ¬± SEM. (b) Categorical data are expressed as percentages (number). (c) Radiological lesions: mild, patients with a single lobe involved and without visible cavities; moderate, patients presenting unilateral involvement of two or more lobes with cavities, if present, reaching a total diameter no greater than 4 cm; severe, bilateral disease with massive affectation and multiple cavities. Abbreviations: AFB, acid-fast bacilli. Thus, present IGRAs might be complemented by using antigens to which patients with active TB do not respond. Here, we describe an immune-stimulatory function of the 16-kDa conserved protein coded by the M. tuberculosis open reading frame Rv2626c, one of the most abundant proteins under low-oxygen conditions in M. tuberculosis lysates (Rosenkrands et al., 2002). Variable results using Rv2628 and Rv2626c latency antigens were reported by others, although the disparities could be related to differences in stimulation schemes, antigen concentrations, ethnic backgrounds, and BCG vaccination history (Schuck et al., 2009; Chegou et al., 2012; Lin et al., 2007; Black et al., 2009; Riano et al., 2012; Hozumi et al., 2013; Mensah et al., 2014). Bashir et al. reported that Rv2626c induced higher IFN-Œ≥ levels in TB patients than in HD, but LTBI individuals were not analyzed in this study (Bashir et al., 2010). Leyten et al. demonstrated that TST+ individuals strongly recognized more latency antigens (e.g.,: Rv2628 and Rv2626c) as compared to cured patients or patients undergoing anti-TB treatment (Leyten et al., 2006). The study by Goletti et al. evaluated T-cell IFN-Œ≥ responses to M. tuberculosis latency antigens in a mainly European population composed by subjects at different stages of TB infection, concluding that remote LTBI individuals showed signiÔ¨Åcantly higher IFN-Œ≥ responses to Rv2628 antigen than recent LTBI subjects, active TB and controls (Goletti et al., 2010). However, we studied a very different population: 20% of QFT-GIT+ and 12% of QFT-GIT‚àí subjects were from other Latin American countries (Table 1), while the rest were Argentinean. In addition, the genetic mixture of Argentinean people is comprised by 79.9% European, 15.8% Native American, and 4.3% African contributions, which would imply that there was a fairly diverse genetic background in the population studied (Corach et al., 2010; Avena et al., 2006). Since some HD subjects produced low levels of IFN-Œ≥ against Rv2626c (Fig. 2), we searched for Rv2626c immunodominant epitopes to which only LTBI individuals responded. None of the mentioned studies that tested latency antigens investigated the immunodominance of peptide pools to discriminate LTBI individuals (Schuck et al., 2009; Leyten et al., 2006; Chegou et al., 2012; Lin et al., 2007; Black et al., 2009; Riano et al., 2012; Hozumi et al., 2013; Mensah et al., 2014; Bashir et al., 2010; Goletti et al., 2010). We observed that pools C, B, D, E and F all allowed improving the discernment between LTBI and HD (Table 3), but pools B, D and F resulted the most speciÔ¨Åc and immunogenic ones. These Ô¨Åndings reinforced our data showing that the response to Rv2626c might allow discriminating between LTBI and HD. Additionally, a signiÔ¨Åcantly higher cumulative IFN-Œ≥ response to the immunogenic peptide pools B, D and F was detected in LTBI individuals as compared to HD (Table 3). Together, we demonstrated that speciÔ¨Åc immunogenic peptide pools from Rv2626c might improve the discrimination of LTBI individuals from HD in a BCG-vaccinated population. Interestingly, when we analyzed the response to Rv2626c both in PBMCs and in whole blood from LTBI, HD and TB patients, we observed that only LTBI individuals secreted signiÔ¨Åcant amounts of IFN-Œ≥ against this antigen (Fig. 2A and B), indicating that the use of Rv2626c or its immunogenic peptides might comprise a new tool to differentiate LTBI individuals from both HD and patients with active disease. Thus, the use of Rv2626c in combination with ESAT-6 and CFP-10 could improve current kits to allow the differential diagnosis of LTBI and active TB subjects. D. Pe√±a et al. / EBioMedicine 2 (2015) 882‚Äì888 A) CFP-10 + ESAT-6 in PBMC 2500 p<0.001 Rv2626c in PBMC p<0.001 p=0.03 p<0.001 p<0.001 p=0.36 2000 IFN-Œ≥ (pg/ml) IFN-Œ≥ (pg/ml) 2500 2000 1500 800 600 400 1500 1000 500 200 0 0 HD B) TB LTBI Rv2626c in whole blood 4000 3500 3000 500 p<0.001 p=0.99 p<0.001 HD TB LTBI CFP-10 + ESAT-6 in whole blood 4000 p=0.005 p=0.008 p=0.96 IFN-Œ≥ (pg/ml) IFN-Œ≥ (pg/ml) 887 400 300 200 3000 2000 1000 100 0 0 HD TB LTBI HD TB ROC curve for the detection of infected individuals (LTBI or TB) ROC curve for the detection of LTBI individuals C) LTBI D) Rv2626c 80 Sensitivity % 80 Sensitivity % 100 100 60 40 AUC: 0,8579 P < 0.0001 20 CFP-10 + ESAT-6 60 40 AUC: 0,8469 P < 0.0001 20 0 0 0 20 40 60 100% - Specificity% 80 0 20 40 60 80 100% - Specificity% Fig. 2. IFN-Œ≥ production against M. tuberculosis antigens in PBMCs and whole blood from non-TB individuals and TB patients. (A) Peripheral blood mononuclear cells (PBMCs) from healthy donors (HD; N = 45), patients with tuberculosis (TB; N = 56) and latently M. tuberculosis infected individuals (LTBI; N = 56) were cultured with Rv2626c or CFP-10 + ESAT-6 for 5 days. Then, cell free supernatants were recovered and IFN-Œ≥ production was evaluated by ELISA. (B) Whole blood from HD (N = 28), TB patients (N = 20) and LTBI individuals (N = 25) was cultured with Rv2626c or CFP-10 + ESAT-6 for 24 h. Afterwards, plasma samples were collected and IFN-Œ≥ production was evaluated by ELISA. (A‚ÄìB) Bars represent the Mean ¬± SEM. Mann‚ÄìWhitney test for unpaired samples. (C‚ÄìD) ROC curve analyses for evaluation of the predictive value of whole blood IFN-Œ≥ levels produced in response to (C) Rv2626c or (D) CFP-10 + ESAT-6, for differentiating (C) LTBI individuals from non-LTBI individuals (HD or TB), or (D) infected individuals (LTBI or active TB) from HD. Abbreviations: ROC, receiver operating characteristic; AUC, area under the ROC curve. Additionally, Rv2626c or its immunogenic peptide pools might be candidates to replace potential vaccine antigens present in current IGRAs. In conclusion, the vast genetic mixture of Argentinean people makes the use of Rv2626c or its immunogenic peptide pools promising to discriminate individuals with active and latent M. tuberculosis infections in BCG-vaccinated people. However, future studies should determine whether Rv2626c might be also useful discriminating LTBI from TB patients and healthy donors in non-BCG-vaccinated populations. provided expert advice. All authors contributed to data gathering and interpretation, and revision of the report. Contributors Acknowledgments VEG and HEC designed the study. MG, RMM and DJP were in charge of patient recruitment, diagnosis and sample collection. DP, AIR, and REH were responsible for processing samples and performing ELISA and Ô¨Çow cytometry analysis. JP, NLT, NOA and AR contributed with standard laboratory work. DP, VP, JI, HEC and VEG did the data management and analysis. DP and VEG wrote the manuscript. MMG and JI This investigation received Ô¨Ånancial support from the Agencia Nacional de Promoci√≥n Cient√≠Ô¨Åca y Tecnol√≥gica (ANPCyT, PAE-PID00127, PAE-PICT-207-02332 and PICT-0240 to VEG); University of Buenos Aires (20020100100221 and 20020130100236BA to VEG; 20020130100720BA to HEC), and Consejo Nacional de Investigaciones Cient√≠Ô¨Åcas y Tecnol√≥gicas (CONICET; PIP 2012‚Äì2014 to VEG. and HEC). Declaration of Interests DP, AIR, VP, MMG, JI, HEC and VEG have a patent application for the use of the peptide pools for the LTBI detection (patent application number No. EP14306329.5). 888 D. Pe√±a et al. / EBioMedicine 2 (2015) 882‚Äì888 The ANPCyT, the University of Buenos Aires, and CONICET had no role in study design, data gathering, analysis, and interpretation, or writing of the report. VEG had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. We thank Dr. Peter Barnes for the insightful discussions. Appendix A. Supplementary Data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ebiom.2015.05.026. References American Thoracic Society (ATS) and the Centers for Disease Control and Prevention (CDC), 2000. Targeted tuberculin testing and treatment of latent tuberculosis infection. Am. J. Respir. Crit. Care Med. 161 (4 Pt 2), S221‚ÄìS247. World Health Organization, 2014. Global Tuberculosis Report Geneva. Avena, S.A., Goicoechea, A.S., Rey, J., Dugoujon, J.M., Dejean, C.B., Carnese, F.R., 2006. Gene mixture in a population sample from Buenos Aires City. Med. (B Aires) 66 (2), 113‚Äì118. Bashir, N., Kounsar, F., Mukhopadhyay, S., Hasnain, S.E., 2010. Mycobacterium tuberculosis conserved hypothetical protein rRv2626c modulates macrophage effector functions. Immunology 130 (1), 34‚Äì45. Black, G.F., Thiel, B.A., Ota, M.O., et al., 2009. Immunogenicity of novel DosR regulonencoded candidate antigens of Mycobacterium tuberculosis in three high-burden populations in Africa. Clin. Vaccine Immunol. 16 (8), 1203‚Äì1212. Cervantes-Villagrana, A.R., Hernandez-Pando, R., Biragyn, A., et al., 2013. Prime-boost BCG vaccination with DNA vaccines based in beta-defensin-2 and mycobacterial antigens ESAT6 or Ag85B improve protection in a tuberculosis experimental model. Vaccine 31 (4), 676‚Äì684. Chegou, N.N., Hoek, K.G., Kriel, M., Warren, R.M., Victor, T.C., Walzl, G., 2011. Tuberculosis assays: past, present and future. Expert Rev. Anti-Infect. Ther. 9 (4), 457‚Äì469. Chegou, N.N., Black, G.F., Loxton, A.G., et al., 2012. Potential of novel Mycobacterium tuberculosis infection phase-dependent antigens in the diagnosis of TB disease in a high burden setting. BMC Infect. Dis. 12, 10. Corach, D., Lao, O., Bobillo, C., et al., 2010. Inferring continental ancestry of Argentineans from autosomal, Y-chromosomal and mitochondrial DNA. Ann. Hum. Genet. 74 (1), 65‚Äì76. Davila, J., McNamara, L.A., Yang, Z., 2012. Comparison of the predicted population coverage of tuberculosis vaccine candidates Ag85B-ESAT-6, Ag85B-TB10.4, and Mtb72f via a bioinformatics approach. PLoS One 7 (7), e40882. Diel, R., Goletti, D., Ferrara, G., et al., 2011. Interferon-gamma release assays for the diagnosis of latent Mycobacterium tuberculosis infection: a systematic review and meta-analysis. Eur. Respir. J. 37 (1), 88‚Äì99. Dye, C., Scheele, S., Dolin, P., Pathania, V., Raviglione, M.C., 1999. Consensus statement. Global burden of tuberculosis: estimated incidence, prevalence, and mortality by country. WHO Global Surveillance and Monitoring Project. JAMA 282 (7), 677‚Äì686. Ganguly, N., Siddiqui, I., Sharma, P., 2008. Role of M. tuberculosis RD-1 region encoded secretory proteins in protective response and virulence. Tuberculosis (Edinb.) 88 (6), 510‚Äì517. Goletti, D., Butera, O., Vanini, V., et al., 2010. Response to Rv2628 latency antigen associates with cured tuberculosis and remote infection. Eur. Respir. J. 36 (1), 135‚Äì142. Hozumi, H., Tsujimura, K., Yamamura, Y., et al., 2013. Immunogenicity of dormancyrelated antigens in individuals infected with Mycobacterium tuberculosis in Japan. Int. J. Tuberc. Lung Dis. 17 (6), 818‚Äì824. Leyten, E.M., Lin, M.Y., Franken, K.L., et al., 2006. Human T-cell responses to 25 novel antigens encoded by genes of the dormancy regulon of Mycobacterium tuberculosis. Microbes Infect. 8 (8), 2052‚Äì2060. Lin, M.Y., Geluk, A., Smith, S.G., et al., 2007. Lack of immune responses to Mycobacterium tuberculosis DosR regulon proteins following Mycobacterium bovis BCG vaccination. Infect. Immun. 75 (7), 3523‚Äì3530. Mensah, G.I., Addo, K.K., Tetteh, J.A., et al., 2014. Cytokine response to selected MTB antigens in Ghanaian TB patients, before and at 2 weeks of anti-TB therapy is characterized by high expression of IFN-gamma and Granzyme B and inter-individual variation. BMC Infect. Dis. 14, 495. Pai, M., Denkinger, C.M., Kik, S.V., et al., 2014. Gamma interferon release assays for detection of Mycobacterium tuberculosis infection. Clin. Microbiol. Rev. 27 (1), 3‚Äì20. Perkins, N.J., Schisterman, E.F., 2006. The inconsistency of ‚Äúoptimal‚Äù cutpoints obtained using two criteria based on the receiver operating characteristic curve. Am. J. Epidemiol. 163 (7), 670‚Äì675. Reither, K., Katsoulis, L., Beattie, T., et al., 2014. Safety and immunogenicity of H1/IC31(R), an adjuvanted TB subunit vaccine, in HIV-infected adults with CD4+ lymphocyte counts greater than 350 cells/mm3: a phase II, multi-centre, double-blind, randomized, placebo-controlled trial. PloS One 9 (12), e114602. Riano, F., Arroyo, L., Paris, S., et al., 2012. T cell responses to DosR and Rpf proteins in actively and latently infected individuals from Colombia. Tuberculosis 92 (2), 148‚Äì159. Rosenkrands, I., Slayden, R.A., Crawford, J., Aagaard, C., Barry III, C.E., Andersen, P., 2002. Hypoxic response of Mycobacterium tuberculosis studied by metabolic labeling and proteome analysis of cellular and extracellular proteins. J. Bacteriol. 184 (13), 3485‚Äì3491. Sauzullo, I., Vullo, V., Mastroianni, C.M., 2015. Detecting latent tuberculosis in compromised patients. Curr Opin Infect Dis 28 (3), 275‚Äì282. Schuck, S.D., Mueller, H., Kunitz, F., et al., 2009. IdentiÔ¨Åcation of T-cell antigens speciÔ¨Åc for latent Mycobacterium tuberculosis infection. PLoS One 4 (5), e5590. Sester, M., Sotgiu, G., Lange, C., et al., 2011. Interferon-gamma release assays for the diagnosis of active tuberculosis: a systematic review and meta-analysis. Eur. Respir. J. 37 (1), 100‚Äì111. Shingadia, D., Novelli, V., 2008. The tuberculin skin test: a hundred, not out? Arch. Dis. Child. 93 (3), 189‚Äì190. van Dissel, J.T., Soonawala, D., Joosten, S.A., et al., 2011. Ag85B-ESAT-6 adjuvanted with IC31(R) promotes strong and long-lived Mycobacterium tuberculosis speciÔ¨Åc T cell responses in volunteers with previous BCG vaccination or tuberculosis infection. Vaccine 29 (11), 2100‚Äì2109. Vukmanovic-Stejic, M., Reed, J.R., Lacy, K.E., Rustin, M.H., Akbar, A.N., 2006. Mantoux test as a model for a secondary immune response in humans. Immunol. Lett. 107 (2), 93‚Äì101. Youden, W.J., 1950. Index for rating diagnostic tests. Cancer 3 (1), 32‚Äì35. Zerbini, E.V., 2013. Programa Nacional de Control de la Tuberculosis: Normas T√©cnicas. Instituto Nacional de Enfermedades Respiratorias Dr. Emilio Coni, Santa Fe, Argentina. ˇ˛Articles Anastrozole for prevention of breast cancer in high-risk postmenopausal women (IBIS-II): an international, double-blind, randomised placebo-controlled trial Jack Cuzick, Ivana Sestak, John F Forbes, Mitch Dowsett, Jill Knox, Simon Cawthorn, Christobel Saunders, Nicola Roche, Robert E Mansel, Gunter von Minckwitz, Bernardo Bonanni, Tiina Palva, Anthony Howell, on behalf of the IBIS-II investigators* Summary Background Aromatase inhibitors effectively prevent breast cancer recurrence and development of new contralateral tumours in postmenopausal women. We assessed the efficacy and safety of the aromatase inhibitor anastrozole for prevention of breast cancer in postmenopausal women who are at high risk of the disease. Methods Between Feb 2, 2003, and Jan 31, 2012, we recruited postmenopausal women aged 40 70 years from 18 countries into an international, double-blind, randomised placebo-controlled trial. To be eligible, women had to be at increased risk of breast cancer (judged on the basis of specific criteria). Eligible women were randomly assigned (1:1) by central computer allocation to receive 1 mg oral anastrozole or matching placebo every day for 5 years. Randomisation was stratified by country and was done with blocks (size six, eight, or ten). All trial personnel, participants, and clinicians were masked to treatment allocation; only the trial statistician was unmasked. The primary endpoint was histologically confirmed breast cancer (invasive cancers or non-invasive ductal carcinoma in situ). Analyses were done by intention to treat. This trial is registered, number ISRCTN31488319. Findings 1920 women were randomly assigned to receive anastrozole and 1944 to placebo. After a median follow-up of 5 0 years (IQR 3 0 7 1), 40 women in the anastrozole group (2%) and 85 in the placebo group (4%) had developed breast cancer (hazard ratio 0 47, 95% CI 0 32 0 68, p<0 0001). The predicted cumulative incidence of all breast cancers after 7 years was 5 6% in the placebo group and 2 8% in the anastrozole group. 18 deaths were reported in the anastrozole group and 17 in the placebo group, and no specific causes were more common in one group than the other (p=0 836). Interpretation Anastrozole effectively reduces incidence of breast cancer in high-risk postmenopausal women. This finding, along with the fact that most of the side-effects associated with oestrogen deprivation were not attributable to treatment, provides support for the use of anastrozole in postmenopausal women at high risk of breast cancer. Funding Cancer Research UK, the National Health and Medical Research Council Australia, Sanofi-Aventis, and AstraZeneca. Lancet 2014; 383: 1041 48 Published Online December 12, 2013 http://dx.doi.org/10.1016/ S0140-6736(13)62292-8 This online publication has been corrected. The corrected version first appeared at thelancet.com onMarch21,2014 See Comment page 1018 Copyright © Cuzick et al. Open Access article distributed under the terms of CC BY *Listed in the appendix Centre for Cancer Prevention, Wolfson Institute of Preventive Medicine, Queen Mary University of London, London, UK (Prof J Cuzick PhD, I Sestak PhD, J Knox MSc); Australian New Zealand Breast Cancer Trials Group, Calvary Mater Newcastle, University of Newcastle, Waratah, NSW, Australia (Prof J F Forbes MD); Academic Department of Biochemistry (ProfMDowsettPhD)and Breast Unit (N Roche MD), The Royal Marsden NHS Trust, London, UK; Breast Care Centre, Southmead Hospital, Bristol,UK(SCawthornMD); UniversityofWestern Australia,Crawley,WA, Australia (Prof C Saunders MD); DepartmentofSurgery, UniversityofWalesCollegeof Medicine,Cardiff,UK (ProfREManselMD); German Breast Group, Neu-Isenburg, Germany (G von Minckwitz MD); University Women s Hospital, Frankfurt, Germany (G von Minckwitz); Division of Chemopreventionand Genetics,EuropeanInstituteof Oncology, Milan, Italy (BBonanniMD);Pirkanmaa Cancer Society, Tampere, Finland (T Palva MD); and GenesisBreastCancer Prevention Centre, Manchester, UK (Prof A Howell MD) Introduction Breast cancer is the most common form of cancer in women,with1 4millionnewcasesreportedworldwidein 2008.1 Its incidence is rapidly increasing, largely because of an ageing population, rising socioeconomic status, increasesinobesity,andseverallifestylechanges,suchas decreases in physical activity, later age at first childbirth, andreductionsinbreastfeeding.Althoughimprovements in lifestyle are an important part of breast cancer prevention, as they are for cardiovascular disease, prophylactic treat- ment is also likely to have an important role, especially for women at high risk (ie, 10-year risk of 5% or more). Oestrogen is a key factor in breast cancer carcino- genesis, and reductions in its synthesis can decrease breast cancer risk. Oestrogen production is driven by the aromatase enzyme, which converts androgens to oestrogens. Trials in the adjuvant setting have shown that aromatase inhibitors more effectively prevent breast cancer recurrence2 4 and also development of new contralateral tumours3,5 in postmenopausal women than does tamoxifen. In a meta-analysis,6 tamoxifen and three other selective oestrogen receptor modulators were shown to reduce the frequency of oestrogen-receptor- positive tumours by 51% overall, but no effect was reported for oestrogen-receptor-negative tumours. The reduction in contralateral tumours has proved an important surrogate for the preventive effects of tamoxifen6,7 and has been confirmed in a trial of the aromatase inhibitor exemestane,8 but whether this reduction extends to other agents is unclear. One study of the preventive effects of an aromatase inhibitor has been done in high-risk women without breast cancer: in the MAP.3 trial,8 exemestane was compared with placebo in postmenopausal women. Exemestane significantly reduced the incidence of all breast cancer by 53% and invasive breast cancer by 65% after a median follow-up of 3 years.8 No serious side- effects of exemestane were recorded, but median follow-up was fairly short for detection of any serious adverse events.8 Here, we report the first results from the International Breast cancer Intervention Study II (IBIS-II), in which the efficacy and safety of the aromatase www.thelancet.com Vol 383 March 22, 2014 1041 Articles Correspondence to: Prof Jack Cuzick, Centre for Cancer Prevention, Wolfson Institute of Preventive Medicine, Queen Mary University of London, London EC1M 6BQ, UK j.cuzick@qmul.ac.uk See Online for appendix inhibitor anastrozole for prevention of breast cancer are being compared with placebo. Methods Study design and participants IBIS-II is an international, double-blind, randomised placebo-controlled trial. Between Feb 2, 2003, and Jan 31, 2012, postmenopausal women aged 40 70 years were recruited in 153 centres in 18 countries (appendix). Women were deemed to be postmenopausal when they were aged 60 years or older; had had a bilateral oophorectomy; were younger than 60 years, but had a uterus and had had amenorrhoea for at least 12 months; or were aged less than 60 years, had no uterus, and had a concentration of follicle stimulating hormone of greater than 30 IU/L. Entry criteria were designed to include women aged 45 60 years who had a relative risk of breast cancer that was at least two times higher than in the general population, those aged 60 70 years who had a risk that was at least 1 5 times higher, and those aged 40 44 years who had a risk that was four times higher. Full eligiblity criteria are listed in the appendix; to be eligible, women had to meet at least one of the criteria. Women who did not meet other eligibility criteria were included if the Tyrer-Cuzick model indicated a 10-year risk of breast cancer of more than 5%.9 Exclusion criteria were: premenopausal status; any previous diagnosis of breast cancer (except for oestrogen- receptor-positive ductal carcinoma in situ diagnosed less than 6 months previously and treated by mastectomy); any invasive cancer in the previous 5 years (except for non-melanoma skin cancer or cervical cancer); present or previous use of selective oestrogen receptor modulators for more than 6 months (unless as part of IBIS-I and treatment was completed at least 5 years before study entry); intention to continue hormone replacement therapy; prophylactic mastectomy; evidence of severe osteoporosis (T score < 4 or more than two vertebral fractures); life expectancy of fewer than 10 years; psychologically or physiologically unfit for the study; or a history of gluten or lactose intolerance, or both. The trial was approved by the UK North West Multi- centre Research Ethics Committee and was done in accordance with the Declaration of Helsinki, under the principles of good clinical practice. Participants provided written informed consent. Randomisation and masking Eligible women were randomly assigned (1:1) by central computer allocation to either anastrozole or matching placebo. Randomisation was stratified by country and was done with randomly chosen randomisation blocks (size six, eight, or ten) to maintain balance. All IBIS-II personnel, participants, and clinicians were masked to treatment allocation; only the trial statistician had access to unblinded data. Procedures Women received 1 mg oral anastrozole or matching placebo every day for 5 years. The primary endpoint was histologically confirmed breast cancer (invasive cancers or non-invasive ductal carcinoma in situ). Secondary endpoints were oestrogen-receptor-positive breast cancer, breast cancer mortality, other cancers, cardiovascular disease, fractures, adverse events, and deaths not due to breast cancer. 3864 women underwent randomisation 1920 assigned to anastrozole group 1944 assigned to placebo group 6 were ineligible 3 premenopausal 2 had >2 spinal fractures 1 ER-negative ductal carcinoma in situ 1914 received anastrozole 1937 received placebo 7 were ineligible 3 premenopausal 2 had previous diagnosis of breast cancer 1 ER-negative ductal carcinoma in situ 1 adopted* Figure 1: Trial profile ER=oestrogen receptor. *Could not give family history to establish whether high risk. Age (years) Parous Age at first child birth (years) Height (cm) Body-mass index (kg/m2) <25 25 30 >30 Use of hormone replacement therapy within previous 12 months Two or more first-degree or second-degree relatives with breast or ovarian cancer One first-degree relative with bilateral breast cancer Oestrogen-receptor-positive ductal carcinoma in situ treated by mastectomy within 6 months Data are median (IQR) or n (%). Anastrozole group (n=1920) 59 5 (55 0 63 5) 1601 (83%) 24 0 (21 0 27 0) 162 0 (158 0 166 0) 581 (30%) 699 (36%) 640 (33%) 128 (7%) 956 (50%) 164 (9%) 160 (8%) Placebo group (n=1944) 59 4 (55 1 63 3) 1637 (84%) 24 0 (21 0 27 0) 162 2 (158 0 167 0) 568 (29%) 732 (38%) 644 (33%) 152 (8%) 938 (48%) 141 (7%) 166 (9%) Age at menarche (years) 13 0 (1 2 14 0) 13 0 (12 0 14 0) Age at menopause (years) 50 0 (45 0 52 0) 49 0 (45 0 52 0) Weight (kg) 71 8 (64 0 82 2) 72 1 (64 0 83 5) Previous use of hormone replacement therapy 893 (47%) 910 (47%) Hysterectomy 631 (33%) 656 (34%) One first-degree relative with breast cancer at age 50 years or younger 675 (35%) 653 (34%) Lobular carcinoma in situ or atypical hyperplasia 154 (8%) 190 (10%) 10-year Tyrer-Cuzick risk (%) 7 6% (5 8 9 9) 7 8 (5 1 10 2) Table 1: Baseline characteristics 1042 www.thelancet.com Vol 383 March 22, 2014 Articles Women visited local clinics at baseline, 6 months, and 12 months, and then annually until the 5-year follow-up point. At baseline after enrolment but before random- isation women had a mammogram and physical breast examination to exclude any pre-existing breast cancer, unless they had undergone these procedures within 12 months before enrolment. Mammograms were then done at least every 2 years. Women also had a dual energy x-ray absorptiometry scan and two spinal radiographs in the lateral dimensions at baseline to assess bone density, unless they had undergone these procedures within 2 years before enrolment. Follow-up after 5 years varied and consisted of a mixture of clinic visits, annual questionnaires, and also record linkage systems in the UK. Blood samples were taken at baseline, after 1 year, and after 5 years for assessment of potential biomarkers. A detailed exploration of changes in bone mineral density, fractures, and use of bisphosphonates will be reported elsewhere. Statistical analysis Analyses were done on an intention-to-treat basis. A secondary per-protocol sensitivity analysis was done after some enrolled women were subsequently identified as ineligible. Initial assumptions for power calculation were based on an incidence of six cases of breast cancer per 1000 women per year, and a compliance-adjusted reduction in incidence of breast cancer of 50% with anastrozole. This calculation led to a sample size of 4000 women. However, interim figures indicated that incidence of breast cancer was higher than predicted: the overall event rate was 6 6 cases of breast cancer per 1000 women per year, which, with a 50% reduction in the anastrozole group, would translate to nine cases of breast cancer per 1000 women per year for placebo. Therefore, the sample size was reduced to 3500 women. The expected number of new cancers after a median of 5 years of follow-up for a total trial size of 3500 women was 78 in the placebo group and 39 in the anastrozole group, leading to a power in excess of 90% for a 5% significance level. Analyses of the efficacy endpoints were based on hazard ratios (HRs). Cox proportional hazards models10,11 were used to derive HRs with 95% CIs. Survival curves were estimated with the Kaplan-Meier method.12 Results are presented for predefined or common (affecting at least 5% of participants) adverse events, or those for which a significant difference between groups was recorded (with an ± of 0 02). Side- effects and secondary endpoints were compared with relative risks. Adherence was calculated by the Kaplan- Meier method, with censoring at breast cancer occur- rence, death, 5 years of follow-up, or the cutoff date. Fisher s exact tests were used to compare frequency of adverse events when appropriate. All p values were two sided. All analyses were done in Stata (version 12.1). This trial is registered, number ISRCTN31488319. Figure 2: Analyses by type of breast cancer Numbers in subgroups do not match totals because of missing data. ER=oestrogen receptor. Number of women Hazard ratio (95% CI) 0 50 (0 32 0 76) 0 42 (0 25 0 71) 0 78 (0 35 1 72) 0 30 (0 12 0 74) 0 47 (0 32 0 68) p value 0 001 0 001 0 538 0 009 <0 0001 All invasive cancers Invasive ER-positive cancers Invasive ER-negative cancers Ductal carcinoma in situ All Anastrozole group (n=1920) 32 (2%) 20 (1%) 11 (1%) 6 (<1%) 40 (2%) Placebo group (n=1944) 64 (3%) 47 (2%) 14 (1%) 20 (1%) 85 (4%) 0 1 0 2 0 5 1 2 5 Hazard ratio 10 Placebo group Invasive ER-positive cancers in placebo group Anastrozole group Invasive ER-positive cancers in anastrozole group 5 5 6% 3 3% 2 8% 1 4% Number at risk 0 01234567 Time since randomisation (years) Placebo group 1944 1927 1645 1445 1241 975 706 506 Anastrozole group 1920 1909 1654 1463 1264 978 720 516 www.thelancet.com Vol 383 March 22, 2014 1043 Figure 3: Cumulative incidence of all breast cancers and of invasive ER-positive breast cancers ER=oestrogen receptor. Role of the funding source The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. IS had full access to all the data in the study, and JC, JFF, and AH had final responsibility for the decision to submit for publication. Results 3864 women underwent randomisation (figure 1). Median age at entry was 59 5 years (IQR 55 0 63 5) and 695 (18%) were older than 65 years. 1803 women (47%) had previously used hormone replacement therapy, and 1287 (33%) had had a hysterectomy (table 1). 1894 (49%) had two or more first-degree relatives who had had breast or ovarian cancer, and 1328 (34%) had one first-degree relative who had had breast cancer when aged 50 years or younger (table 1, appendix). 326 women (8%) had been diagnosed with oestrogen- receptor-positive ductal carcinoma in situ within the previous 6 months and been treated by mastectomy, and 344 (9%) had a benign lesion with a diagnosis of lobular carcinoma in situ or atypical hyperplasia (table 1). 13 women were shown to be ineligible after Cumulative incidence (%) Articles Grade* Low Intermediate High Nodal status Positive Negative Tumour size d"10 mm 10 20 mm >20 mm Oestrogen-receptor status* Positive Negative Progesterone-receptor status* Positive Negative All 7 8 16 29 9 26 12 16 18 44 11 19 8 28 13 17 20 47 11 14 9 28 15 22 32 64 0 86 (0 31 2 38) 0 55 (0 30 1 01) 0 35 (0 16 0 74) 0 75 (0 35 1 58) 0 41 (0 23 0 70) 0 58 (0 27 1 21) 0 28 (0 13 0 62) 0 76 (0 37 1 56) 0 42 (0 25 0 71) 0 78 (0 35 1 72) 0 32 (0 15 0 67) 0 68 (0 35 1 31) 0 50 (0 32 0 76) 0 01 0 08 0 1 0 08 0 2 Number of women Anastrozole group Placebo group (n=1920) (n=1944) Hazard ratio ptrend (95% CI) 0 1 0 2 0 5 1 2 5 Hazard ratio 7-year risk in placebo group* Number of women who developed breast cancer Hazard ratio (95% CI) 0 47 (0 28 0 80) 0 46 (0 27 0 78) 0 43 (0 20 0 90) 0 49 (0 26 0 94) 0 47 (0 26 0 85) 0 52 (0 31 0 78) 0 31 (0 12 0 84) 0 47 (0 31 0 71) 0 44 (0 17 1 15) 0 36 (0 20 0 62) 0 61 (0 37 1 03) 0 30 (0 08 1 07) Age (years) Anastrozole group (n=1920) 20 20 10 14 16 35 5 34 6 17 23 3 Placebo group (n=1944) 44 41 23 28 34 66 19 72 13 47 38 12 d"60 4 1% >60 6 7% Body-mass index (kg/m2) <25 4 7% 25 30 5 2% >30 6 9% Lobular carcinoma in situ or atypical hyperplasia No 4 9% Yes 12 1% Ductal carcinoma in situ No 5 2% Yes 9 7% Previous use of hormone replacement therapy No 6 0% Yes 5 3% Less than 12 months before enrolment 9 0% All 0 2 0 5 1 2 Hazard ratio 1044 www.thelancet.com Vol 383 March 22, 2014 Figure 4: Analyses by invasive breast cancer characteristics Numbers in subgroups do not match totals because of missing data. *Assessed at local laboratories. Figure 5: Subgroup comparisons *Cumulative risk calculated with Cox proportional hazards model. randomisation (figure 1) and were excluded from a The cutoff date for analysis was May 15, 2013. Median secondary per-protocol analysis. No new cancers follow-up was 5 0 years (IQR 3 0 7 1). 19399 women- occurred in this group and the omission of these women did not change the results (data not shown). years of follow-up had been accrued (9727 in the anastrozole group vs 9672 in the placebo group). At the time of data Articles Breast cancer Cerebrovascular accident or stroke Other Data are n (%). Table 2: Causes of death Anastrozole group (n=1920) 2 (<1%) 2 (<1%) 4 (<1%) Placebo group (n=1944) 0 2 (<1%) 4 (<1%) Other cancer 7 (<1%) 10 (1%) Cardiac arrest 3 (<1%) 1 (<1%) Total 18 (1%) 17 (1%) Anastrozole group (n=1920) 14 (1%) 10 (1%) 4 (<1%) Endometrial cancer Thyroid cancer Cancer of the nervous system Ovarian cancer Carcinomatosis 1 Data are n (%), unless otherwise stated. *p=0 005. Placebo group (n=1944) 27 (1%) 20 (1%) 7 (<1%) 5 (<1%) 2 (<1%) 0 7 (<1%) 1 (<1%) Risk ratio (95% CI) 0 53 (0 28 0 99) 0 51 (0 24 1 08) 0 58 (0 17 1 97) 0 61 (0 15 2 54) 0 58 (0 17 1 97) 1 01 (0 06 16 18) Skin cancer Non-melanoma Melanoma Gastrointestinal cancer 4 (<1%) 12 (1%) 0 34 (0 11 1 04) Colorectal 3 (<1%) 11 (1%) 0 28 (0 08 0 99) 3 (<1%) 0 3 (<1%) 4 (<1%) Leukaemia, lymphoma, or myeloma 4 (<1%) 7 (<1%) 0 58 (0 17 1 97) Cancer of the urinary tract 2 (<1%) 5 (<1%) 0 41 (0 08 2 08) Lung cancer 4 (<1%) 4 (<1%) 1 01 (0 25 4 04) Vaginal cancer 1 (<1%) 0 (<1%) Total* 40 (2%) 70 (4%) 0 58 (0 39 0 85) Table 3: Frequency of cancers other than breast cancer lock, 979 women (51%) in the anastrozole group and 975 (50%) in the placebo group had completed 5 years of treatment. We estimated full 5-year adherence to be 68% in the anastrozole group versus 72% in the placebo group (p=0 0047; appendix). The main reasons for treatment discontinuation were adverse events (375 [20%] in the anastrozole group; 298 [15%] in the placebo group) and patient refusal (94 [5%] in the anastrozole group; 98 [5%] in the placebo group). At the cutoff date, 357 women (19%) in the anastrozole group and 450 (23%) in the placebo group were continuing with treatment. Significantly more breast cancers (including ductal carcinoma in situ) were recorded during follow-up in the placebo group than in the anastrozole group (HR 0 47, 95% CI 0 32 68; p<0 0001; figure 2). The predicted cumulative incidence of all breast cancers after 7 years in the placebo group was double that in the anastrozole group (figure 3), suggesting that 36 women (95% CI 33 44) would need to be treated with anastrozole to prevent one cancer in 7 years of follow-up. Invasive oestrogen-receptor- positive tumours were also significantly more common in the placebo group than in the anastrozole group (figures 2, 3), but no significant benefit was recorded for invasive oestrogen-receptor-negative tumours (figure 2). We noted no evidence of heterogeneity for invasive cancers (p=0 3). Anastrozolereducedfrequencyofhigh-gradetumours significantly more effectively than it reduced frequency of low-grade tumours (figure 4). We recorded no sig- nificant heterogeneity in the effect of anastrozole in different subgroups, but larger differences were noted for oestrogen-receptor-positive, progesterone-receptor- positive, and node-negative tumours (figure 4). When models were adjusted for age, body-mass index, previous use of hormone replacement therapy, and smoking status, we recorded similar HRs as for univariate analyses (data not shown). Further details for ductal carcinoma in situ according to treatment allocation are shown in the appendix. Further exploratory analyses did not show any hetero- geneity according to subgroups divided by age, body-mass index, previous use of hormone replacement therapy, and ductal carcinoma in situ, although non-significantly larger effects were recorded for women with lobular carcinoma in situ or atypical hyperplasia and those who had not previously used hormone replacement therapy (figure 5). In the placebo group, the highest 7-year cumulative inci- dences were recorded for lobular carcinoma in situ or atypical hyperplasia (12 1%), followed by ductal carcinoma in situ (9 7%), and none of these lesions (4 1%). 35 deaths had been reported by data cutoff (table 2). No specific causes were more common in one group than in the other (p=0 836; table 2). Overall frequency of cancers other than breast cancer was significantly higher in the placebo group than in the anastrozole group (table 3). Notably, gastrointestinal cancers (p=0 05) and skin cancers overall (p=0 05) were more common in the placebo group than in the anastrozole group (table 3). Many adverse events were reported (table 4). Total number of fractures and number of fractures in specific sites did not differ significantly by group (table 4). 627 (16%) women were taking a bisphosphonate during the trial and concomitant use was similar between treatment groups (330 [17%] in anastrozole group vs 297 [15%] in placebo group). Musculoskeletal adverse events were reported in significantly more women in the anastrozole group than in the placebo group (p=0 0001; table 4). We recorded no significant difference between groups for mild (p=0 9) or severe (p=0 06) arthralgia, but moderate arthralgia was more common with anastro- zole than with placebo (p=0 01; table 4). Carpal tunnel syndrome and joint stiffness were both significantly more common in the anastrozole group than in the placebo group (table 4). Vasomotor symptoms were common in both groups, but significantly more frequent with anastrozole than placebo (p<0 0001; table 4). Signifi- cantly more women taking anastrozole than those taking placebo reported dry eyes (table 4). Vaginal or uterine www.thelancet.com Vol 383 March 22, 2014 1045 Articles Any Musculoskeletal Arthralgia* Mild Moderate Severe Joint stiffness Pain in hand or foot Carpal tunnel syndrome or nerve compression Gynaecological Vaginal dryness Haemorrhage or bleeding Vaginal or uterine prolapse Vulvovaginal pruritus Eye Dry eyes Conjunctivitis Glaucoma Cataract Anastrozole group (n=1920) 1709 (89%) 1226 (64%) 972 (51%) 385 (20%) 422 (22%) 151 (8%) 143 (7%) 178 (9%) 67 (3%) 460 (24%) 357 (19%) 65 (3%) 13 (1%) 40 (2%) 348 (18%) 83 (4%) 12 (1%) 12 (1%) 90 (5%) Placebo group (n=1944) 1723 (89%) 1124 (58%) 894 (46%) 386 (20%) 363 (19%) 123 (6%) 96 (5%) 147 (8%) 43 (2%) 423 (22%) 304 (16%) 81 (4%) 31 (2%) 60 (3%) 335 (17%) 58 (2%) 5 (<1%) 24 (1%) 95 (5%) Risk ratio (95% CI) 1 00 (0 98 1 03) 1 10 (1 05 1 16) 1 10 (1 03 1 18) 1 01 (0 89 1 15) 1 18 (1 04 1 33) 1 24 (0 99 1 56) 1 51 (1 17 1 94) 1 23 (0 99 1 51) 1 58 (1 08 2 30) 1 10 (0 98 1 24) 1 19 (1 03 1 37) 0 82 (0 60 1 13) 0 42 (0 22 0 81) 0 68 (0 45 1 00) 1 05 (0 92 1 21) 1 45 (1 04 2 01) 2 43 (0 86 6 88) 0 51 (0 25 1 00) 0 96 (0 72 1 27) Fractures 164 (9%) 149 (8%) 1 11 (0 90 1 38) Arm 66 (3%) 61 (3%) 1 10 (0 78 1 54) Leg 65 (3%) 57 (3%) 1 15 (0 81 1 64) Rib, spine, or collarbone 23 (1%) 18 (1%) 1 29 (0 70 2 39) Pelvic or hip 9 (<1%) 10 (1%) 0 91 (0 37 2 24) Skull 1 (<1%) 1 (<1%) 1 01 (0 06 16 18) Vasomotor* 1090 (57%) 961 (49%) 1 15 (1 08 1 22) Mild 550 (29%) 504 (26%) 1 10 (1 00 1 22) Moderate 390 (20%) 330 (17%) 1 20 (1 05 1 37) Severe 150 (8%) 127 (7%) 1 20 (0 95 1 50) Vascular 152 (8%) 127 (7%) 1 27 (0 97 1 52) Hypertension 89 (5%) 55 (3%) 1 64 (1 18 2 28) Myocardial infarction or cardiac failure 8 (<1%) 9 (<1%) 0 90 (0 35 2 32) Thrombosis or embolism 19 (1%) 17 (1%) 1 13 (0 59 2 17) Phlebitis 9 (<1%) 8 (<1%) 1 14 (0 44 2 95) Cerebrovascular accident 3 (<1%) 6 (<1%) 0 51 (0 13 2 02) Infections 230 (12%) 217 (11%) 1 07 (0 90 1 28) Influenza 25 (1%) 12 (1%) 2 11 (1 06 4 19) Otitis media 18 (1%) 6 (<1%) 3 04 (1 21 7 64) Data are n (%), unless otherwise stated. Details of any reported adverse event were recorded at every follow-up visit. Adverse events shown here are those that were predefined, common (affecting at least 5% of participants), or differed significantly (p <0 02) between groups. *Assessments of severity broadly based on Common Terminology Criteria for Adverse Events, but some discretion used by clinicians. Hot flushes or night sweats. Table 4: Adverse events of any severity reported at any time 1046 www.thelancet.com Vol 383 March 22, 2014 prolapse and vaginal pruritus were also significantly reduced with anastrozole (table 4). Hypertension was significantly increased with anastrozole, but we recorded no significant differences in frequencies of thrombo- embolic events, cerebrovascular events, or myocardial infarction (table 4). Discussion We have shown that anastrozole substantially reduces incidence of breast cancer in the first 7 years of follow-up in women at high risk. Our results are similar to those recorded with exemestane in the MAP.3 trial.8 The reduction in incidence that we have reported is greater than that recorded for selective oestrogen receptor modulators such as tamoxifen.6 The effect of tamoxifen has been shown to persist for at least 10 years,6,13 and further follow-up is needed to establish whether anastrozole has such a sustained effect. We noted reductions in frequency of breast cancer in most subgroups of participants, although anastrozole s effect seemed to be increased in women with lobular carcinoma in situ or atypical hyperplasia. This increased effect was also shown in two prevention trials of tamoxifen.14,15 An intriguing finding in our study was that anastrozole s effect seemed to be greatest for high-grade tumours. Although highly significant, this finding could have been a result of chance, because other indicators of aggressive or fast growing tumours (eg, node positivity and large tumour size) were not differentially affected. As in MAP.3,8 we recorded no significant differences between groups for cardiovascular events, but musculo- skeletal and vasomotor symptoms were increased with anastrozole. Additionally, frequency of carpal tunnel syndrome was significantly higher with anastrozole, as was noted in the ATAC trial,16 although the disorder was still fairly rare. The high frequency of musculoskeletal and vasomotor symptoms in the placebo group is notable, because they are usually linked with an aromatase inhibitor in non-randomised comparisons.17 We have also confirmed an increase in frequency of hypertension with anastrozole, as was first reported in the ATAC trial.18 A new exploratory finding is the significant increase in frequency of dry eyes with anastrozole, although the total number of events was small. Mixed findings relating to dry eyes in the menopause and hormone replacement therapy have been reported.19 Oestrogenic and androgenic receptors are located on corneal and conjunctival epithelia,19,20 but possible effects of aromatase inhibitors on vision have been previously linked with retinal changes.21,22 We know of only two uncontrolled reports in which dry eyes have previously been associated with aromatase inhibitors.21,23 In one,23 sicca syndrome of the eyes and mouth was associated with anastrozole in patients with probable Sjogren s syndrome. However, in our study, only four cases of Sjogren s syndrome were reported three with anastrozole and one with placebo. Further validation of the increased frequency of dry eyes in women taking an aromatase inhibitor is merited. The reduced frequency of cancers other than breast cancer recorded in the anastrozole group is surprising, especially for colorectal cancers, in which hormone replacement therapy is known to be protective24 and for which the ATAC trial suggested a non-significant increase with anastrozole compared with tamoxifen in the adjuvant setting.3 Likewise, the reduction in non-melanoma skin cancer has not been reported previously with aromatase inhibitors, although the skin is known to be a site of aromatase activity.25 It is also interesting that incidence of endometrial cancer did not reduce, because increased oestrogen concentrations are a strong risk factor for this disease.26 Additionally, a substantially decreased risk of endometrial cancer with anastrozole was recorded in the ATAC trial,3 although the comparator was tamoxifen which is known to increase risk of endometrial cancer.14,27,28 Strengths of this study are the large number of breast cancer events recorded and the median follow-up of 5 years, which is longer than for previous trials. Further follow-up is needed to fully assess the value of anastrozole in the prevention setting. Although a wide range of entry criteria were used in this trial, we recruited few women because of their breast density, which is a strong risk factor for the identification of high-risk women.29,30 Establishment of whether an aromatase inhibitor is effective in such a population is needed. We have shown that anastrozole reduces the risk of invasive oestrogen-receptor-positive breast cancer and ductal carcinoma in situ by more than 50%, but that it has little effect on oestrogen-receptor-negative cancers. The reported reductions are larger than are those reported for tamoxifen or raloxifene.5 Therefore, anastrozole is an attractive option for postmenopausal women at increased risk of breast cancer. Although many side-effects recorded have been associated with oestrogen deprivation, they were only slightly more frequent in the anastrozole group than in the placebo group, indicating that most of these symptoms are not drug related. No additional side-effects have been recorded with anastrozole after treatment completion in the adjuvant setting,3 which is likely to be true in the preventive setting as well. Full adherence for 5 years was 70% overall and only slightly lower in the anastrozole group than in the placebo group. Overall adherence at 3 years was 75%, which is similar to that in the MAP.3 trial,8 which had 85% overall adherence at 35 months. Adherence in our study was slightly better than for tamoxifen in IBIS-I,14 but our findings emphasise the need to understand and Articles Panel: Research in context Systematic review We searched PubMed before our study began for reports published in English between Jan 1, 1980, and Dec 31, 2001. We used the search terms breast cancer , prevention , and aromatase inhibitor . We identified no other trials of breast cancer prevention with an aromatase inhibitor. However, we identified several adjuvant trials in which contralateral tumours were reported.5 Before the planned analysis, we used the same criteria to search PubMed again for reports published before May 30, 2013. Only one other prevention trial with exemestane had been reported,8 and updated or new results for contralateral tumours had been reported for some of the adjuvant trials. We also identified an overview of selective oestrogen receptor modulators for breast cancer prevention.6 Finally, we identified two large trials in which aromatase inhibitors are being assessed for prevention of ductal carcinoma in situ (ISRCTN37546358 and NCT00053898), but results have not been reported. Interpretation Overall, our data suggest that aromatase inhibitors are the most effective agents available for breast cancer prevention. Follow-up in our trial was longer than that in the MAP.3 prevention trial8 and adjuvant trials, and we recorded substantially more events. Equally important is the finding that most side-effects associated with oestrogen deprivation were not attributable to the treatment; most were also increased in the placebo group. Because anastrozole and exemestane have greater efficacies than do tamoxifen and raloxifene, and have a different but generally decreased side-effect profile, anastrozole or exemestane emerge as the treatments of choice for risk reduction in most postmenopausal women at high risk of breast cancer. www.thelancet.com Vol 383 March 22, 2014 1047 1048 www.thelancet.com Vol 383 March 22, 2014 Articles minimise dropout. Dissemination of the fact that most side-effects are not treatment related could help. In the USA, the American Society of Clinical Oncology task force has recommended that exemestane be con- sidered for prevention in addition to tamoxifen and raloxifene,31 and in the UK, the National Institute for Health and Care Excellence has recommended that tamoxifen and raloxifene be offered to women at high risk of breast cancer.32 Our results strongly support the use of anastrozole for preventive treatment of high-risk postmenopausal women (panel). Contributors JC, JFF, MD, SC, CS, NR, REM, GvM, BB, TP, and AH designed the study. JC, JFF, SC, CS, NR, REM, GvM, BB, TP, and AH collected data. JC and IS analysed data and wrote the report. JC, IS, JFF, MD, SC, CS, NR, REM, GvM, BB, TP, and AH interpreted data. JK managed the project. Conflicts of interest JC received funding for IBIS-II from Sanofi-Aventis and AstraZeneca, and is a paid member of a speaker s bureau for AstraZeneca. JFF has received grant support from Novartis. MD has received grant support from and is a paid member of a speakers bureau for AstraZeneca. The other authors declare that they have no conflicts of interest. Acknowledgments This study was funded by Cancer Research UK (C569/A5032), the National Health and Medical Research Council Australia (GNT300755, GNT569213), Sanofi-Aventis, and AstraZeneca. Sanofi-Aventis and AstraZeneca provided anastrozole and matching placebo. The study sponsor was Queen Mary University of London. EBioMedicine 1 (2014) 64‚Äì71 Contents lists available at ScienceDirect EBioMedicine journal homepage: www.ebiomedicine.com Original Article An Endotoxin Tolerance Signature Predicts Sepsis and Organ Dysfunction at Initial Clinical Presentation Olga M. Pena a,1, David G. Hancock b,1, Ngan H. Lyle a, Adam Linder c, James A. Russell a, Jianguo Xia a, Christopher D. Fjell a,c, John H. Boyd c, Robert E.W. Hancock a,d,‚Åé a Centre for Microbial Diseases and Immunity Research, 2259 Lower Mall Research Station, University of British Columbia, Vancouver, British Columbia V6T 1Z4, Canada Flinders University Medical School, GPO Box 2100, Adelaide 5001, South Australia, Australia Division of Critical Care Medicine, University of British Columbia at St Paul's Hospital, P3311 1081 Burrard Street, Vancouver, British Columbia V6Z 1Y6, Canada d Wellcome Trust Sanger Institute, Cambridgeshire, United Kingdom b c a r t i c l e i n f o Article history: Received 5 July 2014 Received in revised form 6 October 2014 Accepted 7 October 2014 Available online 7 October 2014 Keywords: Sepsis Severe sepsis Diagnosis Cellular reprogramming Endotoxin tolerance Signature Immune dysfunction a b s t r a c t Background: Sepsis involves aberrant immune responses to infection, but the exact nature of this immune dysfunction remains poorly deÔ¨Åned. Bacterial endotoxins like lipopolysaccharide (LPS) are potent inducers of inÔ¨Çammation, which has been associated with the pathophysiology of sepsis, but repeated exposure can also induce a suppressive effect known as endotoxin tolerance or cellular reprogramming. It has been proposed that endotoxin tolerance might be associated with the immunosuppressive state that was primarily observed during late-stage sepsis. However, this relationship remains poorly characterised. Here we clarify the underlying mechanisms and timing of immune dysfunction in sepsis. Methods: We deÔ¨Åned a gene expression signature characteristic of endotoxin tolerance. Gene-set test approaches were used to correlate this signature with early sepsis, both newly and retrospectively analysing microarrays from 593 patients in 11 cohorts. Then we recruited a unique cohort of possible sepsis patients at Ô¨Årst clinical presentation in an independent blinded controlled observational study to determine whether this signature was associated with the development of conÔ¨Årmed sepsis and organ dysfunction. Findings: All sepsis patients presented an expression proÔ¨Åle strongly associated with the endotoxin tolerance signature (p b 0.01; AUC 96.1%). Importantly, this signature further differentiated between suspected sepsis patients who did, or did not, go on to develop conÔ¨Årmed sepsis, and predicted the development of organ dysfunction. Interpretation: Our data support an updated model of sepsis pathogenesis in which endotoxin tolerance-mediated immune dysfunction (cellular reprogramming) is present throughout the clinical course of disease and related to disease severity. Thus endotoxin tolerance might offer new insights guiding the development of new therapies and diagnostics for early sepsis. ¬© 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/3.0/). 1. Introduction Sepsis continues to be the major infection-related cause of death globally. In the United States alone, more than 120,000 persons die of sepsis each year (Martin et al., 2003). Despite modern medical advances including new antibiotics and vaccines, best practice treatments, and wellequipped intensive care units (Angus et al., 2001), sepsis mortality rates often remain high at ~30% (Lyle et al., 2014; Jawad et al., 2012). Bacterial endotoxins, such as lipopolysaccharide (LPS), are potent inducers of ‚Åé Corresponding author at: Centre for Microbial Diseases and Immunity Research, 2259 Lower Mall Research Station, University of British Columbia, Vancouver, British Columbia V6T 1Z4, Canada. E-mail address: bob@hancocklab.com (R.E.W. Hancock). 1 These authors contributed equally to this work. inÔ¨Çammation and have been suggested as triggers for the observed hyper-inÔ¨Çammation in sepsis, as well as the early life-threatening cytokine storm causing septic shock (Salomao et al., 2012). However despite these inÔ¨Çammatory components of sepsis, more than 30 clinical trials testing anti-inÔ¨Çammatory agents for the treatment of sepsis have shown no beneÔ¨Åt (Lyle et al., 2014; Hotchkiss et al., 2013). This has contributed to a shift in our understanding of sepsis, from a condition of hyper-inÔ¨Çammation to one characterised by phases of inÔ¨Çammation and immune dysfunction/immunosuppression (Hotchkiss et al., 2013). Whilst our understanding of the immune dysfunction phase in sepsis remains limited (Lyle et al., 2014; Hotchkiss et al., 2013), one of the many hypotheses attempting to characterise the immune state in sepsis has suggested a role for endotoxin tolerance in the later stages of this process (Cavaillon et al., 2005; Otto et al., 2011; Schefold et al., 2008). Endotoxin tolerance, also termed cell reprogramming, can be deÔ¨Åned as the severely http://dx.doi.org/10.1016/j.ebiom.2014.10.003 2352-3964/¬© 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/3.0/). O.M. Pena et al. / EBioMedicine 1 (2014) 64‚Äì71 reduced capacity of a cell to respond to LPS during a second exposure to this stimulus and represents an immune amnesia rather than an antiinÔ¨Çammatory response (Cavaillon and Adib-Conquy, 2006). Other bacterial products can similarly induce reprogramming (Buckley et al., 2006). Despite some similarities in the cytokine production proÔ¨Åles of cells isolated from sepsis patients at late-stage sepsis time-points and in cells from in vitro endotoxin tolerance models (Cavaillon et al., 2005; Otto et al., 2011; Schefold et al., 2008), a robust link has yet to be made between sepsis and endotoxin tolerance. Obtaining a clear understanding of the inÔ¨Çammatory and immunosuppressive phases, including the clinical timepoints at which each occurs or predominates, is likely crucial to improving sepsis outcomes. Here, we applied robust bioinformatics approaches to in-house and previously-published cohorts of early-stage sepsis patients. We showed that sepsis is characterised by an endotoxin tolerance phenotype that occurs very early during the clinical course of disease and is linked to disease severity. This places endotoxin tolerance as a novel therapeutic and diagnostic target in sepsis that may be used to predict the development of sepsis and organ failure in critically ill patients. 2. Materials and Methods 2.1. Gene Signature DeÔ¨Ånition and Analysis Endotoxin tolerance and inÔ¨Çammatory gene signatures were derived from our previously-published (Pena et al., 2011) microarray analyses of human PBMC identifying differentially-expressed genes compared to control PBMCs (GSE22248), with the inÔ¨Çammatory signature being delimited to genes that overlapped with a human volunteer endotoxin challenge (Calvano et al., 2005), as described in Fig. 1. Gene lists for the signatures are found in Supplementary Tables 1 and 2. Analysis of the presence or absence of the endotoxin tolerance and inÔ¨Çammatory signatures in patients and controls was performed using the well-established, statistically-rigorous gene set test ROAST (Wu et al., 2010), that asks whether a given set of genes/signature is enriched in a dataset. The ROAST method increases the strength of the test by additionally allowing the consideration of the direction of gene expression (Wu et al., 2010). The ROAST test is designed for any linear modelled data and was therefore suitable for both the microarray and RNA-Seq 65 data, evaluated using the linear model in the Limma package (Wu et al., 2010; Smyth, 2004), ROAST, which delivers a speciÔ¨Åc p-value for the association of a given gene set with a particular condition (e.g. sepsis), was run with 99,999 rotations and so the lowest possible p-value arising from this test is 0.00001. 2.2. Meta-Analysis Datasets A search without Ô¨Ålter restrictions was performed in the US National Library of Medicine (PubMed) Database, the public repositories National Centre for Biotechnology Information-Gene Expression Omnibus (NCBIGEO) and the European Bioinformatics Institute (EBI-ArrayExpress). The search terms used individually or in combination, include ‚Äúsepsis‚Äù, ‚Äúseptic shock‚Äù, ‚Äúsepticaemia‚Äù, ‚Äúbacteraemia‚Äù, ‚Äúmicroarrays‚Äù, ‚ÄúRNA-Seq‚Äù, ‚Äúcluster analysis‚Äù, ‚Äútranscription proÔ¨Åling‚Äù, ‚Äúgene expression‚Äù, ‚ÄúLPS‚Äù, ‚Äúendotoxin‚Äù, ‚ÄúinÔ¨Çammation‚Äù, and ‚Äúinfection‚Äù. Datasets were also manually searched in review articles. The Ô¨Ånal selection of datasets was based on the inclusion and exclusion criteria described in Supplementary Table 3. Within datasets, selected samples were excluded if they did not meet the inclusion criteria (e.g. SIRS patients that could not be classiÔ¨Åed as having sepsis due to lack of clinical information). In addition, exclusion criteria were: 1) studies only analysing a small number of genes (e.g.: RT-qPCR); 2) single-nucleotide polymorphism studies; 3) studies analysing only a single gene or pathway; 4) studies using resident immune cells (e.g. alveolar macrophages); and 5) studies using solid organ tissues. It is important to mention that we found several datasets (especially with adult populations) using SIRS patients as controls. However, for this meta-analysis we selected datasets using healthy controls, as these allowed us to observe even minor transcriptional changes present in sepsis despite the stage of the disease (i.e. early or late stage). Parallel with this, we performed an initial in-house clinical study of 22 sepsis patients by analysing gene expression responses using RNA-Seq. 2.3. Patient Selection and New Clinical Study Design In a blinded, observational, controlled cohort study, patients with suspected sepsis were identiÔ¨Åed when the attending physician activated Fig. 1. DeÔ¨Ånition of the ‚Äòendotoxin tolerance signature‚Äô. Schematic representation of the method used to deÔ¨Åne the endotoxin tolerance signature and the inÔ¨Çammatory signature. The endotoxin tolerance signature was obtained from our previously published dataset (Pena et al., 2011) and deÔ¨Åned as 99 genes uniquely differentially expressed in endotoxin-tolerant human PBMCs (treated twice with LPS), but not inÔ¨Çammatory human PBMCs (treated once with LPS), as compared to controls (fold change N 2, p-value b0.05) (please see ref. Pena et al., 2011 for more details regarding this dataset). The inÔ¨Çammatory signature was reduced from the 178 genes uniquely differentially expressed in inÔ¨Çammatory human PBMCs to a common 93 gene signature, by selecting genes that were consistently differentially expressed in an in vivo human volunteer endotoxin challenge dataset (Calvano et al., 2005). 66 O.M. Pena et al. / EBioMedicine 1 (2014) 64‚Äì71 the Institutional Severe Sepsis Order Set (Supplementary Fig. 1). Patients were enrolled from St. Paul's Hospital, Vancouver Canada, at the time of the Ô¨Årst microbiological culture drawn for suspected sepsis. To determine the appropriate sample size for this study we used a standard power calculation for adequate sensitivity (Jones et al., 2003). To achieve a sensitivity of at least 0.9 at a 95% conÔ¨Ådence level, we estimated a required sample size of 35 sepsis patients and 70 patients total (assuming that 50% of patients with a suspicion of sepsis actually have sepsis). We recruited 72 total patients which proved subsequently to include 37 sepsis patients. The inclusion criteria for this study were the suspicion of sepsis by the attending physician, activation of the Institutional Severe Sepsis Order Set, and successfully obtaining samples for microbial cultures. The majority of patients (83%) were enrolled from the emergency room. As shown in Supplementary Table 4, these individuals were heterogeneous. Our UBC ethical approval protocol enabled deferred consent allowing early patient recruitment in cohorts that spanned from non-infected to septic shock. As controls, we recruited consented healthy individuals, with no evidence of infection, who were scheduled for non-urgent surgery. Blood was collected in EDTA tubes at the time of initial blood culture, and immediately placed on ice. Plasma and buffy coat were separated and two 1-ml aliquots transferred into bar-coded cryovials at ‚àí20 ¬∞C until they were transferred to a secure, alarmed ‚àí80 ¬∞C freezer. Study identiÔ¨Åcation numbers were assigned to the secured enrolment forms and used during all subsequent analyses; thus researchers analysing gene expression in these patients were blinded as to patient identity or clinical course, which was only revealed during Ô¨Ånal data analysis. Clinical data was stored in an ORACLE-based database on a Ô¨Årewalled, RSS encrypted server at St Paul's Hospital. Clinical data was collected retrospectively by physician researchers blinded to the RNA-Seq data. Sepsis was retrospectively deÔ¨Åned as suspected or proven infection in addition to at least two of the following assessments: Initial WBC, b4000 or N 12,000 per Œºl; Triage Temperature b36C or N 38 ¬∞C; and Triage Heart Rate, N 90 bpm. New organ dysfunction was deÔ¨Åned as outlined in Supplementary Table 5 and based on laboratory values collected in the electronic medical record system. Initial vital signs were retrospectively extracted from the paper records. 2.4. Ethical Conduct of Research All studies were performed under UBC ethics approval [IDs H1100505 for patient sample collection and H08-00293 for RNA-Seq analysis]. Our UBC ethical approval protocol H11-00505 enabled deferred consent allowing early patient recruitment in cohorts that spanned from non-infected to septic shock and subsequent approval by patients. 2.5. RNA-Seq Transcriptomic analysis was performed by the high throughput sequencing of cDNAs (RNA-Seq). cDNA libraries were prepared from total RNA using the TruSeq Stranded Total RNA Sample Prep Kit with a Ribo-Zero sample preparation guide (Illumina). RNA-Seq was performed on a GAIIx instrument (Illumina), using a single read run of 63 bp-long sequence reads (+ adapter/index sequences). A standard analysis protocol was used whereby raw basecall data was converted to FASTQ sequence Ô¨Åles using Off-Line Basecaller 1.9.4 (Illumina) and a custom Perl script. Reads were aligned to the hg19 human genome with TopHat version 2.06 and Bowtie2 2.0.0-beta6 (Trapnell et al., 2009) and mapped to Ensembl transcripts. Raw data was deposited into NCBI GEO. 2.6. Data Analysis All data processing was performed in R using Bioconductor modules (Gentleman et al., 2004). For the meta-analysis, normalised datasets were downloaded from NCBI GEO using the Bioconductor package GEOquery (Davis and Meltzer, 2007). An additional quantile normalisation step was included if the data required further normalisation. For the RNA-Seq analysis, data was normalised using the Voom function in the Limma package which converts read counts to weighted log base 2 counts per million. For both the metaanalysis and RNA-Seq analyses, data was summarised using the linear model in the Limma package (Smyth, 2004; Gentleman et al., 2004). 2.7. ClassiÔ¨Åcation Analysis Each dataset was split into training (containing 2/3 of sepsis patients and controls) and test (containing 1/3 of sepsis patients and controls) sets, using random sampling. A model was deÔ¨Åned on the training set and then assessed on the test set using the randomForest package (Liaw and Wiener, 2002) with ntree set to 1000. The procedure was repeated 100 times, and the average AUC values were recorded for each dataset. Note that AUC (also known as AUROC) = Area Under Receiver Operator Curve (TPR (Sensitivity) vs. FPR (1-SpeciÔ¨Åcity) curve) is an indicator of the accuracy of diagnosis, based on the Endotoxin Tolerance Signature, for the compared groups. AUC values were calculated using the ROCR package (Sing et al., 2005). 3. Results 3.1. Endotoxin Tolerance Signature in Early Sepsis To assess whether endotoxin tolerance (cellular reprogramming) contributes to the immune dysfunction observed in sepsis, we Ô¨Årst deÔ¨Åned genetic signatures of endotoxin tolerance and inÔ¨Çammation (Fig. 1). The ‚Äòendotoxin tolerance signature‚Äô (Supplementary Table 1) comprised 99 genes that were uniquely differentially expressed in endotoxin-tolerant PBMCs, but not inÔ¨Çammatory PBMCs, as compared to non-LPS-exposed controls (Pena et al., 2011). An ‚ÄòinÔ¨Çammatory signature‚Äô (Supplementary Table 2) was deÔ¨Åned based on genes differentially regulated in inÔ¨Çammatory PBMCs but not in endotoxin-tolerant PBMCs by combining dysregulated genes present in our published dataset (Pena et al., 2011) and an experimental human endotoxinchallenge dataset (Calvano et al., 2005). We Ô¨Årst recruited an initial cohort of 22 adult patients with conÔ¨Årmed sepsis at various timepoints (1 to 3 days after the initial clinical suspicion of sepsis) throughout the early clinical course of disease. To increase the number of sepsis patients in this analysis and to reduce the limitations of a single-centre study, we also performed a retrospective global meta-analysis on 10 published, independent and blinded clinical sepsis cohorts, encompassing 571 early sepsis patients (1 or 3 days post-ICU admission) and 160 healthy controls (Supplementary Table 3). Healthy controls were used as the basis for comparison to allow for the detection of smaller changes in gene expression and to limit study‚Äìstudy variability in the control population used as baseline. To assess the relative expression of the Endotoxin Tolerance and inÔ¨Çammatory signatures in sepsis patients versus healthy controls, we used a gene-set test approach, which examines whether there is differential enrichment of a given signature (gene-set) between groups (Wu et al., 2010). We found that sepsis patients in all 11 cohorts (in-house and meta-analysis) showed an immunological expression proÔ¨Åle strongly associated with the endotoxin tolerance signature when compared to controls (Fig. 2). Whilst the inÔ¨Çammatory signature was significantly associated with eight of the datasets, this association was consistently weaker than for the endotoxin tolerance signature (Supplementary Fig. 2). In contrast to previous reports associating endotoxin tolerance only with late stage sepsis (Cavaillon et al., 2005; Otto et al., 2011; Schefold et al., 2008), the association with the ‚Äòendotoxin tolerance signature‚Äô was present in sepsis patients as early as Day 1 post-ICU admission, and was maintained on Day 3, consistent with the early development of a ‚Äòstable‚Äô endotoxin tolerance proÔ¨Åle in O.M. Pena et al. / EBioMedicine 1 (2014) 64‚Äì71 Fig. 2. Sepsis patients from published datasets showed a strong association with the ‚Äòendotoxin tolerance signature‚Äô. A gene-set test approach, ROAST (Wu et al., 2010), testing the statistically signiÔ¨Åcant presence of a signature (collection) of genes, was used to characterise the enrichment of ‚ÄòEndotoxin Tolerance‚Äô in sepsis patients versus controls from a small study performed by us and 10 previously published datasets. All datasets contained sepsis patients recruited at day 1 or 3 post-ICU admission and were compared to ‚Äòhealthy‚Äô controls. The ROAST gene-set test was run with 99,999 rotations so the most signiÔ¨Åcant p-value resulting from this test is 0.00001. p-Values from the ROAST gene-set test were graphed as log (1/p-value), and untransformed p-values are shown for ease of visualization. sepsis patients (Fig. 2). Thus the early immune dysfunction in Sepsis appeared to be characterised by endotoxin tolerance/cellular reprogramming. 3.2. Endotoxin Tolerance Signature at First Clinical Presentation We next sought to better understand the timing of endotoxin tolerance development in early sepsis. To do this, we recruited a unique blinded, prospective, observational cohort of patients at the earliest possible stage of clinical disease. Patients were recruited immediately after clinical suspicion of sepsis (i.e. prior to diagnosis), based on the attending physician's physical examination and request for microbial culture testing. RNA-Seq was performed on RNA isolated from the initial blood sample taken for cultures to aid in sepsis diagnosis/microbial identiÔ¨Åcation. We recruited 72 very early suspected sepsis patients (sufÔ¨Åcient power to achieve 90% sensitivity, see the Materials and Methods), as well as 11 control patients recruited prior to elective surgery with no underlying morbidities (Supplementary Table 4). Since only a subset of the patients in this ‚Äòcritically ill‚Äô patient cohort would go on to develop true sepsis, this cohort represented a clinically-challenging cohort of patients who initially presented with variable serious derangements in physiology (potentially caused by sepsis). Based on secondary clinical assessments following sample isolation (Supplementary Table 4), patients were retrospectively classiÔ¨Åed as ‚ÄòSepsis‚Äô (n = 37), or ‚ÄòNo Sepsis‚Äô (n = 35), consistent with current sepsis diagnostic criteria (see the Materials and Methods) (Lyle et al., 2014; Bone et al., 1992; Vincent et al., 2013; Levy et al., 2003). Strikingly, even at the earliest stage of clinical sepsis, the endotoxin tolerance signature was signiÔ¨Åcantly enriched in patients who were subsequently 67 conÔ¨Årmed to have sepsis (‚ÄòSepsis‚Äô group), but not in those with other diagnoses (‚ÄòNo Sepsis‚Äô group) (Fig. 3a). Whilst the inÔ¨Çammatory signature did not reach statistical signiÔ¨Åcance in the ‚ÄòNo Sepsis‚Äô group, the contrasting relative enrichment of the Endotoxin Tolerance and inÔ¨Çammatory signatures in the 2 groups may indicate a fundamental difference in the balance of endotoxin tolerance and inÔ¨Çammation unique to sepsis patients (Fig. 3a). The endotoxin tolerance signature was also enriched in the ‚ÄòSepsis‚Äô group when directly compared to the ‚ÄòNo Sepsis‚Äô group (Fig. 3b), which supports the speciÔ¨Åcity of endotoxin tolerance to sepsis and not just to ‚Äòill‚Äô patients. To further exclude the possibility that the signature was only detecting critically ill patients, we also determined that the endotoxin tolerance signature was not signiÔ¨Åcantly enriched in acute kidney transplant rejection (p value = 0.213) or myocardial infarction (p value = 0.433) patients (Kurian et al., 2014; Silbiger et al., 2013), compared to healthy controls. Together these data suggest that endotoxin tolerance is present throughout the initial clinical course of sepsis, detectable before ‚Äòdiagnosis‚Äô, and can be used to differentiate patients who develop sepsis in a cohort of patients where there was a suspicion of sepsis. Current deÔ¨Ånitions of sepsis (Lyle et al., 2014; Bone et al., 1992; Vincent et al., 2013; Levy et al., 2003) refer to suspected or conÔ¨Årmed infection together with other systemic abnormalities such as initial WBC count, triage temperature and heart rate, whilst severe sepsis usually involves one or more organ failures. All patients here were suspected to have infection upon Ô¨Årst clinical presentation which was conÔ¨Årmed in 19 of the 37 individuals who were eventually diagnosed with sepsis (Supplementary Table 4). Nevertheless, similar levels of signiÔ¨Åcance of association of the endotoxin tolerance signature with sepsis were observed for this group as a whole, and for that subset of the group with conÔ¨Årmed infections (p b 0.01). To further highlight this concept, we compared signature enrichment in the ‚ÄòSepsis‚Äô and ‚ÄòNo Sepsis‚Äô groups following further separation based on microbial culture results. Separating the groups based on culture results did not change the overall associations between endotoxin tolerance and the ‚ÄòNo Sepsis‚Äô group (Fig. 3d). Indeed amongst the ‚ÄòSepsis‚Äô group, the endotoxin tolerance signature was signiÔ¨Åcantly enriched in the culture positive group, although there was a trend towards enrichment in the culture negative group (Fig. 3c). This is consistent with the sensitivity issues of bacterial culturing methods and the challenge in diagnosing true infectionpositive patients in those with suspected infection. Moreover, since our RNA-Seq analysis was performed on the same blood samples used for diagnostic microbial cultures, the strong association between sepsis and our endotoxin tolerance signature suggests that this signature might provide a more sensitive tool for diagnosis than microbial culture. It has been commonly postulated that the immunosuppressive state in sepsis is related to the end-stage development of organ dysfunction, and so we sought to address this using our clinical cohort. Subsequent organ dysfunction development (cardiovascular, coagulation, kidney, liver, and respiratory, Supplementary Tables 4, 5) was assessed up to 48 h after study enrolment, with patients retrospectively grouped into organ-dysfunction positive and negative groups, independent of sepsis diagnosis. These groups were then subjected to the same gene-set test analysis, as above. Interestingly, the endotoxin tolerance signature was found to be signiÔ¨Åcantly associated with the development of several individual and multiple (3 +) organ dysfunction(s) (Fig. 4a). Although ICU admission may depend on the inherent subjectivity of hospital practice, such as space or number of beds available in each department, patients that are moved to the ICU are generally in a deteriorating condition with an increased risk of mortality. Therefore, we also assessed the requirement for ICU admission as a second, less precise measure of disease severity and showed that endotoxin tolerance signature was again associated with this indicator of increased disease severity (Fig. 4b). These results indicated that endotoxin tolerance appears to be associated with sepsis severity and speciÔ¨Åcally with the downstream development of organ failure. 68 O.M. Pena et al. / EBioMedicine 1 (2014) 64‚Äì71 Fig. 3. The ‚Äòendotoxin tolerance signature‚Äô was strongly associated with sepsis patients at Ô¨Årst clinical presentation. A gene-set test approach (Wu et al., 2010) was used to characterise the enrichment, cf. controls as well as non-sepsis critically ill patients, of the ‚ÄòEndotoxin Tolerance‚Äô and ‚ÄòInÔ¨Çammatory‚Äô signatures in prospective sepsis patients from a unique in-house cohort recruited on Ô¨Årst clinical suspicion of sepsis (i.e. generally in the emergency ward and before ICU admission cf. the studies described in Fig. 2 that were post-ICU admission). Patient groups were subsequently deÔ¨Åned based on retrospective clinical characteristics as ‚ÄòSepsis‚Äô or ‚ÄòNo Sepsis‚Äô consistent with the current sepsis criteria3,15,16,36 (Supplemental Table 4). Analyses were performed comparing (a) the ‚ÄòSepsis‚Äô and ‚ÄòNo Sepsis‚Äô groups vs. controls and (b) the ‚ÄòSepsis‚Äô and the ‚ÄòNo Sepsis‚Äô groups to each other. Additionally, enrichment of the signature was also analysed based on microbial culture results within (c) the ‚ÄòSepsis‚Äô group and (d) the ‚ÄòNo Sepsis‚Äô group. 3.3. Prognostic Potential Given the strong association between endotoxin tolerance and sepsis across more than 600 patients from 11 independent datasets, we hypothesised that our signature would be a useful tool in sepsis diagnosis. Whilst the full 99 gene, endotoxin tolerance signature was useful for characterising the immune dysfunction in sepsis, a smaller number of genes would be of more use in a diagnostic test. Thus we Ô¨Årst sought to reduce our 99-gene signature to its essential component, prior to testing its diagnostic utility. To do this, we selected genes that showed greater than 1.5 fold differential expression between sepsis patients and controls across the majority (7 +) of the 10 literature datasets. This identiÔ¨Åed a core-set of 31 genes from the original 99 gene endotoxin tolerance signature (Fig. 5). It is worth noting that many of the genes in our endotoxin tolerance signature are individually dysregulated in other disease states; however the combination appeared speciÔ¨Åc for sepsis. We then used the classiÔ¨Åcation algorithm randomForest to preliminarily assess the diagnostic utility of our 31 gene core-set to classify sepsis patients. We divided each dataset (external and internal) into training and test sets and performed randomForest classiÔ¨Åcation independently on each dataset. The core-set showed excellent performance when separating sepsis patients from controls across all datasets (average AUC of 96.1%), and when separating patients who subsequently developed sepsis or individual/combined organ failure in our cohort of patients with a suspicion of sepsis (AUCs from 74.1 to 84.7%, with liver failure lower at 54.1%) (Table 1). The core-set showed improved performance when classifying Sepsis patients with conÔ¨Årmed infection (70.4%), requiring ICU admission (73.6%), and with at least 1 organ failure (75.2%) suggesting that our endotoxin tolerance signature shows Fig. 4. The ‚Äòendotoxin tolerance signature‚Äô was strongly associated with sepsis patients at Ô¨Årst clinical presentation and was associated with the severity of the disease. A gene-set test approach (Wu et al., 2010) was used to characterise the enrichment, cf. surgical controls, of the ‚ÄòEndotoxin Tolerance‚Äô and ‚ÄòInÔ¨Çammatory‚Äô signatures in prospective sepsis patients from a unique in-house cohort recruited on Ô¨Årst clinical suspicion of sepsis (i.e. generally in the emergency ward prior to ICU admission). (a) Patients were grouped into individual-, combined- (3+), individual type of organ failure and no-organ failure groups. NB. No patients were observed with sepsis-associated encephalopathy. (b) Patients were also grouped into those requiring and those not-requiring transfer to the ICU. O.M. Pena et al. / EBioMedicine 1 (2014) 64‚Äì71 69 Fig. 5. A core-set of endotoxin tolerance genes characteristic of sepsis patients. A core-set of 31 of the 99 genes from the ‚Äòendotoxin tolerance signature‚Äô was determined based on the most frequently differentially expressed genes observed literature sepsis datasets. For better visual comparison across different studies, each individual dataset was further transformed by dividing gene expression values into six equal bins. Data is presented as a heatmap with blue and red representing relatively low and high expression, respectively. increased diagnostic performance (and clinical relevance) when identifying patients who go on to develop deÔ¨Ånitive sepsis and more serious disease (Table 1). Similar results were obtained with the full 99 gene endotoxin tolerance signature (Table 1). The strong performance of our classiÔ¨Åer across multiple distinct datasets and at a clinically relevant timepoint (blood draw for diagnostic cultures) supports the potential use of our classiÔ¨Åer in the diagnosis of sepsis. 4. Discussion The association between the endotoxin tolerance signature and conÔ¨Årmed sepsis was strong and statistically signiÔ¨Åcant in a total of 12 distinct datasets (Figs. 2, 3) and as such was independent of sample size, location, method, gender, age and ethnicity. These results are consistent with our hypothesis that the endotoxin tolerance signature is robustly associated with very early sepsis. The endotoxin tolerance signature was also associated with disease severity measured primarily by the development of organ dysfunction. Therefore, we propose here an updated model of sepsis pathogenesis mediated by an endotoxin tolerance-mediated immune dysfunction. This is consistent with but further clariÔ¨Åes a recent study (Hotchkiss et al., 2013) that suggested that early sepsis was associated with coincident inÔ¨Çammatory and anti-inÔ¨Çammatory/immunosuppressive responses. It is worth mentioning that endotoxin tolerance is not an anti-inÔ¨Çammatory state per se but rather a cellular reprogramming (which also occurs with Gram positive bacteria) that leads to immune amnesia, disabling responses to agonists like endotoxin (Cavaillon and Adib-Conquy, 2006; Buckley et al., 2006; Pena et al., 2011). We also demonstrated that this immune dysfunction could be detected at a clinically relevant ‚Äòdiagnostic‚Äô time-point, providing unique information regarding the patients' functional immune status. In the future, our genetic classiÔ¨Åer could help to deÔ¨Åne a subset of patients who might beneÔ¨Åt from immunomodulation (e.g. antiendotoxin tolerance) and supportive therapies. Sepsis has been traditionally classiÔ¨Åed as an early stage excessive inÔ¨Çammatory state followed by a transition to a late stage antiinÔ¨Çammatory/immunosuppressive state (e.g. endotoxin tolerance) (Cavaillon et al., 2005; Otto et al., 2011; Schefold et al., 2008). However, a model of concurrent immunosuppression and hyperinÔ¨Çammation has 70 O.M. Pena et al. / EBioMedicine 1 (2014) 64‚Äì71 Table 1 Diagnostic potential of the endotoxin tolerance signature. Each dataset was split into training (containing 2/3 of sepsis patients and controls) and test (containing 1/3 of sepsis patients and controls) sets using random sampling. Datasets GSE13015 and GSE11755 were omitted from this analysis due to low numbers of controls (N = 3) in each dataset. For each of the remaining 8 datasets, the model was deÔ¨Åned on the training set and then assessed on the test set using the randomForest package (Xiu and Jeschke, 2013) with ntree set to 1000. The procedure was repeated 1000 times, and the average AUC (Area Under Receiver Operator Curve) values (representing the accuracy of the diagnosis), recorded for each dataset. This analysis was repeated on our dataset to classify patients with an initial suspicion of sepsis who did or did not go on to develop sepsis or organ failure. Variable AUC using 31 gene endotoxin tolerance core-set AUC using 99 gene endotoxin tolerance signature Sepsis (patient numbers in brackets) vs. controls In-house Sepsis study #1 (22) vs. controls In-house Sepsis study #2 (37) vs. controls GSE28750 study (30) vs. controls GSE9692 study (45) vs. controls GSE13904 study (227) vs. controls GSE26440 study (130) vs. controls GSE4607 study (84) vs. controls GSE8121 study (71) vs. controls GSE26378 study (70) vs. controls GSE54514 study (35) vs. controls Mean 78.4% 98.1% 100% 99.4% 97.8% 99.2% 98.4% 98.1% 100% 91.5% 96.1% 77.8% 97.1% 100% 99.3% 97.9% 99.1% 98.3% 98.1% 100% 94.4% 95.9% Sepsis vs. No Sepsis ‚Äî study #2 Sepsis vs. No Sepsis Sepsis (Positive Culture vs. No Sepsis) Sepsis (ICU vs. No Sepsis) Sepsis (1+ Organ Failure vs. No Sepsis) 63.9% 70.4% 73.6% 75.2% 66.4% 75.7% 75.6% 74.5% Organ Failure vs. No Organ Failure ‚Äî study #2 Respiratory Cardiovascular Liver Acute kidney injury Coagulation Combined (3+) 84.7% 84.1% 54.1% 76.7% 74.4% 78.4% 84.8% 82.0% 55.3% 79.4% 77.3% 77.6% been recently hypothesised to explain the complex pathogenesis of sepsis at various timepoints (Hotchkiss et al., 2013). The current study indicates that immunosuppression is being driven by endotoxin tolerance that occurs at a much earlier stage of clinical disease than previously appreciated, consistent with the failure of immunosuppressive treatments in more than 30 clinical trials (Lyle et al., 2014). In agreement with the concurrent inÔ¨Çammation/immunosuppression model, we observed signiÔ¨Åcant enrichment of both the endotoxin tolerance and inÔ¨Çammatory signatures at all early-stages of disease, although the endotoxin tolerance signature dominated. If there is an immunological phase characterised solely by excessive inÔ¨Çammation, our data would suggest that this occurs pre-clinically (Supplementary Fig. 3). Although ‚Äòpre-clinical‚Äô disease can represent a relatively large time frame, it is less clinically relevant from a therapeutic/intervention perspective. One possible explanation for the concurrent occurrence of both inÔ¨Çammation and immunosuppression may lie at the immune cell population level. Due to their continuous replenishment from the bone marrow (Summers et al., 2010), neutrophils, which show very little gene expression, are potential drivers of pro-inÔ¨Çammatory cytokine responses, even as longer-lived monocyte/macrophage populations and other perhaps antigen presenting cells (dendritic cells and B lymphocytes) are being locked into and driving the endotoxin tolerance/cellular reprogramming response (Parnell et al., 2013). Therefore, at a systemic level, sepsis appears to be typiÔ¨Åed by a combination of neutrophilic inÔ¨Çammation, driving vascular leakage, coagulation, lymphocyte death, etc. (Hotchkiss and Karl, 2003; de Jong et al., 2010), and monocytic/macrophage reprogramming, impairing immune responses to primary and secondary infections (Hotchkiss et al., 2013; Leentjens et al., 2013; Xiu and Jeschke, 2013). Future studies aimed at assessing the development of cellular reprogramming over time (instead of a single timepoint) and at a cellular level will help clarify these processes. In addition, it will be interesting to examine patients with severe trauma (Hotchkiss et al., 2013) to see if they also have the endotoxin tolerance signature, despite the possibility that they are non-infectious. However, performing longitudinal studies on suspected sepsis patients recruited with deferred consent is logistically challenging. The other important observation in relation to the balance between inÔ¨Çammation and endotoxin tolerance, is the relative enrichment of each signature in critically ill patients who do or do not develop sepsis (Fig. 3). From a biological perspective, our observations may suggest that in individuals with localized infections (e.g. patients in the No Sepsis group), when an initial insult occurs, the brief inÔ¨Çammatory response quickly subsides to balance inÔ¨Çammation and bring the system to homeostasis. However, in sepsis, where there is an uncontrolled source of infection, and possible contributing genetic factors (Murkin and Walley, 2009), the immunological balance between inÔ¨Çammation and endotoxin tolerance becomes detrimentally unbalanced towards a state increasingly dominated by endotoxin tolerance (Supplementary Fig. 3). This model is supported by our observed association between the endotoxin tolerance signature and disease severity/organ dysfunction (Fig. 4). Organ dysfunction is considered the main factor contributing to patient deterioration and ultimately death. Importantly, the endotoxin tolerance signature was present up to 48 h prior to the development of organ dysfunction, suggesting that this signature might be additionally used as a screening method to assess which patients are at a higher risk for developing a worsening condition. Moreover, network analysis (Xia et al., 2014) of the Endotoxin Tolerance genes revealed that 53 of the genes formed a very tight protein‚Äìproteininteraction sub-network suggesting that the signature may identify key genes related to immune dysfunction (and possibly susceptibility to infection) in sepsis patients (Supplementary Fig. 4). Whilst the current study was primarily focused on classifying the immune dysfunction in sepsis, the extremely strong association between endotoxin tolerance and sepsis leads us to explore the potential use of our signature in diagnosis. Indeed, both the gene-set and classiÔ¨Åcation analyses support this use (Table 1, Figs. 1‚Äì3). However, one limitation of this analysis was the relatively low patient numbers in some of the datasets (including our in-house datasets). Whilst the ROAST geneset test (Wu et al., 2010) was designed for much lower group numbers, classiÔ¨Åcation tests (such as randomForest (Xiu and Jeschke, 2013)) are optimally used with larger patient numbers. Thus larger cohort studies, speciÔ¨Åcally designed for assessing the potential of this signature as a O.M. Pena et al. / EBioMedicine 1 (2014) 64‚Äì71 diagnostic tool (especially for the prediction of organ failure in sepsis patients), will likely be required to conÔ¨Årm its potential as a diagnostic signature. However, as the association of endotoxin tolerance and sepsis was robustly identiÔ¨Åed in 12 distinct datasets, endotoxin tolerance is likely to be clinically relevant (as a potential key factor in sepsis aetiology and/or therapeutic target), regardless of the diagnostic utility of our speciÔ¨Åc signatures. In this regard there are a number of agents in development for unlocking an M2 macrophage state (Sica and Mantovani, 2012) that drives endotoxin tolerance (Pena et al., 2011). In conclusion, we have provided a description of a unique endotoxin tolerance gene expression proÔ¨Åle, present very early in the course of sepsis, and linked to sepsis pathogenesis and the risk of developing organ dysfunction. The results of this study should be further tested prospectively in a large multicenter cohort of patients with sepsis using current deÔ¨Ånitions for infection, sepsis, severe sepsis, and multiple system organ failure. Author Contributions OMP was involved in deÔ¨Åning the initial concept and designing the study, discovering the endotoxin tolerance signature, and writing the manuscript; DGH performed most of the Ô¨Ånal bioinformatics and participated in writing the manuscript; NHL was involved in retrieving the clinical data from the medical records and editing the manuscript; AL was involved in retrieving the clinical data from the medical records and critiquing the manuscript; JAR was involved in discussing and critiquing the Ô¨Åndings throughout the study and editing the manuscript; JX was involved in initial bioinformatic analyses, especially network analyses, and provided essential advice regarding determining diagnostic accuracy; CDF was involved in bioinformatic analyses including analysis of raw data from RNA-Seq, and downloading of relevant studies from GEO; JHB was involved in discussing the initial concepts and conducting the in-house clinical study as well as editing the manuscript; and REWH was involved in deÔ¨Åning the initial concept and design of the study, supervising all aspects of the research, and in extensively editing the manuscript. Role of the Funding Source The funding source, Canadian Institutes of Health Research (MOP74493 and MCT-44152), provided support for this research but was not involved in any aspect pertinent to the study. Competing Interests The authors declare no relevant conÔ¨Çicts of interest. The study depicted here has been Ô¨Åled for a US provisional patent application 61/953,458 (inventors Robert Hancock, Olga Pena, David Hancock and John Boyd) which claims the use of the signature for diagnosing sepsis and the use of molecules that reverse the endotoxin tolerant state of immune cells as a therapeutic strategy. Acknowledgements We would like to acknowledge the Canadian Institutes of Health Research (MOP-74493 and MCT-44152) for funding this study. We thank Reza FalsaÔ¨Å and Mihai Cirstea for the excellent technical support in the processing of samples. OP held a Vanier-CIHR Doctoral Scholarship, NHL an AMMI/PÔ¨Åzer Post Residency and UBC Clinician Investigator fellowship, JX a CIHR and UBC Killam postdoctoral fellowship, CDJ a CIHR postdoctoral fellowship, and REWH a Canada Research Chair. Appendix A. Supplementary Data Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ebiom.2014.10.003. 71 References Angus, D., Linde-Zwirble, W., Lidicker, J., Clermont, G., Carcillo, J., Pinsky, M., 2001. Epidemiology of severe sepsis in the United States: analysis of incidence, outcome, and associated costs of care. Crit. Care Med. 29 (7), 1303‚Äì1310. Bone, R.C., Balk, R.A., Cerra, F.B., et al., 1992. DeÔ¨Ånitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine. Chest 2009 136 (5 Suppl.), e28. Buckley, J.M., Wang, J.H., Redmond, H.P., 2006. Cellular reprogramming by gram-positive bacterial components: a review. J. Leukoc. Biol. 80 (4), 731‚Äì741. Calvano, S.E., Xiao, W., Richards, D.R., et al., 2005. A network-based analysis of systemic inÔ¨Çammation in humans. Nature 437 (7061), 1032‚Äì1037. Cavaillon, J., Adib-Conquy, M., 2006. Bench-to-bedside review: endotoxin tolerance as a model of leukocyte reprogramming in sepsis. Crit. Care Med. 10 (5), 233. Cavaillon, J., Adrie, C., Fitting, C., Adib-Conquy, M., 2005. Reprogramming of circulatory cells in sepsis and SIRS. J. Endotoxin Res. 11 (5), 311‚Äì320. Davis, S., Meltzer, P.S., 2007. GEOquery: a bridge between the Gene Expression Omnibus (GEO) and BioConductor. Bioinformatics 23 (14), 1846‚Äì1847. de Jong, H.K., van der Poll, T., Wiersinga, W.J., 2010. The systemic pro-inÔ¨Çammatory response in sepsis. J. Innate Immun. 2 (5), 422‚Äì430. Gentleman, R.C., Carey, V.J., Bates, D.M., et al., 2004. Bioconductor: open software development for computational biology and bioinformatics. Genome Biol. 5 (10), R80. Hotchkiss, R.S., Karl, I.E., 2003. The pathophysiology and treatment of sepsis. N. Engl. J. Med. 348 (2), 138‚Äì150. Hotchkiss, R.S., Monneret, G., Payen, D., 2013. Immunosuppression in sepsis: a novel understanding of the disorder and a new therapeutic approach. Lancet Infect. Dis. 13 (3), 260‚Äì268. Jawad, I., Luksic, I., Rafnsson, S.B., 2012. Assessing available information on the burden of sepsis: global estimates of incidence, prevalence and mortality. J. Glob. Health 2 (1), 010404. Jones, S.R., Carley, S., Harrison, M., 2003. An introduction to power and sample size estimation. Emerg. Med. J. 20 (5), 453‚Äì458. Kurian, S.M., Williams, A.N., Gelbart, T., et al., 2014. Molecular classiÔ¨Åers for acute kidney transplant rejection in peripheral blood by whole genome gene expression proÔ¨Åling. Am. J. Transplant. 14 (5), 1164‚Äì1172. Leentjens, J., Kox, M., van der Hoeven, J.G., Netea, M.G., Pickkers, P., 2013. Immunotherapy for the adjunctive treatment of sepsis: from immunosuppression to immunostimulation. Time for a paradigm change? Am. J. Respir. Crit. Care Med. 187 (12), 1287‚Äì1293. Levy, M.M., et al., 2003. 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis DeÔ¨Ånitions Conference. Intensive Care Med. 29, 530‚Äì538. Liaw, A., Wiener, M., 2002. ClassiÔ¨Åcation and regression by randomForest. R News 2 (3), 18‚Äì22. Lyle, N.H., Pena, O.M., Boyd, J.H., Hancock, R.E.W., 2014. Barriers to the effective treatment of sepsis: antimicrobial agents, sepsis deÔ¨Ånitions and host directed therapies. Ann. N. Y. Acad. Sci. 1323 (1), 101‚Äì114. Martin, G.S., Mannino, D.M., Eaton, S., Moss, M., 2003. The epidemiology of sepsis in the United States from 1979 through 2000. N. Engl. J. Med. 348 (16), 1546‚Äì1554. Murkin, J.M., Walley, K.R., 2009. Genetic susceptibility to inÔ¨Çammatory injury and various adverse outcomes. J. Extra Corpo. Technol. 41 (1), 43‚Äì49. Otto, G.P., Sossdorf, M., Claus, R.A., et al., 2011. The late phase of sepsis is characterized by an increased microbiological burden and death rate. Crit. Care 15 (4), R183. Parnell, G.P., Tang, B.M., Nalos, M., et al., 2013. Identifying key regulatory genes in the whole blood of septic patients to monitor underlying immune dysfunctions. Shock 40 (3), 166‚Äì174. Pena, O.M., Pistolic, J., Raj, D., Fjell, C.D., Hancock, R.E.W., 2011. Endotoxin tolerance represents a distinctive state of alternative polarization (M2) in human mononuclear cells. J. Immunol. 186 (12), 7243‚Äì7254. Salomao, R., Brunialti, M.K., Rapozo, M.M., Baggio-Zappia, G.L., Galanos, C., Freudenberg, M., 2012. Bacterial sensing, cell signaling, and modulation of the immune response during sepsis. Shock 38 (3), 227‚Äì242. Schefold, J.C., Hasper, D., Volk, H.D., Reinke, P., 2008. Sepsis: time has come to focus on the later stages. Med. Hypotheses 71 (2), 203‚Äì208. Sica, A., Mantovani, A., 2012. Macrophage plasticity and polarization: in vivo veritas. J. Clin. Invest. 122 (3), 787‚Äì795. Silbiger, V.N., Luchessi, A.D., Hirata, R.D., et al., 2013. Novel genes detected by transcriptional proÔ¨Åling from whole-blood cells in patients with early onset of acute coronary syndrome. Clin. Chim. Acta 421, 184‚Äì190. Sing, T., Sander, O., Beerenwinkel, N., Lengauer, T., 2005. ROCR: visualizing classiÔ¨Åer performance in R. Bioinformatics 21 (20), 3940‚Äì3941. Smyth, G.K., 2004. Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat. Appl. Genet. Mol. Biol. 3 (Article3). Summers, C., Rankin, S.M., Condliffe, A.M., Singh, N., Peters, A.M., Chilvers, E.R., 2010. Neutrophil kinetics in health and disease. Trends Immunol. 31 (8), 318‚Äì324. Trapnell, C., Pachter, L., Salzberg, S.L., 2009. TopHat: discovering splice junctions with RNA-Seq. Bioinformatics 25 (9), 1105‚Äì1111. Vincent, J.L., Opal, S.M., Marshall, J.C., Tracey, K.J., 2013. Sepsis deÔ¨Ånitions: time for change. Lancet 381, 774‚Äì775. Wu, D., Lim, E., Vaillant, F., Asselin-Labat, M.L., Visvader, J.E., Smyth, G.K., 2010. ROAST: rotation gene set tests for complex microarray experiments. Bioinformatics 26 (17), 2176‚Äì2182. Xia, J., Benner, M.J., Hancock, R.E.W., 2014. NetworkAnalyst ‚Äî integrative approaches for protein‚Äìprotein interaction network analysis and visual exploration. Nucleic Acids Res. 42, W167‚ÄìW174. Xiu, F., Jeschke, M.G., 2013. Perturbed mononuclear phagocyte system in severely burned and septic patients. Shock 40 (2), 81‚Äì88. An exploration of management practices in hospitals * K. John McConnella, , , * Anna Marie Changb, * Thomas M. Maddoxc, * Douglas R. Wholeyd, * Richard C. Lindroothe Show more doi:10.1016/j.hjdsi.2013.12.014 Get rights and content Abstract Background Management practices, including, for example, “Lean” methodologies originally developed at Toyota, may represent one mechanism for improving healthcare performance. Methods We surveyed 597 nurse managers at cardiac units to score management on the basis of poor, average, or high performance on 18 practices across 4 dimensions (Lean operations, performance measurement, targets, and employee incentives). We assessed the relationship of management scores to hospital characteristics (size, non-profit status) and market level variables. Results Our findings provide concrete examples of the high degree of management proficiency of some hospitals, as well as wide variation in management practices. Although the exact ways in which these tools have been implemented vary across hospitals, we identified multiple examples of units that use standardization in their care, track performance on a frequent basis and display data in a visual manner, and set aggressive goals and communicate them clearly to their staff. Regression models indicate that higher management scores are associated with hospitals in more competitive markets, teaching hospitals, and hospitals with a higher net income from patient services (p<0.05). Conclusions High quality management practices have been successfully adopted by some hospitals in the US, but the ways in which these practices have been implemented may vary, reflecting the specific context or environment of the hospital. The adoption of modern management practices may be driven in part by market pressure. Implications An improved understanding of key management practices may assist researchers and policy-makers in identifying mutable hospital characteristics that can drive efficiency, safety, and quality. Keywords * Management; * Lean; * Cardiac units; * Hospitals; * Quality improvement 1. Introduction It is well documented that the U.S. healthcare system has problems with inefficiency and variable quality of care.1, 2 and 3 In efforts to transform their care, a number of hospitals have invested in management practices, including, for example, “Lean” methodologies originally developed at Toyota.4, 5, 6 and 7 These practices typically use a combination of tools with a focus on streamlining workflow, incorporating a mindset of continuous improvement, and eliminating different types of waste. Management practices are not strictly limited to Lean; hospitals are also experimenting with ways to use data to drive change or engage and incentivize their employees. This article has three goals. First, we present a relatively concise framework to measure “management.” By management, we do not mean “disease” or “care” management, which are typically focused on narrower clinical issues. Rather, we investigate 18 organizational questions across four management dimensions: (1) operations; (2) performance monitoring; (3) target setting; and (4) employee incentives. Hospitals across the country have invested heavily in the approaches that parallel these management dimensions. Second, using a large survey of cardiac units across the U.S., we provide concrete examples of management approaches in use today. By providing actual examples of these tools, we hope to improve the potential for evaluating and spreading beneficial practices. Third, having scored management practices as part of previous work, we quantitatively test the association between management scores (as a dependent variable) and hospital characteristics. Overall, we aim to open the black box of hospital operations and provide a more material understanding of how and why management practices are disseminated. 2. Materials and methods 2.1. Survey design To measure management, we took an approach originally developed by economists to assess management in the manufacturing setting8 and 9 and adapted it to the cardiac inpatient setting. We surveyed cardiac units on 18 questions covering four dimensions of management for which a variety of tools and rudimentary empirical evidence has surfaced: (1) Lean operations; (2) performance measurement; (3) target setting; and (4) employee incentives. This framework has been validated through survey work in more than 6000 manufacturing firms8 and 9 as well as a large-scale field experiment in India10and serves as the basis for the newly introduced Management and Organizational Practice Survey (MOPS) component of the US Census.11 Our management surveys used several steps to allow flexibility in measuring management while avoiding the tendency of respondents to answer questions in a way that they believed might be “correct.” First, surveys were conducted by phone, and respondents were not told in advance that they were being scored. Second, to elicit scores, the interview was based on a series of structured open-ended questions (e.g., “Can you tell me how you promote your employees?”), rather than closed questions (e.g., “Do you promote your employees on tenure [yes/no]?”). For each practice, the first questions were broad, continuing with detailed follow-up questions until the interviewer could make an accurate assessment of the organization?s practice, providing a score from 1 to 5, with higher score representing better practices. Table 1 provides a brief description of these 4 groupings and 18 practices, including examples of the questions we asked and an indication of the scoring approach. Additional technical details of the survey have been described previously12 and are available in e-Appendix. Table 1. Measurement and definition of management practices. Practice Examples of questions asked Low score (1) High score (5) Standardizing care/lean operations (1) Admitting the patient How do you determine what types of information will be collected on patients entering the unit? How are admission order sets determined? Little standardization and few protocols exist Protocols/standardized order sets are used by all staff and regularly monitored for compliance (2) Standardization within the unit How standardized are the main clinical processes? Little standardization and few protocols exist Protocols exist for all patients, and are regularly followed up on through some form of monitoring or oversight (3) Coordination on handoffs As a patient is moved to various sites – ED, Cath Lab, CCU, Telemetry Units, etc. – are there protocols for nurse-to-nurse transitions within the cardiac unit? There are no systems in place to verify information in handoffs; little coordinated information is passed on to the next site Inter-area protocols are known and used by all staff and regularly followed up on; communication is interactive and contingency plans are in place (4) Communication among staff How would you describe the ability for nurses to provide input on physician care? Who participates in rounds? Nurses and physicians have little willingness to see issues outside of each (specific) work domain Physicians and nurses communication is interactive; multiple channels exist for soliciting concerns (5) Patient focus What systems are in place to capture patient or family questions or concerns about their care at any point during their visit? Patient satisfaction is rarely measured; no systems in place to capture patient concerns Multiple means in place to capture patient/family concerns; systems capture and respond to feedback (6) Discharging the patient What type of discharge assessments and order sets are used? Who is responsible for the patient after they are discharged from the hospital? Little is known about the patient after discharge Standardized discharge process for all patients, including a formal handoff to follow-up physician Performance measurement (7) Technology adoption Can you take me through the rationale for introducing a new device, drug or therapy in your unit? Adoption of new drugs or technologies is ad hoc or does not follow a regular process Systematic approach in the adoption of new technologies, including review of evidence on costs and effectiveness. (8) Monitoring errors/safety Can you tell me about your strategies for avoiding medical errors? Can you give me an example of how a process changed when a “near miss” occurred? Staff recognize importance of avoiding errors but safety depends on individual efforts Strategies for avoiding errors are in place and monitored; near-misses are seen as opportunities for improvement (9) Continuous improvement How do problems typically get exposed and fixed? Can you talk me through the process for a recent problem that you faced? Process improvements are made only when problems occur, or only involve one staff group Exposing problems in a structured way is integral to individuals? responsibilities (10) Performance review How do you review your main performance indicators? Can you tell me about a recent review meeting? Performance is reviewed infrequently or in an un-meaningful way (e.g., only success or failure is noted) Performance is continually reviewed, based on the indicators tracked; all aspects are followed up to ensure continuous improvement (11) Performance dialog How are these performance review meetings structured? How is the agenda determined? Do you find that you have enough information for review? The right information for a constructive discussion is often not present or quality is low Conversations focus on problem solving and addressing root causes; meetings an opportunity for constructive feedback Targets (12) Target balance What types of targets are set for your unit by the unit itself or by the hospital (i.e., not by the Joint Commission or other regulators)? Goals are focused only on government targets and achieving the budget Goals are a balanced set of targets covering multiple dimensions: (e.g., financial balance, quality, employee satisfaction, patient satisfaction); interplay of all goals is understood by all staff groups (13) Target inter-connection How are these goals cascaded down to the different staff groups or to individual staff members? Goals do not cascade down the organization Goals increase in specificity as they cascade, ultimately defining individual expectations for all staff groups (14) Target stretch How tough are your targets? Do you feel pushed by them? Goals are either too easy or impossible to achieve, at least in part because they are set with little clinician involvement Goals are genuinely demanding for all parts of the organization and developed in consultation with senior staff Employee incentives (15) Rewarding high performers How does your appraisal/review system work? Can you tell me about your most recent round? Are bonuses provided to the group or the individual or a blend? Staff members are rewarded in the same way irrespective of their level of performance There is an evaluation system for the awarding of performance related rewards, including personal financial rewards and shared team rewards (16) Removing poor performers How long is under-performance tolerated? How difficult is it to terminate a nurse/ clinician? Poor performers are rarely removed from their positions We work hard to identify weaknesses and improve or remove poor performers. (17) Managing talent How do you ensure you have enough staff/nurses of the right type in the hospital? Senior staff do not communicate that attracting, retaining and developing talent throughout the organization is a top priority Senior staff are evaluated and held accountable on the strength of the talent pool they actively build (18) Retaining talent If you had a top performing manager, nurse or clinician that wanted to leave, what would the hospital do? We do little to try and keep our top talent We do whatever it takes to retain our top talent across all staff groups Table options The survey was conducted in summer, 2010. Interviews were conducted by a team of 9 individuals with professional and educational experience in healthcare and management, using a standard interview guide. Interviews were scored by two members of the interview team, with one member asking questions and scoring responses, and the second member listening and scoring responses in parallel. We used the American Hospital Association (AHA) Guide to identify hospitals with interventional cardiac catheterization laboratories and to determine hospital contact information. We excluded federal (Veterans Administration) hospitals and hospitals with fewer than 25 annual Medicare discharges with a primary diagnosis of AMI. Following our previous work, we converted the management scores from the original 1- to 5-point scale to z scores (mean, 0; SD, 1). This process mediated scaling differences between questions (e.g., interviewers may have consistently given higher scores for some questions and lower scores for others.) The average management score, used in our regressions, is defined as the average of these z-scores. “Best” practices – those receiving a score of a “5” – represented the interviewer?s assessment of a unit whose performance was at the high end of the scale. We collected multiple descriptions of “average” and “best” practices and present selected examples in this study. 2.2. Administrative data For our regression analyses, we obtained hospital administrative data (i.e., number of beds, teaching status) from the AHA Guide13 and Medicare?s Provider of Service file. A list of accredited Master of Business Administration (MBA) schools was obtained from the Association to Advance Collegiate School of Business (AACSB). Competition measures and AMI volume were based on the 2010 Medicare Provider Analysis and Review file. 2.3. Statistical analysis In order to assess the relationship between management scores and hospital characteristics, we conducted regression analyses that used the average of the management z-scores as a dependent variable. Our independent variables included a measure of competition, the predicted Herfindahl-Hirschman Index (HHI – described in more detail below), net income reported by the hospital for patient services, distance to the nearest AACSB-accredited MBA school, region, ownership, AMI volume (25–75, 76–125, 126–250, and >250 discharges annually), licensed beds (less than 151, 151 to 374, and more than 374), rural vs. urban, teaching status, open heart surgery capability, and hospital system membership. We also included a number of prespecified “interviewer” controls designed mitigate biases across interviewers and types of interviewees.12These included indicator variables for interviewer, interviewee job position (e.g., nurse manager vs. unit director), interviewee location (e.g., intensive care unit vs. telemetry), and the duration, day, and week of the interview HHIs are indices of the competitiveness of the market and were calculated based on a hospital choice model. Following Dranove and colleagues,14 we used HHIs based on predicted zip code-level patient flows instead of actual patient flows, because predicted flows more closely approximated a measure of market pressure, rather than the patients actually referred to the hospital. Patient flows were estimated using a grouped McFadden choice model, with choice of hospital modeled as a function of hospital level variables interacted with the drive time between zip code centroids and all hospitals in the market, patient age range, and per capita income for the patient?s zip code.14 and 15 HHIs can range between 0 and 1, with higher values representing a more concentrated, less competitive market. The study protocol was approved by the institutional review board of Oregon Health & Science University. 3. Results We attempted to contact and schedule interviews with all hospitals identified from administrative data as having interventional cardiology and at least 25 annual AMI discharges. We scored management practices in 597 hospitals, representing a 51.5% response rate. Table 2 displays key hospital characteristics. Surveyed hospitals were slightly more likely to be not-for-profit hospitals and offer cardiac surgery. Table 2. Hospital characteristics and survey response. Characteristic Responded to survey Did not respond P-value N (597) % N (562) % Location 0.07 Rural 85 14 60 11 Urban 512 86 502 89 Ownership 0.03 Public 72 12 53 9 Nonprofit 459 77 419 75 For profit 66 11 90 16 Hospital type 0.62 Teaching 114 18 101 19 Nonteaching 483 82 461 81 Cardiac facilities 0.05 Catheterization only 110 18 130 23 Open heart surgery 487 82 432 77 Licensed beds 0.51 <=150 77 13 82 15 151–374 335 56 321 57 >=375 185 31 159 28 Annual AMI volumea 0.20 25–50 142 24 161 29 51–99 257 43 240 43 100–199 173 29 139 25 >=200 25 4 22 4 P-values designate the statistical significance of the difference between a characteristic of the hospital for hospitals that responded to the survey and those that did not. a AMI volume based on Medicare FFS visits. Table options Table 3 provides an indication of the distribution of specific management practices across our surveyed units, as well as the average score across all practices. Overall, hospitals performed best on Question 10 (Performance Review) and most poorly on Question 18 (Retaining Talent). All of the 18 questions suggest considerable room for improvement: fewer than 50% of hospitals scored a 4 or 5 on all but 2 questions (Q5, Patient Focus, and Q10, Performance Review). Table 3. Distribution of management practice scores in cardiac units. Practice Percent of units receiving score Average across units Score=1 (poor practice) (%) Score=2 (%) Score=3 (typical/average practice) (%) Score=4 (%) Score=5 (best practice) (%) (1) Admitting the patient 2 14 44 31 9 3.3 (2) Standardization within the unit 2 10 43 35 11 3.4 (3) Coordination on handoffs 2 17 54 22 5 3.1 (4) Communication among staff 2 16 46 29 7 3.2 (5) Patient focus 1 12 36 35 16 3.5 (6) Discharging the patient 3 18 50 26 4 3.1 (7) Technology adoption 7 15 44 27 8 3.1 (8) Monitoring errors/safety 2 11 39 36 13 3.5 (9) Continuous improvement 4 17 29 35 15 3.4 (10) Performance review 1 11 31 41 15 3.6 (11) Performance dialog 3 15 33 40 9 3.4 (12) Target balance 7 20 36 25 11 3.1 (13) Target inter-connection 6 25 35 25 10 3.1 (14) Target stretch 8 28 37 23 4 2.9 (15) Rewarding high performers 19 20 40 17 4 2.7 (16) Removing poor performers 9 30 36 22 3 2.8 (17) Managing talent 6 28 38 24 5 2.9 (18) Retaining talent 22 40 28 8 2 2.3 Table options Table 4 provides examples of management approaches in hospitals in the US. We provide an example of an “average” practice (as scored by our interview team) and two examples of “best” practices – as identified by our interview team – for each of the 18 questions. Table 4. Examples of “Average” and “Best” management practices in cardiac units. Practice “Average” practice Best practice – example 1 Best practice – example 2 Standardizing care/lean operations 1. Admitting the patient Predefined admission order sets exist for STEMIa patients but not NSTEMIbpatients. The unit does not monitor how often the standardized order sets are used. Hospital has computerized standardized admission order sets for all patients and clinical pathways. Physicians may deviate from standardized order sets but a stopgap measure is in place that requires immediate approval of the exception for the admission process to continue. Hospital has standardized admission order sets for all patients with chest pain.EMS can initiate “Code STEMIa” is initiated (an out-of-hospital and in-hospital team is alerted and responds immediately). Every STEMIa is reviewed within 24 h; ten random chest pain admission cases are audited a week. 2. Standardization and protocols within the unit Unit has protocols for key procedures (e.g., peripherally inserted catheters). Clinicians receive training when hired and have annual competency checks. Unit has adopted protocols for all procedures from published best practice guidelines. Hospital-wide teams perform specific procedures. All staff complete extensive education and competency on required skills is continually reviewed. Unit monitors key procedures through planned observation and/or submission of checklists reviewed by the quality department. Main clinical processes are standardized and regularly monitored. Bundles exist for all key clinical procedures. Each bundle has an associated checklist that is audited regularly for compliance. Staff must pass competency exams on all unit processes and procedures quarterly. Staff must attend regular practice update and skills review meetings. 3. Coordination on handoffs Handoff report paperwork supplements verbal report. Situation, Background Assessment, Recommendation (SBAR) protocol is used for most but not all handoffs. Nurses will follow-up with staff when there are reports of inappropriate handoffs. Unit uses an automated phone report system; the nurse calls in SBAR report; the handoff nurse receives a return confirmation phone call when the receiving nurse has listened to the report; nurses communicate additional information at this time. Patients are only transferred if electronic chart indicates protocol completion. Managers audit these recorded reports. A protocol, based on SBAR, is used for handoffs between units. Face-to- face two-way communication is required. A pre-printed form is used to document each handoff, accompanies each patient, and becomes a part of the patient?s permanent record. Management monitors the written component and reports compliance along with random observational audits. 4. Communication among staff Nurses are receptive to physician suggestions but find physicians uncooperative. Nurses lead the daily patient rounds; physicians attend at their convenience. The unit uses an intensivist model. An interdisciplinary team rounds daily. Checks and balances are in place on all procedures that allow for interruptions when there is a patient safety concern. Anonymous incident and patient safety reporting systems are in place. Nurses and physicians do not hesitate to speak up or make suggestions. Concerns can be reported anonymously to a clinical safety officer. Patient rounds are multidisciplinary, occurring daily with nurses presenting. Peer reports of communication are incorporated into annual performance reviews. 5. Patient focus The unit encourages feedback from patients; nurse manager conducts customer service rounds to ensure that patients are happy with their care. All patient rooms have a whiteboard that list their care team names and contact information, patient preferences and concerns, and the current care plan, updated every 12 h. Follow-up phone calls are made to every patient within 48 h of discharge; they are asked to complete a survey on their care experience. A patient relations professional meets with each patient during admission and is available during their stay to address any questions or concerns. The manager rounds daily and the primary nurse rounds hourly. Hospitals have a “Condition H” program; enables patient and family members to call for immediate help if they feel they are not receiving the medical attention they need. 6. Discharging the patient Medication reconciliation is standardized but little is done to educate the patient beyond providing discharge instructions. Unit calls high-risk patients within 48 h of discharge. Primary care providers in the hospital?s system can access electronic medical records immediately but others are faxed the patient summary. Discharge begins as soon as patient is admitted. They round on patients every day and try to determine barriers or challenges that would exist for the patient when they go home. Consultation includes pharmaceutical, respiratory, vascular, and related services. Patients are screened for depression before they are discharged. The unit works to identify a PCP for patients (even if none is currently established) and set up an appointment. Discharge planning begins on arrival. The case manager looks at the home environment to assess what aid is needed. The interdisciplinary team determines a care plan for each patient. Pharmacy fills prescriptions before patients are discharged and can arrange monetary assistance if necessary. The unit sets up follow-up appointments before discharge with and faxes records to the patient?s PCP. The case manager verifies that the patient went to the follow-up appointment. Performance measurement 7. Technology adoption Ad hoc committees are formed to evaluate requests for new drugs/devices. The potential profit margin is the greatest driver for adoption. The hospital is a member of a 10-hospital learning collaborative that identifies new drug and devices. The group uses their combined resources to conduct a systematic review of the item. Adoption of the item is based on the results of this review. The products committee reviews all new drug and device requests. The drug/device must meet evidence-based guidelines for safety and efficacy and be comparatively more cost-effective than existing treatments to be approved. The hospital sends representatives to academic research conferences to learn about new drugs/devices. 8. Monitoring errors Double identifiers, name/DOB verification and time-outs protocols are used during all procedures. Nurse Manager performs random audits to ensure compliance. Unit has a non-punitive process supported by computerized system that allows anyone to anonymously enter a safety concern or near miss. All managers meet with risk managers and other departments (e.g., pharmacy) on a weekly basis to review these records. A near miss leads to a team review of processes/systems that can be changed. All high-risk procedures have an associated protocol that is regularly monitored for compliance. Adverse events and near misses are tracked with a Performance Improvement Scorecard. The hospital is contributing member to a learning collaborative that regularly communicates errors and necessary system changes that are relevant to other group members. 9. Continuous improvement Every staff member must complete a quarterly process feedback survey; the quality department collects information on problems in the unit and suggestions for improvement. Each care team “huddles” every 4 h to discuss problems. A member of every care team, on a rotating basis, must attend the weekly unit council meeting and bring forward problems/concerns and work with group to create an action plan. Recommendations must be brought to one of the 7 different hospital staff councils (quality, operations, safety, etc.) for approval. Suggestions for improvement can be written in a “flip chart” available in the staff lounge and these are reviewed weekly. Collected suggestions are the catalyst for changes (e.g., relabeling confusing medications). The nurse manager expects her staff to take ownership of identifying problems and designing solutions. 10. Performance review The nurse manager initiates a meeting when a major problem is brought to her attention. A small group is formed to create the detailed recommendation. The unit monitors and tracks performance indicators based on a balanced scorecard. This is updated and posted daily. The Quality Committee reviews the indicators continuously, and assigns responsibility for creating actionable items and sharing these solutions with all staff and other units. Performance indicators are continuously updated and displayed in the staff lounge and on the unit website. Charts are compiled by the quality department. A weekly multidisciplinary meeting that takes place to discuss changes in performance and organize initiation of a Plan-Do-Check-Act (PDCA) analysis. 11. Performance dialog Review conversations are held with the appropriate data present; objectives of meetings are clear to all participating and a clear agenda is present; conversations do not, drive to the root causes of the problems The leadership has adopted a systematic process for reviewing performance indicators. A data analyst summarizes the results of his report and each team member reads their prepared response. The format of the meeting is very formal and was adopted to ensure that all members are able to speak without domination of one person. Team votes on whether to initiate any performance improvement plans and decision is based on majority opinion. Performance review meetings are structured using the 5 hospital “pillars” or goals. Attendance is mandatory for all staff and feedback is encouraged. The Performance Improvement team addresses low performance, and then smaller staff teams are created to make recommendations for needed changes. The successes and failures of previous improvement strategies are discussed, and used to develop follow-up plans. Targets 12. Target balance Goals include quality (reducing medical errors), service growth, and patient satisfaction Hospital implemented a balanced score card that includes 5 major elements: service, growth, people, quality, and finance. The performance management system is well understood by staff. Strategic plans are developed to respond to conflict when it arises between the hospital?s core values. The hospital and unit follow “key result” areas that incorporate employee and patient satisfaction, finances, quality, safety, and growth. It is often difficult to meet all goals simultaneously but senior managers regularly ask for staff feedback. Goals are continuously being re-evaluated to ensure balance; decision processes underlying these adjustments are transparent to staff. 13. Target inter-connection The medical director develops broad goals from those of the hospital system CEO. Unit goals are communicated to staff in regular and performance meeting. All unit goals have been designed to support a single overarching goal: to provide the best possible patient care. There are five main categories for performance goals with numerous supporting goals within each category. Every individual is responsible for working towards at least one sub-goal within each category and is graded quarterly on their performance. Goals cascade down from the organization, hospital, unit, and staff. The hospital uses forums and regular meetings to keep the staff involved in setting and evaluating targets because they believe staff buy-in is essential for top performance. Individuals are evaluated on the unit and hospital performance as well as meeting their own individual targets. 14. Target stretch The unit meets its goals 75% of the time. There is significant variance in success; some targets are met 100% of the time while others are never met. They struggle with reducing falls to their target level but managers have had no say in adjusting this goal. Each goal has 3 categories of success to encourage stretch: (1) target, (2) expected but difficult and (3) distinguished. The unit reached target or expected but difficult level for 65% of its goals, while only 10% of distinguished level goals. Setting targets levels is a collaborative process supported by leadership and clinical staff. Targets are compared internally and externally to national standards. All units are held to the same standard of excellence. Goals are aggressive for all units in the hospital and are developed by consulting nurse managers in each unit. Targets are continuously adjusted to externally set benchmarks to motivate everyone to provide the best care possible. Employee incentives 15. Rewarding high performers Evaluation is based primarily on attendance and patient satisfaction scores. Merit raises range from 1% to 5% and based on performance rating received on the annual evaluation. There are no group or individual bonuses. The hospital has a pay-for-performance model with awards based on 5 key priorities. There is a Spirit Award (financial bonus) given to the entire hospital if hospital meets its goals. If the unit meets its goals, there is an additional, separate bonus for the unit. Individuals may also be eligible for bonuses if they meet goals that have been defined as linking to the larger unit and hospital goals. Performance evaluations include a self, peer, and manager evaluation of each individual. Eligibility for individual salary is based on the combination of these three scores; top performers receive a 3, 5, or 7% raise if they achieve a customary, above average or excellent rating respectively. The management team chooses a yearly bonus eligible goal. This year if the hospital meets its aggressive patient satisfaction target all employees receive a $1500 bonus. 16. Removing poor performers There is a corrective action process with an associated performance improvement plan. The remediation and termination process is grueling, time consuming and not very effective. Anonymous peer reporting and error tracking is used to identify staff issues. Performance concerns are immediately addressed with the individual: the manager places the employee on a performance improvement plan and assigns him a mentor/skills coach. The hospital has a no tolerance policy for low performers and if timely improvements are not made the employee will be terminated. Employees self-report weaknesses and can opt-in to request skill remediation, peer coaching and reassignment without penalty. Inability to adequately perform is grounds for immediate termination., but systems are focused on identifying problems and correcting them at an early stage. 17. Managing talent Managers identify potential candidates for a hospital leadership program. The program provides tuition reimbursement for classes and advanced training in management. The program is voluntary and managers are not held accountable for their staff?s performance. There is a staff development program; a career development team meets with staff bi-annually to evaluate career goals and provide guidance. Continuing education courses and tuition reimbursement are provided. Employees can participate in the clinical ladder program that provides financial incentives for education and participation in hospital councils and quality improvement efforts. Managers are evaluated on the proportion of their staff holding advanced certifications and degrees. Hospital has a nursing student program that pays students to complete nursing school clinical requirements as a way to attract the best new graduates. The application process is rigorous; candidates are interviewed by two future peers, a manager, and a member of the senior leadership team. Managers and staff receive a recruitment bonus for signing and retaining new employees. The “reaching for the stars” program provides tuition reimbursement and bonuses for completing advanced certifications and degrees. 18. Retaining talent Managers work hard to keep their star performers happy. HR policy allows manager to change schedules and approve extended leaves of absence for top performers. Managers and senior leadership are not held accountable for voluntary turnover. Senior management has access to a fund that may be used to retain star employees. Managers and HR will adjust work schedules and compensation to insure that they are competitive with other hospitals in the community. Managers meet with each staff member quarterly to measure their job satisfaction and elicit concerns. Managers are generally aware of resignations long before they become a reality and work proactively to keep them. Managers have flexibility to offer per diem and part-time positions, tuition reimbursement for further education, salary adjustments, and position reassignments. Senior management?s pay is partially based on voluntary turnover rates. Note: STEMI patients generally require more immediate care than NSTEMI patients. a STEMI refers to ST elevation myocardial infarction. b NSTEMI refers to non-ST elevation myocardial infarction. Table options 3.1. Lean operations Questions on operations follow the typical flow of AMI patients through the hospital: from admission (Q1), to procedures performed on the patient (Q2), to activities that must be conducted during the patient?s stay (Q3-5) through discharge (Q6). The most poorly managed hospitals did not have a high degree of standardization for different stages of patient care (admission, treatment, and discharge). Typical hospitals had some policies in place, but relatively weak systems for monitoring whether the policies were being used to the right effect. However, some hospitals were able to instill a high degree ofstandardization in the ways in which their patients were cared for. As an example of what is possible among very well managed hospitals, some hospitals start the discharge planning process very early (“as soon as the patient is admitted”), have a multidisciplinary team and approach to identifying barriers or challenges for the patient upon discharge, and identify the point of follow-up upon discharge. Some of the most promising examples indicate a movement toward a more streamlined and integrated pathway of care – similar to the pathways and value streams that have become mainstream in manufacturing and technology sectors, 16 wherein the pathway for every product and service is evaluated, with the goal of eliminating steps that are redundant or do not add value. 3.2. Performance measurement Performance measurement is a broad category that is closely related to operations but focuses on the extent to which the unit is effective in using data and in proactively addressing problems. As an example of best practices, some hospitals track data on a very frequent basis and use visual tools to track performance. The data serve as a basis for regular conversations to encourage feedback and problem-solving. Performance review also included questions about the ways in which hospitals used evidence to make purchasing decisions (Q7). For many hospitals, adoption of a new device was based strictly on physician input. However, one high-scoring hospital had a product committee that required new drugs and devices to meet evidence-based guidelines for safety and efficacy and be comparatively more cost-effective than existing treatments. Some hospitals approach errors proactively, seeking to avoid errors rather than responding to errors after they occur. This proactive stance is hallmark of “High-Reliability Organizations” (HROs), organizations that have achieved a high degree of safety or reliability despite operating in complex, high risk settings.17, 18 and 19 3.3. Targets The existence of targets is one way of assessing priorities within the cardiac unit. Known targets are likely to influence ways that resources and efforts are deployed for patient care. When asked what types of targets were set for the unit, many nurse managers had difficulty providing an answer, often settling on “volume.” When asked about their toughest targets, some hospitals were unsure, while other units immediately indicated that they were attempting to achieve a zero percent infection rate. Best practices included targets that tie to the larger organization goals, are based, in part, on input from frontline staff, and have a significant degree of “stretch” – that is, targets are not too easy nor too hard. These hospitals use targets as a way of directing employee attention and challenging employees to achieve ambitious patient care goals. 3.4. Employee incentives Our employee incentive questions focused on the policies for nursing staff. Some hospitals promote solely based on tenure, doing relatively little to reward excellence. In the worst cases, there are few consequences for poor performance. Some hospitals recognize the importance of engaging their staff in additional training as well as using financial bonuses that are tied to performance. Managers may be held accountable for building and retaining an excellent team. Although our questions on incentives are based on the literature on High Performing Work Practices,20, 21, 22, 23, 24, 24, 25 and 26 our definition of best practices for this dimension may be more controversial than other dimensions. In particular, some observers have noted that annual performance reviews may be inferior to real-time, daily reviews, or that performance pay should be exclusively team-based. 3.5. Regression results Results of our models of management as a function of hospital and market level variables are displayed in Table 5. Higher management scores were associated with a more competitive market, higher net income from patient service, teaching hospital status, and proximity to MBA schools. These associations are in line with our expectations. However, the magnitudes of associations are relatively small. For example, a one-standard deviation decrease in HHI (increase in competition) or increase in net income is associated with an approximate increase in the management score of 0.1; relative to non-teaching hospitals, teaching hospitals had management scores that were higher by approximately 0.15. However, these estimates may underestimate the true effect of hospital and market level variables for several reasons, including the relative noise that may exist in our management score. Table 5. Model estimating association of management practice score with hospital & market characteristics. Variables Coefficient 95% CI Hospital HHI ?0.282?? (?0.552, ?0.011) Hospital net income from patient service ($100 M) 0.0457?? (0.001, 0.091) Teaching hospital 0.149?? (0.0168, 0.281) Distance to closest MBA program (100 s of miles) ?0.066? (?0.133, 0.001) Rural location 0.011 (?0.129 to 0.152) Ownership Government/public 0.044 (?0.100 to 0.188) For profit ?0.069 (?0.219, 0.081) System membership 0.051 (?0.049, 0.150) Annual AMI medicare FFS volumea 51–99 cases 0.067 (?0.046, 0.180) 100–199 cases 0.101 (?0.036, 0.238) 200 or More cases 0.153 (?0.097, 0.403) Hospital sizeb 151–374 Licensed beds 0.080 (?0.069, 0.230) 375 or More licensed beds 0.014 (?0.177, 0.206) Facility performs open heart surgery ?0.014 (?0.142, 0.114) Note: Regressions also control for regional and interviewer fixed effects. ? p<0.1. ?? p<0.05. a Reference is AMI volume 50 or fewer cases. b Reference category is Hospital Size 150 or fewer licensed beds. Table options Perhaps surprisingly, we do not find a correlation with other variables that we might expect to be associated with better management, including hospital size, AMI volume, or rural vs. urban status. Generally we might have expected larger, high volume, and urban hospitals to have higher management scores since these characteristics have often been associated with better patient quality.27, 28 and 29 One possible explanation for this negative finding may be that our sample was restricted to relatively high volume hospitals that had the capacity to perform interventional cardiology, and thus the differences in management may be relatively small across this fairly select group of hospitals. In addition, the service line which we studied – cardiac care – may represent a service with relatively good management, reflecting scrutiny from hospital administrators (because of the relative importance of cardiac care to hospital financial performance) and from the public reporting efforts. 4. Conclusions In this paper, we have described a framework for exploring and measuring management in healthcare, as well as providing concrete examples of the types of management practices that hospitals are currently utilizing. These examples may assist researchers and policymakers by providing an indication of the types of policies and practices that hospitals undertake in efforts to improve quality, attract more patients, and potentially reduce inefficiency. Furthermore, our findings suggest that many cardiac units could substantially improve their management approaches in multiple areas. The dimensions in Table 1 and e-Appendix could be used by hospital administrators and clinicians to “self-score” their unit. If a self-assessment is low, our framework may serve as a roadmap toward progress in particular management domains. Although our management practices are not intended to be an exclusive list of management approaches, they do represent a relatively concise set of practices that have been empirically verified in other sectors,8, 9 and 10 serve as the background for national surveys of management,11 and are closely aligned with the ways in which researchers and healthcare leaders have identified top strategies or mechanisms for improving the value of healthcare.17, 19, 20, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40 and 41 To date, a challenge in studying management in healthcare has been the lack of a relatively concise set of measures that translate across organizations. For example, in a review of 185 studies of quality improvement, Alexander and Hearld noted that the majority of studies were difficult to generalize to other organizations due to unique study contexts.42 By focusing on a relatively narrow set of management dimensions and allowing for flexibility in the way that they are measured and implemented, we hope to advance the study and understanding of management in healthcare. We find a wide variation in the dissemination of management practices, with relatively few hospitals achieving “best” practice scores on most dimensions. Fewer than 20% of hospitals score a 4 or a 5 on more than nine measures, and a majority score a 3 or lower on 15 out of 18 measures. Thus, there may be significant gains that can be achieved in healthcare quality and efficiency as hospitals move to adopt more effective management approaches. The association between better management and competition is a consistent with research on management in a number of different settings, countries, and sectors, including manufacturing in the US, France, the United Kingdom (UK), and Germany,9substance abuse treatment centers in the US,43 and hospitals in the UK.44 This finding has important policy implications and may be an area for greater scrutiny, particularly since the effect of competition on hospital behavior appears to be complex. While most research generally supports the finding that competition improves quality,45 and 46 less is known about whether hospitals become more efficient when faced with greater competitive pressures. Policies and research designed to promote a high value healthcare system may benefit from an improved understanding about the ways in which management is affected by regulation and competition, and the ways in which these changes in management may affect quality, safety, and efficiency. Our results suggest future directions for hospital management practices and quality of AMI care. The best practices detailed in our surveys, which are well validated in the manufacturing industry, should be rigorously evaluated through demonstrations, careful evaluations, or experimental studies in order to determine their independent impact on quality and other patient-centered outcomes. In addition, qualitative studies of high performing hospitals? practices and their translation to lower performing hospitals – similar to the approach used to improve STEMI door-to-balloon times28, 29 and 30 – may also provide important improvements in hospital quality of care. Of critical importance are studies that can identify whether management practices, with their heavy focus on the reduction of waste, can actually reduce the overall cost of care without diminishing quality. Many of the policies in the Affordable Care Act – including incentives to slow healthcare spending and to manage complex patients across multiple sites of care – are likely to heighten the role of management practices in driving organizational performance.47 and 48Recent results from the first performance year of its Pioneer Accountable Care Organization (ACO) model indicate that the top performer was the Bellin ThedaCare Health Partners ACO, a longtime proponent and promoter of Lean methodologies.49 The ability to study management may help to establish an evidence-base of organizational practices to assist in meeting the coming challenge to provide high quality, efficient care. CLINICAL RESEARCH STUDY Angiotensin-converting Enzyme Inhibitors and Outcomes in Heart Failure and Preserved Ejection Fraction Marjan Mujib, MD, MPH,a Kanan Patel, MBBS, MPH,b Gregg C. Fonarow, MD,c Dalane W. Kitzman, MD,d Yan Zhang, MS, MSPH,b Inmaculada B. Aban, PhD,b O. James Ekundayo, MD, DrPH,e Thomas E. Love, PhD,f Meredith L. Kilgore, PhD,b Richard M. Allman, MD,b,g Mihai Gheorghiade, MD,h Ali Ahmed, MD, MPHb,g a New York Medical College, Valhalla; bUniversity of Alabama at Birmingham, Birmingham; cUniversity of California, Los Angeles; Wake Forest University, Winston-Salem, NC; eMeharry Medical College, Nashville, Tenn; fCase Western Reserve University, Cleveland, Ohio; gVeterans Affairs Medical Center, Birmingham, Ala; hNorthwestern University, Chicago, Ill. d ABSTRACT BACKGROUND: The role of angiotensin-converting enzyme (ACE) inhibitors in patients with heart failure and preserved ejection fraction remains unclear. METHODS: Of the 10,570 patients aged ’Ü65 years with heart failure and preserved ejection fraction (’Ü40%) in the Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients With Heart Failure (20032004) linked to Medicare (through December 2008), 7304 were not receiving angiotensin receptor blockers and had no contraindications to ACE inhibitors. After excluding 3115 patients with pre-admission ACE inhibitor use, the remaining 4189 were eligible for new discharge prescriptions for ACE inhibitors, and 1706 received them. Propensity scores for the receipt of ACE inhibitors, calculated for each of the 4189 patients, were used to assemble a cohort of 1337 pairs of patients, balanced on 114 baseline characteristics. RESULTS: Matched patients had a mean age of 81 years and mean ejection fraction of 55%, 64% were women, and 9% were African American. Initiation of ACE inhibitor therapy was associated with a lower risk of the primary composite end point of all-cause mortality or heart failure hospitalization during 2.4 years of median follow-up (hazard ratio [HR], 0.91; 95% confidence interval [CI], 0.84-0.99; P œ≠ .028), but not with individual end points of all-cause mortality (HR, 0.96; 95% CI, 0.88-1.05; P œ≠ .373) or heart failure hospitalization (HR, 0.93; 95% CI, 0.83-1.05; P œ≠ .257). CONCLUSION: In hospitalized older patients with heart failure and preserved ejection fraction not receiving angiotensin receptor blockers, discharge initiation of ACE inhibitor therapy was associated with a modest improvement in the composite end point of total mortality or heart failure hospitalization but had no association with individual end point components. Published by Elsevier Inc. ‚Ä¢ The American Journal of Medicine (2013) 126, 401-410 KEYWORDS: Angiotensin-converting enzyme inhibitors; Heart failure; Preserved ejection fraction Approximately half of the estimated 6 million patients with heart failure in the United States have diastolic heart failure or heart failure with preserved ejection fraction.1 Most of Funding: The project described was supported by Grant R01HL097047 from the National Heart, Lung, and Blood Institute/National Institutes of Health (NIH) (Principal Investigator: Dr Ahmed). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Heart, Lung, and Blood Institute or NIH. Dr Ahmed also is supported by a generous gift from Jean B. Morris of Birmingham, Alabama. Dr Allman is supported by NIH/National Center for Research Resources Grant 5UL1 RR025777. OPTIMIZE-HF was funded by GlaxoSmithKline (Principal Investigator: Dr Fonarow). Conflict of Interest: None. 0002-9343/$ -see front matter Published by Elsevier Inc. http://dx.doi.org/10.1016/j.amjmed.2013.01.004 these patients are older adults and prognostically similar to those with systolic heart failure or heart failure with reduced ejection fraction.2,3 Angiotensin-converting enzyme (ACE) inhibitors reduce all-cause mortality in patients with heart Authorship: AA conceived the study hypothesis and design in collaboration with coauthors. AA and MM wrote the first draft. AA, MM, KP and YZ performed statistical analyses in collaboration with IBA and TEL. All authors interpreted the data, participated in critical revision of the manuscript for important intellectual contents, and approved the final version of the article. IBA, AA, MM, KP, and YZ had full access to the data. Requests for reprints should be addressed to Ali Ahmed, MD, MPH, UAB Center for Aging, 1720 2nd Ave South, CH-19, Ste-219, Birmingham, AL 35294-2041. E-mail address: aahmed@uab.edu 402 The American Journal of Medicine, Vol 126, No 5, May 2013 failure and reduced ejection fraction.4-6 Although angiotenMATERIALS AND METHODS sin receptor blockers did not reduce mortality in patients Data Sources and Study Population with heart failure and reduced ejection fraction, they imThe Organized Program to Initiate Lifesaving Treatment in proved outcomes7,8 and are considered the drugs of choice Hospitalized Patients With Heart Failure (OPTIMIZE-HF) for these patients who cannot tolerate ACE inhibitors.9 is a national registry of hospitalHowever, despite evidence of simized patients with heart failure and ilar neurohormonal activation in has been well described in the litheart failure with preserved ejecCLINICAL SIGNIFICANCE erature.20-22 Briefly, charts from tion fraction,10 there is no clear 48,612 hospitalizations due to evidence of the efficacy of renin‚óè The role of angiotensin-converting enheart failure occurring in 259 hosangiotensin system inhibition in zyme (ACE) inhibitors in patients with pitals from 48 states between these patients. heart failure and preserved ejection March 2003 and December 2004 The lack of efficacy of angiofraction remains unclear. were abstracted by trained staff.20 tensin receptor blockers in paA primary discharge diagnosis of ‚óè In older adults with heart failure and tients with heart failure and preheart failure was determined on served ejection fraction has now preserved ejection fraction not receivthe basis of the International Clasbeen well established in 2 large ing angiotensin receptor blockers, initisification of Diseases, 9th Revimulticenter randomized controlled ation of ACE inhibitors therapy was assion codes 428, 402.01, 402.11, trials.11,12 However, the role of sociated with a 9% modest reduction in 402.91, 404.01, 404.03, 404.11, ACE inhibitors is less clear. In the the composite end point of total morand 404.91.22 Of the 48,612 hosPerindopril in Elderly People with tality or heart failure hospitalization, pitalizations, 20,839 were due to Chronic Heart Failure (PEP-CHF) but not in individual end points. heart failure and preserved ejectrial, the only randomized contion fraction ’Ü40%. Extensive trolled trial of ACE inhibitors in data on baseline demographics, heart failure and preserved ejecmedical history, including admission and discharge medition fraction, 850 patients (mean age, 75 years) recruited cations (including ACE inhibitors) and angiotensin receptor from 8 European countries were randomized to receive blockers, hospital course, discharge disposition, and physiperindopril or placebo. During 2.1 years of median fol22 cian specialty, also were collected. Data on contraindicalow-up, perindopril had no effect on the primary end tions to the use of ACE inhibitors also were collected from point of all-cause mortality or heart failure hospitalizapatients not receiving these drugs. Missing values for contion (hazard ratio [HR], 0.92; P œ≠ .545) or all-cause tinuous variables were imputed on the basis of values premortality (HR, 1.09; P œ≠ .665).13 dicted by age, sex, and race. The nonsignificant effect of perindopril was explained Because OPTIMIZE-HF did not collect data on longin part by the unexpected low (45%) event rates and loss term outcomes, we linked OPTIMIZE-HF to Medicare of power (from 90% to 35%) in PEP-CHF and substantial outcomes data up to December 31, 2008, obtained from open-label perindopril use after the first year of followthe Centers for Medicare and Medicaid Services.23 Of the up, before which perindopril tended to reduce the risk of 20,839 heart failure hospitalizations due to heart failure the primary end point (HR, 0.69; P œ≠ .055) and signifand preserved ejection fraction, 13,270 could be linked to icantly reduced the risk of heart failure hospitalization 13 Medicare data. These events occurred in 11,997 unique (HR, 0.63; P œ≠ .033). This early benefit of perindopril patients, 10,889 of whom were aged 65 years or in PEP-CHF is similar to the early benefit of enalapril in older,24,25 of whom 10,570 were discharged alive (Figpatients with heart failure and reduced ejection fraction ure 1). Because angiotensin receptor blockers have not in the Studies of Left Ventricular Dysfunction (SOLVD), been shown to improve outcomes in heart failure and in which enalapril had no effect after the second year of 5 preserved ejection fraction,7,11,12,25 we excluded 1871 follow-up. These observations, taken together with the patients who received angiotensin receptor blockers. Of neurohormonal activation in heart failure with preserved 10 the remaining 8699 patients, 107 without data on disejection fraction, led us to hypothesize that ACE inhibcharge use of ACE inhibitors and another 1288 patients itor use may be associated with improved outcomes in with contraindication to ACE inhibitors were excluded, patients with heart failure and preserved ejection fracleading to a final working sample size of 7304 patients tion, despite the definitive lack of efficacy of angiotensin who would be eligible for a discharge prescription for receptor blockers in these patients. Therefore, the objecACE inhibitors (Figure 1). tive of the current study was to test this hypothesis in a propensity-matched (balanced)14,15 inception cohort (new users)16,17 of restricted (excluding those with contraindications to ACE inhibitors)18,19 patients with heart failure and preserved ejection fraction. Assembly of an Inception Cohort Because prevalent drug use may result in selection bias and left censoring,16,17,26 we assembled an inception cohort of Mujib et al ACE Inhibitors in Heart Failure with Preserved Ejection Fraction 403 OPTIMIZE-HF hospitalizations (n = 48,612) Restrict to ejection fraction 40% (n = 20,839) Linked to Medicare data (n = 13,270) Excluded (2700 patients): 1273 = Multiple hospital admissions 1108 = Age <65 years 319 = Discharged dead Discharged alive (n = 10,570) Excluded (3266 patients): 1871 = Receiving angiotensin receptor blockers 107 = No data on ACE inhibitors 1288 = Contraindications to ACE inhibitors Eligible for ACE inhibitor therapy (n = 7304) Excluded: 3115 patients receiving ACE inhibitors during hospital admission Eligible for new ACE inhibitor therapy Assembly of the Inception Cohort (n = 4189) ACE inhibitor (n = 1706) No ACE inhibitor (n = 2483) 1337 patients receiving ACE inhibitors were matched by propensity scores to 1337 patients not receiving ACE inhibitors Assembly of the Matched Cohort (n = 2674) Figure 1 Flow chart displaying assembly of inception cohort of matched patients with heart failure and preserved ejection fraction. ACE œ≠ angiotensin-converting enzyme; OPTIMIZE-HF œ≠ Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients With Heart Failure. 4189 patients who were not receiving prior ACE inhibitor therapy, and a discharge prescription for ACE inhibitors for these patients would be considered an initiation of therapy. Therefore, we excluded 3115 patients who received ACE inhibitors before hospital admission. Of the 4189 patients with no history of ACE inhibitor use or no contraindication to new ACE inhibitor therapy, 1706 (41%) received a new discharge prescription for ACE inhibitors (Figure 1). Assembly of a Balanced Cohort In a well-designed randomized controlled trial, the probability of receiving a treatment is 50%, regardless of whether a patient is randomized to the treatment or placebo group. However, treatment assignment in the real world is seldom random, and as such, these probabilities in nonrandomized controlled trial studies would vary from 0% to 100%. Often, these probabilities are dictated by various measured and unmeasured patient and care characteristics. In real-world patients with heart failure, the probability of the receipt of an ACE inhibitor may be influenced by age, ejection fraction, blood pressure, serum potassium, serum creatinine, known adverse effects, and perceived or real contraindications. For example, a 75-year-old patient with heart failure and preserved ejection fraction who has low blood pressure and high serum potassium will likely have a low probability of receiving ACE inhibitors, whereas a 45-year-old patient with heart failure and reduced ejection fraction who has normal blood pressure and normal serum potassium will likely have a high probability of receiving these drugs. These probabilities or propensity scores for the receipt of ACE inhibitors are predicted by data and may be similar in 2 patients. However, it is possible that 1 of these 2 patients 404 Age Female African American Nursing home residency Current smoking Le≈å ventricular ejec∆üon frac∆üon Heart failure hospitaliza∆üon (Hx) Coronary artery disease (Hx) Myocardial infarc∆üon (Hx) Coronary revasculariza∆üon (Hx) Hypertension (Hx) Diabetes mellitus (Hx) Atrial Ô¨Åbrilla∆üon (Hx) Ventricular arrhythmia (Hx) Valvular heart disease (Hx) Idiopathic cardiomyopathy (Hx) Implantable cardioverter-deÔ¨Åbrillator(Hx) Biventricular pacemaker (Hx) Stroke/transient ischemic a∆©ack (Hx) Hyperlipidemia (Hx) Liver disease (Hx) Chronic obstruc∆üve pulmonary disease(Hx) Asthma (Hx) Pulmonary hypertension (Hx) Peripheral vascular disease (Hx) Chronic kidney disease(Hx) Acute renal failure (Hx) Dialysis (Hx) Anemia (Hx) Depression (Hx) Thyroid disease (Hx) Ischemic heart disease (PF) Uncontrolled hypertension (PF) Pneumonia (PF) Worsening kidney func∆üon(PF) Arrhythmias (PF) Non-adherence to diet (PF) Non-adherence to drugs (PF) Other causes (PF) The American Journal of Medicine, Vol 126, No 5, May 2013 Beta-blockers(A) Diure∆ücs (A) Aldosterone antagonists (A) Hydralazine (A) Nitrates (A) Digoxin (A) Amlodipine (A) Other calcium channel blockers (A) An∆ü-arrhythmic drugs (A) Warfarin (A) An∆ü-platelet drugs (A) Aspirin (A) Sta∆üns (A) Dyspnea on exer∆üon (A) Fa∆ügue (A) Orthopnea (A) Paroxysmal nocturnal dyspnea(A) Dyspnea at rest (A) Chest pain (A) Palpita∆üon (A) Anorexia (A) Dizziness (A) Weight (A) Weight (D) Pulse (A) Systolic blood pressure(A) Diastolic blood pressure(A) Jugular venous pressure(A) Third heart sound (A) R√¢les (A) R√¢les (D) Hepatomegaly (A) Lower extremity edema (A) Lower extremity edema (D) Serum sodium (A) Serum crea∆ünine (A) Hemoglobin (A) Brain natriure∆üc pep∆üde(A) Troponin eleva∆üon (A) Dobutamine (H) Dopamine (H) Milrinone (H) Nesiri∆üde (H) Cardioversion (H) Dialysis (H) Electrophysiological study (H) Mechanical ven∆üla∆üon (H) Pacemaker, biventricular (H) Pacemaker, others (H) Right heart catheteriza∆üon (H) Coronary angiography (H) Percutaneous coronary interven∆üon (H) Coronary artery bypass gra≈å (H) Pulse (D) Systolic blood pressure(D) Diastolic blood pressure(D) Hospital, bed-size Hospital, academic Hospital, interven∆üonal Hospital, transplant *Hospital, Midwest *Hospital, Northwest *Hospital, South *Hospital, West Length of hospital stay Beta-blockers (D) Diure∆ücs (D) Aldosterone antagonists (D) Hydralazine (D) Nitrates (D) Digoxin (D) Amlodipine (D) Other calcium channel blockers (D) An∆ü-arrhythmic drugs (D) Warfarin (D) An∆ü-platelet drugs (D) Sta∆üns (D) Aspirin (D) Absolute standardized difference (%) Figure 2 Love plot displaying absolute standardized differences comparing 114* baseline characteristics between older patients with heart failure and preserved ejection fraction, receiving a new discharge prescription of ACE inhibitors, before and after propensity score matching. A œ≠ admission; D œ≠ discharge; H œ≠ in-hospital; Hx œ≠ medical history; PF œ≠ precipitating factor. *Four regions entered as single categorical variable in the model. actually received ACE inhibitors and the other patient did not. These 2 patients could then be matched to assemble a pair of patients receiving and not receiving ACE inhibitors who had similar predicted probabilities of receiving ACE inhibitors. In a properly conducted propensity-matched study, patients receiving and not receiving a treatment, such as an ACE inhibitor, would be balanced on all measured baseline characteristics.14,27-31 Of note, this balance can be achieved while remaining blinded to study outcomes, a key feature of a randomized controlled trial.27 We used propensity scores for the receipt of ACE inhibitors to assemble our study cohort so that patients receiving and not receiving these drugs would be balanced on all measured baseline characteristics.31-33 We estimated propensity scores for each of the 4189 patients using a nonparsimonious multivariable logistic regression model.32,33 In the model, the receipt of ACE inhibitors was the dependent variable, and 114 baseline characteristics displayed in Figure 2 were used as covariates. Although propensity scores can be used in regression models or for stratification, matching by propensity scores allows assembly of cohorts in which baseline balance can be estimated and displayed in visually pleasant tabular forms. We used a greedy matching protocol to match 1337 (78%) of the 1706 patients receiving ACE inhibitors with 1337 patients who did not receive ACE inhibitors but had the same propensity or probability to receive them.34,35 The effectiveness of the propensity score model was assessed by estimating absolute standardized differences15,28,36 and presented as a Love plot.37-39 A difference of 0% indicates no residual bias, and values œΩ10% are considered inconsequential. Mortality and Hospitalization The primary outcome of the current analysis was the composite end point of all-cause mortality or heart failure hospitalization.24 Secondary outcomes included all-cause mortality, heart failure, and all-cause hospitalizations. Data on mortality and hospitalization were obtained from the 100% MedPAR File and 100% Beneficiary Summary File between January 1, 2002, and December 31, 2008. Statistical Analysis For descriptive analyses, we used Pearson‚Äôs chi-square and Wilcoxon rank-sum tests for the pre-match data, and McNemar‚Äôs test and paired sample t test for post-match comparisons, as appropriate. Because measured baseline characteristics are balanced in propensity-matched cohorts, we used bivariate Cox proportional hazard models to determine the associations of a new discharge prescription for ACE inhibitors (independent variable) with outcomes (dependent variable) among matched patients during 6 years of follow-up (median, and 25th and 75th percentiles, 2.4, 0.7, and Mujib et al ACE Inhibitors in Heart Failure with Preserved Ejection Fraction 405 Table 1 Baseline Patients and Care Characteristics of Older Patients with Heart Failure and Preserved Ejection Fraction, by a New Discharge Prescription for Angiotensin-Converting Enzyme Inhibitors Variables Mean (SD) or n (%) Age (y) Female African American Left ventricular ejection fraction (%) Precipitating factors for hospital admission Ischemic heart disease Uncontrolled hypertension Worsening renal function Arrhythmia Nonadherence to diet Nonadherence to medications Medical history No prior heart failure hospitalization Coronary artery disease Hypertension Diabetes mellitus Atrial fibrillation Hyperlipidemia Chronic obstructive pulmonary disease Peripheral vascular disease Chronic kidney disease Admission symptoms and signs Dyspnea on exertion Fatigue Orthopnea Paroxysmal nocturnal dyspnea Dyspnea at rest Chest pain Pulse (beats/min) Systolic blood pressure (mm Hg) Diastolic blood pressure (mm Hg) Jugular venous pressure elevation Third heart sound Pulmonary rales Lower-extremity edema Laboratory values Serum sodium (mEq/L) Serum creatinine (mg/dL)* Serum hemoglobin (g/dL) Serum brain natriuretic peptide, (pg/mL)* Serum troponin elevation‚Ä† Length of hospital stay Hospital characteristics Bed size* Academic Interventional Transplant Hospital location by region Midwest Northeast South West Before Propensity Score Matching After Propensity Score Matching Use of ACE Inhibitors Use of ACE Inhibitors No (n œ≠ 2483) Yes (n œ≠ 1706) P Value 81 1611 184 56 (8) (65) (7) (9) 81 1087 174 54 (8) (64) (10) (10) .713 .439 .001 œΩ.001 81 843 127 55 (8) (63) (10) (9) 81 860 116 55 (8) (64) (9) (10) .793 .513 .503 .870 242 166 112 395 58 88 (10) (7) (5) (16) (2) (4) 219 185 45 248 51 108 (13) (11) (3) (15) (3) (6) .002 œΩ.001 .002 .226 .192 œΩ.001 160 107 46 201 34 54 (12) (8) (3) (15) (3) (4) 152 124 41 205 36 72 (11) (9) (3) (15) (3) (5) .676 .252 .668 .872 .904 .117 348 1047 1722 842 972 662 804 334 1608 (14) (42) (69) (34) (39) (27) (32) (14) (65) 321 695 1205 549 585 515 457 201 984 (19) (41) (71) (32) (34) (30) (27) (12) (58) œΩ.001 .357 .374 .243 .001 .013 œΩ.001 .112 œΩ.001 215 551 923 438 489 385 378 155 798 (16) (41) (69) (33) (37) (29) (28) (12) (60) 209 556 939 440 493 394 377 165 795 (16) (42) (70) (33) (37) (30) (28) (12) (60) .786 .875 .538 .968 .906 .729 1.000 .598 .937 1465 567 493 243 1048 510 85 144 73 614 118 1485 1621 (59) (23) (20) (10) (42) (21) (22) (30) (17) (25) (5) (60) (65) 1109 350 469 249 755 361 85 151 77 473 118 1122 1105 (65) (21) (28) (15) (44) (21) (21) (31) (19) (28) (7) (66) (65) œΩ.001 .074 œΩ.001 œΩ.001 .188 .627 .390 œΩ.001 œΩ.001 .030 .003 œΩ.001 .732 846 267 331 168 582 277 85 148 75 370 76 839 873 (63) (20) (25) (13) (44) (21) (22) (31) (18) (28) (6) (63) (65) 837 289 321 168 582 284 84 148 75 351 73 847 866 (63) (22) (24) (13) (44) (21) (21) (29) (18) (26) (6) (63) (65) .747 .317 .686 1.000 1.000 .779 .630 .807 .865 .435 .865 .775 .806 137 1.20 11.9 740 359 6 (10) (0.70) (2.1) (690.82) (15) (5) 137 1.10 12.0 806 272 14 (11) (0.60) (2.1) (773.18) (16) (354) .529 œΩ.001 .067 .003 .187 .242 137 1.20 12.0 753 194 6 (11) (0.70) (2.0) (717.50) (15) (5) 137 1.20 11.9 780 206 6 (10) (0.60) (2.1) (744.68) (15) (4) 350 985 1816 363 (221) (40) (73) (15) 375 784 1334 222 (200) (46) (78) (13) œΩ.001 œΩ.001 œΩ.001 .141 355 604 1030 201 (212) (45) (77) (15) 360 578 1031 175 (207) (43) (77) (13) .790 .334 1.000 .161 702 420 879 482 (28) (17) (35) (19) 560 274 499 373 (33) (16) (29) (22) œΩ.001 416 232 402 287 (31) (17) (30) (22) 419 224 413 281 (31) (17) (31) (21) .776 ACE œ≠ angiotensin-converting enzyme; SD œ≠ standard deviation. *Displayed as median (interquartile range). ‚Ä†Determined by local laboratories. No (n œ≠ 1337) Yes (n œ≠ 1337) P Value .647 .980 .639 .604 .546 .245 406 The American Journal of Medicine, Vol 126, No 5, May 2013 4.5 years, respectively). Log-minus-log survival plots were used to check proportional hazards assumptions. We conducted a formal sensitivity analysis to estimate the degree of hidden bias that could explain away a significant association between ACE inhibitors and the primary composite outcome among our matched patients.40 Subgroup analyses were conducted to determine the homogeneity of association between the use of ACE inhibitors and the composite primary outcome. Because an older cohort with a long follow-up will ultimately have 100% mortality, an estimation of the number needed to treat using absolute risk difference may be less useful. Therefore, using a formula proposed for survival anal- Table 2 Procedures and Treatment in Older Patients with Heart Failure and Preserved Ejection Fraction, by a New Discharge Prescription for Angiotensin-Converting Enzyme Inhibitors Variables Mean (SD) or n (%) Admission medication Beta-blockers Aldosterone antagonists Angiotensin receptor blockers* Diuretics Digoxin Hydralazine Nitrates Amlodipine Non-amlodipine calcium channel blockers Antiarrhythmic drugs Warfarin Antiplatelet drugs Aspirin Statins In-hospital treatment/procedure Dobutamine Dopamine Milrinone Nesiritide Right heart catheterization Coronary angiography Coronary artery bypass grafting Percutaneous coronary intervention Electrophysiological study Cardioversion Pacemaker-biventricular Dialysis Discharge medication Beta-blockers Aldosterone antagonists Angiotensin receptor blockers* Diuretics Digoxin Hydralazine Nitrates Amlodipine Non-amlodipine calcium channel blockers Antiarrhythmic drugs Warfarin Antiplatelet drugs Aspirin Statins Before Propensity Score Matching After Propensity Score Matching Use of ACE Inhibitors Use of ACE Inhibitors No (n œ≠ 2483) Yes (n œ≠ 1706) P Value No (n œ≠ 1337) Yes (n œ≠ 1337) P Value 1168 104 ‚Äî 1538 453 73 467 230 480 233 584 267 869 638 (47) (4) (44) (3) (62) (18) (3) (19) (9) (19) (9) (24) (11) (35) (26) 745 47 ‚Äî 838 276 23 266 137 259 113 322 166 590 437 (49) (16) (1) (16) (8) (15) (7) (19) (10) (35) (26) .031 .014 ‚Äî œΩ.001 .083 .001 .007 .166 .001 .001 œΩ.001 .285 .782 .954 634 47 ‚Äî 732 223 20 219 103 196 101 284 141 479 365 (55) (17) (2) (16) (8) (15) (8) (21) (11) (36) (27) 621 44 ‚Äî 724 231 20 220 112 210 95 286 140 478 356 (54) (17) (2) (17) (8) (16) (7) (21) (11) (36) (27) .639 .828 ‚Äî .778 .717 1.000 1.000 .573 .476 .711 .963 1.000 1.000 .730 20 51 9 190 48 117 12 18 17 30 9 82 (1) (2) (0.4) (8) (2) (5) (0.5) (1) (1) (1) (0.4) (3) 21 26 8 114 53 132 13 20 9 16 12 35 (1) (2) (0.5) (7) (3) (8) (0.8) (1) (1) (1) (0.7) (2) .169 .210 .594 .235 .015 œΩ.001 .250 .133 .525 .409 .125 .016 12 16 6 88 35 77 8 14 8 14 8 29 (1) (1) (0.4) (7) (3) (6) (0.6) (1) (1) (1) (1) (2) 9 17 6 93 38 85 6 13 8 14 10 32 (1) (1) (0.4) (7) (3) (6) (0.4) (1) (1) (1) (1) (2) .664 1.000 1.000 .762 .813 .578 .791 1.000 1.000 1.000 .815 .788 1318 179 ‚Äî 1932 543 94 562 223 500 283 669 298 996 646 (53) (7) 1603 155 ‚Äî 1399 394 27 412 122 198 149 447 234 850 512 (62) (9) œΩ.001 .028 ‚Äî .001 .349 œΩ.001 .254 .034 œΩ.001 .005 .594 .102 œΩ.001 .005 790 99 ‚Äî 1079 290 27 329 95 167 137 366 178 616 398 (59) (7) 783 112 ‚Äî 1084 304 26 318 110 182 124 366 176 605 381 (59) (8) .811 .385 ‚Äî .842 .553 1.000 .650 .311 .410 .434 1.000 .955 .687 .487 (78) (22) (4) (23) (9) (20) (11) (27) (12) (40) (26) (82) (23) (2) (24) (7) (12) (9) (26) (14) (50) (30) ACE œ≠ angiotensin-converting enzyme; SD œ≠ standard deviation. *Patients receiving angiotensin receptor blockers on admission and during discharge were excluded. (47) (4) (81) (22) (2) (25) (7) (13) (10) (27) (13) (46) (30) (46) (3) (81) (23) (2) (24) (8) (14) (9) (27) (13) (45) (29) All-cause mo ortality or heart failure hos h spitalization Mujib et al ACE Inhibitors in Heart Failure with Preserved Ejection Fraction 63% were women, and 9% were African American. Before matching, patients receiving a new prescription for ACE inhibitors were more likely to be symptomatic but had a lower prevalence of comorbidities, such as atrial fibrillation and chronic kidney disease. These and other pre-match imbalances were balanced after matching (Tables 1 and 2, and Figure 2). Absolute standardized differences for all 114 baseline characteristics between the 2 treatment groups were œΩ10% (mostly œΩ5%), suggesting substantial bias reduction. 1.0 No ACE inhibitors 0.8 0.6 ACE inhibitors 0.4 0.2 HR=0.91 95% CI=0.84‚Äì0.99; p=0.028 0.0 0 1 2 3 4 5 6 Follow-up (years) No. at risk No ACE inhibitors 1337 711 ACE inhibitors i hibit 1337 748 521 572 391 424 287 331 407 83 101 Figure 3 Kaplan-Meier plot for primary composite end point of all-cause mortality or heart failure hospitalization in a propensity-matched inception cohort of older patients with heart failure and preserved ejection fraction receiving and not receiving a new discharge prescription for ACE inhibitors. ACE œ≠ angiotensin-converting enzyme; CI œ≠ confidence interval; HR œ≠ hazard ratio. yses, we estimated the number needed to treat with ACE inhibitors to prevent 1 primary composite end point event.41 All statistical tests were 2-tailed, and 95% confidence intervals (CIs) were constructed. Finally, we examined the association of ACE inhibitors with outcomes among pre-match patients using multivariable Cox regression models adjusting for all 114 baseline characteristics used in the propensity model and propensity scores. All data analyses were performed using SPSS for Windows version 18 (SPSS Inc, Chicago, Ill). RESULTS New Prescriptions for Angiotensin-Converting Enzyme Inhibitors and Outcomes During 2.4 years of median follow-up, the primary composite end point of all-cause mortality or heart failure hospitalization occurred in 80% (1076/1337) and 83% (1112/1337) of matched patients with heart failure and preserved ejection fraction receiving and not receiving a new discharge prescription for ACE inhibitors, respectively (HR when the use of ACE inhibitors was compared with their nonuse, 0.91; 95% CI, 0.84-0.99; P œ≠ .028; Figure 3 and Table 3). An estimated 71 patients (95% CI, 40-646) will need to be treated over a median 2.4 years of follow-up to prevent 1 primary composite end point event. The association between new ACE inhibitor use and the primary composite end point was homogeneous across various subgroups of patients (Figure 4). ACE inhibitor use had no significant association with individual end point components of allcause mortality and hospitalization (Table 3). Among the 4189 pre-match patients, the primary composite end point occurred in 79% (1351/1706) and 84% (2079/2483) of patients receiving and not receiving a new discharge prescription for ACE inhibitors, respectively (HR, 0.84; 95% CI, 0.78-0.90; P œΩ.001). Multivariable-adjusted and propensity-adjusted HRs for the primary composite end point associated with ACE inhibitor use were 0.93 (95% CI, 0.86-1.00; P œ≠ .049) and 0.94 (95% CI, 0.87-1.01; P œ≠ .098), respectively. Baseline Characteristics Matched patients (n œ≠ 2674) had a mean (œÆ standard deviation) age of 81 (œÆ8) years and mean (œÆ standard deviation) left ventricular ejection fraction of 55% (œÆ9), DISCUSSION Findings from our study demonstrate that a new discharge prescription for ACE inhibitors was associated with a sta- Table 3 Outcomes by a New Discharge Prescription for Angiotensin-Converting Enzyme Inhibitors in a Propensity-Matched Inception Cohort of Older Patients with Heart Failure and Preserved Ejection Fraction Events (%) Outcomes Combined end point of all-cause mortality or heart failure hospitalization All-cause mortality Heart failure hospitalization All-cause hospitalization No ACE Inhibitors (n œ≠ 1337) ACE Inhibitors (n œ≠ 1337) HR* (95% CI) P Value 1112 (83%) 1076 (80%) 0.91 (0.84-0.99) .028 951 (71%) 564 (42%) 1155 (86%) 930 (70%) 558 (42%) 1165 (87%) 0.96 (0.88-1.05) 0.93 (0.83-1.05) 0.97 (0.89-1.05) .373 .257 .401 ACE œ≠ angiotensin-converting enzyme; CI œ≠ confidence interval; HR œ≠ hazard ratio. *HRs comparing patients receiving ACE inhibitors with those not receiving ACE inhibitors calculated using Cox regression model. 408 The American Journal of Medicine, Vol 126, No 5, May 2013 Figure 4 Association of a new discharge prescription of ACE inhibitors with primary composite end point of all-cause mortality or heart failure hospitalization in subgroups of propensity-matched inception cohort of older patients with heart failure and preserved ejection fraction. ACE œ≠ angiotensin-converting enzyme; CI œ≠ confidence interval; GFR œ≠ glomerular filtration rate. tistically significant modest 9% lower risk of the composite end point of all-cause mortality or heart failure hospitalization in a wide spectrum of propensity-matched older patients with heart failure and preserved ejection fraction who were balanced on more than 100 potential confounders. Similar multivariable-adjusted or propensity-adjusted associations were observed when traditional regression-based risk adjustment models were used in the pre-match cohort. However, ACE inhibitors had no significant association with individual end point components of all-cause mortality or heart failure hospitalization. Findings from the current rigorously conducted propensity-matched inception cohort study based on nationally representative real-world patients provide evidence that the use of ACE inhibitors may be associated with a modest improvement in the long-term composite end point of total mortality or heart failure hospitalization in older patients with heart failure and preserved ejection fraction. The 9% reduction in the composite end point in our study is substantially smaller than the 26% reduction in the same end point in younger patients with systolic heart failure in the SOLVD trial.5 In the SOLVD trial, enalapril seemed to have a more robust effect on heart failure hospitalization than on mortality, which in part also may explain the overall benefit of ACE inhibitors in patients with heart failure with preserved ejection fraction. The effect of ACE inhibitors also may be mediated by their beneficial effect on aortic stenosis, the prevalence of which would be expected to be high in older patients with heart failure with preserved ejection fraction. The inhibition of the renin-angiotensin system has been shown to be associated with improved outcomes in patients with aortic stenosis.42 The lack of significant association with all-cause mortality in our study may in part be explained by the different modes of death in patients with heart failure with preserved versus reduced ejection fraction. Findings from a major randomized controlled trial of ACE inhibitors in systolic heart failure suggest that these drugs had no significant effect on sudden cardiac death but had a robust effect on death due to pump failure.5,6 Although sudden death accounts for 40% to 50% Mujib et al ACE Inhibitors in Heart Failure with Preserved Ejection Fraction of cardiovascular deaths in patients with heart failure regardless of ejection fraction, death due to pump failure is less common in those with preserved ejection fraction, accounting for 24% of cardiovascular deaths (vs 41% in those with reduced ejection fraction).43 This may in part explain the lack of an effect of ACE inhibitors on mortality in patients with heart failure and preserved ejection. Most randomized controlled trials of ACE inhibitors in heart failure excluded those with preserved ejection fraction. The overall direction and magnitude of the associations with the primary end point observed in our study (9% reduction) are consistent with those from PEP-CHF (8% reduction).13 A recent propensity-matched study of ACE inhibitors or angiotensin receptor blockers based on the Swedish Heart Failure Registry reported mortality reduction in patients with heart failure and preserved ejection fraction.44 This association seems inflated because approximately 25% of patients in that study were receiving angiotensin receptor blockers, with proven lack of effect on mortality.11,12 In addition, in PEP-CHF, perindopril had no effect on all-cause mortality, not even during the first year of follow-up, when it reduced heart failure hospitalization, suggesting lack of efficacy on mortality.13 That study, based on the Swedish Heart Failure Registry, also was limited by biases due to lack of restriction to patients without contraindications,18,19 lack of exclusion of prevalent drug users,16,17 and incomplete matching,45 because more than one quarter of 43 variables used in propensity matching were imbalanced after matching.44 Despite potential overestimation of the association in the Swedish Heart Failure Registry, findings from PEP-CHF and our study suggest that ACE inhibitor therapy may be associated with a modest improvement in the long-term clinical outcomes in patients with heart failure and preserved ejection fraction. However, given the lack of benefit of angiotensin receptor blockers in those patients,7,11,12,25 these findings need to be interpreted with caution and be replicated in other restricted propensitymatched inception cohorts. Study Limitations Our study has several limitations. Findings from our sensitivity analysis suggest that this association could be explained away by a hidden covariate that would increase the odds for the receipt of ACE inhibitors by approximately 1%. However, to act as a confounder, an unmeasured covariate must be a near-perfect predictor of outcome and not be strongly correlated with any of the 114 measured baseline covariates used in our study, which is unlikely. We were able to match approximately 80% of patients receiving ACE inhibitors, thus minimizing any effect on external validity. We had no data on names and doses for individual ACE inhibitors. We also had no data on the use of ACE inhibitors after discharge.46 Substantial crossover may result in regression dilution47 and may explain the modest associations observed in our study. Lack of data on aortic stenosis is another limitation. The clinical data for the analyses were collected from the medical record and depended on the 409 accuracy and completeness of the clinical documentation. Although this study is confined to fee-for-service Medicare, patient and hospital participation in OPTIMIZE-HF were voluntary and limited to all those hospitals participating in a quality improvement registry, and this may limit the generalizability of the results. However, Medicare-linked OPTIMIZE-HF patients have been shown to have similar characteristics and outcomes as patients with heart failure in the general Medicare population.48 CONCLUSIONS In hospitalized older patients with heart failure and preserved ejection fraction who were not receiving angiotensin receptor blockers, a new discharge prescription for ACE inhibitors was associated with a modest improvement in the composite end point of total mortality or heart failure hospitalization but had no association with the individual components of mortality and heart failure hospitalization. Findings from this rigorously conducted propensity-matched inception cohort study need to be interpreted in the context of inconclusive findings from the PEP-CHF trial and proven lack of efficacy of angiotensin receptor blockers in these patients. Additional well-designed prospective studies are needed. References 1. Roger VL, Go AS, Lloyd-Jones DM, et al. Heart disease and stroke statistics‚Äî2011 update: a report from the American Heart Association. Circulation. 2011;123:e18-e209. 2. Ahmed A, Perry GJ, Fleg JL, Love TE, Goff DC Jr, Kitzman DW. Outcomes in ambulatory chronic systolic and diastolic heart failure: a propensity score analysis. Am Heart J. 2006;152:956-966. 3. Fonarow GC, Stough WG, Abraham WT, et al. Characteristics, treatments, and outcomes of patients with preserved systolic function hospitalized for heart failure: a report from the OPTIMIZE-HF Registry. J Am Coll Cardiol. 2007;50:768-777. 4. The CONSENSUS Trial Study Group. Effects of enalapril on mortality in severe congestive heart failure. Results of the Cooperative North Scandinavian Enalapril Survival Study (CONSENSUS). N Engl J Med. 1987;316:1429-1435. 5. The SOLVD Investigators. Effect of enalapril on survival in patients with reduced left ventricular ejection fractions and congestive heart failure. N Engl J Med. 1991;325:293-302. 6. Garg R, Yusuf S. Overview of randomized trials of angiotensinconverting enzyme inhibitors on mortality and morbidity in patients with heart failure. Collaborative Group on ACE Inhibitor Trials. JAMA. 1995;273:1450-1456. 7. Cohn JN, Tognoni G, Valsartan Heart Failure Trial I. A randomized trial of the angiotensin-receptor blocker valsartan in chronic heart failure. N Engl J Med. 2001;345:1667-1675. 8. Granger CB, McMurray JJ, Yusuf S, et al. Effects of candesartan in patients with chronic heart failure and reduced left-ventricular systolic function intolerant to angiotensin-converting-enzyme inhibitors: the CHARM-Alternative trial. Lancet. 2003;362:772-776. 9. Hunt SA, Abraham WT, Chin MH, et al. 2009 focused update incorporated into the ACC/AHA 2005 Guidelines for the Diagnosis and Management of Heart Failure in Adults: a report of the American College of Cardiology Foundation/American Heart Association Task Force on Practice Guidelines: developed in collaboration with the International Society for Heart and Lung Transplantation. Circulation. 2009;119:e391-479. 410 10. Kitzman DW, Little WC, Brubaker PH, et al. Pathophysiological characterization of isolated diastolic heart failure in comparison to systolic heart failure. JAMA. 2002;288:2144-2150. 11. Massie BM, Carson PE, McMurray JJ, et al. Irbesartan in patients with heart failure and preserved ejection fraction. N Engl J Med. 2008;359: 2456-2467. 12. Yusuf S, Pfeffer MA, Swedberg K, et al. Effects of candesartan in patients with chronic heart failure and preserved left-ventricular ejection fraction: the CHARM-Preserved Trial. Lancet. 2003;362:777-781. 13. Cleland JG, Tendera M, Adamus J, Freemantle N, Polonski L, Taylor J. The perindopril in elderly people with chronic heart failure (PEPCHF) study. Eur Heart J. 2006;27:2338-2345. 14. Rosenbaum PR, Rubin DB. The central role of propensity score in observational studies for causal effects. Biometrika. 1983;70:41-55. 15. Rosenbaum PR, Rubin DR. Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. Am Stat. 1985;39:33-38. 16. Ray WA. Evaluating medication effects outside of clinical trials: new-user designs. Am J Epidemiol. 2003;158:915-920. 17. Danaei G, Tavakkoli M, Hernan MA. Bias in observational studies of prevalent users: lessons for comparative effectiveness research from a meta-analysis of statins. Am J Epidemiol. 2012;175:250-262. 18. Feenstra H, Grobbee RE, in‚Äôt Veld BA, Stricker BH. Confounding by contraindication in a nationwide cohort study of risk for death in patients taking ibopamine. Ann Intern Med. 2001;134:569-572. 19. Psaty BM, Siscovick DS. Minimizing bias due to confounding by indication in comparative effectiveness research: the importance of restriction. JAMA. 2010;304:897-898. 20. Fonarow GC, Abraham WT, Albert NM, et al. Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure (OPTIMIZE-HF): rationale and design. Am Heart J. 2004;148:43-51. 21. Gheorghiade M, Abraham WT, Albert NM, et al. Systolic blood pressure at admission, clinical characteristics, and outcomes in patients hospitalized with acute heart failure. JAMA. 2006;296:2217-2226. 22. Fonarow GC, Abraham WT, Albert NM, et al. Association between performance measures and clinical outcomes for patients hospitalized with heart failure. JAMA. 2007;297:61-70. 23. Hammill BG, Hernandez AF, Peterson ED, Fonarow GC, Schulman KA, Curtis LH. Linking inpatient clinical registry data to Medicare claims data using indirect identifiers. Am Heart J. 2009;157:995-1000. 24. Zhang Y, Kilgore ML, Arora T, et al. Design and rationale of studies of neurohormonal blockade and outcomes in diastolic heart failure using OPTIMIZE-HF registry linked to Medicare data. Int J Cardiol. 2011 Nov 24. [Epub ahead of print] doi.org/10.1016/j.ijcard.2011. 10.089. 25. Patel K, Fonarow GC, Kitzman DW, et al. Angiotensin receptor blockers and outcomes in real-world older patients with heart failure and preserved ejection fraction: a propensity-matched inception cohort clinical effectiveness study. Eur J Heart Fail. 2012;14:1179-1188. 26. Hernan MA, Hernandez-Diaz S, Robins JM. A structural approach to selection bias. Epidemiology. 2004;15:615-625. 27. Rubin DB. Using propensity score to help design observational studies: application to the tobacco litigation. Health Serv Outcomes Res Methodol. 2001;2:169-188. 28. Austin PC. Primer on statistical interpretation or methods report card on propensity-score matching in the cardiology literature from 2004 to 2006: a systematic review. Circ Cardiovasc Qual Outcomes. 2008;1:62-67. 29. Heinze G, Juni P. An overview of the objectives of and the approaches to propensity score analyses. Eur Heart J. 2011;32:1704-1708. The American Journal of Medicine, Vol 126, No 5, May 2013 30. Michels KB, Braunwald E. Estimating treatment effects from observational data: dissonant and resonant notes from the SYMPHONY trials. JAMA. 2002;287:3130-3132. 31. Ahmed A, Husain A, Love TE, et al. Heart failure, chronic diuretic use, and increase in mortality and hospitalization: an observational study using propensity score methods. Eur Heart J. 2006;27:1431-1439. 32. Ahmed A, Fonarow GC, Zhang Y, et al. Renin-angiotensin inhibition in systolic heart failure and chronic kidney disease. Am J Med. 2012; 125:399-410. 33. Ahmed A, Rich MW, Zile M, et al. Renin-angiotensin inhibition in diastolic heart failure and chronic kidney disease. Am J Med. 2013; 126:150-161. 34. Filippatos GS, Ahmed MI, Gladden JD, et al. Hyperuricaemia, chronic kidney disease, and outcomes in heart failure: potential mechanistic insights from epidemiological data. Eur Heart J. 2011;32:712-720. 35. Guichard JL, Desai RV, Ahmed MI, et al. Isolated diastolic hypotension and incident heart failure in older adults. Hypertension. 2011;58: 895-901. 36. Normand ST, Landrum MB, Guadagnoli E, et al. Validating recommendations for coronary angiography following acute myocardial infarction in the elderly: a matched analysis using propensity scores. J Clin Epidemiol. 2001;54:387-398. 37. Aronow WS, Ahmed MI, Ekundayo OJ, Allman RM, Ahmed A. A propensity-matched study of the association of peripheral arterial disease with cardiovascular outcomes in community-dwelling older adults. Am J Cardiol. 2009;103:130-135. 38. Ekundayo OJ, Dell‚ÄôItalia LJ, Sanders PW, et al. Association between hyperuricemia and incident heart failure among older adults: a propensity-matched study. Int J Cardiol. 2010;142:279-287. 39. Wahle C, Adamopoulos C, Ekundayo OJ, Mujib M, Aronow WS, Ahmed A. A propensity-matched study of outcomes of chronic heart failure (HF) in younger and older adults. Arch Gerontol Geriatr. 2009;49:165-171. 40. Rosenbaum PR. Sensitivity to hidden bias. In: Rosenbaum PR, ed. Observational Studies. Vol 1. New York: Springer-Verlag; 2002:105-170. 41. Altman DG, Andersen PK. Calculating the number needed to treat for trials where the outcome is time to an event. BMJ. 1999;319:1492-1495. 42. Nadir MA, Wei L, Elder DH, et al. Impact of renin-angiotensin system blockade therapy on outcome in aortic stenosis. J Am Coll Cardiol. 2011;58:570-576. 43. Zile MR, Gaasch WH, Anand IS, et al. Mode of death in patients with heart failure and a preserved ejection fraction: results from the Irbesartan in Heart Failure With Preserved Ejection Fraction Study (IPreserve) trial. Circulation. 2010;121:1393-1405. 44. Lund LH, Benson L, Dahlstrom U, Edner M. Association between use of renin-angiotensin system antagonists and mortality in patients with heart failure and preserved ejection fraction. JAMA. 2012;308:2108-2117. 45. Rosenbaum PR, Rubin DB. The bias due to incomplete matching. Biometrics. 1985;41:103-116. 46. Butler J, Arbogast PG, Daugherty J, Jain MK, Ray WA, Griffin MR. Outpatient utilization of angiotensin-converting enzyme inhibitors among heart failure patients after hospital discharge. J Am Coll Cardiol. 2004;43:2036-2043. 47. Clarke R, Shipley M, Lewington S, et al. Underestimation of risk associations due to regression dilution in long-term follow-up of prospective studies. Am J Epidemiol. 1999;150:341-353. 48. Curtis LH, Greiner MA, Hammill BG, et al. Representativeness of a national heart failure quality-of-care registry: comparison of OPTIMIZE-HF and non-OPTIMIZE-HF Medicare patients. Circ Cardiovasc Qual Outcomes. 2009;2:377-384. CLINICAL RESEARCH STUDY An International Model to Predict Recurrent Cardiovascular Disease Peter W.F. Wilson, MD,a Ralph D‚ÄôAgostino Sr, PhD,b Deepak L. Bhatt, MD, MPH,c Kim Eagle, MD,d Michael J. Pencina, PhD,e Sidney C. Smith, MD,f Mark J. Alberts, MD,g Jean Dallongeville, MD,h Shinya Goto, MD, PhD,i Alan T. Hirsch, MD,j Chiau-Suong Liau, MD, PhD,k E. Magnus Ohman, MD,l Joachim R√∂ther, MD,m Christopher Reid, PhD,n Jean-Louis Mas, MD,o Ph. Gabriel Steg, MDp; for the REACH Registry a Atlanta VA Medical Center and Cardiology Division, Emory University School of Medicine, Atlanta, Ga; bStatistics and Consulting Unit, Boston University, Boston, Mass; cVA Boston Healthcare System, Brigham & Women‚Äôs Hospital, and Harvard Medical School, Boston, Mass; dUniversity of Michigan Cardiovascular Center, Ann Arbor; eStatistics and Consulting Unit, Department of Biostatistics, Boston University, Mass; fUniversity of North Carolina at Chapel Hill; gNorthwestern University Medical School, Chicago, Ill; hInstitut Pasteur de Lille, Lille, France; iDepartment of Medicine, Tokai University School of Medicine, Kanagawa, Japan; jDivision of Epidemiology and Community Health, University of Minnesota School of Public Health and Minneapolis Heart Institute Foundation, Minneapolis; kDepartment of Internal Medicine, National Taiwan University Hospital and School of Medicine, Taipei, Taiwan; l Division of Cardiology, Duke University, Durham, NC; mDepartment of Neurology, Klinikum Minden, Minden, Germany; nMonash University, Victoria, Australia; oService de Neurologie, Centre Raymond Garcin, H√¥pital Sainte-Anne, Paris, France; pINSERM U-698 et Universit√© Paris VII‚ÄìDenis Diderot, H√¥pital Bichat-Claude Bernard, Paris, France. ABSTRACT BACKGROUND: Prediction models for cardiovascular events and cardiovascular death in patients with established cardiovascular disease are not generally available. METHODS: Participants from the prospective REduction of Atherothrombosis for Continued Health (REACH) Registry provided a global outpatient population with known cardiovascular disease at entry. Cardiovascular prediction models were estimated from the 2-year follow-up data of 49,689 participants from around the world. RESULTS: A developmental prediction model was estimated from 33,419 randomly selected participants (2394 cardiovascular events with 1029 cardiovascular deaths) from the pool of 49,689. The number of vascular beds with clinical disease, diabetes, smoking, low body mass index, history of atrial fibrillation, cardiac failure, and history of cardiovascular event(s) œΩ1 year before baseline examination increased risk of a subsequent cardiovascular event. Statin (hazard ratio 0.75; 95% confidence interval, 0.69-0.82) and acetylsalicylic acid therapy (hazard ratio 0.90; 95% confidence interval, 0.83-0.99) also were significantly associated with reduced risk of cardiovascular events. The prediction model was validated in the remaining 16,270 REACH subjects (1172 cardiovascular events, 494 cardiovascular deaths). Risk of cardiovascular death was similarly estimated with the same set of risk factors. Simple algorithms were developed for prediction of overall cardiovascular events and for cardiovascular death. CONCLUSIONS: This study establishes and validates a risk model to predict secondary cardiovascular events and cardiovascular death in outpatients with established atherothrombotic disease. Traditional risk factors, burden of disease, lack of treatment, and geographic location all are related to an increased risk of subsequent cardiovascular morbidity and cardiovascular mortality. ¬© 2012 Elsevier Inc. All rights reserved. ‚Ä¢ The American Journal of Medicine (2012) 125, 695-703 KEYWORDS: Acute coronary syndromes; Cardiovascular disease; Cerebrovascular disease/stroke; Coronary disease; Mortality; Peripheral vascular disease; Risk factors Funding: See last page of article. Conflict of Interest: See last page of article. Authorship: See last page of article. Requests for reprints should be addressed to Peter W.F. Wilson, MD, 0002-9343/$ -see front matter ¬© 2012 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2012.01.014 Atlanta VAMC and Emory Clinical Cardiovascular Research Institute, 1462 Clifton Road, Suite 500, Atlanta, GA 30322. E-mail address: peter.wf.wilson@emory.edu 696 The American Journal of Medicine, Vol 125, No 7, July 2012 The prediction of atherothrombotic events (coronary artery consisted of one or more of the following criteria: stable disease, cerebrovascular disease, and peripheral arterial disangina with documented coronary artery disease, history of ease) in the outpatient setting has focused on the occurrence unstable angina with documented coronary artery disease, of first events,1-3 and Framingham Heart Study data or history of percutaneous coronary intervention, history of similar observational study data have been used to predict coronary artery bypass graft surgery, or previous myocarvascular disease risk in people. dial infarction. Documented carLess cardiovascular prediction rediovascular disease consisted of a search has focused on patients hospital or neurologist report with CLINICAL SIGNIFICANCE with existing coronary artery disthe diagnosis of transient ischemic ease.4 The estimation of vascular attack or ischemic stroke. Docu‚óè Risk of recurrent cardiovascular disease disease risk among people known mented peripheral artery disease in survivors of myocardial infarction can to have atherothrombotic disease consisted of one or both of the be estimated from outpatient clinic at baseline would facilitate effollowing criteria: current interinformation. forts by patients, clinicians, and mittent claudication with ankle‚óè The key determinants of recurrent carpublic health officials to prevent brachial index of œΩ0.9 or a histhese morbid and mortal events tory of intermittent claudication diovascular disease are age, male sex, in a real-world setting. Multitogether with a previous and recurrent smoking, diabetes mellitus, low variable analysis of risk for related intervention such as angiobody mass index, number of vascular current cardiovascular events plasty, stenting, atherectomy, beds involved, CVD event in the past could provide the underpinnings peripheral arterial bypass graft, year, history of heart failure, atrial fiof an individualized risk scoring or other vascular intervention, brillation, not taking statins, and not system and help to identify indiincluding amputation. taking anti-platelet therapy. viduals at higher risk for more Data were collected using a intensive investigation, treatstandardized, international case ment, and follow-up. report form that was completed Largely, risk-assessment scores for initial cardiovascular at the study visit. Body mass index was defined as weight event prediction have been developed from specific geo(kilograms) divided by the square of the height (meters), graphic regions, such as the US,1,5,6 Germany,7 Italy,8 and and subjects were categorized using commonly used defthe UK.9 Projects that have tested the applicability of risk initions of obesity and overweight. Current smoking was assessment scores across large regions such as the US10 or defined as ’Ü5 cigarettes per day on average within the Europe11 have been undertaken as post hoc analyses. The last month before entry into the Registry, and former REduction of Atherothrombosis for Continued Health smoking was defined as ’Ü5 cigarettes per day on average (REACH) Registry was established in 2003 to recruit and more than 1 month before entry into the Registry. Chofollow a large cohort of outpatients with known atherolesterol levels were transcribed from the clinical record thrombotic disease or at high risk of developing atheroand lipids were not measured in a standard manner in the thrombosis.12 This design provides a contemporary data set registry participants. Because of differences in high-denthat includes individuals across the world and allows the sity lipoprotein cholesterol measurement methods around assessment of risk for cardiovascular events, including inthe world, and lack of standardization for this assay in vestigation of the potential role of vascular disease burden, many locales, information concerning high-density lipoprotective effect of specific treatments, and effects accordprotein cholesterol levels was not obtained. ing to geographic region. Physician selection was performed at the country level in The purpose of this report is to identify the determinants collaboration with the respective national coordinators, and of secondary vascular ischemic events and cardiovascular was designed to provide a representative distribution across death in the study participants, to develop a vascular disease specialties, geographic regions, types of practice (officerisk algorithm from this experience, and to develop a robust based, public or private hospital-based, dispensary only, risk prediction model that has been internally validated. multiple, or other structures), and locations (urban, suburban, or rural). Each physician recruited a maximum of 15 patients (20 patients in the US). METHODS In each country, 10% of the investigator sites that enThe study design, selection of physicians,12 and baseline rolled at least one patient underwent quality control assessand follow-up experience of participants in the REACH ment. Six percent of these sites were chosen randomly, and Registry13-15 have been published previously. The current an additional 4% were chosen in response to queries about study included outpatients aged ’Ü45 years with established missing data requests. For each of the sites undergoing coronary artery disease, cerebrovascular disease, or periphmonitoring, 100% of case report forms for patients enrolled eral arterial disease, with enrollment by their physician over at that site were monitored for source documentation and a 7-month period on a worldwide basis between December accuracy and information related to risk factor levels, med2003 and June 2004. Documented coronary artery disease Wilson et al Recurrent Vascular Disease Prediction ical history items, medication use, and cardiovascular diagnoses. Follow-up Participants were followed for the development of a subsequent cardiovascular event and were invited to a baseline clinical examination and follow-up evaluation at 12 and 24 months after the baseline. At the follow-up visits, data were collected regarding interim development of clinical outcomes according to self-report and medical records available, vascular and endovascular procedures, employment status, weight and current smoking status, and medications used since baseline. The current report is based on a database lock of June 15, 2007 for analysis of the 2-year followup. This follow-up interval was used for the development of multivariable risk estimation because of the high participation rate at this time and because only a single baseline examination was available in the study. Model performance was assessed at 20 months to avoid unstable predictions occurring at the very end of follow-up. Statistical Analysis A large sample was available, and we desired to validate the results of prediction models. A priori we randomly split the cohort 2:1 and developed the predictive models in two thirds of the participants and validated it in the remaining one third. Risk analyses were performed using Cox proportionalhazard regression.16 Model development included clinical judgment and careful consideration of well-accepted traditional variables for vascular disease risk assessment. Stepwise selection models were run with age, sex, and geographic region forced into the model. Treating each region as a separate stratum did not improve model performance, and we collapsed geographic region membership into higher (Eastern Europe and Middle East), intermediate (North America/Western Europe‚Äîreferent), and lower (Japan or Australia)-risk locations. Separate analyses were performed for the cardiovascular event and cardiovascular death end points, and significance levels were set at 0.05 to identify candidate risk factors. In an effort to retain the same variables in both models, we included only variables that were significant in both models. Pair-wise interactions with sex and age were assessed at the 0.01 level to account for multiple testing and to avoid weak interaction signals that would not transport well with application of the function to new cohorts. The traditional candidate variables considered for this analysis included systolic and diastolic blood pressure, blood pressure treatment, cholesterol level, diabetes mellitus, current smoking, and body mass index modeled as a continuous variable as well as in categories (œΩ20 kg/m2, 20-30 kg/m2, œæ30 kg/m2). Variables that characterized the vascular disease burden of participants at baseline included information concerning the number of vascular beds affected, occurrence of a cardiovascular event in the year 697 before the baseline evaluation, history of congestive heart failure, and history of atrial fibrillation. Treatment with acetylsalicylic acid and statins also was included, but information on the type or dose of statin or the dose of acetylsalicylic acid was not available. To assess the role of previously diagnosed vascular disease, we used ordinal variables to test for the effects of 1, 2, or 3 vascular beds involved. Only data from subjects with complete outcome and covariate information for a given end point were used to obtain the rates for that end point. Statistical analysis was performed using SAS Version 9 software (SAS Institute Inc., Cary, NC). Discrimination was assessed using C-statistics calculated for each of these models; the error associated with C-statistic estimates was estimated using published methods.17 Calibration was assessed using decile plots of mean predicted and observed risks, and D‚ÄôAgostino and Nam18 modified Hosmer-Lemeshow chi-squared statistics with 9 degrees of freedom. Performance was assessed at 20 months of follow-up (95th percentile of follow-up for events) on both development and validation data sets. Simplified risk calculator was developed for 20-month risk prediction for cardiovascular death and next cardiovascular event using methods described by Sullivan et al.19 Event rates of cardiovascular death, myocardial infarction, stroke, and cardiovascular hospitalization were calculated. End points were not adjudicated. Cardiovascular death included fatal stroke, fatal myocardial infarction, or other cardiovascular death. Other cardiovascular death included other death of cardiac origin; pulmonary embolism; any sudden death including unobserved, and unexpected death (eg, death while sleeping) unless proven otherwise by autopsy, death following a vascular operation, vascular procedure, or amputation; death attributed to heart failure; death following a visceral or limb infarction; and any other death that could not be definitely attributed to a nonvascular cause or hemorrhage. Any myocardial infarction or stroke followed by a death, whatever the cause, in the next 28 days, was considered to be a fatal myocardial infarction or fatal stroke. Cardiovascular hospitalization consisted of hospitalization for unstable angina, transient ischemic attack, worsening of claudication related to peripheral artery disease, other ischemic arterial event, coronary artery bypass grafting, coronary angioplasty/stenting, carotid surgery, carotid angioplasty/stenting, amputation affecting lower limbs, peripheral bypass graft, or angioplasty/stenting for peripheral artery disease.13 RESULTS The baseline characteristics of the participants are shown in Table 1. Participants were followed for a maximum of 24 months, with mean follow-up of 19 and median of 21 months. There were 55,814 persons in the REACH Study who were eligible for these data analyses, and approximately 89% (n œ≠ 49,689) from the 5473 participating study sites in 44 countries had full information on covariates at 698 The American Journal of Medicine, Vol 125, No 7, July 2012 Table 1 Baseline Characteristics of Participants in the Development and Validation Cohorts Characteristic Traditional risk factors Male, % Age, y, mean œÆ SD Current smoking, % Diabetes, % Body mass index Body mass index œæ30 kg/m2, % Body mass index œΩ 20 kg/m2, % Blood pressure, mm Hg, mean œÆ SD Systolic Diastolic Cholesterol level, mg/dL Atrial fibrillation, % Cardiovascular burden, % Coronary artery disease Cerebrovascular disease Peripheral arterial disease Number of vascular beds: 1, 2, or 3 Cardiovascular event in past year* Congestive heart failure* Cardiovascular treatment, % Statins Hypertension treatment Acetylsalicylic acid Geographic Region North America Latin America Western Europe Eastern Europe Middle East Asia Australia Japan Development Cohort (n œ≠ 33,419) Validation Cohort (n œ≠ 16,270) 66.9 68.4 œÆ 10.1 14.6 36.9 27.7 œÆ 5.4 27.2 66.8 68.3 œÆ 10.2 13.8 36.5 27.7 œÆ 5.3 27.0 3.8 3.5 136.8 œÆ 19.3 78.2 œÆ 11.2 191.4 œÆ 45.9 11.7 136.6 œÆ 19.5 78.2 œÆ 11.2 191.9 œÆ 46.7 11.4 72.4 33.8 14.9 72.5 33.7 14.5 80.8, 17.2, 2.0 81.1, 17.1, 1.8 31.5 32.3 15.2 15.5 69.2 91.0 71.5 68.5 91.1 71.4 36.4 3.1 26.6 10.4 1.3 9.4 4.6 8.2 36.3 3.1 26.6 10.9 1.3 9.6 4.6 7.6 *Based on information obtained at baseline evaluation. baseline and complete follow-up. By random selection, two thirds of the participants were assigned to a developmental cohort (n œ≠ 33,419), and the remaining one third were assigned to the validation cohort (n œ≠ 16,270). The baseline characteristics for these participants are shown in Table 1. The study participants were predominantly male (67%), and the mean age was 68 years at entry. Approximately 53% lived in Europe or Western Europe; 21% lived in Australia, Japan, or other parts of Asia, and 12% lived in Eastern Europe or the Middle East. Geographic clusters were determined according to proximity of the regions and similarities in cardiovascular disease incidence reported within the REACH Registry over 1 year of follow-up.13 Approximately 72% had coronary artery disease, 34% had cerebrovascular disease, and 14% had peripheral artery disease. Overall, 81% of the participants had vascular disease affecting one arterial bed, 17% had 2 arterial beds involved, and 2% had 3 arterial beds involved. The occurrence of a cardiovascular event in the year before entry into the registry was common, and 32% of the participants were affected. A history of congestive heart failure at entry was less frequent, and affected 15%. Treatment with statins or acetylsalicylic acid was reported in 71% of the participants, and hypertension therapy was reported in 91%. Among the 33,419 participants with vascular disease at baseline, there were 2394 cardiovascular events, and 1029 of these were cardiovascular deaths. Traditional risk factors (sex, age, smoking, diabetes, systolic blood pressure level, diastolic blood pressure level, low body mass index), the cardiovascular disease burden at baseline (number of vascular beds affected, cardiovascular event in the past year, history of congestive heart failure, and history of atrial fibrillation), cardiovascular treatment (hypertension therapy, statin use, acetylsalicylic acid use), and 3 geographic regions (1 ‚Äì North America or Western Europe, 2 ‚Äì Eastern Europe or Middle East, 3 ‚Äì Japan or Australia) were evaluated for inclusion in the stepwise prediction models. The variables selected in our stepwise models included age, diabetes, smoking status, number of vascular beds affected, history of atrial fibrillation, cardiac failure, history of cardiovascular event(s) œΩ1 year before the baseline examination, body mass index œΩ20 kg/m2, geographic region, statin use, and acetylsalicylic acid therapy. The results of the multivariable prediction of recurrent cardiovascular events and cardiovascular death with the hazard ratios and statistical significance are shown in Table 2. All variables in the table were significant at P œΩ.05 level in both models. Candidate variables in the stepwise regression model that did not make it to final model included systolic blood pressure, diastolic blood pressure, and cholesterol level. To test the ability of the prediction model to discriminate the risk of recurrent cardiovascular events in other regions, we used a validation set of 16,270 REACH participants who had not been included in the developmental data set. These participants met the same entry criteria and had the same duration of follow-up for CV events. In the validation data set, there were 1172 recurrent cardiovascular events and 494 cardiovascular deaths during the follow-up interval. The C-statistic for the prediction of a next cardiovascular event was 0.67 (95% confidence interval [CI], 0.66-0.68) for both the developmental and the validation datasets. The corresponding C-statistic for the prediction of cardiovascular death was 0.74 (95% CI, 0.73-0.76) for the developmental dataset and 0.75 (95% CI, 0.73-0.77) for the internal validation dataset. The D‚ÄôAgostino and Nam18 chi-squared value was 15.14 in the validation set for prediction of next cardiovascular events, indicating good calibration for this outcome. The corresponding Wilson et al Table 2 Recurrent Vascular Disease Prediction 699 Multivariable Prediction of Recurrent Cardiovascular Events* in the Development Cohort Type of Cardiovascular Event Factor Traditional risk factors Male sex Age, years Smoking (current vs other) Diabetes (yes/no) BMI œΩ20 kg/m2 Cardiovascular disease burden at baseline Number of vascular beds 1 2 3 Cardiovascular event* in past year (yes/ no) Congestive heart failure (yes/no) Atrial fibrillation (yes/no) Cardiovascular treatment Statins (yes/no) Acetylsalicylic acid (yes/no) Adjustment for geographic region North America or Western Europe Eastern Europe or Middle East Japan or Australia Cardiovascular Death Hazard Ratio (95% CI) Next Cardiovascular Event Hazard Ratio (95% CI) 1.28 1.05 1.36 1.59 1.74 (1.12-1.46) (1.04-1.06) (1.13-1.64) (1.40-1.80) (1.34-2.24) 1.11 1.03 1.27 1.46 1.37 (1.02-1.21) (1.03-1.04) (1.13-1.43) (1.34-1.58) (1.14-1.64) 1.00 1.28 1.64 1.31 (referent) (1.15-1.43) (1.32-2.06) (1.15-1.49) 1.00 1.35 1.83 1.46 (referent) (1.26-1.46) (1.58-2.13) (1.35-1.59) 2.46 (2.15-2.82) 1.53 (1.32-1.78) 1.68 (1.53-1.85) 1.30 (1.17-1.46) 0.80 (0.70-0.91) 0.84 (0.73-0.95) 0.75 (0.69-0.82) 0.90 (0.83-0.99) 1.00 (referent) 1.30 (1.07-1.57) 0.52 (0.41-0.66) 1.00 (referent) 1.32 (1.17-1.48) 0.73 (0.64-0.84) BMI œ≠ body mass index; CI œ≠ confidence interval. *A cardiovascular event is defined as an occurrence of myocardial infarction, cerebrovascular disease, or cardiovascular death. D‚ÄôAgostino and Nam18 chi-squared for the prediction of cardiovascular death also was acceptable at 19.45. Decile-based calibration plots suggested satisfactory performance for both outcomes (Figure 1). A risk score sheet for the development of the next cardiovascular event and for cardiovascular death in patients with clinically recognized atherothrombotic cardiovascular disease is shown in Figure 2. Using this approach, the patient information for age, current smoking, diabetes mellitus, number of vascular beds, cardiovascular event in the past year, congestive heart failure, statin therapy, and acetylsalicylic acid therapy are matched with the status for each variable, and the assigned number of points are then summed. The total number of points is then matched with the estimated 20-month risk. Using Figure 2, the number of risk points for a next cardiovascular event in a person with the following characteristics: man (œ©1), 62 years old (œ©8), nonsmoker (œ©0), diabetes present (œ©2), body mass index 25 kg/m2 (œ©0), history of myocardial infarction in the past 3 months (œ©2 for one vascular bed, œ©2 for event in the past year), history of congestive heart failure (œ©3), no history of atrial fibrillation (œ©0), taking statins (œ™2), taking acetylsalicylic acid (œ™1), Eastern Europe residence (œ©2), and the sum is 17 points for a next cardiovascular event, which translates to an absolute risk of 11% over a 20-month interval. Similarly, the number of risk points for cardiovascular death (1 œ© 8 œ© 0 œ© 2 œ© 0 œ© 2 œ© 2 œ© 3 œ© 0 œ™ 2 œ™ 1 œ© 1 œ© 2 œ≠ 18 points), corresponds to a 13% risk over a 20-month interval. Risk calculations can also be made using methods shown in the Appendix. DISCUSSION This study uses multivariable Cox regression analysis to identify factors that contribute to the risk of secondary ischemic events in a large, contemporary registry of patients with established cardiovascular disease. The factors that were statistically significant in the analyses included traditional risk variables for initial cardiovascular disease events (age, smoking, diabetes mellitus, body mass index), variables related to cardiovascular disease burden (number of vascular beds involved, history of a cardiovascular event in the year before the baseline evaluation, history of congestive heart failure, and history of atrial fibrillation), and use of cardiovascular risk reduction treatments (current use of statins or acetylsalicylic acid). This is the first international, prospective study to establish a comprehensive risk model to predict subsequent cardiovascular events and cardiovascular death in outpatients with established atherothrombotic disease. Most of the research concerning prediction of cardiovascular disease has focused on first cardiovascular events, and has reported on the experience of apparently 700 Figure 1 Calibration by decile according to risk for next cardiovascular events and for cardiovascular death. Bars represent observed (Kaplan-Meier; black) and model-based predicted (decile specific means; gray) probabilities of a cardiovascular outcome over 20 months in deciles of model-based predicted probabilities. healthy individuals who are ‚Äúat risk.‚Äù With that approach the variables age, sex, cholesterol level, blood pressure, smoking, and diabetes mellitus have typically been used to predict events.1,5,7 Over the past 3 decades, the treatment of first cardiovascular events has improved, survival has increased, and health care providers are increasingly required to provide effective long-term care for outpatients with established cardiovascular disease. It is helpful to identify, within this relatively large group of patients, those individuals at greatest short-term risk of a subsequent cardiovascular event or cardiovascular death in order to target them for more intensive follow-up, investigation, and treatment. Symptoms, traditional risk factors for initial vascular disease events, the burden of known vascular disease, and the results of specialized testing may affect risk for future vascular disease events.3 In addition to increasing age and male sex, the key factors that have been associated with increased risk of cardiovascular events have been higher blood pressure, cholesterol, lower high-density lipoprotein cholesterol, diabetes mellitus, and cigarette smoking.1,3 The American Journal of Medicine, Vol 125, No 7, July 2012 The initial REACH analyses tested for associations between traditional risk factors used in formulations like the Framingham risk model for initial cardiovascular disease events,20 and recurrent cardiovascular events included age, sex, systolic blood pressure, diastolic blood pressure, diabetes mellitus, and cigarette smoking.1 From this list of variables, age was highly associated with recurrent cardiovascular disease, and in age-adjusted regression models that considered the variables individually, only current smoking and the presence of diabetes mellitus were associated with recurrent cardiovascular disease. As shown in the multivariable analysis presented in Table 2, these variables (age, smoking, diabetes, and low body mass index) were significantly associated with recurrent cardiovascular events in the multivariable prediction model. An adverse effect of low body mass index on risk for future cardiovascular events has been noted for patients with cardiovascular disease and may represent poor nutrition and cachexia in severely ill patients.21 The REACH registry is a global study and allowed for investigation of vascular disease risk according to where the participant resided. In multivariable analysis, risk for recurrent cardiovascular disease was lower in some regions (Japan, Australia) and higher in others (Eastern Europe, Middle East) in comparison with the other sites, where the referent was the combined experience of Europe and North America. Further investigation into the origins of these differences is merited, and public health experts have identified a need to prevent and manage chronic diseases throughout the world.22 Baseline levels of blood cholesterol, systolic pressure, and diastolic pressure were not related to the development of recurrent cardiovascular events. Part of the explanation for this finding may be that medications to treat these conditions were very commonly used at the baseline examination. As shown in Table 1, current statin use at entry into the registry was reported for 74% of the participants. The survey form used in the study only queried whether the cholesterol level was œæ200 mg/dL, and information on the actual cholesterol level at baseline was not available for all participants. This study investigated the potential effect of the burden of cardiovascular disease on the risk of recurrent cardiovascular events. As described above for the traditional variables, the initial approach tested for age- and sex-adjusted associations between the cardiovascular burden variables (shown in Table 1) and cardiovascular death or recurrent cardiovascular disease. In an effort to summarize the potential risk associated with atherothrombotic disease involving multiple vascular beds, we used a composite variable ‚Äúnumber of vascular beds affected by symptomatic vascular disease,‚Äù which was highly associated with recurrent ischemic events. In addition, the occurrence of a cardiovascular event in the year before the baseline examination and a positive history of congestive heart failure were associated with greater risk of recurrent atherothrombotic events. In the multivariable analyses shown in Table 2, each of these Wilson et al Recurrent Vascular Disease Prediction 701 CV Step CV CV death: Next CV event: event Factor death factors and points factors and points points Man Woman Man Woman 1 Sex 1 0 1 20‚Äì24 25-29 30-34 35-39 40‚Äì44 45-49 50‚Äì54 0 2 1 2 3 4 5 7 4 5 0 20‚Äì24 25-29 30-34 35-39 40‚Äì44 45-49 50‚Äì54 0 6 1 2 3 4 5 6 Age, years 55‚Äì59 60-64 65-69 70‚Äì74 75-79 80‚Äì84 85-89 3 points 8 9 10 11 12 55‚Äì59 60-64 65-69 70‚Äì74 75-79 80‚Äì84 85-89 7 13 8 No Yes No Yes 0 2 0 1 No Yes No Yes 0 2 0 2 No Yes No Yes 0 2 0 2 9 10 11 12 13 Smoking Diabetes mellitus BMI < 20 kg/m2 Number of One Two Three One Two Three 6 2 vascular beds 4 6 1 2 No Yes No Yes 0 2 0 1 Congestive heart No Yes No Yes failure 0 3 0 4 No Yes No Yes 0 2 0 2 No Yes No Yes 0 ‚Äì2 0 -1 No Yes No Yes 0 -1 0 -1 Eastern Europe or No Yes No Yes Middle East 0 2 0 1 No Yes No Yes 0 -2 0 -3 CV event in past 3 7 year 8 9 10 11 Atrial fibrillation Statin therapy ASA therapy 12 13 14 Japan or Australia Next CV event points total CV death points total Figure 2 Score sheet to estimate the risk of the next cardiovascular event and of cardiovascular death over 20 months of follow-up, based on separate, multivariable, regression-estimating equations for each outcome. The sheet is used by assigning points for each factor, summing the points, and determining the percentage risk of either the next cardiovascular event or of cardiovascular death. CV œ≠ cardiovascular. variables was associated with the development of recurrent cardiovascular disease. Treatments for cardiovascular disease were considered as potential predictive factors in these analyses, and the factors of blood pressure treatment, lipid-lowering therapy, and use of acetylsalicylic acid were considered. Almost all of the lipid-lowering medications being used were statins, which led to our using that single medica- 702 The American Journal of Medicine, Vol 125, No 7, July 2012 Step 15 Calculation of risk from point score total Next CV event points 0 1 2 8 9 10 11 12 16 17 18 20-month risk of next CV event, % <1 1 1.2 1.4 1.6 1.9 2.2 2.5 3 3.5 4 4.7 5.4 6.3 7.3 8.5 9.8 11 13 Next CV event points 19 20 21 22 23 24 25 26 27 28 ‚â•29 20-month risk of next CV event, % 15 17 20 23 26 30 34 38 43 48 >50 CV death points 0-8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 ‚â•26 20-month risk of CV death, % <1 1.1 1.4 1.8 2.3 3 3.8 4.9 6.2 7.9 10 13 16 20 25 30 37 45 >50 3 4 5 Figure 2 tion as the lipid treatment variable. The use of antiplatelet medications other than acetylsalicylic acid was uncommon (œΩ28%), and those data were not analyzed for potential associations with recurrent cardiovascular outcomes. Initial age- and sex-adjusted analyses showed no relation between blood pressure medication use and the development of recurrent cardiovascular events. On the other hand, both statin medications and acetylsalicylic acid use were associated with a significantly lower risk for recurrent cardiovascular events and cardiovascular death in the multivariable analyses (Table 2). The strengths of this study include its large size, contemporary data collection, international scope, and consideration of several vascular disease event types. The study has limitations. A central laboratory for measurement of the lipid levels was not available, newer riskfactor variables such as inflammatory markers or natriuretic peptides were not included, and nonfatal ischemic event end points were not adjudicated. The risk prediction variables shown to be important determinants for recurrent cardiovascular disease and for cardiovascular death in patients who have already experienced at least one cardiovascular event should be investigated in other population samples. We have provided internally validated estimates of risk for recurrent cardiovascular disease and for cardiovascular death in patients who have already experienced a clinical vascular disease event. We encourage other research groups to use similar strategies to evaluate the risk of vascular disease events in this setting. The REACH Registry reflects the experience of outpatient clinics that were included as part of a research study, and the external validation of our research approach is of interest. We believe that investigating the components of vascular disease risk is probably more important than corroborating absolute risk estimates, which are likely to improve over time with more effective therapies for patients known to have clinical cardiovascular disease. Future research endeavors should consider prognostic factors beyond traditional risk markers that have been used 6 7 13 14 15 Continued to estimate risk for initial cardiovascular events. These factors include elements that identify the burden of ischemic atherothrombotic disease such as polyvascular disease, heart failure, a recent cardiovascular event, history of atrial fibrillation, and the salutary effects of medications such as statins and acetylsalicylic acid. ACKNOWLEDGMENT We thank sanofi-aventis and Bristol-Myers Squibb for their support of the REACH Registry. We also acknowledge statistical support by Alain Richard, PhD, at sanofi-aventis. The REACH Registry enforces a no ghost-writing policy. The first draft was written by Dr. Wilson. We thank the REACH Editorial Support Group for providing editorial help and assistance in preparing this manuscript, including editing, checking content and language, formatting, referencing, and preparing tables and figures. The REACH Registry is endorsed by the World Heart Federation. References 1. Wilson PW, D‚ÄôAgostino RB, Levy D, et al. Prediction of coronary heart disease using risk factor categories. Circulation. 1998;97(18): 1837-1847. 2. Expert Panel on Detection Evaluation and Treatment of High Blood Cholesterol in Adults. Summary of the second report of the National Cholesterol Education Program (NCEP) expert panel on detection, evaluation, and treatment of high blood cholesterol in adults (Adult Treatment Panel II). JAMA. 1993;269:3015-3023. 3. Lloyd-Jones DM. Cardiovascular risk prediction: basic concepts, current status, and future directions. Circulation. 2010;121(15):17681777. 4. Califf RM, Armstrong PW, Carver JR, et al. 27th Bethesda Conference: matching the intensity of risk factor management with the hazard for coronary disease events. Task Force 5. Stratification of patients into high, medium and low risk subgroups for purposes of risk factor management. J Am Coll Cardiol. 1996;27(5):1007-1019. 5. Anderson KM, Wilson PWF, Odell PM, Kannel WB. An updated coronary risk profile. A statement for health professionals. Circulation. 1991;83:357-363. 6. Anderson KM, Odell PM, Wilson PWF, Kannel WB. Cardiovascular disease risk profiles. Am Heart J. 1991;121:293-298. Wilson et al Recurrent Vascular Disease Prediction 7. Assmann G, Cullen P, Schulte H. Simple scoring scheme for calculating the risk of acute coronary events based on the 10-year follow-up of the prospective cardiovascular Munster (PROCAM) study. Circulation. 2002;105(3):310-315. 8. Ferrario M, Chiodini P, Chambless LE, et al. Prediction of coronary events in a low incidence population. Assessing accuracy of the CUORE Cohort Study prediction equation. Int J Epidemiol. 2005; 34(2):413-421. 9. Hippisley-Cox J, Coupland C, Vinogradova Y, et al. Derivation and validation of QRISK, a new cardiovascular disease risk score for the United Kingdom: prospective open cohort study. BMJ. 2007;335 (7611):136. 10. D‚ÄôAgostino RB Sr, Grundy S, Sullivan LM, Wilson P. Validation of the Framingham coronary heart disease prediction scores: results of a multiple ethnic groups investigation. JAMA. 2001;286(2):180-187. 11. Conroy RM, Pyorala K, Fitzgerald AP, et al. Estimation of ten-year risk of fatal cardiovascular disease in Europe: the SCORE project. Eur Heart J. 2003;24(11):987-1003. 12. Ohman EM, Bhatt DL, Steg PG, et al. The REduction of Atherothrombosis for Continued Health (REACH) Registry: an international, prospective, observational investigation in subjects at risk for atherothrombotic events-study design. Am Heart J. 2006;151(4):786.e1786.e10. 13. Steg PG, Bhatt DL, Wilson PW, et al. One-year cardiovascular event rates in outpatients with atherothrombosis. JAMA. 2007;297(11):11971206. 14. Bhatt DL, Steg PG, Ohman EM, et al. International prevalence, recognition, and treatment of cardiovascular risk factors in outpatients with atherothrombosis. JAMA. 2006;295(2):180-189. 15. Bhatt DL, Eagle KA, Ohman EM, et al. Comparative determinants of 4-year cardiovascular event rates in stable outpatients at risk of or with atherothrombosis. JAMA. 2010;304(12):1350-1357. 16. Cox DR. Regression models and life tables. J R Stat Soc Series B Stat Methodol. 1972;34:187-220. 17. Pencina MJ, D‚ÄôAgostino RB. Overall C as a measure of discrimination in survival analysis: model specific population value and confidence interval estimation. Stat Med. 2004;23(13):2109-2123. 18. D‚ÄôAgostino RB Sr, Nam BH. Evaluation of the performance of survival analysis models: discrimination and calibration measures. In: Balakrishnan N, Rao CR, eds. Handbook of Statistics, Volume 23: Advances in Survival Analysis. Amsteram: Elsevier; 2004:1-26. 19. Sullivan LM, Massaro JM, D‚ÄôAgostino RB Sr. Presentation of multivariate data for clinical use: the Framingham Study risk score functions. Stat Med. 2004;23(10):1631-1660. 20. D‚ÄôAgostino RB Sr, Vasan RS, Pencina MJ, et al. General cardiovascular risk profile for use in primary care: the Framingham Heart Study. Circulation. 2008;117(6):743-753. 21. von Haehling S, Lainscak M, Springer J, Anker SD. Cardiac cachexia: a systematic overview. Pharmacol Ther. 2009;121(3):227-252. 22. Beaglehole R, Epping-Jordan J, Patel V, et al. Improving the prevention and management of chronic disease in low-income and middle-income countries: a priority for primary health care. Lancet. 2008;372(9642):940-949. Funding: The REACH Registry is sponsored by sanofi-aventis, Bristol-Myers Squibb, and the Waksman Foundation (Tokyo, Japan). The sponsors provide logistical support. All the publication activity is controlled by the REACH Registry Global Publication Committee (Ph. Gabriel Steg, Deepak L. Bhatt, Mark Alberts, Ralph D‚ÄôAgostino, Kim Eagle, Shinya Goto, Alan T. Hirsch, Chiau-Suong Liau, Jean-Louis Mas, E. Magnus Ohman, Joachim R√∂ther, Sidney C. Smith, and Peter W.F. Wilson). All manuscripts in the REACH Registry are prepared by independent authors who are not governed by the funding sponsors and are reviewed by an academic publication committee before submission. The funding sponsors have the opportunity to review manuscript submissions but do not have authority to change any aspect of a manuscript. 703 Conflict of Interest: Dr Wilson has received research grants from sanofi-aventis within the last 3 years. Prof. D‚ÄôAgostino has received consultancy fees from sanofi-aventis and grant support from Pfizer. Dr Bhatt discloses the following relationships: advisory board: Medscape Cardiology; board of directors: Boston VA Research Institute, Society of Chest Pain Centers; chair: American Heart Association Get With The Guidelines Science Subcommittee; honoraria: American College of Cardiology (Editor, Clinical Trials, Cardiosource), Duke Clinical Research Institute (clinical trial steering committees), Slack Publications (Chief Medical Editor, Cardiology Today Intervention), WebMD (CME steering committees); research grants: Amarin, AstraZeneca, Bristol-Myers Squibb, Eisai, Ethicon, Medtronic, sanofi-aventis, The Medicines Company; unfunded research: FlowCo, PLx Pharma, Takeda. Dr Eagle has received consulting fees or served on paid advisory boards from the National Heart, Lung, and Blood Institute and the R.W. Johnson Foundation, and received grant support from sanofi-aventis. Dr Pencina has received consultancy fees from sanofi-aventis for work on this project. Dr Smith has received honoraria for consulting or Data and Safety Monitoring Board fees from sanofi-aventis, GlaxoSmithKline, Fournier, and AstraZeneca, and lecture fees for speaking at Continuing Medical Education symposia supported by unrestricted educational grants from Bayer, Pfizer, Merck, and sanofi-aventis. Dr Alberts discloses the following relationships: consultant/advisory board‚ÄîAGA Medical, AstraZeneca, Bayer, Bristol-Myers Squibb, Boehringer Ingelheim, Diadexus, Eli Lilly & Co, Genentech, KOS, Medicines Company, Merck, Novo Nordisk, PDL Biopharma Inc, Pfizer, Photo Thera, and sanofi-aventis; research grants‚ÄîAGA Medical, AstraZeneca, Bristol-Myers Squibb, Boehringer Ingelheim, Novo Nordisk, Photo Thera, sanofiaventis, and Schering Plough; speaker‚Äôs bureau‚ÄîAstraZeneca, BristolMyers Squibb, Boehringer Ingelheim, Diadexus, Genentech, Medicines Company, Novo Nordisk, PDL Biopharma Inc, and sanofi-aventis; honoraria‚ÄîAGA Medical, AstraZeneca, Bayer, Bristol-Myers Squibb, Boehringer Ingelheim, Diadexus, Eli Lilly & Co, Genentech, KOS, Medicines Company, Merck, Novo Nordisk, PDL Biopharma Inc, Pfizer, sanofiaventis, TAP Pharmaceuticals-Data and Safety Monitoring Board, and Schering Plough; review panel‚ÄîTAP Pharmaceuticals-Data and Safety Monitoring Board, and Schering Plough. Dr Dallongeville has received consulting fees from Bristol-Myers Squibb, MSD, and sanofi-aventis; and lecture fees from MSD-SP and sanofi-aventis; and grant support from Pfizer. Prof. Goto has received honoraria and consulting fees from Astellas, AstraZeneca, Bayer, Bristol-Myers Squibb, Daiichi-Sankyo, Eisai, GlaxoSmithKline, Kowa, Novartis, Otsuka, sanofi-aventis, Schering-Plough, and Takeda. Prof. Goto also received research grants from Eisai, Ono, sanofi-aventis, AstraZeneca, Kowa, and Pfizer within the past 3 years. Dr Hirsch has received research grants from Bristol-Myers Squibb, sanofi-aventis, AstraZeneca, and the National Heart, Lung, and Blood Institute; and consulting fees from Pfizer, Bristol-Myers Squibb, and sanofi-aventis. Prof. Ohman has received grant support from Bristol-Myers Squibb, The Medicines Company, Eli Lilly, sanofi-aventis, and Schering-Plough; consultancy fees from Abiomed, Datascope, Inovise, Liposcience, Response Biomedical, Savacor, and The Medicines Company; payment for speaker‚Äôs bureau from CV Therapeutics and Schering-Plough within the past 3 years; and is a shareholder of Inovise, Medtronic, and Savacor. Prof. R√∂ther has received payment for speakers‚Äô bureau and consultancy fees from sanofi-aventis, Boehringer Ingelheim, MSD, and Bristol-Myers Squibb. Dr Mas has received consulting fees from sanofiaventis, Servier, and Bristol-Myers Squibb; and lecture fees from sanofiaventis, Bristol-Myers Squibb, and Boehringer Ingelheim. Prof. Steg has received honoraria for advisory board attendance and consulting fees from AstraZeneca, Boehringer Ingelheim, Bayer, Medtronic, GlaxoSmithKline, Merck, Nycomed, sanofi-aventis, Servier, Astellas, and The Medicines Company; and payment for speakers‚Äô bureau from Boehringer Ingelheim, Bristol Myers-Squibb, GlaxoSmithKline, Medtronic, Nycomed, sanofi-aventis, and Servier. None of the other authors reported disclosures. Authorship: The authors had full access to the data and take full responsibility for its integrity. All authors have read and agree to the manuscript as written. 703.e1 The American Journal of Medicine, Vol 125, No 7, July 2012 APPENDIX Table Cox Model Coefficients for Calculation of 20-Month Risk of Recurrent Cardiovascular Events Type of Cardiovascular Event Factor Traditional risk factors Male sex Age, years Smoking (current vs other) Diabetes (yes/no) BMI œΩ 20 kg/m2 Cardiovascular disease burden at baseline Number of vascular beds Cardiovascular event* in past year (yes/no) Congestive heart failure (yes/no) Atrial fibrillation (yes/no) Cardiovascular treatment Statins (yes/no) Acetylsalicylic acid (yes/ no) Aemsustment for geographic region Eastern Europe or Middle East Japan or Australia Cardiovascular Death ‚ê§ Next Cardiovascular Event ‚ê§ 0.24519 0.04966 0.30925 0.46141 0.55132 0.10246 0.03089 0.24121 0.37824 0.31428 0.24928 0.26681 0.30277 0.38168 0.89976 0.51873 0.42705 0.26652 œ™0.22296 œ™0.17968 œ™0.28332 œ™0.10151 0.25934 0.27574 œ™0.65524 œ™0.31604 BMI œ≠ body mass index. *The 20-month risk for cardiovascular death can be calculated as exp(‚å∫‚ê§X ‚Äì 4.03317) , where ‚ê§ is the regression coefficient and X is the 1-0.97749 level for each risk factor; the risk for next cardiovascular event is given as 1-0.93681exp(‚å∫‚ê§X ‚Äì 2.68845). REACH REGISTRY EXECUTIVE COMMITTEE Deepak L. Bhatt, MD, MPH, VA Boston Healthcare System and Brigham and Women‚Äôs Hospital, Boston, MA (chair); Ph. Gabriel Steg, MD, H√¥pital Bichat-Claude Bernard, Paris, France (chair); E. Magnus Ohman, MD, Duke University Medical Center, Durham, NC; Joachim R√∂ther, MD, Klinikum Minden, Minden, Germany; Peter W. F. Wilson, MD, Emory University School of Medicine, Atlanta, GA. REACH REGISTRY GLOBAL PUBLICATION COMMITTEE Mark Alberts, MD, Northwestern University Medical School, Chicago, IL; Deepak L. Bhatt, MD, MPH, VA Boston Healthcare System and Brigham and Women‚Äôs Hos- pital, Boston, MA (chair); Ralph D‚ÄôAgostino, PhD, Boston University, Boston, MA; Kim Eagle, MD, University of Michigan, Ann Arbor, MI; Shinya Goto, MD, PhD, Tokai University School of Medicine, Isehara, Kanagawa, Japan; Alan T. Hirsch, MD, University of Minnesota School of Public Health and Minneapolis Heart Institute Foundation, Minneapolis, MN; Chiau-Suong Liau, MD, PhD, Taiwan University Hospital and College of Medicine, Taipei; JeanLouis Mas, MD, Centre Raymond Garcin, Paris, France; E. Magnus Ohman, MD, Duke University Medical Center, Durham, NC; Joachim R√∂ther, MD, Klinikum Minden, Minden, Germany; Sidney C. Smith, MD, University of North Carolina at Chapel Hill; P. Gabriel Steg, MD, H√¥pital Bichat-Claude Bernard, Paris, France (chair); Peter W. F. Wilson, MD, Emory University School of Medicine, Atlanta, GA. NATIONAL COORDINATORS Australia: Christopher Reid, Victoria. Austria: Franz Aichner, Linz; Thomas Wascher, Graz. Belgium: Patrice Laloux, Mont-Godinne. Brazil: Denilson Campos de Albuquerque, Rio de Janeiro. Bulgaria: Julia Djorgova, Sofia. Canada: Eric A. Cohen, Toronto, ON. Chile: Ramon Corbalan, Santiago. China: Chuanzhen LV, Shanghai; Runlin Gao, Beijing. Denmark: Per Hildebrandt, Frederiksberg. Finland: Ilkka Tierala, Helsinki. France: Jean-Louis Mas, Patrice Cacoub, and Gilles Montalescot, Paris. Germany: Klaus Parhofer, Munich; Uwe Zeymer, Ludwigshafen; Joachim R√∂ther, Minden. Greece: Moses Elisaf, Ioannina. Interlatina (Guatemala): Romulo Lopez, Guatemala City. Hong Kong: Juliana Chan, Shatin. Hungary: Gy√∂rgy Pfliegler, Debrecen. Indonesia: Bambang Sutrisna, Jakarta. Israel: Avi Porath, Beer Sheva. Japan: Yasuo Ikeda, Tokyo. Lebanon: Ismail Khalil, Beirut. Lithuania: Ruta Babarskiene, Kaunas. Malaysia: Robaayah Zambahari, Kuala Lumpur. Mexico: Efrain Gaxiola, Jalisco. The Netherlands: Don Poldermans, Rotterdam. Philippines: Maria Teresa B. Abola, Quezon City. Portugal: Victor Gil, Amadora. Romania: Constantin Popa, Bucharest. Russia: Yuri Belenkov and Elizaveta Panchenko, Moscow. Saudi Arabia: Hassan Chamsi-Pasha, Jeddah. Singapore: Yeo Tiong Cheng. South Korea: Oh Dong-Joo, Seoul. Spain: Carmen Suarez, Madrid. Switzerland: Iris Baumgartner, Bern. Taiwan: ChiauSuong Liau, Taipei. Thailand: Piyamitr Sritara, Bangkok. United Arab Emirates: Wael Al Mahmeed, Abu Dhabi. United Kingdom: Jonathan Morrell, Hastings. Ukraine: Vira Tseluyko, Kharkov. United States of America: Mark Alberts, Chicago, IL; Robert M. Califf, Durham, NC; Christopher P. Cannon, Boston, MA; Kim Eagle, Ann Arbor, MI; Alan T. Hirsch, Minneapolis, MN. A complete list of the REACH Registry Investigators appears in JAMA. 2006;295:180-189. ˇ˛Anorexia, cachexia and nutrition N a stephens K c h Fearon Abstract anorexia, cachexia and malnutrition occur in many chronic disease states including cancer, chronic obstructive pulmonary disease, chronic heart failure, chronic renal failure, chronic liver failure, rheumatoid arthritis and aiDs. Despite the relative frequency of such symptoms and signs, they are often overlooked or managed inadequately. cachexia is characterized by a negative nitrogen and energy balance, with loss of both lean body mass and adipose tissue. it is due to a combination of reduced food intake and metabolic change triggered by pro-inflammatory cytokines and the neuroendocrine stress response. clinical management needs to focus on stabilizing the primary condition whilst treating any reversible factors that might contribute to ongoing weight loss. in the past, cachexia was seen to be an irreversible and terminal event. however, advancements over the last couple of decades have meant that in a substantial propor- tion of patients, therapy can reverse, or at least halt, the progression of wasting with subsequent improvement in quality, if not quantity, of life. Keywords anorexia; cachexia; cancer; cardiac failure; coPD; malnutrition Anorexia, cachexia and malnutrition are important issues in a wide range of disease processes including cancer, chronic heart failure (CHF) and chronic obstructive pulmonary disease (COPD). However, in clinical practice many clinicians do not assess routinely nutritional status, or treat anorexia/cachexia appropriately. This is detrimental for the patient as development of malnutrition and muscle wasting increases morbidity and mortality and impinges on quality of life. Definitions Anorexia is the loss of the desire to eat associated with a reduced food intake. Cachexia is a clinical syndrome characterized by anorexia, early satiety, severe weight loss, weakness, anaemia and oedema. Cachexia is easily recognized in its advanced stage, but much more difficult to predict for the purposes of early intervention. N A Stephens MBChB MRCSEd is a Surgical Research Fellow at the Department of Clinical and Surgical Sciences (Surgery), University of Edinburgh, Edinburgh, UK. Competing interests: none declared. K C H Fearon MBChB(Hons) MD FRCS(Glas) is Professor of Surgical Oncology at the Department of Clinical and Surgical Sciences (Surgery), University of Edinburgh, Edinburgh, UK. Competing interests: none declared. Epidemiology and prevalence Weight loss is seen as the hallmark of cachexia and commonly triggers patients to consult their family physician. At diagnosis, one-third of all cancer patients will have lost >5% of their origi- nal body weight and 20% of cancer-related deaths are directly caused by malnutrition and cachexia due to immobility and car- diorespiratory failure.1 Patients with CHF-related cachexia or COPD with associated malnutrition also have increased mortality. Anorexia is frequently associated with cachexia (Figure 1) and contributes to the development of malnutrition and cachexia since it results in a reduced oral intake. Aetiology and pathogenesis Cachexia Whilst starvation leads to depletion in both fat mass and lean body mass, cachectic patients often suffer disproportionate loss of skeletal muscle mass.2 These changes are attributed to the presence of a chronic inflammatory response perpetuated by proinflammatory cytokines (tumour necrosis factor (TNF)-±, interferon (IFN)-≥, interleukin (IL)-1 and IL-6) and stimulation of the neuroendocrine stress response. Other potential mediators of cachexia include deficiencies of anabolic factors (e.g. testos- terone, insulin-like growth factor (IGF)-1) and an excess of cata- bolic factors (eg. myostatin, glucocorticoids).3 Anorexia Anorexia has a neurohormonal origin due to disturbances of the central mechanisms controlling food intake along with several secondary mechanisms (Figure 2).4 Energy intake is controlled by the hypothalamus where peripheral signals convey informa- tion on energy and adiposity status. The arcuate nucleus within the hypothalamus plays a key role in feeding and metabolism. In particular, the POMC (pro-opiomelanocortin) neurons are Physical Problems Incidence of cachexia/anorexia 100 80 60 40 20 0 Anorexia Cachexia Cancer COPD CHF COPD, chronic obstructive pulmonary disease; CHF, congestive heart failure. meDiciNe 36:2 78 © 2007 Published by elsevier ltd. Figure 1 % affected Physical Problems Pathogenesis of anorexia Neurohormonal Cytokines (IL-1, IL-6, TNF) Peripheral signals: Adiposity signals (leptin, ghrelin) Energy signals (intracellular- malonyl-CoA) Neurotransmitters Peptides Secondary mechanisms Anxiety Depression Intestinal obstruction Nausea Vomiting Persistent pain Treatment side effects Taste alterations Arcuate nucleus Orexigenic pathway NPY/AgRP Anorexigenic pathway POMC/CART Stimulate food intake Inhibit food intake Second order neurones NPY, neuropeptide Y; AgRP, agouti-related peptide; POMC, pro-opiomelanocortin; CART, cocaine and amfetamine-regulated transcript. Behavioural response Anorexia/Malnutrition Pathogenesis of cachexia in chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF) and cancer Disease COPD CHF Pathogenesis Energy/food intake Cancer Oxidative stress NF∫B/UPP Anaerobic metabolism Neuroendocrine stress Cytokines EE ë! Intake ì! EE ë!/î! Intake î! EE ë!/î!/ì! Intake ì! Imbalance of protein degradation/synthesis in favour of catabolism Cachexia NF∫B, nuclear factor-kappa B; UPP, ubiquitin proteasome pathway; EE, energy expenditure. Cytokines Tumour specific factors Figure 2 a target for cytokine activation leading to hyperactivation of the melanocortin 4 receptor (MC4R) causing anorexia.5,6 This has led to interest not only in developing MC4R antagonists to reverse anorexia, but also MC4R agonists in the treatment of obesity.7 Disease-specific pathogenesis (Figure 3) Cardiac cachexia Cardiac cachexia is seen in patients with severe CHF. Muscle atrophy occurs in up to 68% of CHF patients8 and there is a reduction in bone mineral density and fat tissue mass. Blood levels of inflammatory cytokines, in particular TNF±, are raised.9 Certain stimuli may propagate the innate immune response and include: " hypoxia: shown to stimulate TNF± production10 " the myocardium: during failure is capable of producing cytokines11 " endotoxins: due to bacterial translocation through oedema- tous bowel wall.12 meDiciNe 36:2 79 Figure 3 Additionally, there is a general neuroendocrine activation stimu- lated by the sympathetic nervous system (SNS), renin-angiotensin- aldosterone system and the natriuretic peptide system. Renin stimulates release of both norepinephrine and angiotensin II, the latter of which is known to cause anorexia and wasting in animal models (in part, by reducing IGF-1).13 In combination with SNS activation and increased levels of cortisol, this leads to a metabolic shift towards catabolism and a rise in resting energy expenditure. COPD cachexia The first important factor leading to development of muscle wasting in COPD is a change in energy metabolism. There is elevation of resting energy expenditure largely due to pulmonary mechanical inefficiency and increased respiratory muscle work. Anaerobic metabolism is increased leading to less efficient ATP production and there is impaired oxidative phosphorylation.14 Initially, there is a compensatory increase in dietary intake, but when this becomes compromised due to dyspnoea, fatigue and meal-related reduction in oxygen saturation, the patient loses weight. The acute phase response occurs during infective exacer- bations of COPD and, if these recur frequently or there is contin- ued low-grade inflammation, skeletal muscle is broken down and © 2007 Published by elsevier ltd. lost to supply amino acids to support increased hepatic export protein synthesis. The nuclear factor-kappa B (NF∫B) signalling pathway provides a vital role in the control of muscle degrada- tion by enhancing transcription of the proteolytic ubiquitin pro- teasome pathway components.15 TNF± is commonly associated with NF∫B activation, but alternative triggers include systemic inflammation, oxidative stress and tissue hypoxia all of which are more prevalent in COPD patients with low body weight. Cancer cachexia Cancer cachexia results from interactions between the host and the tumour, the nature of which are not fully understood. The host responds to the presence of a tumour by activation of both the acute phase and neuroendocrine stress response. Pro-inflammatory cytokines are produced by host inflammatory cells, although cytokines can also be produced directly by tumour cells or from a combination of the two.16 The tumour itself can express procachectic factors that not only have direct catabolic effects (e.g. proteolysis inducing factor, lipid mobilizing factor) but also may act via amplification of the pro-inflammatory and stress cascades. A net inadequate nutri- tional intake occurs frequently in cancer patients who fail to increase their intake appropriately to cope with increasing resting energy demands. This reduced food intake and metabolic change are vitally important in patients with cancer cachexia. Clinical presentation, examination and investigation Clinical history and examination remain the fundamental tools in diagnosing patients with anorexia, cachexia and nutritional problems. Advanced cachexia is easily recognized, but diagnos- ing cachexia in its early stages remains challenging. Weight loss, anorexia and fatigue are the commonest symptoms reported, but different symptoms predominate in individual patients and may also change with time. Weight loss >5% suggests develop- ing cachexia, while patients with >15% weight losses are well advanced into the cachectic state. Plasma albumin concentra- tion may be low and, if associated with an elevated C-reactive protein or erythrocyte sedimentation rate, reflects an underlying systemic inflammatory response that contributes to the weight- losing process.17 A nutritional assessment should be carried out, and a dieti- tian involved at an early stage. Body weight and body mass index (BMI) are traditional indicators of nutrition, but with the increasing tendency towards obesity in the West, and the possi- ble presence of compounding factors such as oedema or ascites, these markers may be normal even in a patient who is cachectic and malnourished. The Malnutrition Universal Screening Tool (MUST) is perhaps a more useful screening tool for patients at nutritional risk. BMI, percentage weight loss, and effect of dis- ease on food intake are scored to categorize the patient s risk of malnutrition. Appropriate treatment is then instigated based on this risk. Management Given that cachexia is due to both anorexia and metabolic change, it is logical that any treatment modality would aim to treat both of these factors. The first steps in managing anorexia and cachexia should be optimal treatment of the underlying cause, elimination of ongoing inflammation or infection, followed by active reduction of any exacerbating factors, such as diabetes or immobility. Nutrition and exercise programmes Targeted pulmonary rehabilitation programmes in COPD cachexia have been shown to be beneficial in randomized con- trolled trials.18 With just 2"3 weeks of oral/enteral nutritional supplementation, body weight rose and respiratory muscle func- tion improved significantly in COPD patients.19 When daily nutri- tional supplements were combined with an exercise programme, there was a weight gain of 0.4 kg/week.20 Exercise seems to increase the effectiveness of nutritional therapy and stimulate appetite. Improvements in body weight, fat free mass, ventila- tory muscle function, handgrip strength, peak work capacity and health-related quality of life have all been demonstrated in COPD patients managed in this way.18,21 This translates to improved patient survival. In cancer cachexia, adequate nutritional intake is of prime importance. Patients should have small high-energy meals avoid- ing high fat contents, as this will delay gastric emptying. Meals should be eaten in pleasant surroundings and extreme tastes and smells should be avoided. High energy/protein supplements should also be prescribed. Artificial nutritional support, either via the enteral or parenteral route has a place in patients where limited food intake is the main cause and where there is a limited tumour burden and good performance status to justify invasive treatment. Drug therapy Current therapies for cachexia act by either reducing anorexia/ stimulating appetite, attenuating skeletal muscle catabolism or stimulating muscle protein anabolism. Cardiac cachexia Angiotensin-converting enzyme (ACE) inhibitors reduce levels of atrial natriuretic peptide, TNF± and IL6 and reduce the risk of weight loss. Angiotensin II antagonists produce similar effects. The diminished activity of angiotensin II can restore depressed levels of circulating IGF1 leading to increased muscle protein synthesis. ≤-blockers have been found to prevent, and in some cases partially reverse, cardiac cachexia.22 Cancer cachexia Megesterol acetate, a synthetic progestagen, is an appetite stimu- lant which helps reduce anorexia.23 Whilst there are improve- ments in appetite, weight and well-being, it can induce male hypogonadism and adrenal suppression and should thus be used selectively. Non-steroidal anti-inflammatory drugs (NSAIDs) inhibit prostaglandin production by the cyclo-oxygenases (COX) and are commonly used to treat fever and pain. Ibuprofen has been found to decrease C-reactive protein, stimulate body weight gain and may improve survival in cancer.24,25 Eicosapentaenoic acid (an omega-3 fatty acid) has gained widespread interest in the treatment of cancer cachexia as it inhibits IL-1 and TNF± produc- tion. When pancreatic cancer patients with cachexia were given doses 20 40 times the normal dietary intake, their weight was stabilized.26,27 There is also evidence that omega-3 fatty acids Physical Problems have anti-tumour activity themselves and can down-regulate the toxicity of chemotherapy.28,29 Summary Anorexia, cachexia and nutritional deficits are important clinical phenomena, which should be routinely assessed in all patients who are at risk. Therapeutic and preventative measures should be implemented at an early stage and the combined resources of a multidisciplinary team used to attenuate disease progression and improve patient function and quality of life. EBIOM-00176; No of Pages 10 EBioMedicine xxx (2015) xxx‚Äìxxx Contents lists available at ScienceDirect EBioMedicine journal homepage: www.ebiomedicine.com A Novel Assay to Measure the Magnitude of the Inducible Viral Reservoir in HIV-infected Individuals Francesco Andrea Procopio a, R√©mi Fromentin a, Deanna A. Kulpa a, Jessica H. Brehm a, Anne-Gaelle Bebin a, Matthew C. Strain b, Douglas D. Richman b, Una O'Doherty c, Sarah Palmer d,e, Frederick M. Hecht f, Rebecca Hoh f, Richard J.O. Barnard g, Michael D. Miller g, Daria J. Hazuda g, Steven G. Deeks f, RaÔ¨Åck-Pierre S√©kaly a,‚Åé, Nicolas Chomont a,‚Åé‚Åé a Vaccine and Gene Therapy Institute Florida, Port St. Lucie, FL, USA University of California San Diego, La Jolla, California and Veterans Affairs San Diego Healthcare System, San Diego, CA, USA c Department of Pathology and Laboratory Medicine, University of Pennsylvania, Philadelphia, PA, USA d Centre for Virus Research, Westmead Millennium Institute, Westmead, Australia e Sydney Medical School, University of Sydney, Sydney, Australia f Department of Medicine, University of California San Francisco, San Francisco, CA, USA g Infectious Disease, Merck Research Laboratories, West Point, PA, USA b a r t i c l e i n f o Article history: Received 16 June 2015 Received in revised form 23 June 2015 Accepted 23 June 2015 Available online xxxx Keywords: HIV Latency Inducible virus Reservoir Multiply spliced RNA Eradication TILDA a b s t r a c t Background: Quantifying latently infected cells is critical to evaluate the efÔ¨Åcacy of therapeutic strategies aimed at reducing the size of the long-lived viral reservoir, but the low frequency of these cells makes this very challenging. Methods: We developed TILDA (Tat/rev Induced Limiting Dilution Assay) to measure the frequency of cells with inducible multiply-spliced HIV RNA, as these transcripts are usually absent in latently infected cells but induced upon viral reactivation. TILDA requires less than a million cells, does not require RNA extraction and can be completed in two days. Findings: In suppressed individuals on ART, we found the median frequency of latently infected CD4+ T cells as estimated by TILDA to be 24 cells/million, which was 48 times more than the frequency measured by the quantitative viral outgrowth assay, and 6‚Äì27 times less than the frequencies of cells harbouring viral DNA measured by PCR-based assays. TILDA measurements strongly correlated with most HIV DNA assays. The size of the latent reservoir measured by TILDA was lower in subjects who initiated ART during the early compared to late stage of infection (p = 0.011). In untreated HIV disease, the frequency of CD4+ cells carrying latent but inducible HIV largely exceeded the frequency of actively producing cells, demonstrating that the majority of infected cells are transcriptionally silent even in the absence of ART. Interpretations: Our results suggest that TILDA is a reproducible and sensitive approach to measure the frequency of productively and latently infected cells in clinical settings. We demonstrate that the latent reservoir represents a substantial fraction of all infected cells prior to ART initiation. Research in context: In this manuscript, we describe the development of a novel assay that measures the magnitude of the latent HIV reservoir, the main barrier to HIV eradication. This novel assay, termed TILDA for Tat/rev Induced Limiting Dilution Assay, requires only 10 ml of blood, does not necessitate extraction of viral nucleic acids, is highly reproducible, covers a wide dynamic range of reservoir sizes and can be completed in two days. As such, TILDA may represent an alternative to existing assays used to evaluate the efÔ¨Åcacy of therapeutic strategies aimed at reducing the size of the latent HIV reservoir. ¬© 2015 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1. Introduction ‚Åé Correspondence to: R.-P. S√©kaly, Case Western Reserve University, Cleveland, OH 44106, USA. ‚Åé‚Åé Correspondence to: N. Chomont, Current address: Centre de Recherche du CHUM, 900 rue St-Denis, H2X0A9 QC, Canada. E-mail addresses: raÔ¨Åck.sekaly@case.edu (R.-P. S√©kaly), nicolas.chomont@umontreal.ca (N. Chomont). Advances in the treatment of HIV infection have dramatically reduced AIDS mortality and improved the quality of life of many people living with HIV (Palella et al., 1998). The initiation of ART results in a rapid drop in plasma viral load and in a substantial reduction in the number of cells carrying proviral DNA in both blood and tissues http://dx.doi.org/10.1016/j.ebiom.2015.06.019 2352-3964/¬© 2015 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Please cite this article as: Procopio, F.A., et al., A Novel Assay to Measure the Magnitude of the Inducible Viral Reservoir in HIV-infected Individuals, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.019 2 F.A. Procopio et al. / EBioMedicine xxx (2015) xxx‚Äìxxx (Murray et al., 2014; Garrigue et al., 2000). However, ART alone does not eradicate HIV. The virus persists in long-lived latently infected CD4+ T cells that can produce infectious particles following T cell stimulation (Finzi et al., 1997; Wong et al., 1997; Chun et al., 1997a). The magnitude of the latent reservoir is highly variable among virally suppressed individuals and is inÔ¨Çuenced by a variety of factors including the nadir CD4 + T cell count (Boulassel et al., 2012), the CD4/CD8 ratio (Chun et al., 2002), the time between infection and initiation of ART (Josefsson et al., 2013; Strain et al., 2005; Archin et al., 2012; Buzon et al., 2014; Ananworanich et al., 2012; Chomont et al., 2009) and, to a lesser extent, the duration of therapy (Siliciano et al., 2003). Because strategies targeting this latent reservoir are now being tested in clinical trials, well-validated high-throughput assays that quantify this reservoir are urgently needed to evaluate the effectiveness of these interventions. Several assays have been developed to measure HIV persistence in virally suppressed subjects who have received ART for years (Lewin et al., 1999; Li et al., 2010; Siliciano and Siliciano, 2005; Yu et al., 2008; Palmer et al., 2003; Vandergeeten et al., 2014; Pasternak et al., 2008; Eriksson et al., 2013; Brussel and Sonigo, 2003; Rouzioux et al., 2005; Strain et al., 2013; O'Doherty et al., 2002). The quantitative viral outgrowth assay (Q-VOA) performed on puriÔ¨Åed resting CD4+ T cells isolated from patients on suppressive ART is considered the gold standard method to measure the minimum frequency of latently infected cells with replication competent virus (Finzi et al., 1997; Siliciano and Siliciano, 2005; Chun et al., 1997b). However, this assay is timeconsuming, labour intensive, expensive and requires large volumes of blood (120‚Äì180 ml). Recently, a more rapid and simpliÔ¨Åed version of Q-VOA has been described (Laird et al., 2013), but this novel assay still requires large number of cells and has not been adapted to a large scale clinical trial. In addition, Ho et al. identiÔ¨Åed intact proviruses from p24 negative wells of the virus outgrowth assay, which raises the possibility that the Q-VOA underestimates the size of the in vivo latent reservoir (Ho et al., 2013). This could be attributed to the fact that HIV reactivation in this system is inherently stochastic, as recently proposed (Weinberger and Weinberger, 2013). An alternative to the Q-VOA is the use of PCR-based methods that measure the frequency of cells harbouring HIV genomes (either total or integrated HIV DNA) (Yu et al., 2008; Vandergeeten et al., 2014; Brussel and Sonigo, 2003; Strain et al., 2013; O'Doherty et al., 2002). Although these methods can be used in large cohort studies, they are often criticized, as a large number of the viral genomes quantiÔ¨Åed by these assays are not replication-competent. Indeed, total HIV DNA in PBMCs and resting CD4+ T cells generally yields infected cell frequencies that are 2‚Äì3 logs higher than the Q-VOA, reÔ¨Çecting the high occurrence of defective and non-inducible viral genomes (Eriksson et al., 2013). Notwithstanding this limitation, measuring viral DNA, and particularly integrated HIV DNA, has provided crucial information that have contributed to the understanding of the mechanisms of HIV persistence during ART (Chomont et al., 2009; Agosto et al., 2011; Graf et al., 2011; Mexas et al., 2012; Vandergeeten et al., 2013). Levels of HIV DNA predict viral rebound after structured treatment interruption (Williams et al., 2014; Yerly et al., 2004) and integrated HIV DNA is the only assay that appears to correlate with Q-VOA (Eriksson et al., 2013) although it is likely that frequencies of cells bearing HIV DNA greatly overestimates the size of the inducible latent HIV reservoir (Eriksson et al., 2013). These studies thus emphasize the need to develop novel assays that would measure the size of the latent and inducible HIV reservoir in a simple, reproducible and cost-effective manner. An ideal assay would measure the frequency of latently infected CD4+ T cells without relying on the ampliÔ¨Åcation of viral replication, which is difÔ¨Åcult to control and requires at least a week of cell culture. Measuring the production of viral particles in culture supernatants of stimulated cells is attractive (Cillo et al., 2014), but would require ultracentrifugation and RNA extraction steps that are not desirable for a clinical trial scalable assay. Cell associated RNA is an alternative virological marker that can be easily measured in a limiting dilution assay. Low amounts of cell-associated unspliced HIV RNA are frequently detected in PBMCs from virally suppressed subjects on ART (Lewin et al., 1999; Furtado et al., 1999; Fischer et al., 2002; Schmid et al., 2010; Pasternak et al., 2009), as well as in latently infected CD4+ T cells that do not produce HIV particles (Fischer et al., 2004; Peng et al., 1995; Hermankova et al., 2003), and therefore, cannot be used as a surrogate for viral release (Cillo et al., 2014; Kearney, 2015). In contrast, tat/rev multiply spliced RNA (msRNA) reÔ¨Çects the ability of a cell to produce virus (Lewin et al., 1999; Pasternak et al., 2008, 2009; Fischer et al., 2002, 2004; Schmid et al., 2010; Peng et al., 1995; Hermankova et al., 2003; Vesanen et al., 1997; Sonza et al., 2002). Of note, recent data indicate that tat positive feedback alone may be sufÔ¨Åcient in reversing latency, independent of cellular activation (Razooky et al., 2015), suggesting that tat transcripts may be used as a surrogate for productive infection. This is consistent with the fact that direct addition of tat activates viral expression in resting CD4+ T cells without requiring cellular activation (Lassen et al., 2006; Lin et al., 2003). In addition, many defective HIV genomes have deletions that encompass the tat and rev genes (Ho et al., 2013), indicating that transcripts from these genes are very unlikely to be generated in cells harbouring defective proviruses. This provides a rationale for the development of a sensitive assay that can detect tat/rev msRNA upon maximal stimulation. Here, we report the development of TILDA, a novel assay that measures the frequency of cells harbouring viral genomes that produce tat/rev multiply-spliced HIV RNA upon maximal stimulation. By combining ultrasensitive detection of msRNA and maximal activation of CD4+ T cells in a limiting dilution format, this method allowed us to measure the frequency of cells capable of being induced to produce HIV RNA transcripts. SigniÔ¨Åcantly, this assay has potential for application in larger clinical trials and cohort studies as it requires only 10 ml of blood, is robust, sensitive, covers a wide dynamic range of reservoir size and can be completed in two days. 2. Methods 2.1. Participants and Samples Samples from 20 subjects on stably suppressive ART and from 13 viremic subjects not receiving ART at the time of sampling were used in this study. None of the participants under ART experienced any detectable plasma viremia at the time of study, as assessed by viral load measurement using the Ampliprep/Cobas Taqman HIV-1 Test v 2.0 (Roche), with a detection limit of 20 copies/ml of plasma. These subjects had an undetectable viral load for an average of 6 years (range 3‚Äì17 years). All participants underwent leukapheresis to collect large numbers of PBMCs. Subjects characteristics are summarized in Table 1. Data presented in Figs. 2 and 3 were generated by analysing samples obtained from individuals who participated in a previous study aimed at comparing different assays to measure HIV persistence (Eriksson et al., 2013). BrieÔ¨Çy, this study enrolled 30 subjects from two wellestablished cohorts at the University of California San Francisco (UCSF). Samples from 27 of these participants were used for the present study (17 from the SCOPE cohort and 10 from the OPTIONS cohort). All of the 27 subjects had at least 36 months of continuous ART at study entry with no regimen changes in the preceding 24 weeks and maintenance of plasma HIV RNA levels below the limit of detection of conventional assays for at least 36 months (intermittent isolated episodes of detectable low-level viremia were allowed). Primary or recent HIV infection was deÔ¨Åned if one or several of these 3 criteria were met: (Palella et al., 1998) repeated plasma HIV RNA N5000 copies/ml combined with a negative or indeterminate HIV antibody test; (Murray et al., 2014) seroconversion within 6 months of a documented negative HIV antibody test; (Garrigue et al., 2000) a history compatible with primary HIV infection (including no prior positive HIV antibody tests) and laboratory testing consistent with recent infection on the ‚Äúdetuned‚Äù Please cite this article as: Procopio, F.A., et al., A Novel Assay to Measure the Magnitude of the Inducible Viral Reservoir in HIV-infected Individuals, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.019 F.A. Procopio et al. / EBioMedicine xxx (2015) xxx‚Äìxxx 3 Table 1 Participant characteristics. Sexa CD8 count (cells/Œºl) HIV viral load (copies/ml) Duration of HIV infection (years)b Duration of ART (years)c ART regimend Virally suppressed subjects ART1 NA 1177 ART2 M 1183 ART3 M 1112 ART4 M 326 ART5 M 612 ART6 M 1065 ART7 M 1145 ART8 M 1095 ART9 M 876 ART10 F 522 ART11 F 460 ART12 F 355 ART13 M 289 ART14 F 351 ART15 M 734 ART16 M 267 ART17 M 1102 ART18 M 291 ART19 M 1897 ART20 M 387 Median 734 [353‚Äì1107] [IQR] 935 912 594 1085 834 1408 1410 736 1389 238 1879 815 512 344 703 518 875 491 830 625 815 [556‚Äì1010] b20 b20 b20 b20 b20 b20 b20 b20 b20 b20 b20 b20 b20 b20 b20 b20 b20 b20 b20 b20 b20 [b20‚Äì b 20] NA 25.6 12.9 6.2 25.7 15.0 6.1 22.1 5.7 22.1 5.4 21.1 16.2 15.1 7.2 NA NA 10.4 NA NA 15 [6.7‚Äì21.6] 6.5 3.4 12.9 6.1 16.9 8.2 5.8 6.2 5.0 7.5 4.4 4.7 3.0 3.6 2.5 3.8 NA 2.9 6.1 6.3 5.8 [3.7‚Äì6.4] NA ABC/3TC/ATV ABC/3TC/NVP TDF/FTC/EFV NA TDF/FTC/EFV TDF/FTC/LPV/r TDF/FTC/EFV ABC/3TC/EFV TDF/FTC/ATV/r TDF/FTC/RAL/LPV/r TDF/FTC/DRV/r TDF/FTC/LPV/r TDF/FTC/ATV/r TDF/FTC/RAL TDF/FTC/EFV TDF/FTC/ATV ABC/3TC/ATV/r TDF/FTC/EFV ABC/3TC/ATV Viremic subjects VIR1 F VIR2 M VIR3 F VIR4 M VIR5 M VIR6 M VIR7 F VIR8 F VIR9 M VIR10 NA VIR11 M VIR12 M VIR13 M Median [IQR] 470 2497 751 1552 1704 647 605 543 1590 896 1506 948 2040 948 [647‚Äì1590] 51,000 51,000 23,000 3300 400,000 50,000 21,000 15,000 52,915 6077 51,647 2959 8082 23,000 [8082‚Äì51,000] 11.8 2.5 18.0 5.7 5.1 9.0 5.1 8.0 9.9 4.0 0.4 1.1 0.4 5.1 [2.5‚Äì9.0] ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì ‚Äì CD4 count (cells/Œºl) 319 434 548 480 258 185 284 681 499 543 732 506 1400 499 [319‚Äì548] a NA: not available. Duration of HIV infection was calculated from the date of diagnosis of HIV infection. c Time of documented suppression of viremia to b50 copies/ml on ART. d Drug abbreviation: 3TC, lamivudine; ABC, abacavir; ATV, atazanavir; ATV/r, atazanavir boosted with ritonavir; DRV/r, darunavir boosted with ritonavir; EFV, efavirenz; FTC, emtricitabine; LPV/r, lopinavir boosted with ritonavir; NVP, nevirapine; RAL, raltegravir; TDF, tenofovir disoproxil fumarate. b antibody EIA. Characteristics of these participants can be found in Eriksson et al. (2013). This study was approved by the Martin Health System Institutional review board and by the UCSF Committee on Human Research. All study subjects provided written informed consent prior to participation in the study. 2.2. Isolation of CD4 T cells PBMCs were isolated from the leukapheresis product by Ficoll Hypaque density gradient centrifugation and cryopreserved in liquid nitrogen. Upon thawing of the PBMCs, total CD4+ T cells were isolated by negative magnetic selection (Stemcell technologies). The purity of enriched CD4+ T cells was generally greater than 95%, as assessed by Ô¨Çow cytometry. 2.3. TILDA Enriched CD4 + T cells were resuspended at 2 √ó 106 cells/ml in RPMI + 10% foetal bovine serum and rested for 3‚Äì5 h at 37 ¬∞C, 5% CO2. In some experiments, antiretroviral drugs were added to the culture medium (200 nM raltegravir, 100 nM efavirenz and 180 nM AZT). CD4+ T cells were stimulated for 12 h with 100 ng/ml PMA and 1 Œºg/ml ionomycin (both from Sigma). The duration of 12 h was based on kinetic experiments demonstrating maximal production of tat/rev RNA and preserved cell viability at this time point (not shown). After stimulation, cells were washed in RPMI, counted and serially diluted to 18 √ó 106 cells/ml, 9 √ó 106 cells/ml, 3 √ó 106 cells/ml and 1 √ó 106 cells/ml in culture medium. 1 Œºl of the cell suspension from each serial dilution was distributed in 22 to 24 wells of a 96 well plate containing 5 Œºl of 2√ó reaction buffer from the SuperScript III Platinum One-Step qRT-PCR Kit (Life Technologies) corresponding to 18,000, 9000, 3000 and 1000 cells per well. Different cell numbers were used for samples obtained from viremic, untreated HIV-infected individuals (9000; 3000; 1000 and 333 cells per well). In some experiments, a fraction of the isolated CD4+ T cells were immediately distributed in limiting dilutions after thawing to quantify the frequency of cells that spontaneously produce HIV msRNA (no stimulation). Similar frequencies of positive cells were measured by TILDA whether the cells were distributed immediately after thawing or after 12 h of resting (not shown). Pre-ampliÔ¨Åcation was carried out by adding 5 Œºl of a PCR mix containing 0.2 Œºl Superscript III Platinum Taq (Life Technologies), 0.1 Œºl RNase inhibitor (Life Technologies), 0.125 Œºl of each primer (tat1.4 and rev both at 20 ŒºM), 2.2 Œºl Tris‚ÄìEDTA (TE) buffer and 2.25 Œºl H2O to each well (Ô¨Ånal reaction volume = 11 Œºl). The sequences of the oligonucleotides were slightly adapted from Pasternak et al. (2008): tat1.4: 5‚Ä≤-TGG CAG GAA GAA GCG GAG A-3‚Ä≤; rev: 5‚Ä≤-GGA TCT GTC TCT GTC TCT CTC TCC ACC-3‚Ä≤. Pre-ampliÔ¨Åcation was carried out Please cite this article as: Procopio, F.A., et al., A Novel Assay to Measure the Magnitude of the Inducible Viral Reservoir in HIV-infected Individuals, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.019 4 F.A. Procopio et al. / EBioMedicine xxx (2015) xxx‚Äìxxx using the following steps: reverse transcription at 50 ¬∞C for 15 min, denaturation at 95 ¬∞C for 2 min, 24 cycles of ampliÔ¨Åcation (95 ¬∞C 15 s, 60 ¬∞C 4 min) on a C1000 PCR instrument (Biorad). At the end of the preampliÔ¨Åcation, 40 Œºl of TE buffer was added to each well and 1 Œºl of the diluted PCR products was used as template for the tat/rev real-time PCR reaction. This reaction was performed by adding 5 Œºl of the Lightcycler Probe Master buffer (Roche Applied Sciences), 0.2 Œºl of each primer (tat2 and rev, both at 20 ŒºM), 0.2 Œºl of the probe HIV FamZen at 5 ŒºM and 3.4 Œºl H2O to each well (Ô¨Ånal reaction volume = 10 Œºl). Sequence of tat2 and the HIV probe were also adapted from (Pasternak et al., 2008): tat2: 5‚Ä≤- ACA GTC AGA CTC ATC AAG TTT CTC TAT CAA AGC A -3‚Ä≤. Probe HIV: 5‚Ä≤-/56-FAM/TTC CTT CGG /ZEN/GCC TGT CGG GTC CC/3IABkFQ/-3‚Ä≤. All primers and probes were synthesized by IDT. The real-time PCR reaction was carried out in a Light Cycler 480 II (Roche Life Science) using the following programme: Preincubation 95 ¬∞C for 10 min, 45 cycles of 95 ¬∞C 10 s, 60 ¬∞C 30 s, 72 ¬∞C 1 s and a cooling step at 40 ¬∞C fro 30 s. Positive wells at each dilution were counted and the maximum likelihood method was used to calculate the frequency of cells with inducible HIV msRNA (http://bioinf.wehi.edu.au/ software/elda). 2.4. Statistical Analysis Baseline and induced TILDA values were compared using the Wilcoxon matched-pairs signed rank test. For correlations between values obtained with different assays, normality of the log transformed virologic data was tested with the D'Agostino‚ÄìPearson test. With the exception of total HIV DNA in resting CD4+ T cells and residual viremia by single copy assay, all virologic data met the normality test and correlations were performed on log-transformed values using the Pearson test. When the data did not meet the normality test, the Spearman test was used. When the result of an assay was negative (below the limit of detection), an imputed value representing the lower of the limit of detection was used in the calculation of the Pearson correlation coefÔ¨Åcient. In the case of the single-copy assay for HIV-1 RNA in plasma, for which one-third of the measurements was below the limit of detection (0.2 copies/ml), a low imputed value of 0.1 was used to calculate the Spearman rank correlation coefÔ¨Åcient. Virologic data obtained from subjects who started ART during the early and chronic phases of infection were compared using a two-tailed Mann‚ÄìWhitney test. Data analysis was done using Microsoft Excel and Prism 6.0. 3. Results 3.1. Principle of TILDA We sought to develop an assay that detects tat/rev msRNA (adapted from Pasternak et al., 2008) in maximally activated CD4+ T cells, using a limiting dilution format, postulating that this method will allow us to measure the frequency of cells harbouring inducible viruses. We chose to activate the cells with phorbol 12-myristate 13-acetate (PMA), which has been shown to induce a rapid onset of increased transcription from the HIV LTR in latently infected cell lines (Li et al., 1991), induces higher levels of viral production than PHA (Tong-Starkesen et al., 1989) and activates rare T cell subsets (Wang et al., 2013). Ionomycin, a calcium ionophore, is added to trigger calcium release, a requirement for N-FAT signalling. PMA ionomycin has been shown to induce high levels of viral production in several primary cell models of HIV latency (Spina et al., 2013) as well as in CD4 + T cells from virally suppressed individuals (Bullen et al., 2014; Laird et al., 2014). After stimulation, cells are distributed in 24 replicate wells in a 96 well plate and a nested PCR amplifying tat/rev transcripts is directly performed without RNA extraction (Fig. 1A). Using the maximum likelihood method, the frequency of cells producing msRNA is calculated. TILDA can also be used to measure the frequency of CD4+ T cells that spontaneously produce msRNA ex vivo by omitting the stimulation step. 3.2. TILDA is SpeciÔ¨Åc, Sensitive and Reproducible We Ô¨Årst evaluated the sensitivity and speciÔ¨Åcity of TILDA. We infected activated CD4+ T cells with an infectious HIV clone containing a GFP reporter (NL4-3 nef IRES GFP). Two days postinfection, we sorted GFP+ and GFP- cells and used our assay to detect tat/rev msRNA in wells containing single cells (Fig. S1A). Among 48 well containing GFP+ cells, 42 were positive for HIV DNA and GAPDH (88%). 79% (33/42) of these were positive by TILDA. In contrast, none of the 48 wells containing GFP negative cells were positive for msRNA, in spite of a signiÔ¨Åcant fraction of these well containing HIV DNA (42/48, 88%). These results suggested that the presence of msRNA is a speciÔ¨Åc surrogate for the production of viral proteins, at least in this in vitro system of HIV infection. Using samples obtained from 4 virally suppressed subjects, we determined the robustness of TILDA i.e., intra- and inter-assay coefÔ¨Åcients of variation; we measured the frequency of CD4+ T cells producing tat/ rev msRNA upon stimulation with PMA/ionomycin in duplicate plates or in up to 4 independent experiments performed by different operators. The intra and inter-assay coefÔ¨Åcients of variation of TILDA were 0.15 and 0.21, respectively (Fig. S1B) conÔ¨Årming the robustness of this assay. The theoretical dynamic range of TILDA is N 3logs10 with minimal and maximal detectable values of 1.4 and 3185 cells/106 CD4+ T cells, respectively. Of note, all the experimental values determined in this study ranged over 3 logs, a dynamic range of TILDA that is large enough to efÔ¨Åciently measure low frequency of infected cells as is the case in the vast majority of virally suppressed individuals (Fig. S1C). Importantly, the limit of detection of TILDA indicated above (1.4 cells/106 CD4+ T cells) is based on the number of cells plated in a typical experiment (744,000 cells corresponding to 24 √ó 18,000 + 24 √ó 9000 + 24 √ó 3000 + 24 √ó 1000) and can further be lowered by increasing the number of cells assayed in multiple plates. 3.3. TILDA Measures the Frequency of CD4+ T cells With Inducible msRNA in Virally Suppressed Individuals on ART We used TILDA to measure the frequency of CD4+ T cells producing tat/rev msRNA in samples from 20 virally suppressed subjects on suppressive ART for a median time of 6 years (Table 1). As expected, the frequency of cells spontaneously producing msRNA was low (median = 2.7 cells/106 CD4 + T cells, Fig. 1B), and even below the limit of detection in 5 out of 19 samples (26%). CD4+ T cells producing tat/rev RNA were detected in all samples upon stimulation, with a median frequency of cells producing msRNA of 50 cells/106 CD4+ T cells (p b 0.0001). By subtracting the baseline from the induced values, we calculated the frequency of cells carrying inducible HIV. We estimated that only 7.3% of CD4+ T cells with the ability to generate msRNA produced these transcripts spontaneously, whereas the majority (92.7%) of cells were devoid of msRNA ex vivo but produced them upon stimulation with PMA/Ionmycin (Fig. 1C). These results are in line with previous reports demonstrating that the majority of infected cells in individuals on suppressive ART, harbour latent HIV and do not spontaneously produce viral particles (Chun et al., 1997b; Finzi et al., 1999; Chun et al., 2003). These results demonstrated that TILDA can be used to measure the frequency of CD4+ T cells that have the ability to produce HIV msRNA upon stimulation in virally suppressed subjects. 3.4. TILDA Values Correlate With Several Assays Measuring HIV Persistence During ART We next compared the frequency of latently infected CD4+ T cells measured by TILDA with values obtained from a variety of wellestablished assays (Siliciano and Siliciano, 2005; Yu et al., 2008; Palmer et al., 2003; Strain et al., 2013; Yukl et al., 2010). For this purpose, we used samples that were originally used to compare Q-VOA with several PCR-based assays (Table 2) (Eriksson et al., 2013). In these samples obtained from 27 participants of the SCOPE and Options studies, we found Please cite this article as: Procopio, F.A., et al., A Novel Assay to Measure the Magnitude of the Inducible Viral Reservoir in HIV-infected Individuals, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.019 F.A. Procopio et al. / EBioMedicine xxx (2015) xxx‚Äìxxx 5 A Frequency of cells with msHIV RNA baseline Negative selection Ficoll gradient centrifugation 10-20 mL whole blood PBMCs CD4+ T cells Distribute in limiting dilutions Nested RT-PCR for msHIV RNA (24+40 cycles) Maximum likelihood method Frequency of cells with inducible msHIV RNA 12h PMA +ionomycin C Frequency of CD4+ T cells with msRNA (cell/106 CD4+ T cells) p < 0.0001 ART1 ART2 ART3 ART4 ART5 ART6 ART7 ART8 ART9 ART10 1000 100 10 ART11 ART12 ART13 ART14 ART15 ART16 ART17 ART18 ART19 ART20 1 Induced 100 Induced Baseline 80 60 40 20 0 AR T AR 2 T3 AR T AR 4 T5 AR T AR 6 T7 AR AR T9 T AR 10 T1 AR 1 T AR 12 T AR 15 T1 AR 6 T AR 17 T1 8 Baseline Contribution to the pool of cells with inducible msRNA (%) B Fig. 1. TILDA can be used to measure the frequency of persistently infected cells in blood samples from virally suppressed subjects. (A) Principle of TILDA: PBMCs were isolated from 10‚Äì 20 ml of blood and CD4+ T cells were enriched by negative magnetic selection. CD4+ T cells were precisely counted and directly distributed in a 96 well plate or stimulated for 12 h with PMA/ionomycin. Tat/rev msRNA were quantiÔ¨Åed by an ultrasensitive nested RT-PCR and the frequency of cells with inducible msRNA was determined using the maximum likelihood method. (B) Frequency of CD4+ T cells producing msRNA spontaneously (baseline, n = 19) or after 12 h of stimulation (induced, n = 20) in samples obtained from virally suppressed subjects. Horizontal bars indicate median values. For samples in which no positive cells were detected at baseline, the limit of detection (based on cell input) is plotted and represented as a crossed circle. P value was obtained from the Wilcoxon matched-pairs signed rank test. (C) Proportion of CD4+ T cells that produce msRNA spontaneously or after stimulation in virally suppressed subjects on ART. Only samples with detectable values at baseline (panel B) are represented. that the median frequency of latently infected CD4+ T cells measured by TILDA was 24 cells/106 CD4+ T cells (range 1.4‚Äì101 cells/million), which was 48 times higher than frequencies measured by Q-VOA (median = 0.5, range 0.03‚Äì16 cells/106 CD4+ T cells), and 6‚Äì27 times lower than frequencies of infected cells measured by PCR-based assays (Table 2). We then determined correlations between the values obtained by TILDA and the other assays used to measure the size of the HIV reservoir using samples from the same group of subjects. TILDA values were positively correlated with the frequency of PBMCs harbouring total HIV DNA (r = 0.38, p = 0.05, Fig. 2A) and integrated HIV DNA (r = 0.68, p = 0.002, Fig. 2B) as well as with the frequency of rectal CD4+ T cells harbouring total HIV DNA (r = 0.70, p = 0.002, Fig. 2C). Although TILDA values tended to correlate with the frequencies of resting CD4+ T cells harbouring total and integrated HIV DNA, these correlations did not reach statistical signiÔ¨Åcance (œÅ = 0.48, p = 0.083 and r = 0.47, p = 0.076, respectively, Fig. 2D and E). Of note, the number of samples Table 2 Results from the virological assays used in this study. Methoda Sample Unit n samples Median Ratio with TILDAb Max. value Min. value a b c Q-VOA TILDA Tot. HIV DNA in PBMCs Tot. HIV DNA in CD4 Int. HIV DNA in PBMCs Int. HIV DNA in CD4 Tot. HIV DNA in GALT cells Residual viremia plasma Cell culture in limiting dilution Resting CD4+ T cells Stimulation and PCR in limiting dilution Total CD4+ T cells ddPCR ddPCR alu-PCR alu-PCR ddPCR SCA PBMCs Copies/10 cells Resting CD4+ T cells Copies/106 cells GALT CD4+ T cells Copies/106 cells Plasma Copies/10 cells Resting CD4+ T cells Copies/106 cells Copies/ml 27 150 6.3 1858 1 14 307.5 13.4 1200 1 18 234.5 9.0 1248 19 15 767 27.1 2997 108 17 5301 226c 13,388 259 27 0.5 NA 2.2 0.1 Infectious unit/10 cells 27 0.5 0.021 16 0.0315 6 Cells expressing msRNA/106 cells 27 23.8 ‚Äì 100.6 1.4 PBMCs 6 6 ddPCR: droplet digital PCR; SCA: single copy assay. NA: not applicable. Median values were used to calculate the ratio. Samples for which one of the 2 measurements used in each comparison was not performed were excluded. Ratio between GALT DNA and TILDA measured in blood. Please cite this article as: Procopio, F.A., et al., A Novel Assay to Measure the Magnitude of the Inducible Viral Reservoir in HIV-infected Individuals, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.019 F.A. Procopio et al. / EBioMedicine xxx (2015) xxx‚Äìxxx B 1000 100 10 1 r = 0.381 p = 0.050 0.1 1 10 100 1000 100 10 1 Spearman p = 0.083 = 0.482 0.1 1 10 100 1000 TILDA (cells expressing ms HIV RNA per million CD4+ T cells) G r = 0.676 p = 0.002 1 Integrated HIV DNA (copies/106 resting CD4+ T cells) Total HIV DNA (copies/106 resting CD4+ T cells) 1000 100 10 100 100000 10000 1000 r = 0.702 p = 0.002 100 1 1000 F 10000 1000 10 100 1000 TILDA (cells expressing ms HIV RNA per million CD4+ T cells) TILDA (cells expressing ms HIV RNA per million CD4+ T cells) E 10000 1000 10 TILDA (cells expressing ms HIV RNA per million CD4+ T cells) D C 10000 Residual viremia (copies/mL) Total HIV DNA (copies/106 PBMCs) 10000 Integrated HIV DNA (copies/106 PBMCs) A Rectal total HIV DNA (copies/106 GALT CD4+ T cells) 6 10 1 0.10 Spearman p = 0.070 r = 0.472 p = 0.076 = -0.354 0.01 100 1 10 100 1000 TILDA (cells expressing ms HIV RNA per million CD4+ T cells) 1 10 100 1000 TILDA (cells expressing ms HIV RNA per million CD4+ T cells) Viral outgrowth (IUPM) 100 10 1 0.10 r = 0.328 p = 0.095 0.01 1 10 100 1000 TILDA (cells expressing ms HIV RNA per million CD4+ T cells) Fig. 2. TILDA correlates with several measures of HIV persistence. (A) Correlation between TILDA and the droplet digital PCR assay for HIV DNA in PBMCs. (B) Correlation between TILDA and the Alu-PCR assay for HIV DNA in PBMCs. (C) Correlation between TILDA and the droplet digital PCR assay for HIV DNA in rectal CD4+ T cells. (D) Correlation between TILDA and the droplet digital PCR assay for HIV DNA in resting CD4+ T cells. (E) Correlation between TILDA and the Alu-PCR assay for HIV DNA in resting CD4+ T cells. (F) Correlation between TILDA and single copy assay for residual viremia. (G) Correlation between TILDA and Q-VOA in resting CD4+ T cells. Black circles and grey circles denote subjects who started ART during chronic and early infection, respectively. P values were obtained from the Pearson (A, B, C, E, G) or Spearman test (D, F), according to the results of the normality test (see Methods). in which total and integrated HIV DNA were measured in resting CD4+ T cells was limited (n = 14 and 15, respectively, Table 2), which curtailed our capacity to detect signiÔ¨Åcant associations between TILDA and these two assays. TILDA values did not correlate signiÔ¨Åcantly with residual viremia, although a trend for a negative association was noted (œÅ = ‚àí 0.35, p = 0.070, Fig. 2F). The frequency of resting CD4 + T cells harbouring replication competent HIV measured by Q-VOA tended to correlate with TILDA, but this was not statistically signiÔ¨Åcant in this relatively small population of virally suppressed individuals (r = 0.33, p = 0.095, Fig. 2G); this may be attributed to the fact that Q-VOA uses resting CD4+ T cells, whereas TILDA is performed on total CD4+ T cells that include homeostatically proliferating cells. Altogether, these results demonstrate that TILDA can be efÔ¨Åciently used to measure the frequency of CD4+ T cells producing msRNA upon stimulation, a frequency that correlates with several assays that measure HIV persistence. 3.5. Initiation of ART During Early HIV Infection Leads to a Restricted Size of the Reservoir During Therapy Measured by TILDA We measured the frequency of infected cells with inducible msRNA in virally suppressed subjects who initiated ART during recent (n = 10) or chronic (n = 17) infection (Fig. 3). Using TILDA, we observed that the size of the reservoir was signiÔ¨Åcantly smaller in subjects who started ART during recent infection (b 6 months) when compared to those who initiated therapy at a later stage of the disease (median frequencies of 11 and 29 cells/106 CD4+ T cells in early and chronic infection respectively, p = 0.011). This was in line with the results form previous Please cite this article as: Procopio, F.A., et al., A Novel Assay to Measure the Magnitude of the Inducible Viral Reservoir in HIV-infected Individuals, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.019 F.A. Procopio et al. / EBioMedicine xxx (2015) xxx‚Äìxxx Frequency of CD4+ T cells with msRNA (cell/106 CD4+ T cells) 103 p = 0.011 : ART started during early infection : ART started during chronic infection 102 7 the frequency of cells producing msRNA upon stimulation. This indicated that the increase in the frequency of CD4+ T cells producing msRNA is attributable to viral genomes that were already integrated before stimulation, demonstrating that post-integration latency can be detected and measured in viremic HIV-infected subjects who do not receive ART. 101 4. Discussion 100 Fig. 3. Initiation of ART during early HIV infection leads to a restricted size of the reservoir measured by TILDA. The frequency of cells harbouring inducible msRNA was measured by TILDA on CD4+ T cells obtained from subjects who started ART during chronic (n = 17, red circles) or recent (n = 10, blue circles) infection. Horizontal bars indicate median values. P value was obtained from the Mann‚ÄìWhitney test. studies (Strain et al., 2005; Archin et al., 2012; Buzon et al., 2014; Ananworanich et al., 2012, 2015; Eriksson et al., 2013), in which a lower frequency of reservoir cells was measured in subjects who started ART early when using Q-VOA, integrated HIV DNA in PBMCs and total HIV DNA in resting CD4+ T cells. 3.6. TILDA Reveals the Existence of a Large Pool of Latently Infected CD4+ T cells in HIV-infected Individuals who do not Receive ART Our results conÔ¨Årm that the latent HIV reservoir is established early in infection (Strain et al., 2005; Archin et al., 2012; Buzon et al., 2014; Ananworanich et al., 2012; Chomont et al., 2009). However, the relative proportion of productively and latently infected CD4 + T cells during untreated HIV infection is not known. As TILDA can measure the frequency of cells spontaneously producing msRNA as well as the frequency of cells that has the ability to do so upon maximal stimulation (Fig. 1), we used our assay to measure both frequencies in CD4 + T cells obtained from HIV-infected individuals who did not receive ART (Table 1). CD4+ T cells containing tat/rev transcripts at baseline were readily detected in all samples from viremic individuals (Fig. 4A). Overnight stimulation with PMA/ionomycin led to a signiÔ¨Åcant increase in the frequency of CD4 + T cells producing msRNA (median fold increase of 7.4 in the frequency of cells with HIV msRNA, p = 0.0005). The frequency of cells spontaneously producing these transcripts was strongly correlated with plasma viral load (r = 0.81, p = 0.0014, Fig. 4B). Our calculations indicate that only a minor fraction (18.6%) of CD4+ T cells with inducible HIV genomes were spontaneously producing tat/rev transcripts in vivo (Fig. 4C), whereas the majority of infected cells (81.4%) were latently infected and produced msRNA only upon stimulation. Therefore, similar to virally suppressed subjects, ART-na√Øve individuals harbour a large pool of CD4 + T cells in which msRNA can be induced upon stimulation (Fig. 4D). The frequency of latently infected CD4 + T cells in untreated HIV-infected individuals was positively correlated with the duration of HIV infection (r = 0.71, p = 0.0097, Fig. 4E), suggesting that the size of the latent and inducible reservoir gradually increases in untreated viremic subjects, a consequence of continuous seeding of the HIV reservoir. Non-integrated HIV DNA molecules are present at high frequencies in CD4 + T cells isolated from viremic HIV-infected individuals (Chomont et al., 2009; Koelsch et al., 2008). Integration and transcription of these linear genomes may account for the increase in the frequency of CD4+ T cells producing msRNA we observed upon stimulation with PMA/ionomycin. To test for this possibility, we repeated these experiments in the presence of antiretroviral drugs, including the integrase inhibitor raltegravir. We observed that the addition of raltegravir to the culture medium had no impact on the increase in The deÔ¨Ånitive test of an HIV cure will require the interruption of ART. It is necessary, however, to measure the impact of eradication strategies on the size of the viral reservoir even as therapy is continued if clinical trials are to proceed in an efÔ¨Åcient and ethical manner (International ASSWGoHIVC et al., 2012). A large number of assays have been developed to measure the frequency of infected cells that persist in virally suppressed subjects who have received ART for years (Lewin et al., 1999; Li et al., 2010; Siliciano and Siliciano, 2005; Yu et al., 2008; Palmer et al., 2003; Vandergeeten et al., 2014; Pasternak et al., 2008; Eriksson et al., 2013; Brussel and Sonigo, 2003; Rouzioux et al., 2005; Strain et al., 2013; O'Doherty et al., 2002). Each assay measures different parameters (replication competent HIV, total and integrated HIV DNA, cell associated HIV RNA, viral particles) in different cells types (PBMCs, CD4+ T cells and resting CD4+ T cells). These assays provide a frequency of persistently infected cells that can range over 2 logs in the same individual (Eriksson et al., 2013). For example, the size of the reservoir is likely to be overestimated by PCR assays, which can include a large proportion of defective viruses. On the other hand, the reservoir is likely to be underestimated by the Q-VOA assay as it does not capture all cells carrying replication competent HIV (Ho et al., 2013). This prompted us to develop an assay that would not measure the background attributed to defective genomes and that would not rely on the ampliÔ¨Åcation of viral replication, a step that requires at least seven days of cell culture. Unlike cells that produce unspliced and often short HIV gag RNA (Lassen et al., 2004), infected cells that are actively producing msRNA (particularly tat/rev transcripts) can produce virions (Lewin et al., 1999; Pasternak et al., 2008, 2009; Schmid et al., 2010; Fischer et al., 2004; Peng et al., 1995; Hermankova et al., 2003; Vesanen et al., 1997). This well-characterized process suggests that quantiÔ¨Åcation of tat/rev msRNA may prove to be a sensitive and reproducible manner to estimate the frequency of cells harbouring transcriptionally silent ‚Äì nonetheless inducible ‚Äì viruses after maximal stimulation. TILDA does not rely on the ampliÔ¨Åcation of viral replication and offers similar advantages of the DNA PCR assays with regard to simplicity and small blood volume requirement. It does not require RNA extraction and 96-well plates with serially diluted cells can be stored at ‚àí80 ¬∞C for future quantiÔ¨Åcation in batch analyses, making TILDA well-suited for longitudinal clinical studies. Therefore, TILDA has a number of characteristics that make it attractive for clinical trials, particularly in paediatric studies in which the number of cells that can be obtained are often limited. In addition, TILDA could be used to measure the frequency of infection in rare population of cells, which may facilitate the measurement of the HIV reservoir in tissues. Using TILDA, we found the median frequency of latently infected CD4 + T cells during ART to be 24 cells/million, which is 48 times more than the frequency measured by the quantitative viral outgrowth assay. Ho et al. recently reported that the median frequency of intact noninduced proviruses is around 60-fold higher than the frequency of induced proviruses detected in Q-VOA (Ho et al., 2013). Therefore, TILDA may provide an accurate estimate of the size of the latent HIV reservoir, although the precise fraction of the replication competent latent and inducible HIV reservoir captured by TILDA will require further studies. Similarly, evaluating the ability of TILDA to predict viral rebound upon ART interruption will necessitate the development of wellpowered clinical trials and cannot be inferred from our current observations. Please cite this article as: Procopio, F.A., et al., A Novel Assay to Measure the Magnitude of the Inducible Viral Reservoir in HIV-infected Individuals, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.019 F.A. Procopio et al. / EBioMedicine xxx (2015) xxx‚Äìxxx A B VIR1 VIR2 VIR3 VIR4 VIR5 VIR6 VIR7 VIR8 VIR9 VIR10 VIR11 VIR12 VIR13 Frequency of CD4+ T cells with msRNA (cell/106 CD4+ T cells) p =0.0005 1000 100 10 1 Baseline Plasma viremia (HIV RNA copies/mL) 8 1000000 100000 10000 r = 0.81 p = 0.0014 1000 1 Induced 10 100 1000 Baseline TILDA (cells expressing ms HIV RNA per million CD4+ T cells) D p = 0.0073 100 Induced Baseline 80 Productively infected CD4+ T cells 7.3% 92.7% 60 81.4% 18.6% Latently infected CD4+ T cells 40 20 Infection time (years) VI 9 R 1 VI 0 R 1 VI 1 R 1 VI 2 R 13 R 8 ART VIR VI 7 R R VI VI 5 6 R VI 4 R VI R 2 15 VI R VI E VI 1 0 R Contribution to the pool of cells with inducible msRNA (%) C r = 0.71 p = 0.0097 10 5 0 10 100 1000 Induced TILDA (cells expressing ms HIV RNA per million CD4+ T cells) F p = 0.015 p = 0.015 p = NS VIR1 VIR8 VIR9 VIR10 VIR11 VIR12 VIR13 Frequency of CD4+ T cells with msRNA (cell/106 CD4+ T cells) 10000 1000 100 10 1 Baseline Induced - ARVs Induced + ARVs Please cite this article as: Procopio, F.A., et al., A Novel Assay to Measure the Magnitude of the Inducible Viral Reservoir in HIV-infected Individuals, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.019 F.A. Procopio et al. / EBioMedicine xxx (2015) xxx‚Äìxxx TILDA correlated with several PCR-based assays that measure the size of the latent reservoir, including the quantiÔ¨Åcation of total and integrated HIV DNA in PBMCs. TILDA tended to correlate with the size of the reservoir measured by Q-VOA, but this was not statistically signiÔ¨Åcant. Our data are consistent with a lack of strong correlation between these two measures although the lack of statistical signiÔ¨Åcance may reÔ¨Çect, in part, the sample size. The lack of correlation between the two assays may also result from the fact that Q-VOA is performed on puriÔ¨Åed resting CD4+ T cells, whereas total CD4+ T cells are used for TILDA determinations. Total CD4+ T cells include cells undergoing homeostatic proliferation which express some activation markers and are not captured in the Q-VOA assay. Ho et al. reported that Q-VOA does not correlate strongly with the frequency of cells harbouring intact non-induced proviruses (Ho et al., 2013), suggesting that it may not provide an accurate estimate of the frequency of latently infected CD4+ T cells. Whether TILDA could provide a better estimate of the size of the replication competent reservoir remains to be determined by using full-length sequencing approaches, similar to the ones used by Ho et al. and others (Ho et al., 2013; Cohn et al., 2015). Using TILDA, we observed that early initiation of ART leads to a reduced size of the viral reservoir after years of viral suppression. This is in line with several reports that have clearly demonstrated the beneÔ¨Åt of early therapy on the size of the latent HIV reservoir (Strain et al., 2005; Archin et al., 2012; Buzon et al., 2014; Ananworanich et al., 2012, 2015; Chomont et al., 2009). In addition, we demonstrated that before the initiation of ART, the latent HIV reservoir already represents a substantial fraction of all infected CD4 + T cells and that the size of the reservoir increases with time during untreated HIV infection. This observation further argues for early ART intervention to prevent the establishment and continuous seeding of an increasingly large pool of latently infected CD4+ T cells. Although TILDA constitutes an attractive alternative to other methods to measure the size of the latent HIV reservoir, there are potential limitations associated with the use of our assay. First, although all cells that are releasing viral particles are producing tat/rev msRNA, the opposite may not be necessarily true. We cannot exclude the possibility that some cells that generate a positive signal in TILDA do not produce virions. Indeed, very low levels of nuclear msRNA can be detected in resting CD4 + T cells using ultrasensitive approaches (Lassen et al., 2004, 2006). Therefore, TILDA is likely to overestimate the size of the latent and replication competent reservoir. Notwithstanding this limitation, TILDA is more likely to capture the functional reservoir than other existing PCR-based assays: TILDA positive events result from the transcription of integrated HIV genomes that have an intact LTR (to ensure transcription) and that are not deleted in the two most commonly deleted regions identiÔ¨Åed in defective genomes namely tat and rev (Ho et al., 1995). Of note, TILDA measures frequencies of latently infected CD4+ T cells close to those predicted to harbour replication competent virus in the study by Ho et al. (2013). Another potential caveat of TILDA is the fact that it relies on the ampliÔ¨Åcation of a highly variable region of the HIV genome, suggesting that the primers and probes used in our assay may not recognize all viral quasispecies. We anticipate that the use of TILDA on HIV clades other than B will require further validation and optimization. In conclusion, we report the development and validation of TILDA, a novel assay to measure the size of the latent reservoir that is quantitative, sensitive, robust and requires only 10 ml of blood. Moreover, 9 results can be generated in two days. As such, TILDA may represent an alternative to existing assays used to evaluate the efÔ¨Åcacy of therapeutic strategies aimed at reducing the size of the latent HIV reservoir. Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ebiom.2015.06.019. Author Contributions Conceived and designed the experiments: FAP RF RPS NC. Performed the experiments: FAP RF DK JB AGB MCS. Analysed the data: FAP RF JB RPS NC. Contributed reagents/materials/analysis tools: MCS DDR UOD SP FH RJOB MDM DJH SGD. Wrote the paper: FAP RF RPS NC. Financial Disclosure This work was supported by NIH grant 1R21AI113096 (NC), by the Delaney AIDS Research Enterprise (DARE) to Find a Cure 1U19AI096109 and by the Foundation for AIDS Research (amfAR Research Consortium on HIV Eradication (108928-56-RGRL)). R.F. is supported by an amfAR fellowship (108264-51-RFRL). The funders had no role in study design, data collection, data analysis, interpretation and writing of the report. Declaration of Interests RJOB, MDM and DJH are employees and shareholders of Merck & Co., Inc., manufacturer of raltegravir and efavirenz. Acknowledgements The authors thank the study participants for their important contribution to this study. The authors also thank Patrick Yeramian, Moti Ramgopal, Rebeka Bordi, Brenda Jacobs, and Kathyrin Penniman for their recruitment and clinical assistance with study participants, Amanda McNulty and Stephanie Santos for technical assistance and Janet D. Siliciano, Robert F. Siliciano and Jun Lai for Q-VOA measurements. The authors thank the members of the Cleveland Immunopathogenesis Consortium for their advice and helpful discussions. References Agosto, L.M., Liszewski, M.K., Mexas, A., et al., 2011. Patients on HAART often have an excess of unintegrated HIV DNA: implications for monitoring reservoirs. Virology 409 (1), 46‚Äì53. Ananworanich, J., Schuetz, A., Vandergeeten, C., et al., 2012. Impact of multi-targeted antiretroviral treatment on gut T cell depletion and HIV reservoir seeding during acute HIV infection. PLoS One 7 (3), e33948. Ananworanich, J., Dube, K., Chomont, N., 2015. How does the timing of antiretroviral therapy initiation in acute infection affect HIV reservoirs? Curr. Opin. HIV AIDS 10 (1), 18‚Äì28. Archin, N.M., Vaidya, N.K., Kuruc, J.D., et al., 2012. Immediate antiviral therapy appears to restrict resting CD4+ cell HIV-1 infection without accelerating the decay of latent infection. Proc. Natl. Acad. Sci. U. S. A. 109 (24), 9523‚Äì9528. Boulassel, M.R., Chomont, N., Pai, N.P., Gilmore, N., Sekaly, R.P., Routy, J.P., 2012. CD4 T cell nadir independently predicts the magnitude of the HIV reservoir after prolonged suppressive antiretroviral therapy. J. Clin. Virol. 53 (1), 29‚Äì32. Brussel, A., Sonigo, P., 2003. Analysis of early human immunodeÔ¨Åciency virus type 1 DNA synthesis by use of a new sensitive assay for quantifying integrated provirus. J. Virol. 77 (18), 10119‚Äì10124. Bullen, C.K., Laird, G.M., Durand, C.M., Siliciano, J.D., Siliciano, R.F., 2014. New ex vivo approaches distinguish effective and ineffective single agents for reversing HIV-1 latency in vivo. Nat. Med. 20 (4), 425‚Äì429. Fig. 4. TILDA reveals that untreated HIV-infected individuals harbour a large pool of latently infected CD4+ T cells. (A) Frequency of CD4+ T cells producing tat/rev msRNA spontaneously (baseline) or after 12 h of stimulation (induced) in 13 samples obtained from viremic, untreated HIV-infected individuals. Horizontal bars indicate median values. P value was obtained from the Wilcoxon matched-pairs signed rank test. (B) Correlation between the frequency of CD4+ T cells spontaneously producing HIV msRNA and HIV plasma viremia. P value was obtained from the Pearson test. (C) Proportion of CD4+ T cells that produce tat/rev msRNA spontaneously or after stimulation in untreated HIV-infected subjects. (D) Pie charts show the average relative proportions of productively and latently infected CD4+ T cells in virally suppressed (ART) and untreated viremic (VIR) HIV-infected subjects. P value was obtained from the Mann‚ÄìWhitney test. (E) Correlation between the frequency of CD4+ T cells producing HIV msRNA after stimulation and duration of HIV infection. P value was obtained from the Pearson test. (F) Frequency of CD4+ T cells producing HIV msRNA before and after stimulation in the presence (+ARVs) or absence (‚àíARVs) of antiretrovirals (raltegravir, efavirenz and AZT). P values were obtained from the Wilcoxon matched-pairs signed rank test. Please cite this article as: Procopio, F.A., et al., A Novel Assay to Measure the Magnitude of the Inducible Viral Reservoir in HIV-infected Individuals, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.019 10 F.A. Procopio et al. / EBioMedicine xxx (2015) xxx‚Äìxxx Buzon, M.J., Martin-Gayo, E., Pereyra, F., et al., 2014. Long-term antiretroviral treatment initiated in primary HIV-1 infection affects the size, composition and decay kinetics of the reservoir of HIV-1 infected CD4 T cells. J. Virol. 88, 10056‚Äì10065. Chomont, N., El-Far, M., Ancuta, P., et al., 2009. HIV reservoir size and persistence are driven by T cell survival and homeostatic proliferation. Nat. Med. 15 (8), 893‚Äì900. Chun, T.W., Stuyver, L., Mizell, S.B., et al., 1997a. Presence of an inducible HIV-1 latent reservoir during highly active antiretroviral therapy. Proc. Natl. Acad. Sci. U. S. A. 94 (24), 13193‚Äì13197. Chun, T.W., Carruth, L., Finzi, D., et al., 1997b. QuantiÔ¨Åcation of latent tissue reservoirs and total body viral load in HIV-1 infection. Nature 387 (6629), 183‚Äì188. Chun, T.W., Justement, J.S., Pandya, P., et al., 2002. Relationship between the size of the human immunodeÔ¨Åciency virus type 1 (HIV-1) reservoir in peripheral blood CD4+ T cells and CD4+:CD8+ T cell ratios in aviremic HIV-1-infected individuals receiving long-term highly active antiretroviral therapy. J. Infect. Dis. 185 (11), 1672‚Äì1676. Chun, T.W., Justement, J.S., Lempicki, R.A., et al., 2003. Gene expression and viral production in latently infected, resting CD4+ T cells in viremic versus aviremic HIV-infected individuals. Proc. Natl. Acad. Sci. U. S. A. 100 (4), 1908‚Äì1913. Cillo, A.R., Sobolewski, M.D., Bosch, R.J., et al., 2014. QuantiÔ¨Åcation of HIV-1 latency reversal in resting CD4+ T cells from patients on suppressive antiretroviral therapy. Proc. Natl. Acad. Sci. U. S. A. 111, 7078‚Äì7083. Cohn, L.B., Silva, I.T., Oliveira, T.Y., et al., 2015. HIV-1 integration landscape during latent and active infection. Cell 160 (3), 420‚Äì432. Eriksson, S., Graf, E.H., Dahl, V., et al., 2013. Comparative analysis of measures of viral reservoirs in HIV-1 eradication studies. PLoS Pathog. 9 (2), e1003174. Finzi, D., Hermankova, M., Pierson, T., et al., 1997. IdentiÔ¨Åcation of a reservoir for HIV-1 in patients on highly active antiretroviral therapy. Science 278 (5341), 1295‚Äì1300. Finzi, D., Blankson, J., Siliciano, J.D., et al., 1999. Latent infection of CD4+ T cells provides a mechanism for lifelong persistence of HIV-1, even in patients on effective combination therapy. Nat. Med. 5 (5), 512‚Äì517. Fischer, M., Wong, J.K., Russenberger, D., et al., 2002. Residual cell-associated unspliced HIV-1 RNA in peripheral blood of patients on potent antiretroviral therapy represents intracellular transcripts. Antivir. Ther. 7 (2), 91‚Äì103. Fischer, M., Joos, B., Wong, J.K., et al., 2004. Attenuated and nonproductive viral transcription in the lymphatic tissue of HIV-1-infected patients receiving potent antiretroviral therapy. J. Infect. Dis. 189 (2), 273‚Äì285. Furtado, M.R., Callaway, D.S., Phair, J.P., et al., 1999. Persistence of HIV-1 transcription in peripheral-blood mononuclear cells in patients receiving potent antiretroviral therapy. N. Engl. J. Med. 340 (21), 1614‚Äì1622. Garrigue, I., Pellegrin, I., Hoen, B., et al., 2000. Cell-associated HIV-1-DNA quantitation after highly active antiretroviral therapy-treated primary infection in patients with persistently undetectable plasma HIV-1 RNA. AIDS 14 (18), 2851‚Äì2855. Graf, E.H., Mexas, A.M., Yu, J.J., et al., 2011. Elite suppressors harbor low levels of integrated HIV DNA and high levels of 2-LTR circular HIV DNA compared to HIV+ patients on and off HAART. PLoS Pathog. 7 (2), e1001300. Hermankova, M., Siliciano, J.D., Zhou, Y., et al., 2003. Analysis of human immunodeÔ¨Åciency virus type 1 gene expression in latently infected resting CD4+ T lymphocytes in vivo. J. Virol. 77 (13), 7383‚Äì7392. Ho, D.D., Neumann, A.U., Perelson, A.S., Chen, W., Leonard, J.M., Markowitz, M., 1995. Rapid turnover of plasma virions and CD4 lymphocytes in HIV-1 infection. Nature 373 (6510), 123‚Äì126. Ho, Y.-C., Shan, L., Hosmane, N.N., et al., 2013. Replication-competent noninduced proviruses in the latent reservoir increase barrier to HIV-1 cure. Cell 155 (3), 540‚Äì551. International ASSWGoHIVC, Deeks, S.G., Autran, B., et al., 2012. Towards an HIV cure: a global scientiÔ¨Åc strategy. Nat. Rev. Immunol. 12 (8), 607‚Äì614. Josefsson, L., von Stockenstrom, S., Faria, N.R., et al., 2013. The HIV-1 reservoir in eight patients on long-term suppressive antiretroviral therapy is stable with few genetic changes over time. Proc. Natl. Acad. Sci. U. S. A. 110 (51), E4987‚ÄìE4996. Kearney, M., 2015. Analysis of HIV RNA in single cells reveals clonal expansions and defective genomes. CROI 2015. Koelsch, K.K., Liu, L., Haubrich, R., et al., 2008. Dynamics of total, linear nonintegrated, and integrated HIV-1 DNA in vivo and in vitro. J. Infect. Dis. 197 (3), 411‚Äì419. Laird, G.M., Eisele, E.E., Rabi, S.A., et al., 2013. Rapid quantiÔ¨Åcation of the latent reservoir for HIV-1 using a viral outgrowth assay. PLoS Pathog. 9 (5), e1003398. Laird, G.M., Eisele, E.E., Rabi, S.A., Nikolaeva, D., Siliciano, R.F., 2014. A novel cell-based high-throughput screen for inhibitors of HIV-1 gene expression and budding identiÔ¨Åes the cardiac glycosides. J. Antimicrob. Chemother. 69 (4), 988‚Äì994. Lassen, K.G., Bailey, J.R., Siliciano, R.F., 2004. Analysis of human immunodeÔ¨Åciency virus type 1 transcriptional elongation in resting CD4+ T cells in vivo. J. Virol. 78 (17), 9105‚Äì9114. Lassen, K.G., Ramyar, K.X., Bailey, J.R., Zhou, Y., Siliciano, R.F., 2006. Nuclear retention of multiply spliced HIV-1 RNA in resting CD4+ T cells. PLoS Pathog. 2 (7), e68. Lewin, S.R., Vesanen, M., Kostrikis, L., et al., 1999. Use of real-time PCR and molecular beacons to detect virus replication in human immunodeÔ¨Åciency virus type 1-infected individuals on prolonged effective antiretroviral therapy. J. Virol. 73 (7), 6099‚Äì6103. Li, Y.C., Ross, J., Scheppler, J.A., Franza Jr., B.R., 1991. An in vitro transcription analysis of early responses of the human immunodeÔ¨Åciency virus type 1 long terminal repeat to different transcriptional activators. Mol. Cell. Biol. 11 (4), 1883‚Äì1893. Li, P., Ruel, T., Fujimoto, K., et al., 2010. Novel application of locked nucleic acid chemistry for a Taqman assay for measuring diverse human immunodeÔ¨Åciency virus type 1 subtypes. J. Virol. Methods 170 (1‚Äì2), 115‚Äì120. Lin, X., Irwin, D., Kanazawa, S., et al., 2003. Transcriptional proÔ¨Åles of latent human immunodeÔ¨Åciency virus in infected individuals: effects of Tat on the host and reservoir. J. Virol. 77 (15), 8227‚Äì8236. Mexas, A.M., Graf, E.H., Pace, M.J., et al., 2012. Concurrent measures of total and integrated HIV DNA monitor reservoirs and ongoing replication in eradication trials. AIDS 26 (18), 2295‚Äì2306. Murray, J.M., Zaunders, J.J., McBride, K.L., et al., 2014. HIV DNA subspecies persist in both activated and resting memory CD4+ T cells during antiretroviral therapy. J. Virol. 88 (6), 3516‚Äì3526. O'Doherty, U., Swiggard, W.J., Jeyakumar, D., McGain, D., Malim, M.H., 2002. A sensitive, quantitative assay for human immunodeÔ¨Åciency virus type 1 integration. J. Virol. 76 (21), 10942‚Äì10950. Palella Jr., F.J., Delaney, K.M., Moorman, A.C., et al., 1998. Declining morbidity and mortality among patients with advanced human immunodeÔ¨Åciency virus infection. HIV outpatient study investigators. N. Engl. J. Med. 338 (13), 853‚Äì860. Palmer, S., Wiegand, A.P., Maldarelli, F., et al., 2003. New real-time reverse transcriptaseinitiated PCR assay with single-copy sensitivity for human immunodeÔ¨Åciency virus type 1 RNA in plasma. J. Clin. Microbiol. 41 (10), 4531‚Äì4536. Pasternak, A.O., Adema, K.W., Bakker, M., et al., 2008. Highly sensitive methods based on seminested real-time reverse transcription-PCR for quantitation of human immunodeÔ¨Åciency virus type 1 unspliced and multiply spliced RNA and proviral DNA. J. Clin. Microbiol. 46 (7), 2206‚Äì2211. Pasternak, A.O., Jurriaans, S., Bakker, M., Prins, J.M., Berkhout, B., Lukashov, V.V., 2009. Cellular levels of HIV unspliced RNA from patients on combination antiretroviral therapy with undetectable plasma viremia predict the therapy outcome. PLoS One 4 (12), e8490. Peng, H., Reinhart, T.A., Retzel, E.F., Staskus, K.A., Zupancic, M., Haase, A.T., 1995. Single cell transcript analysis of human immunodeÔ¨Åciency virus gene expression in the transition from latent to productive infection. Virology 206 (1), 16‚Äì27. Razooky, B.S., Pai, A., Aull, K., Rouzine, I.M., Weinberger, L.S., 2015. A hardwired HIV latency program. Cell 160 (5), 990‚Äì1001. Rouzioux, C., Hubert, J.B., Burgard, M., et al., 2005. Early levels of HIV-1 DNA in peripheral blood mononuclear cells are predictive of disease progression independently of HIV-1 RNA levels and CD4+ T cell counts. J. Infect. Dis. 192 (1), 46‚Äì55. Schmid, A., Gianella, S., von Wyl, V., et al., 2010. Profound depletion of HIV-1 transcription in patients initiating antiretroviral therapy during acute infection. PLoS One 5 (10), e13310. Siliciano, J.D., Siliciano, R.F., 2005. Enhanced culture assay for detection and quantitation of latently infected, resting CD4 + T-cells carrying replication-competent virus in HIV-1-infected individuals. Methods Mol. Biol. 304, 3‚Äì15. Siliciano, J.D., Kajdas, J., Finzi, D., et al., 2003. Long-term follow-up studies conÔ¨Årm the stability of the latent reservoir for HIV-1 in resting CD4+ T cells. Nat. Med. 9 (6), 727‚Äì728. Sonza, S., Mutimer, H.P., O'Brien, K., et al., 2002. Selectively reduced tat mRNA heralds the decline in productive human immunodeÔ¨Åciency virus type 1 infection in monocytederived macrophages. J. Virol. 76 (24), 12611‚Äì12621. Spina, C.A., Anderson, J., Archin, N.M., et al., 2013. An in-depth comparison of latent HIV-1 reactivation in multiple cell model systems and resting CD4+ T cells from aviremic patients. PLoS Pathog. 9 (12), e1003834. Strain, M.C., Little, S.J., Daar, E.S., et al., 2005. Effect of treatment, during primary infection, on establishment and clearance of cellular reservoirs of HIV-1. J. Infect. Dis. 191 (9), 1410‚Äì1418. Strain, M.C., Lada, S.M., Luong, T., et al., 2013. Highly precise measurement of HIV DNA by droplet digital PCR. PLoS One 8 (4), e55943. Tong-Starkesen, S.E., Luciw, P.A., Peterlin, B.M., 1989. Signaling through T lymphocyte surface proteins, TCR/CD3 and CD28, activates the HIV-1 long terminal repeat. J. Immunol. 142 (2), 702‚Äì707. Vandergeeten, C., Fromentin, R., DaFonseca, S., et al., 2013. Interleukin-7 promotes HIV persistence during antiretroviral therapy. Blood 121 (21), 4321‚Äì4329. Vandergeeten, C., Fromentin, R., Merlini, E., et al., 2014. Cross-clade ultrasensitive PCRbased assays to measure HIV persistence in large-cohort studies. J. Virol. 88 (21), 12385‚Äì12396. Vesanen, M., Markowitz, M., Cao, Y., Ho, D.D., Saksela, K., 1997. Human immunodeÔ¨Åciency virus type-1 mRNA splicing pattern in infected persons is determined by the proportion of newly infected cells. Virology 236 (1), 104‚Äì109. Wang, H., Daniel, V., Sadeghi, M., Opelz, G., 2013. Differences in the induction of induced human CD4(+) CD25(+) FoxP3(+) T-regulatory cells and CD3(+) CD8(+) CD28(‚àí) T-suppressor cells subset phenotypes in vitro: comparison of Phorbol 12myristate 13-acetate/ionomycin and phytohemagglutinin stimulation. Transplant. Proc. 45 (5), 1822‚Äì1831. Weinberger, A.D., Weinberger, L.S., 2013. Stochastic fate selection in HIV-infected patients. Cell 155 (3), 497‚Äì499. Williams, J.P., Hurst, J., Stohr, W., et al., 2014. HIV-1 DNA predicts disease progression and post-treatment virological control. eLife 3, e03821. Wong, J.K., Hezareh, M., Gunthard, H.F., et al., 1997. Recovery of replication-competent HIV despite prolonged suppression of plasma viremia. Science 278 (5341), 1291‚Äì1295. Yerly, S., Gunthard, H.F., Fagard, C., et al., 2004. Proviral HIV-DNA predicts viral rebound and viral setpoint after structured treatment interruptions. AIDS 18 (14), 1951‚Äì1953. Yu, J.J., Wu, T.L., Liszewski, M.K., et al., 2008. A more precise HIV integration assay designed to detect small differences Ô¨Ånds lower levels of integrated DNA in HAART treated patients. Virology 379 (1), 78‚Äì86. Yukl, S.A., Gianella, S., Sinclair, E., et al., 2010. Differences in HIV burden and immune activation within the gut of HIV-positive patients receiving suppressive antiretroviral therapy. J. Infect. Dis. 202 (10), 1553‚Äì1561. Please cite this article as: Procopio, F.A., et al., A Novel Assay to Measure the Magnitude of the Inducible Viral Reservoir in HIV-infected Individuals, EBioMedicine (2015), http://dx.doi.org/10.1016/j.ebiom.2015.06.019 A novel community-based delivery model to combat cancer disparities * Laura Tesler Waldmana, , , * Ludmila Svobodaa, * Brian F. Youngb, * Gregory A. Abela, * Suzanne Berlina, * Aymen A. Elfikya, * Rachel A. Freedmana, * Mark Drewsc, * Laura Hollandc, * Christopher S. Lathana Show more doi:10.1016/j.hjdsi.2013.09.004 Get rights and content Abstract Implementation lessons • It is hypothesized that this delivery model can decrease wait times for diagnosis and treatment of cancer, increase awareness and knowledge of cancer prevention and treatment, and foster trust with providers and patients from vulnerable communities. • Involving oncologists in clinical diagnosis at community health centers can link specialty care more closely to vulnerable communities. • Funding for this type of clinical innovation is currently limited to institutional and philanthropic sources. A shift in the academic and public sector funding paradigms may be required to enable implementation on a broader level. Keywords * Disparities; * Cancer care delivery; * Quality of care; * Health care access; * Intervention; * Community partnerships 1. Introduction from the editors In this special edition of Into Practice, Healthcare is featuring two case studies that highlight innovative approaches for addressing racial disparities in cancer care. The first case, titled “A Novel Community-Based Delivery Model to Combat Cancer Disparities,” describes a model for combining cancer prevention, diagnosis, education, and treatment services within a community health center. The second case study, titled “Using FastTrack to Implement an Academic Medical Center and Community Health Center Collaborative for Cancer Care Delivery” provides a more granular look at the logistics of integrating medical, administrative, financial, and social support services between the three key groups enabling this initiative: a community health center, a cancer center, and a tertiary academic medical center. Taken together, these two case studies provide an important roadmap for the delivery strategies necessary to tackle racial disparities in specialty care. 2. Background Race and class have a devastating impact on cancer incidence, treatment, and mortality.1, 2 and 3 Disparities in incidence and mortality may be due to a variety of factors including socioeconomic, biologic, and neighborhood factors.4 and 5 With regard to the question of poorer access to care, or lower-quality care, the 2002 Institute of Medicine report, Unequal Treatment, found evidence that racial and ethnic minorities tend to receive lower-quality health care than whites even when controlling for factors such as insurance status and income. 6 Research has documented widespread disparities in cancer care across the spectrum of prevention, diagnosis and treatment. Due to a multitude of factors, patients most in need have the greatest difficulty in further benefitting from high-level tertiary center cancer care.2, 7 and 8 The systemic ripple effect not only affects excluded patients, but results in a disparate enrichment of the clinical trials and translational research that does not include them.9 Given the above realities and challenges, sustained positive results will remain difficult to achieve without a dedicated multi-level systemic approach.6, 10, 11 and 12 Research by Onega et al. has suggested that the effects of racial disparities for African Americans, particularly mortality risk for lung, breast, colorectal, and prostate cancers, were attenuated for patients at National Cancer Institute (NCI) cancer centers.11Committed to research to improve outcomes across the spectrum from prevention to treatment, institutions with an NCI-cancer center designation share in common a robust transdisciplinary research program, often including inter-institutional collaborations; they provide clinical and scientific training in oncology-related fields; and they offer education and outreach programs focusing on cancer prevention and screening. The vast majority of these institutions are also treatment centers, providing cutting edge care to patients and the opportunity to participate in clinical trials.13 In January 2012, Dana-Farber Cancer Institute (DFCI), an NCI-designated Comprehensive Cancer Center, initiated a clinical outreach program as part of a comprehensive cancer Cancer Care Equity Program. The goal of this initiative is to improve local outcomes for the underserved across the spectrum of cancer-related disease by facilitating clinical access to preventive medicine, treatment, and clinical trials at the cancer center (Fig. 1). Through these efforts, the initiative aims to broaden access for vulnerable patient populations and join community partners in the quest for equitable care across the spectrum of cancer related disease. Fig. 1. Mission and goals of the clinical outreach program. Source: Developed by authors. Figure options 3. Problem Existing strategies to address cancer-related health disparities have primarily focused on screenings, education about screenings, prevention,14 and 15 tobacco cessation and, to a lesser extent, follow up for abnormal screening results and referrals to specialty providers.14, 15, 16, 17, 18, 19, 20, 21 and 22 Successful delivery models have included the establishment of screening clinics, screening visits, screenings administered by trained primary care providers, and referral tracking systems.17, 23, 24 and 25 Several interventions have utilized patient navigation delivery models.26, 27, 28 and 29 While these models have improved the rates of, and time to, diagnostic resolution for underserved populations, evidence of their impact on treatment rates and outcomes is lacking, and they are likely to be most effective when combined with other strategies.23 and 29 The impetus for this intervention came from clinical observations made during routine clinical practice. Of greatest concern was the paucity of patients of racially underserved populations being seen at DFCI, as evidenced in lower clinical trial enrollments and the lack of integrated cancer care spanning the continuum from prevention to survivorship. These concerns were bolstered by a needs assessment commissioned by the DFCI Community Benefits office in 2009, which found that racially underserved populations in the area had access to cancer screening, yet continued to suffer from higher mortality rates. Subsequent Internet searches of other NCI cancer center websites to learn about their health disparities initiatives revealed that there were few clinical cancer outreach programs in U.S. urban communities. The programs that did exist were generally restricted to a limited number of cancer types, and primarily focused on screenings, education, tobacco cessation, patient navigation, clinical trial accrual efforts, or survivorship programs. Only three clinical outreach programs offered diagnostic services30, 31 and 32; one of these also provided treatment, but only for three select cancer diagnoses.30 Given this need, delivery models incorporating collaborative care between oncology specialists and primary care providers offer the potential to improve patient care in several areas, including continuity of care, managing co-morbidities, and supportive care.33 A wide range of primary-specialist collaborative care models exist: affiliations between community health centers (CHCs) and specialty practices or hospitals; telephone and videoconference provider–provider consultations; visiting specialist services/outreach clinics in primary health care settings; co-locations of independent primary and specialty care practices; specialists from other institutions stationed at CHCs; specialists employed within a CHC; CHCs that provide primary and specialized care for specific patient populations (e.g., behavioral health, substance abuse); and specialty practices that employ primary care providers.34, 35, 36, 37, 38 and 39 Although the literature on patient-centered medical homes, co-location, integrated care and coordinated care describes successful collaborations in other disease areas,39, 40, 41,42 and 43 and calls have been made for establishing such collaborations for oncology,33 and 44 no program providing a comprehensive approach to cancer care integrated within a primary health care center was identified. A key challenge in integrating cancer care into the CHC setting is the role of referral pathways. Typically, cancer-related care may be obtained through multiple pathways, depending on the presence and severity of symptoms, whether a patient has an established primary care provider, and subsequent evaluation findings (Fig. 2). In each pathway, delays in diagnosis and treatment may occur due to missed appointments, time lags in appointments with specialists, lack of follow up by primary or specialist care providers, or lack of patient follow up. A limited body of research has documented racial and socioeconomic disparities in referrals to oncologists or oncology subspecialists.45,46 and 47 Delays and under-referral to these specialists are problematic given evidence that the care they provide yields improved outcomes.45, 48, 49 and 50 Fig. 2. Cancer diagnosis pathway. Source: Developed by authors based on clinical observations. Figure options Fig. 3. Conceptual model for clinical outreach program. Source: Developed by authors. Figure options 4. Solution 4.1. Intervention development and implementation To address these issues, the decision was made to establish a clinical outreach facility within a community primary health care center. As described in the broader literature, there are several advantages to specialist clinical outreach programs in primary health care settings. The combination of the direct access to the underserved communities, the continuity of care provided in these centers, and the position of the CHC as a community center allows for an interaction that is superior to that of outpatient hospitals or physician offices.51 Furthermore, co-location fosters effective communication and collaborations between primary care providers and specialists, especially in the case of interventions that include multiple opportunities for interactions such as educational sessions, seminars and joint consultations.34 and 36 Indeed, there are well-documented problems relating to incomplete transfer of medical information from primary care to specialists and back again, especially with respect to referral for oncology and hematology.52 By directly involving and encouraging contact between oncologists and the clinicians that serve the community it was anticipated that the intervention would establish trust and comfort that would encourage referrals and collaborations. Patients, in turn, could experience the convenience of having oncology appointments in the same building where they receive their primary care and avoid travel to another neighborhood and the unfamiliar terrain of the cancer center. Finally, having visible DFCI oncology providers in vulnerable communities necessarily conveys a commitment to treatment equity. For the pilot project, DFCI decided to utilize its relationships with its tertiary care academic medical center partner Brigham and Women's Hospital (BWH), and Whittier Street Health Center (WSHC), a CHC located in an underserved community, to establish a cancer outreach facility that combined prevention, diagnosis, education and treatment services. A longstanding relationship between DFCI and the WSHC, as well as the integrated cancer care services already offered through an existing collaboration between DFCI and BWH, helped to lay a foundation for this partnership. During the implementation phase, the development and integration of services among the three participating institutions was achieved by incorporating clinical and administrative stakeholders from each institution into working group sessions through a structured facilitation process.53 4.2. Program design The intervention established a community cancer care facility located within WSHC and staffed by DFCI oncologists, a program nurse and other personnel. This arrangement enables the oncologists and primary care providers to work closely together to coordinate patient care through expedited referrals, ongoing provider–provider communications, and a shared medical records database. At the same time, patients experience the convenience and comfort of receiving cancer evaluation services on-site at their primary care health care center. The conceptual model details the patient referral pathways and services across institutions (Fig. 3), and the roles of the outreach program personnel are described in a companion diagram (Fig. 4). Fig. 4. Integrated evaluationservices: staffroles. Source: Developed by authors Figure options Fig. 5. Staffing requirements for the clinical outreach program. Source: Developed by authors. Figure options 4.2.1. Clinical evaluation Primary care providers refer patients to the on-site cancer care facility for diagnostic evaluation of suspected cancers, abnormal screening, and selected long term follow-up. They may also refer patients with previous cancer histories to re-establish connections with oncology. Integrated evaluation services are provided on-site at the community cancer care facility in a designated clinical space. The facility includes a dedicated staff of DFCI clinical oncologists and a DFCI program nurse/nurse navigator. The areas of expertise among the oncologists include breast, gynecologic, thoracic, genitourinary, and hematologic cancers, but each functions as a generalist clinical oncologist at the outreach facility and sees patients of all cancer types on their rotation days. This arrangement requires that each medical oncologist discuss cases that are outside their area of subspecialty and that cases are “handed off” in a group practice manner. The process of patient evaluation and referral focuses on expediting diagnostic evaluations and treatment initiation. Prior to the initial visit with a clinical oncologist, the program nurse performs the initial patient assessment and triage. After conducting an initial examination and consultation with the patient, the clinical oncologist decides on the appropriate follow-up examinations and referral pathway in consultation with the other clinical oncology physicians. The program nurse then oversees scheduling for appointments, screenings and referrals, and coordinates communications among the primary care provider, clinical oncologist, specialists, and patient. Once appointments are scheduled, the program nurse acts as a patient navigator to follow up with patients both at the outreach facility and referral locations, providing appointment reminders, assisting with transportation, and arranging for additional patient services with access management, case managers, social workers, and interpreters across participating institutions. To facilitate this process, she maintains a patient navigation tracking database specifically developed for the intervention which includes the initial appointment information, details on each phone call made to or by the patient, additional patient services requested, tests ordered, follow up appointments, referrals, and the patient's attendance for each appointment across the three participating institutions. In addition, the program nurse connects patients to prevention and survivorship programs, aids in the formation of oncology clinical support and survivor programs, and collaborates with primary health center staff to augment existing survivor programs with additional resources. Finally, she coordinates educational programs for providers and conducts educational events in the community. 4.2.2. Expedited referrals and access Patients with an active cancer-related issue are referred to DFCI and/or BWH. To avoid losses in follow-up and delays in receiving care, an expedited referral system has been implemented across the three participating institutions, including access management processes for patients whose insurance is not accepted at the tertiary care centers. During the implementation phase of the intervention, senior level representatives from each institution's financial services departments worked together to develop procedures to ensure coverage for patients whose insurance was not accepted by all three institutions. As a result, insurance approvals for all institutions are secured prior to the patient's initial appointment at the outreach facility. The program nurse also works with patients to identify and resolve additional issues that could result in missed appointments. 4.2.3. Multidisciplinary communication The intervention incorporates several features to facilitate communications on multiple levels. Ongoing formal and informal provider–provider consultations occur by phone, email and in-person. Being located in the same building has permitted ad-hoc joint consultations among the primary care provider, oncologist and patient. Each morning that the clinical outreach program is seeing patients, the program nurse briefly visits each of the primary care provider office areas, and throughout the day primary care providers frequently call or stop by the clinical outreach facility for informal curbside consultations. In addition, centralized medical records are maintained within the primary health care center's database. This enables the primary care providers to track their patients, and the oncologists to immediately access the patient's full medical history. Communication is further facilitated through bi-weekly question and answer (Q&A) sessions as described below. Perhaps most important, patients begin their oncologic evaluation in the comfort and familiarity of the same building where they receive primary care, which may reduce the extreme anxiety that often accompanies a first oncology visit and the ensuing detriment to physician–patient communication. 4.2.4. Community education Community-level education is provided by the intervention through seminars held at WSHC and other community venues. Provider education is incorporated through didactic sessions led by the oncologists, program nurse, or invited guests, and bi-weekly informal question and answer (Q&A) sessions open to all primary health center staff members. The bi-weekly Q&A sessions typically focus on a specific cancer-related topic, but also provide an opportunity for primary care and specialist providers to present case studies, oncologists to clarify screening and referral guidelines, and primary care providers to educate the oncologists about the patient population. Periodically, a portion of this time is allocated to discuss broader issues related to the functioning of the clinical outreach program. These sessions have also facilitated ad-hoc provider-provider consultations among attendees immediately preceding and following the meeting. 4.3. Challenges encountered Several issues had to be addressed in order for the conceptual model to progress to the implementation stage. 4.3.1. Funding Initial conversations around the development and implementation of the clinical outreach program stalled due to concerns about the feasibility of securing funding. Federal grant funding for this type of clinical innovation was not available, leaving only institutional or philanthropic funding as options. After the business plan was prepared, a philanthropic donor expressed interest to the DFCI Development Office in funding a community oncology program, and the intervention was determined to be a good fit. Once this philanthropic funding was secured the dialog was able to move forward. 4.3.2. Institutional support Support for the program had to be obtained from leadership across all of the major departments in DFCI, including executive management, the clinical directors from each disease center, nursing leadership, and finance leadership. It was also necessary to work closely with External Affairs in order to avoid duplicating already established efforts underway through existing community benefits programs. Since the Community Benefits office had previously established survivorship, education and screening programs at WSHC, it was agreed that the clinical outreach program and Community Benefits team would collaborate to expand and enhance these initiatives rather than develop new ones. Finally, support had to be obtained to provide administrative support and allocate a significant percentage of the faculty director's time to overseeing the program. 4.3.3. Establishing a partnership with the CHC A longstanding relationship between DFCI and WSHC for other outreach programs implemented over the past decade helped to build a foundation for the collaboration entailed in this intervention. However, it was still necessary to demonstrate the value of the clinical outreach program to WSHC. The WSHC providers were initially skeptical about the utility of the program and expressed uncertainty about the cancer center's level of commitment. To address these concerns, multiple meetings were held over a 1? year period in order to establish trust, form a working partnership, demonstrate DFCI's commitment to the WSHC patients, and incorporate input from WSHC staff into the business plan. The business plan stipulated that WSHC would provide support and training for the providers, and the oncologists and program nurse would perform their clinical duties under WSHC's license. The clinical space was leased by DFCI as part of a previous agreement. To help offset the operating expenses for the program, WSHC would pay a flat fee for each day that clinical services are provided, but could recover these costs through billing for patient appointments. The remaining expenses incurred by DFCI, including staffing, would be covered through the philanthropic funding provided for the program. 4.3.4. Staffing To staff the community cancer care facility (Fig. 5), medical oncologists representing a range of clinical expertise with the interest and willingness to commit the time and effort to see patients on-site at WSHC had to be identified, and a percentage of their time from DFCI had to be allocated to the intervention. In addition, since the conceptual model assigns several critical functions to the program nurse/nurse navigator, it was necessary to find a highly experienced nurse with the appropriate clinical, language and cultural skills. 5. Lessons for the field Discussions around the clinical outreach intervention with institutional leadership and providers revealed that there was indeed interest and demand for this type of service, both from the cancer center and the CHC. Of equal importance, this conceptual model shifts the paradigm of oncology to assist with diagnosis rather than treatment only. Doing so brings the oncology practitioners closer to vulnerable communities and demonstrates the commitment of the cancer center to local communities with limited access to high quality cancer treatment, enabling us to assist the community in a more demonstrative manner. The intervention's impact is currently being evaluated according to several metrics, including time from initial appointment at the clinical outreach facility to resolution (treatment plan established, surveillance plan established, or return to primary care provider for negative diagnoses); utilization of DFCI by under-represented patients documenting an increase in absolute numbers over time; and no-show rates and the points at which they occur in the referral pathway. It is hypothesized that this model of care can decrease wait times for diagnosis and treatment of cancer, increase awareness and knowledge of cancer prevention and treatment, and foster trust with providers and patients from vulnerable communities. An upward trend in total clinical trial accrual numbers for unrepresented minority patients at DFCI is also expected through referrals of patients enrolled in the pilot program and will be monitored. Once performance measures are met, implementation of the intervention at other primary health care centers will be considered. Funding for this type of clinical innovation, however, is currently limited to institutional and philanthropic sources; therefore, a shift in the academic and public sector funding paradigms may also be required in order for the conceptual model to be implemented on a broader level, and to be sustainable in the future. The impact of these sorts of collaborations on health care costs and access in the future is unclear. The Affordable Care Act offers some incentive to work with community health centers around cancer, but how this will manifest remains uncertain.54 The majority of patients are able to access the program, with very few insurance obstacles. Much of this has to do with the unique aspects of the Massachusetts health care system, as well as engaging on this issue early with leadership from all institutions. The DFCI Cancer Care Equity Program initiative forms the basis of an intervention-based program aimed at uniting the research and outreach efforts in disparities with clinical impact. The formation of this model demonstrates collaboration between tertiary care academic centers and a primary health care center, focused on eliminating cancer disparities and emphasizing a cohesive clinical service in a novel manner. A Novel Dermato-Pulmonary Syndrome Associated With MDA-5 Antibodies Report of 2 Cases and Review of the Literature Neal F. Chaisson, MD, Julie Paik, MD, Ana-Maria Orbai, MD, Livia Casciola-Rosen, PhD, David Fiorentino, MD, PhD, Sonye Danoff, MD, PhD, and Antony Rosen, MD Abstract: Melanoma differentiation-associated protein 5 (MDA-5) is a novel autoantibody frequently characterized by interstitial lung disease and a distinct cutaneous phenotype with palmar papules, ulceration, and rash. Virtually all patients have underlying dermatomyositis, but many lack the characteristic clinical myopathy associated with it. In the setting of amyopathic disease, the absence of clinically available biomarkers or clear pathologic diagnosis can complicate effective prognostic and therapeutic intervention. Until recently the presence of MDA-5 antibody associated dermato-pulmonary syndrome was described only in Asian populations. We present 2 cases of MDA-5-associated dermato-pulmonary syndrome and provide a comprehensive review of available literature. (Medicine 2012;91: 220Y228) Abbreviations: ANA = antinuclear antibody, ANCA = antineutrophil cytoplasmic antibody, CADM = clinically amyopathic dermatomyositis, CK = creatine kinase, CT = computed tomography, DLCO = diffusing capacity of lung for carbon monoxide, DM = dermatomyositis, ESR = erythrocyte sedimentation rate, FEV1 = forced expiratory volume in the first second, FVC = forced vital capacity, IFN = interferon, ILD = interstitial lung disease, IV = intravenous, IVIG = intravenous immunoglobulin, MDA-5 = melanoma differentiation-associated protein 5, RF = rheumatoid factor, RIG-I = retinoic acid-inducible gene 1, RLR = RIG-I-like receptor, TSH = thyroid-stimulating hormone, VATS = video-assisted thorascopic surgery. INTRODUCTION D ermatomyositis (DM) is an autoimmune inflammatory myopathy that involves the muscle, skin, and lung to various extents in different patients. At one end of the spectrum, myopathy is prominent with weakness of skeletal muscle, and even muscles of swallowing and respiration. At the other end of the spectrum, there is prominent skin disease, but minimal or absent muscle disease (clinically amyopathic DM ECADM^).4 Patients From the Division of Pulmonary and Critical Care Medicine (NFC, SD) and Division of Rheumatology (JP, A-MO, LC-R, AR), Johns Hopkins University School of Medicine, Baltimore, Maryland; and Department of Dermatology (DF), Stanford University School of Medicine, Palo Alto, California. Financial support and conflicts of interest: This work was supported by National Institutes of Health grants AR44684 (to LC-R), P30 AR053503 (to AR), T32 AR048522 (to A-MO and JP), and T32 HL07534 (to NFC). The authors have no conflicts of interest to disclose. Reprints: Antony Rosen, MD, Division of Rheumatology, Johns Hopkins University School of Medicine, Johns Hopkins Bayview Medical Center, Mason F. Lord Building Center Tower, 5200 Eastern Avenue, Baltimore, MD 21224 (e-mail: arosen@jhmi.edu). Copyright * 2012 by Lippincott Williams & Wilkins ISSN: 0025-7974 DOI: 10.1097/MD.0b013e3182606f0b 220 www.md-journal.com with either overt myopathy or clinically amyopathic disease can manifest interstitial lung disease (ILD) as part of their clinical phenotype. The ILD is also variable in its course and severity.6 Only 50%Y70% of patients with DM and CADM have identifiable myositis specific antibodies,12 leaving a significant proportion of patients with DM who are apparently seronegative. In the setting of clinically amyopathic disease and/or nondiagnostic clinical and pathologic findings in the skin, a definitive diagnosis can be difficult to obtain, which can affect classification and institution of therapy. In 2005, Sato et al27 identified a novel autoantibody recognizing a 140 kDa protein in patients with DM and CADM (initially termed ‚Äò‚ÄòCADM-140‚Äô‚Äô). The 140 kDa autoantigen was subsequently identified as melanoma differentiationassociated protein 5 (MDA-5).28 In the initial studies in Japanese cohorts, most patients had clinically amyopathic disease, and some patients had rapidly progressive ILD.27,28 The cutaneous findings were not well described in these cohorts. Although prior reports had linked both unusual cutaneous findings of ulceration19 and palmar papules16 with ILD, it was Fiorentino et al7 who suggested that the link between all of these clinical findings was autoimmunity to MDA-5. Thus these skin findings, which include skin ulceration and palmar papules, along with ILD and MDA-5 antibodies, comprise what we now term a ‚Äò‚Äòdermato-pulmonary syndrome.‚Äô‚Äô We present 2 severely ill patients with progressive ILD and cutaneous findings not diagnostic for classic DM, absent myositis, negative autoantibodies as determined using clinically available assays, and without diagnostic findings on skin and lung biopsy. Both patients had MDA-5 autoantibodies. We detail the characteristics of these 2 cases, and review the literature describing the typical presentation of the MDA-5 autoantibody-associated dermato-pulmonary syndrome. CASE REPORTS Case 1 A 54-year-old white man developed an erythematous rash surrounding both eyes while vacationing in Florida. Within 2 weeks, he noted shortness of breath with exertion and painful fissuring of the distal aspects of his fingertips bilaterally. He was initially treated with a course of levofloxacin for presumed pneumonia. No improvement was seen, so a 12-day taper of prednisone starting with 30 mg daily was prescribed. He had no improvement with this intervention either, and his shortness of breath and rash persisted. Two months after initial symptom onset, a computed tomography (CT) scan (Figure 1A) of his chest showed subtle changes consistent with early ILD. Two weeks after the CT scan, the dyspnea progressed rapidly and he presented to the emergency department with acute hypoxemia. A second CT scan was performed (Figure 1B), which showed no pulmonary embolism but significant progression of bilateral Medicine & Volume 91, Number 4, July 2012 Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine & Volume 91, Number 4, July 2012 MDA-5 Autoantibody and Dermato-Pulmonary Syndrome FIGURE 1. A, CT scan of chest with intravenous contrast 10 days before admission. The representative slice shows areas of patchy consolidation in the posterior portions of the lower lobes bilaterally. B, CT scan without intravenous contrast, taken on the day of admission, showing significant consolidation in the posterior aspects of the bilateral lower lobes, surrounded by patchy ground-glass infiltrates. C, Fingers of Case 1 showing dry, crackling skin at the fingertips, typical of mechanic‚Äôs hands. D, Hematoxylin and eosin stain of the left lingula showing areas of extensive interstitial fibroplasia and marked pneumocyte atypia and interstitial inflammation consistent with organizing diffuse alveolar damage with associated bronchopneumonia. interstitial and ground-glass opacities, consolidation in the bilateral lower lobes, and mediastinal lymphadenopathy (largest 2.5 √Ç 1.4 cm). Review of his systems was unremarkable for fever or chills, arthralgia, or myalgia. Past medical history was significant only for well-controlled diabetes on metformin. He had never smoked and did not use alcohol or illicit drugs. His family history was unremarkable. On presentation, he was afebrile, normotensive, breathing 26 times/min, and was saturating 95% on 100% non-rebreather mask. Pulse was 105 beats/min. Pulmonary exam revealed tachypnea and diffuse crackles bilaterally. Cardiac and abdominal exam were unremarkable. Skin exam showed evidence of periorbital erythema, fissuring of the distal fingers, and small papular lesions on the palmar surface of his hands (Figure 1C). No additional rash or edema was noted. No evidence of Raynaud or abnormal nailfold capillaries were seen on nailfold capillaroscopy. Comprehensive musculoskeletal exam showed no evidence of erythema, deformity, or synovitis in any joints. Muscle strength was 5/5 in proximal and distal muscle groups. Bulk and tone were preserved in all extremities. Initial blood gas showed a pH of 7.16, pCO2 of 101, and pO2 of 77 on 100% non-rebreather mask. White blood cell count was 9730. Electrolyte panel was unremarkable. Erythrocyte sedimentation rate (ESR) was 60 mm/h. Thyroid-stimulating hormone (TSH), complement C4 and C3, creatine kinase (CK), aldolase, urine protein to creatinine ratio, human immunodeficiency virus (HIV) ELISA, and serum protein electrophore* 2012 Lippincott Williams & Wilkins sis were all normal. Lactate dehydrogenase was moderately elevated at 542 IU/L (normal range, 118Y273 IU/L), antinuclear antibody (ANA) was negative but showed a granular cytoplasmic immunofluorescence staining pattern at a titer of 1:320. Antineutrophil cytoplasmic antibody (ANCA), antitopoisomerase (Scl70) antibody, anti-Smith antibody (Sm), rheumatoid factor (RF), anticyclic citrullinated protein (CCP), and anti-ribonucleoprotein (RNP) antibody were also negative. Given the patient‚Äôs respiratory distress, he was admitted directly to the medical intensive care unit and intubated. He was started on empiric ceftriaxone and azithromycin for communityacquired pneumonia. On hospital day 1, he underwent a videoassisted thorascopic surgery (VATS) biopsy of the lingula. Histology was consistent with organizing diffuse alveolar damage with areas of patchy bronchopneumonia (Figure 1D). No evidence of infection was identified from sputum, blood, or tissue culture. Biopsy of the periorbital rash showed only dilated blood vessels in the superficial dermis. Biopsies of the right deltoid and right quadriceps muscle showed severe type 2 muscle atrophy without inflammatory changes. On hospital day 2, the patient was started on methylprednisolone 250 mg every 6 hours for 3 days and then continued on prednisone 75 mg daily for the duration of his hospital course. One week following admission, another chest CTwas performed that showed further progression of infiltrates. A trial of intravenous immunoglobulin (IVIG) was initiated for 5 days. The patient‚Äôs hypoxemia progressed on mechanical ventilation using ARDS-NET protocol,1 and a trial of high frequency oscillator www.md-journal.com Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 221 Medicine Chaisson et al FIGURE 2. Confirmation of anti-MDA-5 antibodies in the sera of Cases 1 and 2 using an immunoprecipitation assay. 35 S-methionine-labeled MDA-5, generated by in vitro transcription/translation, was immunoprecipitated using serum from Cases 1 and 2 (lanes 2 and 3) or 2 different normal controls (lanes 4 and 5). Input 35S-methionine-labeled MDA-5 (not subjected to immunoprecipitation) is shown in lane 1. These data demonstrate that Cases 1 and 2 have anti-MDA-5 antibodies. ventilation failed. He was transitioned to comfort care and died shortly after ventilator support was withdrawn. Autopsy revealed bilateral diffuse alveolar damage and acute interstitial pneumonitis. There was also focal intraalveolar hemorrhage, edema, and acute inflammation in the left upper lobe. No pulmonary emboli were identified. Paratracheal lymphadenopathy was noted with the largest lymph node measuring 4.5 cm in greatest dimension, benign on histologic examination. No other significant pathologic findings were seen on autopsy. Postmortem, immunoprecipitation was performed on the patient‚Äôs serum to evaluate the presence of MDA-5 antibody, given the constellation of clinical findings and the absence of another cause of rapidly progressive lung disease. Immunoprecipitations performed in our research laboratory using in vitro transcription/translated 35S-methionine-labeled MDA-5 confirmed the presence of MDA-5 antibodies in serum from this patient (Figure 2, Case 1). Cause of death was ultimately determined to be the result of rapidly progressive ILD, likely secondary to MDA-5-associated dermato-pulmonary syndrome. Case 2 A 60-year-old healthy white woman began noticing increasing fatigue, followed 2 weeks later by acute onset of night & Volume 91, Number 4, July 2012 sweats, chills, nonproductive cough, and dyspnea. She was treated with 2 courses of antibiotics over 3 weeks, with improvement in the dyspnea and chills, but continued fatigue. Two months later, she noticed skin changes including periorbital edema, and erythema on her forehead, nose, chin, elbows, and axillae, and she developed symmetric inflammatory arthritis involving her hands, wrists, shoulders, knees, and ankles. She was started on a steroid taper by her primary care physician and referred for further evaluation. Complete blood count and electrolyte panel were unremarkable. ESR was 19 mm/h. TSH, CK, and aldolase were also unremarkable. Serologies, including ANA, ANCA, RF, antiphospholipid antibody, antihistidyl tRNA synthetase (Jo-1) antibody, cryofibrinogen, and cryoglobulins, were negative. Mammogram, Pap smear, and colonoscopy were normal. Steroids were reinitiated at this point with improvement in the arthralgia, but persistence of the rash. Two months later, the patient was admitted to the hospital with worsening dyspnea and chest tightness. Chest CT showed multifocal areas of ground-glass infiltrates. A third course of antibiotics was prescribed with improvement in her dyspnea. A prednisone taper starting at 40 mg daily was prescribed given the persistent rash, arthralgia, myalgia, and fatigue. Only the arthralgia improved with prednisone, and pulmonary symptoms worsened. Repeat CT of the chest showed improved bilateral ground-glass infiltrates compared to the initial scan, but worsening peripheral, patchy airspace opacities. She also had developed new papules of the palms and painful fissures at the lateral aspects of the interphalangeal joints of the hands. Painful ulcers developed on her wrists, forearms, arms and buttocks. She continued to have active arthritis. She was started on methotrexate. Six months after symptom onset, she had lost 60 lb, and her exercise tolerance was limited to walking a few steps in her house. She was hospitalized and transferred to our institution for progressive dyspnea. A third chest CT scan showed multilobar consolidation (Figure 3A). Pulmonary function tests demonstrated a restrictive pattern with severely impaired gas transfer (forced vital capacity EFVC^ 2.34 L, 69% predicted; forced expiratory volume in the first second EFEV1^ 1.87 L, 70% predicted; total lung capacity ETLC^ 3.79 L, 69.6% predicted; diffusing capacity of lung for carbon monoxide EDLCO^ 8.75, 41.3% predicted). On examination the patient was afebrile, normotensive, breathing 18 breaths/min, and was saturating 95% on room air. Her pulse was 98 beats/min. She had diffuse nonscarring alopecia, upper eyelid edema, livedoid rash on her FIGURE 3. A, Thoracic CT images with intravenous contrast enhancement showing scattered peripheral interstitial thickening throughout both lungs and multiple subcentimeter ground-glass pulmonary nodules throughout all lobes of both lungs. B, Cutaneous ulceration of the dorsal third metacarpophalangeal joint of the left hand. 222 www.md-journal.com * 2012 Lippincott Williams & Wilkins Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine & Volume 91, Number 4, July 2012 chin and nasolabial folds, and V-neck erythema. Erythema was present over the dorsum of the second to fifth metacarpophalangeal joints bilaterally with an ulceration on the dorsum of the third left metacarpophalangeal joint (Figure 3B). She had exquisitely painful ulcers at the tips and lateral borders of the fingers, as well as a livedoid palmar rash. Palmar papules were found overlying the second metacarpophalangeal joint as well as thenar and hypothenar eminences. She had wrist and elbow ulcers and bilateral painful ulcers on her buttocks. She had no clinical evidence of weakness. An electromyogram was not performed. White blood cell count was 4270. Electrolyte panel was unremarkable. Sedimentation rate was elevated at 45 mm/h. Complement C3, complement C4, serum protein electrophoresis and urine protein to creatinine ratio were normal. Repeat serologies, including a myositis panel (Quest Diagnostics assay), were unremarkable with exception of ANA of 1:80 in speckled pattern and a positive anti-Ro antibody (ELISA assay). Ferritin was elevated at 713 ng/mL. Skin biopsies from multiple sites (chest, ulcers, palms, buttock ulcer) were nonspecific. A skin biopsy from the chest rash showed rare dying keratinocytes, diffuse pigment-laden macrophages, and a mild, superficial perivascular and interstitial infiltrate. Wedge biopsies of the lung (lingula, left superior lower lobe, left inferior lower lobe) showed interstitial fibrosis with honeycombing. She was treated with prednisone taper starting at 60 mg daily and azathioprine 75 mg daily, escalated to 150 mg daily. After 5 months on therapy her energy level improved and she returned to activities in her home. Pulmonary function tests showed preserved lung function (FVC 2.29 L, 69% predicted; FEV1 1.9 L, 71% predicted; TLC 3.55 L, 65% predicted; DLCO 10.75, 44% predicted). The lesions on her fingers and the ulcerative lesions of her wrists and elbows persisted despite therapy, and were complicated by wound infection necessitating surgical debridement and intravenous antibiotic therapy. No new skin lesions appeared. The patient gave written informed consent for further research antibody testing given her constellation of symptoms and progressive skin and lung disease. She was found to have antiMDA-5 antibody by immunoprecipitation assay (see Figure 2, Case 2). METHODS We reviewed the MEDLINE database (National Library of Medicine, Bethesda, MD) for publications between 1966 and July 6, 2011, using the following search terms: ‚Äò‚Äòamyopathic dermatomyositis,‚Äô‚Äô ‚Äò‚Äòclinically amyopathic dermatomyositis,‚Äô‚Äô ‚Äò‚Äòdermatomyositis,‚Äô‚Äô ‚Äò‚Äòinterstitial lung disease,‚Äô‚Äô ‚Äò‚ÄòMDA-5,‚Äô‚Äô ‚Äò‚Äòmelanoma differentiation-associated protein 5,‚Äô‚Äô ‚Äò‚Äòanti-CADM-140,‚Äô‚Äô and ‚Äò‚Äòanti-CADM-140 antibody.‚Äô‚Äô We reviewed only articles written in English, and limited our search to humans. Additionally, articles were excluded if they did not contain discussion of clinical features of MDA-5 antibody, and/or only reviewed prior studies. One study was excluded for insufficient data.36 Abstract and full text reviews were performed independently by 3 authors (NFC, JP, A-MO). Data extraction was performed by the same 3 authors. Where possible, we attempted to identify publications that may have used the same patients in duplicate studies. Immunoprecipitations to confirm the presence of MDA-5 antibodies were performed as described2 using 35S-methioninelabeled MDA-5 that was generated by coupled in vitro transcription/translation from full-length cDNA encoding human MDA-5 (Origene, Rockville, MD). One KL of serum was used in each immunoprecipitation, and the immunoprecipitates were * 2012 Lippincott Williams & Wilkins MDA-5 Autoantibody and Dermato-Pulmonary Syndrome visualized by fluorography after electrophoresis on 10% SDSpolyacrylamide gels. RESULTS We identified 11 studies that met our inclusion criteria (Table 1). Most studies were case series or cross-sectional analyses. One case report was included. No clinical trials or prospective analyses were identified. Two pairs of studies likely included data from similar patient cohorts. All studies but 1 evaluated Japanese or Korean patients; the other study involved American patients. In all studies but 1, the presence of MDA-5 antibodies was identified only in patients with DM or CADM. In the remaining study, a patient with systemic sclerosis was MDA-5 antibody positive.15 Among MDA-5 antibody-positive patients, 8 studies included data regarding presence of myositis.8,11,13,15,23,27,28,32 Accounting for studies that included overlapping patient cohorts, 80% of MDA-5-positive individuals were clinically amyopathic (CK G300 and clinically insignificant muscle weakness). It is notable that MDA-5 is a cytoplasmic protein, whose expression is interferon (IFN)-regulated. To our knowledge there are no systematic analyses of whether patients with MDA-5 autoantibodies also express other autoantibodies. The first patient described here was ANA negative, and the second patient had low ANA titer and a positive anti-Ro antibody. There was a significant prevalence of ILD (60%Y100%) among MDA-5 antibody-positive individuals, although criteria (radiographic changes or restriction on pulmonary function test) for defining ILD varied by study (see Table 1). A significant portion of patients (22%Y100%) was classified as having acute or subacute ILD, characterized by evidence of progression of ILD within 1 or 3 months, respectively. This was significantly higher than MDA-negative patients (4%Y20%).8,23,32 One study evaluated differences in chest CT scan findings between MDA-5 antibody-positive and -negative patients with DM-associated ILD.32 Patients with MDA-5 antibody tended to have regions of ground-glass attenuation in random distribution or focally in the lower lobes of the lung, whereas MDA-5 antibody-negative patients had predominately reticular changes in the lower lobes. Significant skin findings in patients with MDA-5 antibodies included skin ulceration, arthritis, mechanic‚Äôs hands, palmar papules, heliotrope rash, Gottron papules, periungual erythema, and periungual telangiectasias (Table 2). Evidence from a systematic study of patients referred to a dermatology clinic7 indicates that the first 4 clinical findings are significantly more prevalent in MDA-5 antibody-positive individuals than in MDA5 antibody-negative DM patients. Six studies reported data on treatment, although most did not include specific dosing or mention treatment algorithms employed (Table 3). In all but 1 of the 6 studies,15 prednisone or prednisolone was used. Other immunosuppressants included cyclophosphamide, cyclosporine, azathioprine, IVIG, and, in 1 study, methotrexate. Only 1 study reported individual treatment data.13 In that study, all MDA-5 antibody-positive patients received prednisolone initially, but no specific preference for treatment could be identified beyond this. Mortality data were provided in 8 studies.8,10,11,13,15,18,23,32 Accounting for studies that included overlapping patient cohorts, overall mortality was 38% for MDA-5 antibody-positive patients, over follow-up periods ranging from 6 months to 14 years. Two studies reported mortality data at 6 months8,23 that were much higher in MDA-5 antibody-positive patients compared with MDA-5 antibody-negative patients (57% Erange, 46%Y75%^ vs. 9% Erange, 8%Y9%^). One study that followed patients an www.md-journal.com Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 223 224 www.md-journal.com Nested case-control Multicenter cross-sectional Case series Case series Case series Cross-sectional 27 13 23 11 15 8 Japanese Japanese Japanese Japanese Japanese Japanese Korean Japanese American Japanese Japanese DM, CTD DM DM, CTD, SSc with ILD DM DM and PM MDA-5 DM DM with ILD DM, CTD, IPF, healthy controls DM, CTD, IPF, healthy controls DM, CTD Population Studied 30‚Ä† 192 65 135 376‚Ä† 255‚Ä† 49 2 77 25 294‚Ä† 30‚Ä† 13 65 82 376‚Ä† 42‚Ä† 38 2 77 25 67‚Ä† 8‚Ä† (1/7) 13 (3/10) 14 (3/11) 21 (4/17) 43‚Ä† (9/34) 8‚Ä† (2/6) 9 (4/5) 2 (0/2) 10 (3/7) 12 (4/8) 27‚Ä†‚Ä° (NA) 100 (NA) NA NA 100 (50) 100 (92) 100 (100) 100 (77) 100 (84) 100 (53) 95 (91) 100 (75) 45.8 T 16 47.5 T 24.7 51.5 T 8.8 53.5 T 9.4 NA 44.5 T 12.7 53 T NA 57.3 T 9.4 43.6 T 14.6 47 T 15 60.5 T 10.9 75 46 63 14 44 NA 33 50 NA 58 NA 100 92 100* 95 93 88 66 100 60 100* NA Total Total Patients No. Patients Age at % With % Patients With DM With MDA-5 Onset (yr) CADM or DM With Studied or CADM (M/F) (Mean T SD) (CADM Only) Deaths (%) ILD 100 54 71 79 NA 50 66 100 22 42 NA % With ILD and A/SIP 20 4 NA NA NA NA NA NA 5 8 NA % With ILD and A/SIP MDA-5-Negative Patients Abbreviations: A/SIP = acute or subacute interstitial pneumonitis, CTD = connective tissue disease, IPF = idiopathic pulmonary fibrosis, NA = not available; PM = polymyositis, SD = standard deviation, SSc = systemic sclerosis/scleroderma. *All patients had ILD as part of the inclusion criteria. ‚Ä†Indicates possible duplication of patients between studies. ‚Ä°Number of patients with positive result by immunoprecipitation. Cross-sectional Case report Case series Cross-sectional Cross-sectional Study Type 18 10 7 32 28 Reference Population Demographic Anti-MDA-5-Positive Patients TABLE 1. Demographic Characteristics for Both Anti-MDA-5 Antibody-Positive and -Negative Patients, Previous Reports Chaisson et al Medicine & Volume 91, Number 4, July 2012 * 2012 Lippincott Williams & Wilkins Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine & Volume 91, Number 4, July 2012 MDA-5 Autoantibody and Dermato-Pulmonary Syndrome TABLE 2. Skin Features and Biopsy Results Among Patients Who Were Anti-MDA-5 Positive by Immunoprecipitation Findings Gottron Periungual Periungual Skin Palmar Compatible Heliotrope Papules Erythema Telanciectasias Ulceration Arthritis Papules Mechanic‚Äôs Fever With DM on Reference Rash (%) (%) (%) (%) (%) (%) (%) Hands (%) (%) Skin Biopsy (%) 10 7 27 13 23 15 8 100 70 50 56 15 68 38 100 70 75 86 92 68 25 NA NA NA 70 NA NA 63 NA 90 NA NA NA NA NA 50 80 NA 30 NA NA NA NA 70 50 42 69 NA NA NA 60 NA NA NA NA 63 NA 60 NA NA NA NA NA NA NA NA 74 69 NA NA 0 NA 100 NA NA NA NA Abbreviations: See previous table. average of 588 days reported 7 of 12 patients with MDA-5 antibody died compared to none in the MDA-5 antibody-negative cohort. In all 8 studies, the majority of deaths in MDA-5 antibodypositive patients were attributable to progressive ILD. DISCUSSION DM encompasses a wide spectrum of disease, almost universally involving the skin, and frequently affecting muscle and lung. The disease phentoypes are complex, with the combination of tissues involved in different individuals varying considerably.30 The disease associated with MDA-5 autoantibodies is characterized by striking ILD and a distinct cutaneous phenotype (especially ulcers and vasculopathy), but minimal to absent muscle inflammation and damage. This constellation of damage is suggestive of a common vascular target. The cutaneous and pulmonary involvement in the absence of myositis leads us to posit that this unusual dermato-pulmonary syndrome has distinct underlying causes and mechanisms. Although recent studies have suggested that muscle disease can occur in MDA-5 antibodypositive patients, it is unusual to have significant muscle weakness or elevation of muscle enzymes in this setting.29,32 Detection of commercially available autoantibodies is also rare in patients with anti-MDA-5. In the 2010 study by Gono et al,11 only 21% of patients with MDA-5 antibodies had a positive ANA. However, the presence of cytoplasmic staining by immunofluorescence was noted in both of our patients and is TABLE 3. Referenced Therapies Used in Patients With Anti-MDA-5 Antibodies* Reference 10 32 13 11 15 8 Therapy Pulsed MP, MP, CsA, Tac, IVIG, AZA MP, CsA, Cyc Pulsed MP, MP, CsA, Cyc, IVIG MP, Cyc, CI Pulsed MP, CsA, Cyc, AZA, MTX MP, immunosuppressants not otherwise defined Abbreviations: AZA = azathioprine, CI = calcineurin inhibitor not otherwise specified, CsA = cyclosporin A, Cyc = cyclophosphamide, IVIG = intravenous immunoglobulin, MP = methylprednisolone, MTX = methotrexate, Tac = tacrolimus. *Specific algorithms were not addressed in any study, and individual data were provided in only 1 study (reference 13). * 2012 Lippincott Williams & Wilkins characteristic in patients with MDA-5 antibody.27 In a patient who presents with ILD, palmar papules, painful skin ulcers and absent myopathy, absent nuclear staining and presence of cytoplasmic immunofluorescence pattern should raise suspicion for MDA-5 antibodies. Cutaneous Manifestations of MDA-5 The hallmark cutaneous manifestations of the DM spectrum include violaceous papules or plaques located on the dorsal aspect of the metacarpophalangeal or interphalangeal joints, called Gottron papules; erythematous or violaceous macules in the same distribution, called Gottron macules; and heliotrope rash, which is a periorbital violaceous erythema. These findings, which are present in over 70% of individuals with DM,20 occur with similar prevalence in both MDA-5 antibody-positive and -negative patients. MDA-5 specific skin manifestations include 1) painful, erythematous papules, macules, or both on the palmar surfaces of the metacarpophalangeal and interphalangeal joints that can have a central, ivory coloration, ulceration, and sometimes manifest as 2 distinct papules on either side of the joint; 2) cutaneous ulceration, found in 80% of these patients compared to G20% of anti-MDA-5-negative DM patients. Ulcers can be found on the elbows and knees, lateral nailfolds, and within Gottron papules. Patients with MDA-5 antibodies also have a higher prevalence of tender gums and/or oral erosions compared with MDA-5 antibody-negative patients.7,13 Our 2 patients both had palmar papules, and Case 2 had painful ulceration of the hands, wrists, elbows, and buttocks. In neither case could a clinical diagnosis of classic DM be made with certainty. Additionally, skin biopsies in both patients were nondiagnostic. MDA-5-Associated Interstitial Lung Disease The spectrum of ILD in patients with DM can range from the asymptomatic individual with incidental radiographic or pulmonary function abnormalities, to rapidly progressive and fatal disease. The reported prevalence of ILD among all patients with an idiopathic inflammatory myopathy varies widely.5,21 In all studies of MDA-5-associated disease we identified, the lung was frequently involved and was often severely affected, highlighting the lung as a central focus in this syndrome. Studies indicate that ILD progresses more rapidly in those patients with MDA-5 antibodies, with approximately 40% of MDA-5 antibody-positive patients manifesting rapidly progressive ILD. In the largest group of MDA-5 antibody-positive patients reported to our knowledge to date,13 93% had ILD, using results from chest radiography, chest CT, and pulmonary function tests. Patients were classified www.md-journal.com Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 225 Medicine Chaisson et al as having rapidly progressive ILD if they had worsening dyspnea, hypoxemia, and progression of interstitial changes on radiography within 1 month of onset of symptoms. In general, MDA-5 antibody-positive patients who have coexistent muscle disease have a lower mortality than those without muscle disease.11 Both of our patients initially described pulmonary symptoms close to the time of onset of skin manifestations, with pulmonary symptoms preceding skin disease in 1 patient, and following skin disease in the other. We found only 1 study that looked at specific radiographic findings in patients with MDA-5 antibodies.32 The findings of ground-glass attenuation in MDA-5 antibody-positive patients versus reticular patterns in MDA-5 antibody-negative patients remains very nonspecific and is unlikely to change management without further verification of underlying disease pathology by the clinician. Additionally, these data need to be validated in a large, blinded cohort before firm conclusions can be made. Pathologic data in patients with MDA-5 antibodies have not been well described to date either. In 1 of our patients, wedge biopsy revealed usual interstitial pneumonia pattern, while in the other a diffuse alveolar damage pattern was present. Findings of diffuse alveolar damage are uncommon in patients with CADM and DM, and in these cases, patients generally have poor response to therapy and dismal prognosis.22,25 MDA-5 Autoantibodies and Disease Pathogenesis MDA-5 belongs to the family of retinoic acid-inducible gene I (RIG-I)-like receptors (RLRs), a family of RNA helicases that functions as cytoplasmic sensors of pathogen-associated molecular patterns within viral RNA. To date, 3 RLR members have been identified: RIG-I, MDA-5, and LGP2 (laboratory of genetics and physiology 2). While all are broadly expressed in most tissues, levels are typically low in resting cells but are greatly increased with IFN exposure and after virus infection.17,34,35 The RLRs drive type 1 IFN production and antiviral gene expression, which mediates the intracellular immune response to control virus infection. Recent studies suggest that RLRs may also activate other inflammatory signaling pathways, including the inflammasome.26 Different RLRs appear to detect infection preferentially with different viruses. Interestingly, MDA-5 detects several RNA viruses, including picornaviruses (like polio, coxsackie, and rhinovirus), flaviviruses (Dengue and West Nile), as well as vaccinia (DNA virus). The striking vascular character of the skin phenotype in MDA-5 antibody-positive patients and the rapidly progressive lung disease suggest that blood vessels are the primary target of the immune response, potentially initially targeted during virus infection, and driven later by type I IFN pathways. Defining the mechanisms underlying the initiation and propagation of the anti-MDA-5 immune response is a high priority. Elucidating the sites of MDA-5 expression in the target tissue (and especially blood vessels at these sites) will be important in this regard. It has been noted that resistance to type 1 diabetes in humans is associated with polymorphisms that impair the function of the mda5 gene, suggesting that the relative strength of induction of type I IFNs may be important in establishing and propagating autoimmunity.3,24 Since uncontrolled viral infection on the one hand (inadequate viral sensing by MDA-5) or heightened type I IFN-induced inflammation on the other (strong MDA-5 function) both have the capacity to render an individual susceptible to tissue damage, it will be of interest to define whether mda5 polymorphisms are present in these patients, which either augment or impair MDA-5 expression and function, and thereby regulate downstream type I IFN effects. If augmented MDA-5 signaling and type I IFN activity play a role in 226 www.md-journal.com & Volume 91, Number 4, July 2012 the amplified ILD and vascular phenotype seen in this dermatopulmonary syndrome, this might have therapeutic implications for newly available agents that inhibit type I IFN signaling. Alternatively, if these pathways are decreased in MDA-5 antibodypositive patients, the possibility that increased virus-induced damage might play a role in initiating lung disease in these patients could prompt a different set of approaches. Response to Therapy Effective therapeutic modalities to treat the MDA-5-associated dermato-pulmonary syndrome have not yet been identified, and no data exist comparing the effectiveness of 1 therapy over another. This is likely because this entity has only recently been described, and it is difficult to diagnose given the current lack of a widely available MDA-5 antibody assay. While no firm guidelines can be drawn from the literature regarding treatment of MDA-5 antibody-positive individuals, many experts continue to treat patients similarly to those who have DM and are MDA-5 antibody negative, usually beginning with corticosteroid therapy. Limited reports in the literature of patients with DM and rapidly progressive ILD suggest cyclophosphamide and cyclosporine early in the disease may be beneficial.14,33 Cyclosporine has also been used to treat ulcerative lesions in a patient with CADM and intractable skin ulceration,31 but these observations were made before the recognition of the MDA-5-associated entity. The poor outcomes associated with MDA-5 antibodies strongly suggest that the optimal therapy has not yet been defined. It is possible that rapid, aggressive institution of immunosuppression might avoid irreversible lung injury if diagnosis is prompt. On the other hand, standard approaches to DM and associated conditions may not be appropriate for this condition; instead, more specific therapy targeting type I IFNs, or potentially antiviral therapy if a pathogen is defined, may be more appropriate. In the current study, 1 patient died (Case 1) despite a trial of IVIG, pulse dose prednisone, and cyclophosphamide. The other patient (Case 2) received high-dose steroids and was maintained on prednisone and azathioprine since her diagnosis, but continued to have significant disease activity. Outcomes Data from previous studies show that the presence of MDA-5 antibody is associated with increased mortality in patients with DM and CADM compared to those who are MDA-5 negative.8,23,32 Although many studies did not describe the cause of death, those that did suggested that the majority of individuals succumbed to ILD refractory to immunosuppressive therapy. The presence of MDA-5 antibodies clearly has prognostic value for those who present with this characteristic dermatopulmonary syndrome. Although the numbers remain small, approximately 40% of MDA-5 antibody-positive patients died, mostly within the first year. This was very different from MDA-5 antibody-negative patients, where 1-year mortality was 0%, and overall mortality was G10%. In 2010, Gono et al9 reported findings in individuals with CADM and ILD suggesting that increased ferritin levels correlated with faster progression of ILD. Presence of MDA-5 antibody was not measured in that study, but studies of individuals with MDA-5 antibodies support the observation that the higher ferritin levels might be associated with increased ILD progression.10,11,32 Conclusion In the current report, we describe 2 patients with a progressive dermato-pulmonary syndrome that defied diagnosis using clinically available modalities. Unusual skin features included * 2012 Lippincott Williams & Wilkins Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine & Volume 91, Number 4, July 2012 violaceous papules on the palmar surfaces of the metacarpophalangeal and interphalangeal joints, and painful ulcerative skin lesions. Inflammatory muscle disease was absent. Standard diagnostic studies, including autoantibodies, skin biopsy, and VATS lung biopsy, failed to reveal the underlying process. Antibodies to MDA-5 were present, strongly suggesting that autoimmunity to MDA-5 plays a role in generating this specific phenotype. In these settings, when extensive evaluation is otherwise nondiagnostic, clinical suspicion for MDA-5 antibodies should be high, and further testing should be aggressively pursued. REFERENCES 1. Anonymous. Ventilation with lower tidal volumes as compared with traditional tidal volumes for acute lung injury and the acute respiratory distress syndrome. The Acute Respiratory Distress Syndrome Network. N Engl J Med. 2000;18:1301Y1308. 2. Casciola-Rosen LA, Pluta AF, Plotz PH, Cox AE, Morris S, Wigley FM, Petri M, Gelber AC, Rosen A. The DNA mismatch repair enzyme PMS1 is a myositis-specific autoantigen. Arthritis Rheum. 2001;44:389Y396. 3. Downes K, Pekalski M, Angus KL, Hardy M, Nutland S, Smyth DJ, Walker NM, Wallace C, Todd JA. Reduced expression of IFIH1 is protective for type 1 diabetes. PLoS One. 2010;5:e12646. 4. Euwer RL, Sontheimer RD. Amyopathic dermatomyositis (dermatomyositis sine myositis). Presentation of six new cases and review of the literature. J Am Acad Dermatol. 1991;24:959Y966. 5. Fathi M, Dastmalchi M, Rasmussen E, Lundberg IE, Tornling G. Interstitial lung disease, a common manifestation of newly diagnosed polymyositis and dermatomyositis. Ann Rheum Dis. 2004;56:297Y301. 6. Fathi M, Lundberg IE, Tornling G. Pulmonary complications of polymyositis and dermatomyositis. Semin Respir Crit Care Med. 2007;28:451Y458. 7. Fiorentino D, Chung L, Zwerner J, Rosen A, Casciola-Rosen L. The mucocutaneous and systemic phenotype of dermatomyositis patients with antibodies to MDA5 (CADM-140): a retrospective study. J Am Acad Dermatol. 2011;65:25Y34. 8. Fujikawa K, Kawakami A, Kaji K, Fujimoto M, Kawashiri S, Iwamoto N, Aramaki T, Ichinose K, Tamai M, Kamachi M, Nakamura H, Ida H, Origuchi T, Ishimoto H, Mukae H, Kuwana M, Kohno S, Takehara K, Sato S, Eguchi K. Association of distinct clinical subsets with myositis-specific autoantibodies towards anti-155/140-kDa polypeptides, anti-140-kDa polypeptides, and anti-aminoacyl tRNA synthetases in Japanese patients with dermatomyositis: a single-centre, cross-sectional study. Scand J Rheumatol. 2009;38:263Y267. 9. Gono T, Kawaguchi Y, Hara M, Masuda I, Katsumata Y, Shinozaki M, Ota Y, Ozeki E, Yamanaka H. Increased ferritin predicts development and severity of acute interstitial lung disease as a complication of dermatomyositis. Rheumatology (Oxford). 2010;49:1354Y1360. 10. Gono T, Kawaguchi Y, Ozeki E, Ota Y, Satoh T, Kuwana M, Hara M, Yamanaka H. Serum ferritin correlates with activity of anti-MDA5 antibody-associated acute interstitial lung disease as a complication of dermatomyositis. Mod Rheumatol. 2011;21:223Y227. 11. Gono T, Kawaguchi Y, Satoh T, Kuwana M, Katsumata Y, Takagi K, Masuda I, Tochimoto A, Baba S, Okamoto Y, Ota Y, Yamanaka H. Clinical manifestation and prognostic factor in anti-melanoma differentiation-associated gene 5 antibody-associated interstitial lung disease as a complication of dermatomyositis. Rheumatology (Oxford). 2010;49:1713Y1719. 12. Gunawardena H, Betteridge ZE, McHugh NJ. Myositis-specific autoantibodies: their clinical and pathogenic significance in disease expression. Rheumatology (Oxford). 2009;48:607Y612. * 2012 Lippincott Williams & Wilkins MDA-5 Autoantibody and Dermato-Pulmonary Syndrome 13. Hamaguchi Y, Kuwana M, Hoshino K, Hasegawa M, Kaji K, Matsushita T, Komura K, Nakamura M, Kodera M, Suga N, Higashi A, Ogusu K, Tsutsui K, Furusaki A, Tanabe H, Sasaoka S, Muro Y, Yoshikawa M, Ishiguro N, Ayano M, Muroi E, Fujikawa K, Umeda Y, Kawase M, Mabuchi E, Asano Y, Sodemoto K, Seishima M, Yamada H, Sato S, Takehara K, Fujimoto M. Clinical correlations with dermatomyositis-specific autoantibodies in adult Japanese patients with dermatomyositis: a multicenter cross-sectional study. Arch Dermatol. 2011;147:391Y398. 14. Hirakata M, Nagai S. Interstitial lung disease in polymyositis and dermatomyositis. Curr Opin Rheumatol. 2000;12:501Y508. 15. Hoshino K, Muro Y, Sugiura K, Tomita Y, Nakashima R, Mimori T. Anti-MDA5 and anti-TIF1-gamma antibodies have clinical significance for patients with dermatomyositis. Rheumatology (Oxford). 2010;49: 1726Y1733. 16. Kameda H, Nagasawa H, Ogawa H, Sekiguchi N, Takei H, Tokuhira M, Amano K, Takeuchi T. Combination therapy with corticosteroids, cyclosporin A, and intravenous pulse cyclophosphamide for acute/ subacute interstitial pneumonia in patients with dermatomyositis. J Rheumatol. 2005;32:1719Y1726. 17. Kang DC, Gopalkrishnan RV, Lin L, Randolph A, Valerie K, Pestka S, Fisher PB. Expression analysis and genomic characterization of human melanoma differentiation associated gene-5, mda-5: a novel type I interferon-responsive apoptosis-inducing gene. Oncogene. 2004;23:1789Y1800. 18. Kang EH, Nakashima R, Mimori T, Kim J, Lee YJ, Lee EB, Song YW. Myositis autoantibodies in Korean patients with inflammatory myositis: anti-140-kDa polypeptide antibody is primarily associated with rapidly progressive interstitial lung disease independent of clinically amyopathic dermatomyositis. BMC Musculoskelet Disord. 2010;11:223. 19. Kawachi Y, Maruyama H, Furuta J, Fujisawa Y, Nakamura Y, Takahashi T, Otsuka F. Cutaneous deep necrosis with dermatomyositis: correlation with interstitial pneumonia. Eur J Dermatol. 2007;17:345Y346. 20. Kovacs SO, Kovacs SC. Dermatomyositis. J Am Acad Dermatol. 1998;39:899Y922. 21. Labirua A, Lundberg IE. Interstitial lung disease and idiopathic inflammatory myopathies: progress and pitfalls. Curr Opin Rheumatol. 2010;22:633Y638. 22. Lee CS, Chen TL, Tzen CY, Lin FJ, Peng MJ, Wu CL, Chen PJ. Idiopathic inflammatory myopathy with diffuse alveolar damage. Clin Rheumatol. 2002;21:391Y396. 23. Nakashima R, Imura Y, Kobayashi S, Yukawa N, Yoshifuji H, Nojima T, Kawabata D, Ohmura K, Usui T, Fujii T, Okawa K, Mimori T. The RIG-I-like receptor IFIH1/MDA5 is a dermatomyositis-specific autoantigen identified by the anti-CADM-140 antibody. Rheumatology (Oxford). 2010;49:433Y440. 24. Nejentsev S, Walker N, Riches D, Egholm M, Todd JA. Rare variants of IFIH1, a gene implicated in antiviral responses, protect against type 1 diabetes. Science. 2009;324:387Y389. 25. Parambil JG, Myers JL, Ryu JH. Diffuse alveolar damage: uncommon manifestation of pulmonary involvement in patients with connective tissue diseases. Chest. 2006;130:553Y558. 26. Poeck H, Bscheider M, Gross O, Finger K, Roth S, Rebsamen M, Hannesschlager N, Schlee M, Rothenfusser S, Barchet W, Kato H, Akira S, Inoue S, Endres S, Peschel C, Hartmann G, Hornung V, Ruland J. Recognition of RNA virus by RIG-I results in activation of CARD9 and inflammasome signaling for interleukin 1 beta production. Nat Immunol. 2010;11:63Y69. 27. Sato S, Hirakata M, Kuwana M, Suwa A, Inada S, Mimori T, Nishikawa T, Oddis CV, Ikeda Y. Autoantibodies to a 140-kd polypeptide, CADM-140, in Japanese patients with clinically amyopathic dermatomyositis. Arthritis Rheum. 2005;52:1571Y1576. www.md-journal.com Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 227 Medicine Chaisson et al 28. Sato S, Hoshino K, Satoh T, Fujita T, Kawakami Y, Fujita T, Kuwana M. RNA helicase encoded by melanoma differentiation-associated gene 5 is a major autoantigen in patients with clinically amyopathic dermatomyositis: association with rapidly progressive interstitial lung disease. Arthritis Rheum. 2009;60:2193Y2200. 29. Sato S, Kuwana M. Clinically amyopathic dermatomyositis. Curr Opin Rheumatol. 2010;22:639Y643. 30. Shamim EA, Rider LG, Pandey JP, O‚ÄôHanlon TP, Jara LJ, Samayoa EA, Burgos-Vargas R, Vazquez-Mellado J, Alcocer-Varela J, Salazar-Paramo M, Kutzbach AG, Malley JD, Targoff IN, Garcia-De la Torre I, Miller FW. Differences in idiopathic inflammatory myopathy phenotypes and genotypes between Mesoamerican Mestizos and North American Caucasians: ethnogeographic influences in the genetics and clinical expression of myositis. Arthritis Rheum. 2002;46:1885Y1893. 31. Shimojima Y, Ishii W, Kato T, Hoshi K, Matsuda M, Hashimoto T, Tanaka Y, Ikeda S. Intractable skin necrosis and interstitial pneumonia in amyopathic dermatomyositis, successfully treated with cyclosporin A. Intern Med. 2003;42:1253Y1258. 32. Tanizawa K, Handa T, Nakashima R, Kubo T, Hosono Y, Watanabe K, Aihara K, Oga T, Chin K, Nagai S, Mimori T, Mishima M. 228 www.md-journal.com & Volume 91, Number 4, July 2012 HRCT features of interstitial lung disease in dermatomyositis with anti-CADM-140 antibody. Respir Med. 2011;105:1380Y1387. 33. Yamasaki Y, Yamada H, Yamasaki M, Ohkubo M, Azuma K, Matsuoka S, Kurihara Y, Osada H, Satoh M, Ozaki S. Intravenous cyclophosphamide therapy for progressive interstitial pneumonia in patients with polymyositis/dermatomyositis. Rheumatology (Oxford). 2007;46:124Y130. 34. Yoneyama M, Kikuchi M, Matsumoto K, Imaizumi T, Miyagishi M, Taira K, Foy E, Loo YM, Gale M Jr, Akira S, Yonehara S, Kato A, Fujita T. Shared and unique functions of the DExD/H-box helicases RIG-I, MDA5, and LGP2 in antiviral innate immunity. J Immunol. 2005;175: 2851Y2858. 35. Yoneyama M, Kikuchi M, Natsukawa T, Shinobu N, Imaizumi T, Miyagishi M, Taira K, Akira S, Fujita T. The RNA helicase RIG-I has an essential function in double-stranded RNA-induced innate antiviral responses. Nat Immunol. 2004;5:730Y737. 36. Zahn S, Barchet W, Rehkamper C, Hornung T, Bieber T, Tuting T, Wenzel J. Enhanced skin expression of melanoma differentiation-associated gene 5 (MDA5) in dermatomyositis and related autoimmune diseases. J Am Acad Dermatol. 2011;64:988Y989. * 2012 Lippincott Williams & Wilkins Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. ˇ˛Abstract European Journal of Internal Medicine 18 (2007) 314 320 Original article www.elsevier.com/locate/ejim Anti- and pro-oxidant factors and endothelial dysfunction in chronic cigarette smokers with coronary heart disease E. Rocchi a,N , F. Bursi c, P. Ventura b, A. Ronzoni a, C. Gozzi a, G. Casalgrandi a, L. Marri a, R. Rossi a, M.G. Modena a a Dept. Medicine e Specialita Mediche, Chair of Terapia Medica, Universita di Modena e Reggio Emilia, Italy b Dept. Medicine e Specialita Mediche, Chair of Medicina II, Universita di Modena e Reggio Emilia, Italy c Dept. Emergenza-Urgenza, Chair of Cardiologia, Universita di Modena e Reggio Emilia, Italy Received 22 April 2006; received in revised form 18 September 2006; accepted 12 October 2006 Background: Endothelial dysfunction in cigarette smokers has been ascribed to increased oxidative damage. The aims of the present study were to compare the endothelial function of normotensive smokers with that of non-smokers and to examine its relation to some parameters representative of oxidative damage and of antioxidant capacity. Methods: We investigated 32 chronic smokers (15 30 cigarettes daily) affected by coronary heart disease, ranging from acute myocardial infarction to instable angina pectoris, and 28 matched non-smokers without any definite risk factors. All subjects underwent assessment of nitric oxide (NO)-dependent endothelial function, measured as brachial artery vasodilatation in response to reactive ischemia, using a standardized echographic method. Plasma and urinary levels of NO were also measured in all subjects, as were urinary 15-isoprostane F2t, plasma serum lipids, homocysteine (Hcy), ascorbic acid, retinol, tocopherol, and alpha- and beta-carotene (by high-performance liquid chromatography). Results: Smokers showed a significantly lower NO-mediated vasodilatation response (3.50% vs. 6.18%, pb0.001) and higher levels of urinary NO metabolites and 15-isoprostane F2t. They also had higher levels of Hcy ( p b 0.001); these values were significantly and inversely related to NO serum levels (r="0.512, pb0.001). Moreover, smokers had a significant and corresponding reduction in circulating levels of ascorbic acid, tocopherol, and alpha- and beta-carotene. Conclusions: The present study shows a clear relation between endothelial dysfunction (NO production impairment) and cigarette smoking, especially in the presence of high levels of LDL-cholesterol. It also defines some markers of both oxidative damage and antioxidant protective capacity in this condition. The monitoring of these factors may be advisable in order to assess the amount of endothelial damage. © 2007 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. Keywords: Antioxidants; Endothelial dysfunction; Cigarette smoking; Homocysteine; Isoprostane; Nitric oxide 1. Introduction Cigarette smoking represents an independent risk factor for cardiovascular disease [1]. However, the relative role of nicotine [2 5] and of other factors (hypertension, hypercho- lesterolemia) [6,7] in inducing vascular damage, and partic- N Corresponding author. University of Modena and Reggio Emilia, Dept. Medicine e Specialita Mediche, Chair of Terapia Medica, Via del Pozzo 71, Policlinico di Modena, Italy. Tel.: +39 059 4222151; fax: +39 059 4224363. E-mail address: rocchi.emilio@unimore.it (E. Rocchi). ularly endothelial dysfunction, is not well-defined. The latter, which is widely considered to be an important hallmark of atherosclerosis, is often the final result of different complex mechanisms, involving the production of reactive oxygen species (free radicals and peroxylradicals) [8] and leading to reduced nitric oxide (NO) endothelial availability. In the vasculature, endothelial NO (EDNO) plays a key role in maintaining the blood flow by modulating physio- logical vasodilatation [9 12]. Another important physiolog- ical reaction of NO is the formation of thionitrites or S- nitrosothiols (RSNOs) [13]. These adducts form in the 0953-6205/$ - see front matter © 2007 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.ejim.2006.10.006 Groups Non-smokers (n = 28) Smokers (n=32) Poultry 1.92 ± 0.15 1.66±0.11 Meat 1.64 ± 0.15 1.75±0.16 Leafy vegetables 2.31 ± 0.16 2.17±0.13 Tomatoes and carrots 2.08 ± 0.16 2.03±0.14 Nuts 0.95 ± 0.15 1.05±0.14 Milk 2.24 ± 0.23 2.00±0.20 Dairy products 2.31 ± 0.12 2.55±0.14 E. Rocchi et al. / European Journal of Internal Medicine 18 (2007) 314 320 315 Table 1 Dietary report in the groups studied, based on an empirical scoring system for the different classes of nutrients Fruit 2.54 ± 0.16 2.52±0.16 Values from 0 to 3, corresponding, respectively, to an absence of the class in the weekly diet to an nutritional board. Mean ± S.E. optimal dietary intake, were assigned according to a local presence of a thiol compound and oxygen and have, in turn, an important role in reducing vascular tone and platelet aggregation [14,15]. Among the different kinds of RSNOs, the transformation of homocysteine (Hcy), a well-known independent risk factor for atherothrombosis, into S-nitroso- homocysteine seems relevant. This substance, unlike native Hcy, does not support hydrogen peroxide genera- tion and, hence, cell toxicity, but shows, on the contrary, well-defined and important vasodilating and inhibiting properties of platelet aggregation and adhesion of leuko- cytes to vessel walls [16,17]. In the presence of an excess of reactive oxygen species, there is a reduction in the availability of endothelial NO, due to a greater formation of peroxynitrites, which induces many different damaging effects to lipid, protein, and nucleic acid metabolism. The decrease in NO availability induces different pro-atherogenic events, such as hyper-expression of vascular cell adhesion molecules (V-CAM; I-CAM), an increase in Hcy in native (and hence toxic) form, and a reduction in vasodilator response [17 19]. For these reasons, an assessment of pro- and antioxidant factors that are able to influence endothelial function, and particularly NO availability, seems relevant. In the present study of patients affected by coronary heart disease (CHD) and stratified by the presence/absence of smoking habit, we compared the flow-mediated dilatation (FMD) in response to reactive ischemia and the corresponding biochemical mar- kers of both oxidative damage and antioxidant status, in order to establish which of these markers may be more useful for clinical, diagnostic use and control interventions [20]. 2. Materials and methods 2.1. Patients The study was conducted in accordance with the prin- ciples of the Declaration of Helsinki. All participants gave their written informed consent before the study. Sixty subjects (all Caucasian) were recruited for the study. Thirty-two (aged 50 ± 9 years, mean ± SD, 13 females) were chronic cigarette smokers and 28 (aged 52 ± 8 years, 15 females) were non-smokers. All of the subjects had pre- viously been admitted to the cardiology unit of our institu- tion and diagnosed as having coronary heart disease (CHD), ranging from acute myocardial infarction (AMI) to instable angina pectoris (IAP). Smoking habit (15 30 cigarettes daily) was the single or largely predominant risk factor in the first group; no other definite risk factors were present in the second group. No other disease (diabetes in particular) was observed in either group and all subjects had stopped using drugs (diuretics, nitrates, ACE inhibitors, when present, in some cases) in the last 48 h prior to the study. The dietary habits of the subjects are presented in Table 1. 2.2. Assessment of brachial arterial reactivity A previously validated ultrasound study of brachial artery reactivity was performed in all patients upon entering the study and 6 months later using an Acuson 128 XP/10 mainframe (Acuson, Mountain View, California, USA) with a 7.0 MHz linear-array transducer. Images were stored on a super VHS videotape recorder for further analysis. The technique for assessing brachial artery FMD has already been described in detail elsewhere [21,22]. The methodology has an inter-observer variability in diameter measurements of 0.45 ± 0.25%, yielding a coefficient of variation of 1.34% and a coefficient of repeatability of 0.8%. The study was performed early in the morning and the smokers were ad- vised not to smoke. 2.3. Chemicals All chemicals, of HPLC grade, were purchased from Farmitalia (Milan, Italy). Ascorbic acid, retinol, D-alpha- tocopherol, alpha-carotene, beta-carotene, tocopherol ace- tate, retinol acetate (the last two as internal standards), Table 2 Main laboratory analytes in each group investigated Groups Age (years) Non-smokers 52 ± 1.5 (n=28) Smokers 50 ± 1.7 (n=32) Mean ± S.E. RBC◊106 4.57 ± 0.10 4.49 ± 0.09 Hb g/dl 13.32 ± 0.29 13.29 ± 0.34 LDL mg/dl 143.9 ± 10.6 174.7 ± 6.9 HDL mg/dl 50.1 ± 3.4 48.9 ± 3.25 TGL mg/dl 135.6 ± 23.4 134.8 ± 14.3 TSL mg/dl 519 ± 35 572 ± 23 Glucose mg/dl 110.6 ± 5.8 109.3 ± 6.1 Creatinine mg/dl 1.0 ± 0.04 0.85 ± 0.03 Albumin g/dl 4.05 ± 0.10 4.03 ± 0.07 316 E. Rocchi et al. / European Journal of Internal Medicine 18 (2007) 314 320 Table 3 Plasma ascorbic acid, tocopherol (vitamin E) to total serum lipids ratio (E/TSL), retinol, alpha- and beta-carotene in the groups investigated Groups Non-smokers (n=28) Age years EGTA, gluthatione, ortho-phenylendiamine, nitrate reduc- tase, and NADH were obtained from Sigma (Milan, Italy). Ascorbate oxydase was obtained from Roche (Mannheim, West Germany). Cadmium granules were obtained from Aldrich (Milan, Italy). Echinenone (as an internal standard for carotenoids) was a generous gift from Hoffmann-La Roche (Basel, Switzerland). 2.4. Sample preparation All subjects, after an overnight fasting, underwent blood withdrawal and urine collection. The blood samples were analyzed for liposoluble vitamins (retinol, tocopherol), alpha- and beta-carotene, ascorbic acid, nitric oxide, Hcy, and other routine analytes. 2.5. Determination of vitamins and carotenoids Ascorbic acid in plasma was analyzed following the method of Speek et al. [23]. The ascorbic acid concentra- tion was calculated from peak heights with the standard working solution as a reference. Plasma vitamins and carotenoids, as well as the ratio between vitamin E and the sum of total serum lipids, usually expressed in mg/dl (E/ TSL), were assessed following a previously published method [24]. 2.6. Determination of nitric oxide NO was determined using Green's method, which relies on the measurement of nitrites and nitrates since they are obviously the most stable metabolites [25]. 2.7. Determination of urine 15-isoprostane F2t The determination of 15-isoprostane F2t (also known as 8- iso-prostaglandin F2± or 8-epi-prostaglandin F2± ) in urine was made according to a commercially available enzyme- linked immunoassay (ELISA; MED.DIA, S. Germano Vercellese, VL, Italy). Briefly, urine samples are mixed with an enhancing reagent that essentially eliminates interferences with 15-isoprostane F2t, conjugated to horse- radish peroxydase (HRP) for binding to a polyclonal antibody specific for 15-isoprostane F2t coated on the microplates. The HRP activity results in color development when substrate is added, with the intensity of the color proportional to the amount of 15-isoprostane F2t in the samples or standards [26 30]. 52±1.5 50 ± 1.7 Ascorbic acid mmol/L 41.2±3.1 E/TSL ratio 2.51±0.25 1.92N N N ± 0.11 Retinol mmol/L 3.08±0.19 2.76 ± 0.10 Alpha-carotene nmol/L 34.0±6.8 22.3N N ± 2.9 Beta-carotene nmol/L 173±29.1 105N N N ± 11.8 Smokers (n = 32) Mean±S.E. Statistical analysis (ANOVA): N N pb0.01, N N N pb0.001. 36.5N N ± 2.6 2.8. Determination of homocysteine Briefly, bound Hcy is reduced by dithiothreitol to free Hcy, which is then enzymatically converted to S-adenosyl-L- homocysteine (SAH) by SAH-hydrolase. Sample and tracer (SAH fluorescein) compete for a mouse monoclonal anti- SAH antibody, and the intensity of polarized light is mea- sured. The concentration of Hcy in plasma is inversely re- lated to the fluorescence polarization [31]. Hcy was measured both at baseline (fasting) and after an oral methionine load (PMLHcy; 6 h after 100 mg/kg methionine). 2.9. Laboratory and statistical analysis Other biochemical indices were determined by standard laboratory methods. Results are expressed as mean±S.E.M. Both inter-group comparisons and comparisons between smokers and non-smokers were performed by one-way ANOVA and linear coefficient correlation (r), employing an IBM PC computer (IBM Corp. Armonk, New York, USA). 3. Results The results are presented in Tables 1 3 and Figs. 1 3. No significant difference in dietary habits was found between the smoking and non-smoking groups (Table 1). A complete overview of the routine analyses performed is presented in Table 2. The biochemical results of the subjects studied, who were stratified according to smoking habit, overlapped quite a bit in both groups. With regard to the parameters expressing the biological antioxidant potential, we observed that plasma ascorbic acid, E/TSL ratio, and alpha- and beta-carotene were significantly lower in the smokers (Table 3). Moreover, the FMD by Fig. 1. Percentage of brachial artery flow-mediated dilatation after upper arm occlusion in the non-smoking and smoking groups (analysis of variance between the two groups). vascular reactivity was almost halved in the smoking group (Fig. 1), and the difference between groups was highly significant ( p b 0.001). In order to confirm the hypothesis of an EDNO-dependent impaired vascular response, we measured NO metabolites in blood and urine, obtaining higher values of NO in plasma and lower values in urine of non-smokers with respect to smokers, although without a clear significant difference (Fig. 2). Moreover, the increase in urinary NO in the smoking group seemed quantitatively related to the number of cigarettes smoked daily. The isoprostane levels were also higher in the urine of smokers (pb0.001), supporting the idea of a possible effect of free oxygen radicals in inducing arachidonic acid metabolites, such as isoprostane. Finally, Hcy, which was not signifi- cantly different between the two groups at baseline, showed a highly significant increase ( p b 0.001) after the methionine load (PMLHcy) in the smoking group (Fig. 2). When plasma PMLHcy was plotted against plasma NO levels, a highly Fig. 2. A) Urinary and serum nitric oxide (NO); B) urinary isoprostane; and C) plasma homocysteine (both base value: Hcy and post-methionine load; PMLHcy) in the non-smoking and smoking groups (analysis of variance between the two groups). Fig. 3. Different linear regression indices between plasma homocysteine post-methionine load (PMLHcy) and nitric oxide (NO), total serum lipids (TSL), and vitamin E to total serum lipid ratio (E/TSL), respectively. significant negative correlation was observed (Fig. 3). Similarly, a significant direct correlation between PMLHcy and TSL and a trend towards an inverse correlation between PMLHcy and E/TSL were observed (Fig. 3). Finally, in order to evaluate the relative role of LDL-cholesterol levels with respect to smoking habit, we stratified both the smokers and non-smokers according to normal or high LDL-cholesterol serum levels, obtaining a significant difference only in plasma NO, which was reduced in proportion to the LDL increase, but not in FMD (as percentage) or other significant parameters, within both the smoking and non-smoking groups. 4. Discussion We compared two groups that were significantly homogeneous with regard to the type of illness and life E. Rocchi et al. / European Journal of Internal Medicine 18 (2007) 314 320 317 318 E. Rocchi et al. / European Journal of Internal Medicine 18 (2007) 314 320 and alimentary habits, with smoking considered as the discriminating parameter. FMD, by vascular reactivity, was significantly reduced in smokers. Similar observations were made with regard to NO, levels of which were higher in non- smokers than in smokers. On the other hand, both the decrease in NO in plasma and the increase in plasma nitrites, that was observed in the smokers, seem to suggest that plasma NO is more quickly metabolized as a result of higher oxidative damage (greater production of reactive oxygen species capable of inactivating NO) in this last group, and this seems quantitatively related to the number of cigarettes smoked daily. We chose to assess isoprostane (15-isoprostane F2t, also known as 8-iso-prostaglandin F2±) in order to better evaluate the amount of oxidative damage, as it is the final and stable product formed as a result of the activation of an alternative pathway (i.e., different from the one leading to thromboxane and prostaglandin production) of arachidonic acid metabo- lism in the presence of reactive oxygen species [26]. Urinary 15-isoprostane F2t was significantly increased (Fig. 3) in smokers. A recent study, addressing this topic indeed, con- firmed its role as an independent risk marker of CHD [32]. With regard to Hcy, which is also considered an independent risk factor of vascular damage, we did not observe any significant difference in the baseline values of this parameter between the two groups; nevertheless, PMLHcy values were significantly higher in smokers, suggesting the existence of a significant alteration in Hcy metabolism in these subjects (Fig. 3). This observation is in agreement with the data concerning a linkage between smoking and derangement of Hcy metabolism [33,34] and that not only moderately high base levels of Hcy, but also high levels of PMLHcy alone, represent an independent risk factor for vascular damage [35]. It is likely that in smokers, where NO availability is reduced, this results in a failure of NO-induced detoxifica- tion of Hcy and of other Hcy metabolites with toxic (i.e., vasoconstrictive) effects [36,37]. This consideration is supported by the inverse and significant correlation we observed between PMLHcy and NO (Fig. 3). NO availability protects from endothelial damage induced by Hcy, as NO supplies a metabolic pathway (through the formation of S- nitroso-homocysteine) that is able to protect against Hcy- induced endothelial oxidative damage. This pathway may become exhausted rather quickly in the presence of ex- cessively high levels of Hcy or excessively low levels of NO, or both [16,38]. The parameters of oxidative damage (particularly urinary isoprostane and PMLHcy) showed significant and corresponding results. Thus, the present study confirms what is known from the literature, but what about the antioxidant capacity? In order to evaluate this issue, which has not been extensively explored in conjunction with the data presented, we assessed the most representative parameters of antioxidant activity, expressing the antioxidant power of both the intra- and extracellular lipid phase (carotenoids and liposoluble vitamins) and of the interstitial aqueous phase (vitamin C). Vitamin A levels seemed barely reduced in smokers, as expected, for it is known that plasma circulating vitamin A, which is bound to retinol-binding protein (RBP), is present in small amounts. In contrast, the vitamin A stored in liver, as palmitic ester (about 80% of the total amount of vitamin A in the body), increases [24]. Circulating vitamin A is probably more indicative of liver protein synthesis capacity (as RBP) than of vitamin A body content [39]. This is different for vitamin A precursors (alpha- and beta-carotene), as these compounds are not stored in tissues. Their plasma levels are good indicators of their real body availability. Our data show a significant decrease in vitamin A precursors in the smoking group, especially with regard to beta-carotene (Table 3). Also, vitamin C levels were reduced in smokers. It is well-known that this compound not only has important, direct antioxidant activity in the interstitial aqueous district [40] but it also exerts an important protective effect on vitamin E, probably the most important antioxidant in the lipid phase [24,33,41,42]. Vitamin E (tocopherol) levels were similar in the two groups; yet, the antioxidant effect of tocopherol is directly related to lipoproteins, by which it is carried, protecting them from oxidation [43]. For this reason, the circulating levels of vitamin E should be considered as referring to circulating lipid levels and expressed as E/TSL ratio. As shown in Table 3, the E/TSL ratio was lower in smokers than in non-smokers [44,45]. The observation of a significant correlation between PMLHcy levels and total lipids (Fig. 3) supports the results of the present study, concerning a significant linkage between different well- known cardiovascular risk factors (Hcy, LDL-cholesterol). NO consumption and, hence, endothelial dysfunction repre- sent, in our opinion, the final common mechanism of damage of all these factors, through free radical generation [36,46]. On the other hand, our findings also suggest a role for antioxidant compounds in the prevention of damage itself, particularly from the inverse trend of vitamin E (expressed as E/TSL ratio) with respect to Hcy levels (Fig. 3). Obviously, further cases could better highlight such a finding. However, the present study does, in fact, confirm the protective effects of both vitamin C and vitamin E against the oxidative damage promoted by Hcy [47,48]. As a practical conse- quence, all of these vitamin factors need to be evaluated in conjunction with the markers of oxidative damage. In summary, the present study, while confirming that the endothelial dysfunction, induced by chronic exposure to cig- arette smoking, is clearly due to oxidative damage, outlines as well the investigations that are of clinical value. This damage may, in fact, be assessed both directly by specific biochemi- cal markers (particularly urinary NO and isoprostane) and indirectly by Hcy levels after a methionine load. It is also associated with a reduction in antioxidant capacity (both hydro- and liposoluble factors). Both of these parameters should be measured in an accurate clinical evaluation in order to define the level of damage and possible strategies for its prevention. 5. Learning points " Thepreviouslyvalidatedultrasoundstudyofbrachialartery reactivity to ischemia appears to be a suitable method for studying endothelial nitric oxide (NO) availability. " Endothelial dysfunction (NO production impairment) has been confirmed in cigarette-smoking patients and in the presence of an increase in LDL-cholesterol. " Urinary isoprostane and urinary levels of homocysteine post-methionine load (PMLHcy) are proposed as useful markers of oxidative damage in clinical studies. " Both hydro- and liposoluble antioxidants are shown to be depleted in cigarette-smoking patients with coronary heart disease. Among them, the ratio between tocopherol (vitamin E) and the sum of total circulating lipids (E/TSL) may be chosen as a sensitive marker of endothelial oxidative damage in order to define preventive strategies in clinical practice. Acknowledgement This work was made possible by financial support from the Italian Ministry of University and Scientific and Technological Research (MURST), Grant no. 60 ROCC02, to the University of Modena and Reggio Emilia. ˇ˛Antibiotic resistance Geoff Scott In an environment containing vast numbers of micro-organisms saturated with or repeatedly exposed to antibiotics, Darwinian theory predicts the inevitable selection of resistant organisms. Fol- lowing the introduction of a new antibiotic, resistant strains may be reported within as little as 1 year, and after a latent interval of some years, resistance increases dramatically and the value of the antibiotic for empirical therapy is reduced. Staphylococcus aureus is now almost universally resistant to penicillin a phenomenon first observed in hospital outbreaks of surgical sepsis in the late 1940s. This organism has shown a remarkable facility to become resistant to every antibiotic intro- duced, even vancomycin. Reduced sensitivity to vancomycin (vancomycin-intermediate S. aureus) is associated with a thick peptidoglycan cell wall. The complex gene cassette for vancomycin resistance in Enterococcus spp. (vanA) can easily be transferred to S. aureus in vitro and has now been detected in methicillin- resistant S. aureus (MRSA), making the organism fully resistant to glycopeptides. It is only a matter of time before strains are seen that are sensitive to only a small number of novel antibiotics in development and perhaps some of the older antibiotics (e.g. tetracycline, chloramphenicol). In contrast, some organisms have not yet acquired resistance despite huge pressures; these include Streptococcus pyogenes, Gram-positive anaerobes such as Clostri- dium spp. and Peptostreptococcus spp. and Neisseria meningitidis (to penicillin). Almost all anaerobes were considered to be sensitive to metro- nidazole until recent reports from Spain, France and the USA suggested that resistant Bacteroides fragilis will soon become a serious problem. Ecology Resistance is driven by antibiotic use in human and agricultural/ veterinary practice. Resistance selection occurs by spontaneous mutations occuring at a rate of 10 9 10 6 driven by the presence of antimicrobials. Resistance elements appear in saprophytic bacteria as a result of antibiotic use for growth promotion and protecting crops. When these are eaten, resistance may be transferred to human bacteria that are occasionally pathogenic. The likelihood of finding resistant organisms in the gut is related to the tonnage of antibiotic use in the country in which the individual resides, Geoff Scott is Consultant in Clinical Microbiology at University College London Hospitals, London, UK. He qualified from the Royal Free Hospital, London, and trained at Northwick Park Hospital, London and University College Hospital. His interests include control of tuberculosis and resistant infections, and changing habits. and depends on the ease with which a resistance mechanism can arise. Novel resistance is usually detected in different places in the world at about the same time, implying that antibiotic pressures are similar worldwide and that bacteria have limited means of dealing with the problem of survival. Following the appearance of one mutant resistant progeny, rapidly replicating bacteria can recolonize carrier sites in less than 24 hours. A resistant mutant of Mycobacterium tuberculosis, which replicates once every 24 hours, can recolonize diseased lung within 2 weeks. Once resistance has been selected, organisms transferred from person to person continue to be resistant. When the antibiotic pressure is removed, novel colonizing and infecting flora tend to revert to the sensitive phenotype. Some bacteria containing large resistance plasmids are considered unfit and seem to disappear from the clinical environment more easily than others. Resistance and choice of antibiotics The term antibiotic resistance implies that a particular antibiotic is ineffective in a clinical infection. This may be because the organism is inherently resistant to the antibiotic or because it is inacces- sible. In vitro, resistance is defined by measuring the minimum inhibitory concentration (MIC) and the minimum bactericidal concentration (MBC) of the antibiotic against an organism, under ideal laboratory conditions, using appropriate controls to define the cut-off points between resistant , intermediate and sensitive . Methods of testing in the laboratory are problematic, however, and any result that does not correlate with clinical experience should be challenged. Low MIC and high MBC imply that an antibiotic is bacteristatic; that is, it inhibits the growth of an organism but is unable to kill it. Even with bactericidal antibiotics, killing in vivo normally requires an intact immune system. Thus, though bacteristatic antibiotics are ineffective in neutropenic sepsis, bactericidal antibiotics may suppress infection only for it to recrudesce when the antibiotic is removed. This observation governs strategies for the management of such infections. Paradoxically, even when an organism is declared resistant to an antibiotic by in vitro tests, it may appear to be effective in clinical use; this may be because MIC/MBC can be exceeded in vivo by giving a sufficiently large dose. This is often seen in lower urinary tract infection and in the treatment of, for example, non-meningeal penicillin-resistant pneumococcal infection with high-dose penicillin. Empirical treatment: management of an infection before bacteri- ology results are available depends on making the correct diagnosis and assessing the antibiotic sensitivities of the suspected organ- ism or organisms from local epidemiological knowledge. Once culture results are known, it is usually another day before in vitro antibiotic sensitivities are available. This is a critical period in the management of an infected patient and explains why, in severe infections, broad-spectrum antibiotics are often chosen initially when the definitive diagnosis is uncertain " and why withholding treatment may be life-threatening. This period is generally more protracted when an organism such as M. tuberculosis is slow to grow in vitro. Delay in the treatment of tuberculous meningitis may lead to permanent neurological damage. Whether the correct antibiotics were chosen initially is known only weeks later. ANTIBIOTICS MEDICINE 33:3 47 © 2005 The Medicine Publishing Company Ltd Mechanisms of resistance Resistance may be inherent (e.g. vancomycin against Gram- negative organisms, nitrofurantoin against Proteus spp.) or acquired via genetic elements encoding three fundamental mechanisms: " production of inactivating enzymes " change in the target site " exclusion of the antibiotic from the target site. The latter may occur by restriction of access through porins (only in Gram-negatives) or by active excretion. Antibiotic-inactivating enzymes may be produced in vast excess, surrounding the organism (e.g. ≤-lactamase from S. aureus), or in limited amounts in the periplasmic space of Gram-negatives. The effect in vivo is similar, but the latter organisms may appear sensi- tive in vitro. Classically, this is seen in Enterobacter spp. resistant to extended-spectrum ≤-lactam antibiotics such as piperacillin and cefotaxime; exposure of the organism to inducers such as penicillin, clavulanate, certain cephalosporins and imipenem may switch on the production of chromosomal (AmpC) ≤-lactamases. The classical TEM (Escherichia coli) and SHV (Klebsiella spp.) ≤- lactamases (which inactivate ampicillin) have shown a remarkable ability to mutate to extended-spectrum forms (which inactivate cefoxitin and cefotetan) and to inhibitor resistance (not inhibited by clavulanate or tazobactam). CTX-M cefotaximase arose by escape from the chromosome of Kluyvera spp. and is now widespread in Enterobacteriaceae including E. coli. Other classical inactivating enzymes include chloramphenicol acetyltransferase and aminoglycoside-modifying enzymes. Target site change may be a structural alteration preventing binding of an antibiotic, or a mechanism whereby the metabolic pathway that is normally inhibited is bypassed by an alternative one. This is seen in sulphonamide resistance and in MRSA, which has acquired a novel penicillin-binding protein from Staphylococcus scuiri. Important target site changes include topoisomerases II and IV (quinolones), ≤ subunit of DNA-dependent RNA polymerase (rifamycins) and methylation of 23S target (14/15-membered macrolides). Exclusion of antibiotic: porins are protein structures embedded in the outer bilipid membrane of Gram-negative organisms. They control what passes into and out of the cell on the basis of molec- ular size and charge. Mutations in the genetic elements encoding porins (permeability mutations) may exclude one antibiotic or, more commonly, multiple antibiotics. Pseudomonas aeruginosa has two outer lipid membranes, produces ≤-lactamases constitutively, and therefore tends to be more resistant than coliforms. Alternatively, organisms may actively excrete antibiotics, notably tetracyclines, macrolides and quinolones. Efflux pumps may be up-regulated. S. aureus resistant to erythromycin by active excretion remains sensitive to clindamycin, but the more common target site mutation affects susceptibility to both antibiotics. Acquisition of resistance Direct mutation of chromosomal genes may lead to resistance, and non-fatal mutations are passed to all progeny. Alternatively, small, mobile, circular pieces of DNA termed plasmids may be passed from one bacterium to another (even to bacteria of different species) by various mechanisms such as direct transfer by type II pili (Figure 1) and phage transfer. Plasmids reproduce each time the organism divides and can probably be lost as easily as gained, given the correct environment. Transposons are large genetic elements often containing mul- tiple genes necessary to confer phenotypic resistance. They may encode pheromones, which attract bacteria to each other and can be transferred on plasmids, generally being incorporated into the chromosomal DNA of the new host. Free DNA from dead bacteria and even mammalian cells can be incorporated into bacterial chromosomes. A remarkable mechanism of resistance is seen in penicillin-resistant Streptococcus pneumoniae, which have incor- porated a mosaic of genes into the chromosome from other resis- tant ±-haemolytic streptococci. Problematic resistant bacteria Gram-positive organisms MRSA has replaced methicillin-sensitive S. aureus as the more common cause of hospital-associated sepsis in most parts of the world. The targets of the ≤-lactam antibiotics are penicillin-binding proteins, carboxypeptidases and transpeptidases, which catalyse bridging of pentapeptide subunits of peptidoglycan. All strains of MRSA already produce penicillinase, but they now have a new penicillin-binding protein (PBP2') that is not inhibited by methi- cillin and its congeners oxacillin and flucloxacillin. They are resist- ant to all ≤-lactam antibiotics. The mecA gene encodes PBP2', but requires expression of several ancillary genes within a transposon for full expression. Thus, some strains with mecA may appear to be sensitive to methicillin in vitro. In addition, a plasmid encodes resistance to a variable number of other antibiotics. Strains with a tendency to spread easily and to predominate in hospitals are termed epidemic MRSA (EMRSA). The current epidemic of EMRSA15 or EMRSA16 in the UK started in 1994. In the laboratory, these are detected by their antibiogram, but other typing methods (e.g. pulsed-field gel electrophoresis of chromo- somal DNA) are needed to show that two strains are indistinguish- able, thus implying that cross-infection has occurred. When a few patients in one or two wards acquire a novel strain, such an outbreak can easily be tracked and then controlled. When EMRSA strains become endemic (Figure 2), more general measures are needed to reduce the risk to patients admitted to the hospital. MRSA may become endemic in nursing homes, creating a pool for novel introduction into hospital; transfer of patients from ward to ward and colonization of staff leads to continued exposure of patients to new strains and helps maintain endemicity. Infection with MRSA is not untreatable. Strains are often (though not predictably) susceptible to gentamicin or other aminoglycosides, rifampicin, co-trimoxazole, chloramphenicol or ciprofloxacin, and sometimes to tetracyclines, macrolides, fusidic acid and pseudomonic acid. (Note that rifampicin and fusidic acid should never be used alone because resistant mutants are selected very rapidly.) MRSA is almost always susceptible to glycopeptides, though strains with reduced sensitivity to vancomycin have been occasionally described in patients with chronic colonization or infection who have been treated for several weeks.1 Some rare strains of S. aureus are dependent on vancomycin to allow them to grow. MEDICINE 33:3 48 © 2005 The Medicine Publishing Company Ltd ANTIBIOTICS Penicillin-resistant S. pneumoniae strains have shown a gradual phase-shift increase in MIC to penicillin over several years. MIC is about 0.001 mg/litre in sensitive strains and 1 mg/litre in resistant strains. This is achieved by changes in penicillin-binding proteins with lower affinities for ≤-lactams. In the UK, such strains currently represent about 4% of those causing invasive infection, but in some areas (e.g. Spain) the rate is as high as 50%. These strains are usually resistant to many other useful antibiotics, including cephalosporins, chloramphenicol and erythromycin. Treatment with high-dose penicillin is effective in pneumonia but not in meningitis caused by resistant strains, some of which are sensitive to second-generation and third-generation cephalo- sporins. The mortality from severe pneumococcal pneumonia remains relatively constant regardless of whether the strain is resistant to penicillin, but meningitis caused by a resistant strain is more likely to be fatal. Splenectomized patients are advised to take life-long low-dose oral penicillin prophylaxis against the rare possibility of over- whelming pneumococcal sepsis. In the future, such prophylaxis may become less effective, but it is difficult to identify an alterna- tive simple and safe regimen. Glycopeptide-resistant enterococci: enterococci have become important causes of nosocomial postoperative infection in the USA. They are likely to cause sepsis and pneumonia only in severely ill patients in the UK, and occasionally strains are resistant to vanco- mycin with (vanA) or without (vanB) teicoplanin. These genes require the action of complex accessory genes for full expression. vanA encodes a structural change in the terminal amino acid of the pentapeptide chain of peptidoglycan, from D-ala to D-lac. This substitution prevents binding of vancomycin, enabling construction of the dipeptide bridges in peptidoglycan to continue. Typing indicates wide heterogeneity of strains, and that carriers usually harbour more than one strain. This implies that the trans- poson encoding vanA and the accessory genes is promiscuous and easily able to enter the host s enterococci. These organisms are of low virulence. Infections may respond to simple antibiotics such as amoxicillin if they appear sensitive (Enterococcus faecalis), but are generally resistant to all but a few new antibiotics. Colonization with enterococci is driven by cephalosporins and fluoroquinolones, to which enterococci are constitutively resistant. Gram-negative organisms In contrast to resistant Gram-positives, against which old anti- biotics or antibiotics in development are occasionally active, some Gram-negative organisms (particularly non-glucose fermenters such as Pseudomonas, Stenotrophomonas and Acinetobacter spp.), especially in ICUs, are resistant to every useful antibiotic and there are no new agents in development. Glucose fermenters: in terms of resistance and endemic/epidemic problems in ill, hospitalized patients, Klebsiella, Enterobacter and Serratia spp. are more troublesome than E. coli. They produce SHV-type ≤-lactamases constitutively, and mutations lead to resist- ance to all ≤-lactams other than the carbapenems (or occasionally aztreonam). They often lose susceptibility to quinolones and ANTIBIOTICS ab cd 1 DNA can be transferred a on direct contact, b,cviatypeIIpilior d by phages (bacterial viruses). (b and c by courtesy of Zeneca.) MEDICINE 33:3 49 © 2005 The Medicine Publishing Company Ltd aminoglycosides and become essentially untreatable. Some strains of Salmonella enterica (particularly typhimurium), and the enteric salmonellae (e.g. S. typhi) have acquired stable broad resistance to all useful antibiotics. These are now pandemic. Non-glucose fermenters: P. aeruginosa has been replaced as a troublesome cause of nosocomial infection by other environ- mental and skin bacteria such as Acinetobacter baumanii var. calcoaceticus. Some strains are resistant to all available antibiotics, others are sensitive only to carbapenems and some aminoglyco- sides (e.g. amikacin). In response to long-term antibiotic use over many years, Burkholderia cepacia tends to replace S. aureus and P. aeruginosa as colonizing flora in patients with cystic fibrosis. It may also show resistance to many antibiotics and can cause troublesome cross-infection. Neisseria spp.: N. gonorrhoeae is interesting in that different strains may exhibit the three mechanisms of resistance to penicil- lin (reduced permeability, changes in penicillin-binding proteins and ≤-lactamase production). Each confers a stepwise increase in MIC, and there is no clear cut-off between sensitive and resistant strains. Many strains have become resistant to other oral drugs such as ciprofloxacin, though these are currently rare in the UK. N. meningitidis has been slow to acquire resistance to penicillin, though a few strains isolated recently in Spain have slightly higher MICs than predicted. The value of sulphonamides was largely lost by the 1970s, and though chloramphenicol remains useful, second- generation cephalosporins (cefotaxime or ceftriaxone) seem to give the best results in clinical treatment. M. tuberculosis: resistance to first-line antituberculosis agents has been a major problem in the re-emergence of tuberculosis. In some countries (e.g. the states of the former USSR), multi-drug- resistant strains (resistant at least to rifampicin and isoniazid) are extremely common, and more so in patients who have been treated previously. In the UK, lone resistance to isoniazid occurs in 5% (except in London, where there is currently an epidemic of one clone resistant to isoniazid that started in 1995 and has led to a resistance rate of 15% in affected areas), and to rifampicin, ethambutol and pyrazinamide in 1% or fewer. Multi-drug-resistant strains account for 1.2%. Strains resistant to all are found through- out the country, but are more commonly seen in London. Most patients can be treated satisfactorily with a standard 6-month regimen, but the results of routine antibiotic sensitivity tests are unknown for 5"16 weeks after sending specimens to the laboratory, and patients with resistant strains need more toxic second-line drugs and prolonged courses of treatment. Rapid liquid culture and detection of genes encoding resistance may improve the speed of laboratory diagnosis. Strategies for reducing the impact of resistance Antibiotic resistance is driven by antibiotic use. When antibiotics are superseded and therefore used less, strains resistant to these tend to disappear. In the community: in the UK, more than 80% of human use of antibiotics occurs in the community, mostly for respiratory tract infections. The Standing Medical Advisory Committee, in its report The path of least resistance, made recommendations to reduce inappropriate prescribing.2 " No antibiotics should be given for simple coughs and colds. " Antibiotics should not be routinely prescribed for sore throats, unless there is evidence of streptococcal infection. " Antibiotics are not routinely required for acute otitis media and sinusitis-like symptoms; if given, courses can be limited to 3 days. In addition, 3 days treatment should suffice in otherwise healthy women with uncomplicated cystitis. This strategy has been useful in reducing prescribing by GPs, but anxiety has resulted from anecdotal reports of an increase in bacterial respiratory infections. In hospital, antibiotic use can be reduced by several means. New patients with methicillin-resistant Staphylococcus aureus at University College London Hospitals, 1991 1999 450 400 350 300 250 200 150 100 50 0 Infections Carriers ANTIBIOTICS 1991 1992 1993 1994 1995 1996 1997 1998 1999 Year MEDICINE 33:3 50 © 2005 The Medicine Publishing Company Ltd 2 Annual incidence ANTIBIOTICS Sites of action of antibiotics and some resistance mechanisms Glycopeptide Glycopeptide too large Aminoglycosides ≤-lactams Mutation in porin Porin Aminoglycosides Macrolides Chloramphenicol RNA Gram- Gram-positive negative Peptidoglycan Supercoiled DNA Topoisomerases Quinolones Folate synthesis Sulphonamides Trimethoprim Substrate change Mutation ≤-lactams Periplasmic space (≤-lactamase activity) DNA ≤-lactamases Chloramphenicol acteyltransferase Aminoglycoside-modifying enzymes Inactivating enzymes Active excretion Macrolides Quinolones Tetracyclines " Routine use of antibiotics for surgical prophylaxis should be reduced to a minimum. " Antibiotics should not be started immediately in all suspected infections. Certain patients (e.g. those with febrile neutropenia or evidence of septicaemia) require urgent antibiotic therapy, but in many other cases (e.g. mild pyrexia postoperatively), it is safe and ultimately preferable to withhold antibiotics until culture results are known or there is clear evidence of bacterial infection. " An alternative is to discontinue antibiotics as soon as infor- mation is available suggesting that the problem is not bacterial or has resolved itself. This is termed an antibiotic-stop policy, and the aim is to encourage doctors to actively review the need for antibiotics after, say, the second day, given information on cultures and surrogate markers that has by then become available. " Certain antibiotics can be withheld from the hospital formulary. This is the main benefit of an agreed antibiotic policy. The choice of restricted antibiotics may be decided on the basis of cost rather than likely selection of resistance. (These antibiotics must be used occasionally, however, when resistance to all other available drugs has been selected.) " In theory, antibiotics can be rotated such that, for example, predominantly penicillins are used at some times, and cephalo- sporins or quinolones at others. There is little clear scientific evidence that this has any effect, and major, complicated studies would be needed to determine the effect of change in use on both normal and infecting flora. However, it has been shown that, in a setting of heavy cephalosporin use, discontinuation of use of this class of drugs leads to a reduction in the risk of colonization with glycopeptide-resistant enterococci and Clostridium difficile- associated diarrhoea. Other strategies: development of resistance in human pathogens might be delayed if antibiotics were not used so widely in animal husbandry, particularly for growth promotion. The future Resistance to antibiotics is one of the greatest threats to the success of modern medicine. It has recently become more serious because we can no longer be sure that any antibiotic chosen empirically will work, and because of the emergence of totally resistant bac- teria. How to reduce resistance without simply discontinuing use of all antibiotics is a dilemma. We do not know to what degree antibiotic use must be reduced to decrease the selective pressure and allow reversion to sensitivity, nor do we know how to protect the few remaining drugs that can be used to treat resistant infec- tions. Doctors and vets should overcome the view that they have an inalienable right to prescribe empirical antibiotics, and should set targets for reduction of their personal prescribing. ˇ˛Antibiotics for acute otitis media: a meta-analysis with individual patient data Maroeska M Rovers, Paul Glasziou, Cees L Appelman, Peter Burke, David P McCormick, Roger A Damoiseaux, Isabelle Gaboury, Paul Little, Arno W Hoes Summary Background Individual trials to test effectiveness of antibiotics in children with acute otitis media have been too small for valid subgroup analyses. We aimed to identify subgroups of children who would and would not benefit more than others from treatment with antibiotics. Methods We did a meta-analysis of data from six randomised trials of the effects of antibiotics in children with acute otitis media. Individual patient data from 1643 children aged from 6 months to 12 years were validated and re-analysed. We defined the primary outcome as an extended course of acute otitis media, consisting of pain, fever, or both at 3 7 days. Findings Significant effect modifications were noted for otorrhoea, and for age and bilateral acute otitis media. In children younger than 2 years of age with bilateral acute otitis media, 55% of controls and 30% on antibiotics still had pain, fever, or both at 3 7 days, with a rate difference between these groups of "25% (95% CI "36% to "14%), resulting in a number-needed-to-treat (NNT) of four children. We identified no significant differences for age alone. In children with otorrhoea the rate difference and NNT, respectively, were "36% ("53% to "19%) and three, whereas in children without otorrhoea the equivalent values were "14% ("23% to "5%) and eight. Interpretation Antibiotics seem to be most beneficial in children younger than 2 years of age with bilateral acute otitis media, and in children with both acute otitis media and otorrhoea. For most other children with mild disease an observational policy seems justified. Lancet 2006; 368: 1429 35 See Comment page 1397 Julius Centre for Health Sciences and Primary Care, University Medical Centre Utrecht, the Netherlands (M M Rovers PhD, C L Appelman MD, R A Damoiseaux MD, Prof A W Hoes MD); Departments of Paediatrics and Otolarynglogy, Wilhelmina Children s Hospital, University Medical Centre Utrecht, the Netherlands (M M Rovers); University of Oxford, Department of Primary Health Care, Institute of Health Sciences, Oxford, UK (Prof P Glasziou MD, P Burke FRCGP); Department of Pediatrics, University of Texas Medical Branch Galveston, Texas, USA (Prof D P McCormick MD); Chalmers Research Group, Children s Hospital of Eastern Ontario Research Institute, Ottawa, Ontario, Canada (I Gaboury); Primary Medical Care, Community Clinical Sciences Division, University of Southampton, Aldermoor Health Centre, Southampton, UK (Prof P Little FRCGP) Correspondence to: Dr Maroeska M Rovers Julius Centre for Health Sciences and Primary Care, Stratenum 7∑109, PO Box 85060, 3508 AB Utrecht, the Netherlands M.Rovers@umcutrecht.nl Introduction Acute otitis media is one of the most common childhood infections, the leading cause of doctors consultations, and the most frequent reason for children to take antibiotics.1 Evidence from systematic reviews, however, suggests that antibiotics provide only marginal benefit.2,3 Furthermore, prescribing antibiotics is known to encourage clinic visits for subsequent episodes, intensify pressure on clinicians to prescribe, increase antibiotic use, and promote antibiotic resistance.4 6 Guidelines therefore recommend selective use of anti- biotics for acute otitis media, especially in children aged 2 years or older. In children younger than 2 years, no consensus has been reached. Some guidelines recommend antibiotics for all these children,7,8 whereas others advise antibiotics only for children under 2 years if they are severely affected or have persistent signs of disease or related comorbidity.9,10 Reliable identification of subgroups of children who do, and do not, benefit from treatment with antibiotics has not been straightforward, because individual trials have been too small for valid and reliable subgroup analyses. A meta-analysis of the individual data from original trials enables the opportunity to identify subgroups that are most likely to benefit. We therefore aimed to identify subgroups that might benefit most from such treatment. Methods Selection of trials We did a systematic search of the Cochrane library, PubMed database, EMBASE, and the proceedings of the international symposia on recent advances in otitis media. We selected trials that (1) used random allocation of children, (2) included children aged 0 12 years with acute otitis media, (3) compared antibiotics with placebo or no treatment, and (4) had pain and fever as an outcome. All trials were assessed for four major quality criteria: proper randomisation methods; degree of follow-up; and blinding of the outcome assessor, patient, and care giver. All trials obtained informed consent and ethics approval. The primary investigators of all selected trials were asked for the raw data of their trials. The data thus obtained were thoroughly checked for consistency, plausibility, integrity of randomisation, and follow-up. A few issues were queried with the responsible trial investigator or statistician, and all were resolved. Outcome variables The primary outcome was an extended course of acute otitis media, which was defined as pain, fever, or both at 3 7 days. We used this composite endpoint since both factors are relevant from clinical and patients (or parental) perspectives. Fever was defined as temperature of 38∞C or higher, and pain was assessed by parents and recorded in diary form (as either yes or no). Both outcome measures were dichotomised, since several trials measured them in this way. Fever and pain were also studied separately (as secondary outcomes). Additionally, the adverse effects of antibiotic treatment mentioned in every trial were analysed. Independent predictors of an extended course of disease had been established in an earlier study within www.thelancet.com Vol368 October21,2006 1429 Articles Articles the same setting (unpublished data).11 We used these independent baseline predictors ie, age (<2 vs e"2 years), fever (yes vs no), and bilateral acute otitis media (yes vs no) to investigate whether those at risk of an extended course had enhanced benefits from treatment with antibiotics. We also examined the effects of concurrent otorrhoea at baseline (yes vs no), both alone and in combination with the identified predictors, since this condition seems to be a clinically relevant outcome that occurs too infrequently to be identified as an independent predictor. Statistical analyses Information was available for 72% of the potential subgroups (range 28 100%) and for 90% of the outcome variables (range 81 98%). To reduce bias and to increase statistical efficiency, we imputed the missing data for all trials using the linear regression method (multivariate analyses) available in SPSS (version 12.0).11 Regression was based on the correlation between individual variables with missing values and all other variables, as estimated from the complete set of data. We imputed missing values only within trials. To decide whether pooling of data for analysis was justified, we assessed heterogeneity between studies using I2, which describes the percentage of variation between studies due to heterogeneity rather than chance.12 The range for I2 lies between between 0% (ie, no observed heterogeneity) and 100%. The resulting I2 was lower than 25% (p>0∑30) indicating that studies were sufficiently similar to justify pooling of data. We calculated relative risks (RR), rate differences (RD), and NNT, with their 95% CI, for both the primary and secondary outcomes. To assess whether the effect of antibiotics was modified by age, bilateral acute otitis media, fever, otorrhoea, or a combination of these factors, we did a fixed-effect logistic regression analysis. In this model, the independent variables were: treatment with antibiotics (yes vs no); the potential-effect modifiers (age, bilateral acute otitis media, fever, otorrhoea, or com- binations of these); and an interaction term (defined as use of antibiotics times potential-effect modifier). We also used a binary dummy variable to identify each study within the regression analysis. Dependent variables were an extended course (primary outcome), fever, and pain at 3 7 days (secondary outcomes). We calculated the c-index (area under the receiver operating curve) to measure the accuracy of each model. If a significant interaction effect was identified, we did stratified analyses of the rate ratios and rate differences within each stratum of the subgroups. The percentages of children with an extended course during each consecutive day within each of the identified subgroups were calculated for the five trials that asked parents to fill out diaries noting signs of the disease. Finally, we did sensitivity analyses, including only those trials that measured the outcomes on the same day, used the same dose regimen, or included placebo. All analyses were performed according to the intention-to-treat principle. Role of the funding source This study was sponsored by the Dutch College of General Practitioners and the Netherlands Organisation for Health Research and Development (grant number 4200.0010). This sponsor had no role in study design, Ref 22 Ref 24 Ref 26 Number of patients 121 240 512 Participants Children aged 6 months to 12 years visiting a GP with recurrent AOM Children aged 6 months to 2 years visiting a GP with AOM Children aged 6 months to 5 years presenting to clinics or the emergency department with AOM Interventions Amoxicillin with clavulanate vs placebo Amoxicillin vs placebo Amoxicillin vs placebo Duration of intervention 7 days 10 days 10 days Outcomes Fever after 3 days Pain after 3 days Otorrhoea Otoscopy and tympanometry after 1 month Symptoms at day 4 assessed by a GP (including fever and earpain) Otoscopy and tympanometry after 6 weeks and 3 months Telephone follow-up at day 1, 2, 3, and between 10 and 14 days (including fever) Tympanometry at 1 and 3 months Ref 23 232 Children aged 3 to 10 years with AOM Amoxicillin vs placebo 7 days Symptoms noted by parents (including fever and ear pain) Home visits by researcher after 24 h and 5 7 days Otoscopy and tympanometry after 1 and 3 months Ref 25 315 Children aged 6 months to 10 years visiting a GP with AOM Immediate antibiotics (amoxicillin) vs delayed treatment 7 days Symptoms noted by parents (including fever and earpain) Absence from school Consumption of paracetamol Ref 27 223 Children aged 6 months to 12 years with AOM Immediate antibiotics (amoxicillin) vs delayed treatment 10 days Symptoms noted by parents (including fever and earpain) Analgesic consumption Nasopharyngeal carriage Adverse events Absence from school Tympanometry after 12 and 30 days AOM=acute otitis media; GP=general practitioner. Table 1: Characteristics of the six trials included in our meta-analysis 1430 www.thelancet.com Vol368 October21,2006 Articles Age <2 years Recurrent AOM Winter season Passive smoking! Coughing! Ear pain Bilateral AOM** Perforation! ! Bulging tympanic membrane 287 (35%) 429 (52%) 620 (75%) 218 (33%) 476 (72%) 724 (88%) 220 (33%) 19 (7%) 342 (42%) 567 (35%) 831 (51%) 1243 (76%) 432 (34%) 936 (72%) 1447 (88%) 456 (34%) 39 (7%) 685 (42%) Antibiotics (n=819) 280 (34%) 402 (49%) 623 (76%) 214 (34%) 460 (72%) 723 (88%) 236 (35%) 20 (8%) 343 (42%) Controls Total (n=824) (n=1643) Male sex 411 (50%) 411 (50%) 822 (50%) Siblings* 455 (76%) 472 (78%) 927 (77%) Being breastfed 244 (64%) 255 (64%) 499 (64%) Cryingß 407 (83%) 413 (83%) 820 (83%) Runny nose∂ 428 (77%) 429 (78%) 857 (78%) Fever|| 282 (40%) 287 (41%) 569 (40%) Otorrhoea 51 (19%) 65 (23%) 116 (21%) Red tympanic membrane 751 (92%) 754 (92%) 1505 (92%) Data are number (%). AOM=acute otitis media. Overall number=*1207, 778, 1299, ß984, ∂1105, ||1411, **1328, ! ! 555. Percentages do not always add to 100% because of missing data for some characteristics. Table 2: Baseline characteristics of patients in the six trials data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Results Our search strategy identified nineteen trials that investigated the effectiveness of antibiotics in children with acute otitis media. After screening, nine trials were excluded, because randomisation was inadequate, the control group received another treatment, information about our selected outcomes was not available, or because they focused on special study populations, such as children with ventilation tubes.13 21 Of the ten eligible trials, six research groups provided us with their data22 27 and four did not.28 31 The methodological quality of the six remaining studies was generally high. Five used adequate concealed allocations (blinded randomisations) and outcome assessments. Loss to follow-up was less than 10%. Table 1 shows the main characteristics of the six trials. The mean age of the children was 3∑4 years (range 0 11 ); half were boys; about half had recurrent acute otitis media; and about a third had bilateral acute otitis media (table 2). Our meta-analysis showed that, relative to placebo, overall RR for an extended course of acute otitis media at 3 7 days with antibiotics was 0∑83 (95% CI 0∑78 0∑89). The rate difference between the control group and the antibiotics group was 13% (9 17), resulting in a NNT of eight children. Overall RR of fever at 3 7 days was 0∑95 (0∑92 0∑98); the rate difference was 5% (2 8) and NNT was 20 children. The corresponding figures for children who had pain at 3 7 days were 0∑86 (0∑81 0∑91); 11% (7 15); and ten children, respectively. Our analyses showed that the effect of antibiotics was modified by age and bilateral disease, and by otorrhoea, notably for the primary outcome of pain, fever, or both at 3 7 days (table 3). In children aged less than 2 years with bilateral acute otitis media, more than half the control group and less than a third of the antibiotics group still had pain, fever, or both at 3 7 days, with a rate difference of about 25%. In children aged 2 years or older with bilateral disease the rate difference was about 12%. For age alone no differences were identified. The c-indices, calculated to gauge the accuracy of each model, were 0∑63, 0∑58 and 0∑61, respectively, for age and bilaterality, age alone, and bilaterality alone. About 60% of children with otorrhoea in the control group had pain, fever, or both at 3 7 days, whereas only about 25% of those given antibiotics had protracted illnesses. The rate difference, of about 36%; was much greater than that for those without otorrhoea, which was about 14%. Other factors, in combination with otorrhoea, such as age, bilateral disease, or both did not substantially alter this pattern ie, children with otorrhoea seemed to benefit most from treatment with antibiotics, irrespective of other characteristics. With pain alone as the primary outcome, the effect of antibiotics was modified by age and bilateral disease together (p-value for interaction 0∑01) (table 3). For children aged less than 2 years with bilateral acute otitis media, twice as many controls still had pain at 3 7 days, compared with those given antibiotics. For age alone no differences were identified. Figure 1 shows the proportion of children with an extended course of disease in the subgroups for which antibiotics were of most benefit ie, children younger than 2 years of age with bilateral disease, and those with otorrhoea. For both these subgroups, symptoms resolved faster in children who received antibiotics than in children randomised to the control group, but this difference disappeared after 4 5 days. Sensitivity analyses, including only those trials that measured the outcome at the same time during follow-up, used the same dose of antibiotics, or included a placebo, were in agreement with the overall results. The most commonly described adverse effect of antibiotic treatment was diarrhoea, which ranged from 2% to 14% in controls and from 4% to 21% in those given antibiotics in each of the six trials that we analysed. Occurrence of rash ranged from 2% to 6% in the control groups, and from 1% to 8% in the antibiotic groups. One child from the control group developed meningitis at day 3,24 but seemed to have received antibiotics at day 2 because of deterioration. No mastoiditis or other serious complications were mentioned in these six trials. www.thelancet.com Vol368 October21,2006 1431 Articles Number (%) Group given antibiotics Control RD (95% CI) group (n=824) NNT RR (95% CI) p value for interaction* (n=819) Pain, fever, or both at 3 7 days Age <2 years 567 (35%) 91 (33%) 137 (48%) "15% ("23 to "7) 7 0∑77 (0∑68 0∑89) e"2 years 1076 (65%) 107 (20%) 166 (31%) "11% ("16 to "6) 10 0∑86 (0∑80 0∑93) 0∑83 Bilateral AOM No 872 (66%) 104 (24%) 132 (30%) "6% ("12 to 0) 17 0∑92 (0∑85 1∑00) Yes 456 (34%) 64 (27%) 104 (47%) "20% ("28 to "11) 5 0∑72 (0∑62 0∑84) 0∑021 Age and bilateral AOM <2 years+bilateral AOM 273 (20%) 42 (30%) 74 (55%) "25% ("36 to "14) 4 0∑64 (0∑62 0.80) <2 years+unilateral AOM 261 (20%) 45 (35%) 53 (40%) "5% ("17 to 7) 20 0∑92 (0∑76 1∑11) e"2 years+bilateral AOM 183 (14%) 20 (23%) 30 (35%) "12% ("25 to 1) 9 0∑84 (0∑70 1∑02) e"2 years+unilateral AOM 611 (46%) 59 (19% 79 (26%) "7% ("14 to 0) 15 0∑92 (0∑85 1∑01) 0∑022 Otorrhea Yes 116 (21%) 12 (24%) 39 (60%) "36% ("53 to "19%) 3 0∑52 (0∑37 0∑73) 0∑039 No 439 (89%) 61 (28%) 94 (42%) "14% ("23 to "5%) 8 0∑80 (0∑70 0∑92) Pain at 3 7 days Age, years < 2 years 567 (35%) 77 (28%) 115 (40%) "12% ("20 to -4%) 9 0.83 (0∑73 0∑93) e" 2 years 1076 (65%) 86 (16%) 142 (26%) "10% ("15 to -5%) 10 0.88 (0∑82 0∑93) 0∑76 Bilateral AOM No 872 (66%) 85 (20%) 102 (23%) "3% ("8 to -2%) 34 0.96 (0∑89 1∑03) Yes 456 (34%) 48 (20%) 88 (40%) "20% ("28 to -12%) 5 0.75 (0∑66 0∑85) 0∑005 Age and bilateral AOM < 2 years+bilateral AOM 273 (20%) 32 (23%) 62 (46%) "23% ("34 to -12%) 5 0.70 (0∑58 0∑84) < 2 years+unilateral AOM 261 (20%) 41 (31%) 42 (33%) "2% ("13 to 9%) 50 0.99 (0∑84 1∑17) e" 2years+bilateralAOM 183 (14%) 16 (17%) 26 (30%) "13% ("25 to 1%) 8 0.83 (0∑71 0∑99) e" 2years+unilateralAOM 611 (46%) 44 (15%) 59 (19%) "4% ("10 to 2%) 25 0.95 (0∑88 1∑02) 0∑009 AOM=acute otitis media. RD=rate difference. RR=rate ratio. NNT=number needed to treat. *p value for the interaction term (antibiotics x subgrouping variable) in the fixed effect regression analysis. Table 3: Subgroup analyses with both the rate differences and rate ratios 1432 www.thelancet.com Vol368 October21,2006 Discussion Our meta-analyses of individual patient data showed that antibiotics are more beneficial in children aged less than 2 years with bilateral acute otitis media, and in those with both acute otitis media and otorrhoea ie, in these groups three to four children have to be treated to prevent an extended course of the disease in one child. Although none of the trials included in this meta-analysis have had adequate power to produce precise effect estimates in clinically relevant subgroups, both McCormick27 and Appelman22 and their colleagues had suggested that children younger than 2 years might benefit most from antibiotics for otitis media. The results of our fixed-effect logistic regression analysis, however, showed that the effects of antibiotic treatment were not significantly modified by either age or bilateral disease alone. Additionally, the NNT was lower for the combined model than for individual components, indicating that targeting of both age and bilaterality would increase the benefits of antibiotic therapy. Moreover, the subgroups studied were based on a multivariate prognostic model, which showed that age and bilaterality were both independent predictors of an extended course of disease (unpublished data). Although we need to understand the causal mech- anism of the subgroups effects before final conclusions can be drawn, we can postulate that, in children aged less than 2 years with bilateral acute otitis media and in those with otorrhoea, the infection is more often bacterial than viral. Indeed, Palmu and co- workers32 have shown that culture-positive cases of acute otitis media are more often bilateral than are culture-negative events; middle ear effusion samples obtained through tympanic membranes with known pre-existing perforations were more likely to be culture- positive than were samples obtained through an intact membrane. Furthermore, perforations are more often caused by an infection with Streptococcus pneumoniae than with Haemophilus influenzae or Moraxella catar- rhalis.32 S pneumoniae is most common in young children.32 Figure 1: Proportion of children with an extended course of acute otitis media A: d"2 years with bilateral disease; B: e"2 years with unilateral disease; C: with otorrhoea; and D: without otorrhoea. The main strength of our study was that, by re- analysing the data of six trials, we were able to include 1643 children, which gave us the power to identify subgroups that could benefit most from treatment with antibiotics. Nevertheless, some of our findings deserve further discussion. First, only six of the ten eligible randomised, controlled trials could be included in our meta-analysis. The main characteristics of the four trials for which individual patient data were not available were, however, much the same as those in the six included trials. Moreover, the overall results of our subset of six trials are very similar to the overall results reported by the Cochrane review3 that did include all trials. A funnel plot of the included studies (data not shown) also indicated that publication bias was unlikely. Second, we could not do a pooled analysis with respect to failure rates since these rates were defined and measured differently in each of the six included trials. We did, however, undertake subgroup analyses of failure rate within each trial, and subsequently pooled these results for the six trials. The results were in accord with the pooled results for the subgroups ie, the largest effect of antibiotics was in children aged younger than 2 years with bilateral acute otitis media (rate difference "8%, 95% CI "17% to 0%), and the smallest effect was in children aged 2 years or older with unilateral acute otitis media (rate difference "3%, "7% to 1%). Third, the severity of the pain was estimated by parents and not further quantified in the trials, which could have resulted in an incorrect estimation of the real pain. Analysis with fever alone, however, showed much the same trend. Moreover, the fact that in many children the complaints at days 3 7 were mild should be taken into account in interpretation of reported NNTs and in the decision to initiate antibiotic therapy in individual patients. Fourth, the results are based on child participants, who might not be representative of those visiting general practitioners. For example, the most severely affected children might be under-represented. However, because we had access to raw data from six trials, we had high numbers of children from specific high-risk groups, which are often under-represented in single trials. Furthermore, the children we included seem represen- tative of those with acute otitis media visiting general practitioners, since the percentages of those aged less than 2 years and 2 years or older were much the same as those from a national survey in Netherlands of children with acute otitis media in primary care (ie, 35% vs 33%, and 65% vs 67%, respectively).33 Fifth, the rate of mastoiditis was so low that we could not obtain a precise estimate for risk of this complication. The trials done so far, however, showed that initially withholding antibiotics from children with acute otitis media does not increase suppurative complications. Whether restrictive antibiotic use increases acute mastoiditis at the population level remains unresolved, but the potential increase is only two cases per 100 000 person-years and should be weighed against potential adverse effects.1 Sixth, since not all trials used the most objective diagnostic methods (eg, pneumatic otoscopy or tympanometry) some children in our meta-analysis might not have had ear infections. Sensitivity analyses www.thelancet.com Vol368 October21,2006 1433 Articles 100 90 80 70 60 50 40 30 20 10 0 100 90 80 70 60 50 40 30 20 10 0 AC Control group Antibiotic group BD 0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10 Days of follow-up Proportion of children (%) 1434 www.thelancet.com Vol368 October21,2006 Articles with the three trials that did use these diagnostic methods were, however, in accord with the overall results. Seventh, we did not study all possible subgroups. We selected established predictors of an extended course of disease (unpublished data) and some clinically relevant variables, and did stratified analyses only for those variables that showed a significant p value for the interaction in the fixed regression model. We might therefore have missed a subgroup. Our approach is, however, in agreement with recommendations for study of subgroups.34 The strength of this approach is that our prognostic analyses revealed only a few relevant sub- groups, limiting the number of subgroup analyses and subsequent false-positive findings (type I error) that could be caused by multiple testing. Furthermore, other subgroups that might benefit more from treatment with antibiotics (eg, children with Down syndrome or cleft palate) could not be studied in this meta-analysis of individual patient data, because these subgroups were excluded in the individual trials. The experience of many clinicians that these subgroups of children benefit more from treatment with antibiotics has not yet been evidenced in randomised controlled trials. Eighth, we did not adjust for potential confounding due to differences between trials. We did, however, examine whether such confounding had occurred in our study, and noted that children aged less than 2 years were most likely to have fever and an abnormal tympanic membrane at baseline. We therefore used the Mantel Haenszel technique to adjust for these potential confounders in our subgroup analyses. Since the effect estimates were not altered by adjustments, crude effect estimates are presented. We conclude that antibiotics are beneficial in relieving residual pain or fever at 3 7 days in children younger than 2 years of age with bilateral acute otitis media, and in children with acute otitis media and otorrhoea. For most other children with mild disease an observational policy seems justified. Contributors M M Rovers designed and planned the study, and gathered, analysed, and interpreted the data. P Glasziou and A W Hoes contributed to the initial idea and design of the study, interpreted the data, and super- vised the study. C L Appelman, P Burke, D McCormick, R A Damoiseaux, I Gaboury, and P Little provided the data of the original trials, contributed to the protocol, and interpreted the data. The manuscript was prepared by M M Rovers, and all authors have seen and approved the final version. Conflict of interest statement We declare that we have no conflict of interest. Acknowledgments We thank Anne Schilder for commenting on an earlier version of this paper. CLINICAL RESEARCH STUDY Anticoagulation-associated Adverse Drug Events Gregory Piazza, MD,a Thanh Nha Nguyen, PharmD,b Deborah Cios, PharmD,c Matthew Labreche, PharmD,b Benjamin Hohlfelder,b John Fanikos, RPh, MBA,c Karen Fiumara, PharmD,d Samuel Z. Goldhaber, MDa a Cardiovascular Division, Department of Medicine, Brigham and Women‚Äôs Hospital, Harvard Medical School, Boston, Mass; bVenous Thromboembolism Research Group, cDepartment of Pharmacy, and dCenter for Clinical Excellence, Brigham and Women‚Äôs Hospital, Boston, Mass. ABSTRACT PURPOSE: Anticoagulant drugs are among the most common medications that cause adverse drug events (ADEs) in hospitalized patients. We performed a 5-year retrospective study at Brigham and Women‚Äôs Hospital to determine clinical characteristics, types, root causes, and outcomes of anticoagulant-associated ADEs. METHODS: We reviewed all inpatient anticoagulant-associated ADEs, including adverse drug reactions (ADRs) and medication errors, reported at Brigham and Women‚Äôs Hospital through the Safety Reporting System from May 2004 to May 2009. We also collected data about the cost associated with hospitalizations in which ADRs occurred. RESULTS: Of 463 anticoagulant-associated ADEs, 226 were medication errors (48.8%), 141 were ADRs (30.5%), and 96 (20.7%) involved both a medication error and ADR. Seventy percent of anticoagulantassociated ADEs were potentially preventable. Transcription errors (48%) were the most frequent root cause of anticoagulant-associated medication errors, while medication errors (40%) were a common root cause of anticoagulant-associated ADRs. Death within 30 days of anticoagulant-associated ADEs occurred in 11% of patients. After an anticoagulant-associated ADR, most hospitalization expenditures were attributable to nursing costs (mean $33,189 per ADR), followed by pharmacy costs (mean $7451 per ADR). CONCLUSION: Most anticoagulant-associated ADEs among inpatients result from medication errors and are, therefore, potentially preventable. We observed an elevated 30-day mortality rate among patients who suffered an anticoagulant-associated ADE and high hospitalization costs following ADRs. Further quality improvement efforts to reduce anticoagulant-associated medication errors are warranted to improve patient safety and decrease health care expenditures. ¬© 2011 Elsevier Inc. All rights reserved. ‚Ä¢ The American Journal of Medicine (2011) 124, 1136-1142 KEYWORDS: Adverse drug events; Adverse drug reactions; Anticoagulation; Medication errors Adverse drug events (ADEs), which comprise medication errors and adverse drug reactions (ADRs), represent a major source of harm among hospitalized patients and have been a driving force behind implementation of electronic health Funding: Dr Piazza is supported by a Research Career Development Award (K12 HL083786) from the National Heart, Lung, and Blood Institute (NHLBI). This study was funded, in part, by a clinical research grant from Johnson & Johnson (New Brunswick, NJ). Conflict of Interest: None. Authorship: All authors had access to the data and played a role in writing this manuscript. Requests for reprints should be addressed to Gregory Piazza, MD, Cardiovascular Division, Brigham and Women‚Äôs Hospital, 75 Francis St., Boston, MA 02115. E-mail address: gpiazza@partners.org 0002-9343/$ -see front matter ¬© 2011 Elsevier Inc. All rights reserved. doi:10.1016/j.amjmed.2011.06.009 records, including computerized provider order entry.1-6 One study reported that 6.5 medication-related adverse events occurred per 100 hospitalizations and estimated that more than one quarter of these events were the result of a medication error.7 Medication errors are potentially preventable causes of ADRs that can occur at all stages of the medication process, including prescription, transcription, dispensing, and administration.8-12 Compared with nonpreventable events, potentially preventable ADEs have been shown to double the incremental additional length of stay and health care costs attributable to the event.13 Anticoagulant drugs, including warfarin, unfractionated heparin, and low-molecular-weight heparin, are among the most commonly implicated medications that cause ADEs in Piazza et al Anticoagulation-associated Adverse Drug Events 1137 was noxious and unintended.17 A medication error was hospitalized patients.1,4,6,11,14 Despite implementation of computerized provider order entry, electronic medication defined as an event that caused or led to inappropriate administration records, and improved infusion pump techmedication use or patient harm.17 A medication error that nology (‚Äúsmart pumps‚Äù), medication errors involving antiwas discovered and corrected before reaching the patient coagulant medications remain common.11,12 Elderly15,16 was classified as a near miss. All isolated medication errors and cardiac patients14 represent and ADRs that resulted from medpopulations at particularly high ication errors were considered porisk for suffering anticoagulanttentially preventable. CLINICAL SIGNIFICANCE associated ADRs. We searched the Safety ReTo determine the clinical charporting System using Risk Mon‚óè Most anticoagulant-associated adverse acteristics, types, severity, root itor Pro (rL Solutions, Inc., drug events (70%) are potentially causes, and outcomes of anticoagCambridge, Mass). We evalupreventable. ulant-associated ADEs, we perated the patient characteristics, ‚óè Transcription errors comprise the most formed a 5-year retrospective ADE type and severity, root frequent root cause of anticoagulantstudy of the Safety Reporting Syscause, and outcomes of all antitem at Brigham and Women‚Äôs coagulation-associated medicaassociated medication errors, and medHospital. We reviewed discrete tion errors and ADRs. All patient ication errors are a common root cause ADEs that originated during hosrecords were reviewed for eviof anticoagulant-associated adverse dence of treatment with an antipitalization at Brigham and Womdrug reactions. coagulant medication. Hemoren‚Äôs Hospital. We also conducted ‚óè After anticoagulant-associated adverse rhagic events were classified an analysis of the cost associated drug reactions, most hospitalization exwith hospitalizations in which according to the Global Use of penditures are attributable to nursing ADRs occurred. Strategies to Open Occluded Coronary Arteries (GUSTO) criand pharmacy costs. teria for severe or life-threatenMETHODS ‚óè Efforts to minimize anticoagulant-assoing, moderate, or mild bleedciated medication errors are warranted ing.18 We obtained 30-day Patient Population to improve patient safety and decrease follow-up for 100% of patients Brigham and Women‚Äôs Hospital is hospitalization costs. included in the registry. a 793-bed acute tertiary care facilWe included all anticoagulant ity providing medical and surgical medications administered for procare for patients with general phylaxis and treatment of thrommedical, cardiothoracic, orthopedic, oncologic, neurologic, boembolic events. ADEs associated with the use of the obstetric and gynecologic, neonatal, urologic, and gastroinfollowing anticoagulants were reviewed: unfractionated testinal conditions. Brigham and Women‚Äôs Hospital utilizes heparin, low-molecular-weight heparin, fondaparinux, wara Medical Informatics System that integrates an online medfarin, and direct thrombin inhibitors, such as argatroban, ical record, computerized provider order entry, an electronic bivalirudin, and lepirudin. A multidisciplinary team, includmedication administration record, and an electronic safety ing physicians, pharmacists, and a hospital patient safety reporting system. The Safety Reporting System is a volunofficer, reviewed all reported anticoagulant-related events tary computerized reporting system for ADRs and medicaand evaluated the root case of all ADEs. tion errors, which was instituted in May 2004. We also collected data about the cost associated with hospitalizations in which ADEs occurred. Because medData Collection ication errors that do not culminate in an ADR do not add We reviewed all inpatient anticoagulant-associated ADEs, significant incremental cost to hospitalization, we inincluding ADRs and medication errors, reported at Brigham cluded only ADRs in our cost analysis. ADRs that added and Women‚Äôs Hospital through the Safety Reporting Syssignificant incremental cost to hospitalization were distem from May 2004 to May 2009. Inpatient events for the tinguished from those that did not by consensus of a purpose of our study were required to have originated durphysician (GP), pharmacist (TNN), and patient safety ing hospitalization at Brigham and Women‚Äôs Hospital. We officer (KF). While patients may have suffered multiple excluded reported anticoagulant-associated ADEs that took ADRs during a hospital admission, we analyzed costs place in the Emergency Department, Operating Room, Carincurred after the initial event. We utilized a software diac Catheterization Laboratory, Cardiovascular Recovery program for cost accounting that is capable of tabulating Room, Post-Anesthesia Care Unit, and Neonatal Intensive the cost associated with a patient admission (Transitions Care Unit because medication administration records were System Incorporated, Waltham, Mass). We sorted costs not consistently computerized in these units during the regby the following categories: nursing, pharmacy, blood istry period. We defined an ADR as a response to a drug products, clinical laboratory, radiology, surgery, nonsurnormally used for prophylaxis or therapy of disease and that gical intervention, and anesthesia. We compared total and 1138 The American Journal of Medicine, Vol 124, No 12, December 2011 component costs of hospitalization for admissions in which an ADR occurred and added incremental expense with those for admissions in which an ADR occurred but was not determined to incur incremental cost. Statistical Methods Descriptive statistics including baseline characteristics; variables related to ADEs and those pertaining to outcomes were stratified as continuous or binary. Continuous variables were presented as medians with interquartile ranges. Binary variables were presented as numbers and proportions. Cost variables were presented as means with standard deviations and were compared using a 2-sample t-test. All reported P-values were 2-sided, and a P-value of œΩ.05 was considered statistically significant. All statistical analyses were performed using STATA version 9.2 (STATA Corp., College Station, Tex). Table 2 Characteristics of Adverse Drug Events (ADEs) in Patients Receiving Anticoagulation (n œ≠ 463)* Characteristic of ADE Adverse drug reaction (ADR), n (%) Medication error, n (%) Combined medication error and ADR, n (%) Near miss, n (%) Potentially preventable ADE, n (%) Anticoagulant associated with ADE Unfractionated heparin, n (%) Warfarin, n (%) Low-molecular weight heparin, n (%) Argatroban, n (%) Bivalirudin, n (%) Lepirudin, n (%) Fondaparinux, n (%) 141 226 96 33 322 (30.5) (48.8) (20.7) (7.1) (69.5) 270 96 44 29 18 3 3 (58.3) (20.7) (9.5) (6.3) (3.9) (0.7) (0.7) ADR œ≠ adverse drug reaction. *Patients could have had more than 1 ADE including both medication errors and ADRs. RESULTS Baseline Characteristics Patients who suffered anticoagulant-associated ADEs had a median age of 62 years (Table 1). Their median body mass Table 1 Baseline Characteristics and Medical Conditions in Patients with Anticoagulation-associated Adverse Drug Events (ADEs) (n œ≠ 463)* Baseline characteristic Median age on admission, years (interquartile range) Age œæ75 years, n (%) Male, n (%) Median body mass index, kg/m2 (interquartile range) Median length of stay, d (interquartile range) Medical condition Hypertension, n (%) Surgery in past 2 months, n (%) Ischemic heart disease, n (%) Atrial fibrillation, n (%) Serum creatinine œæ1.5 mg/dL, n (%) Deep vein thrombosis or pulmonary embolism, n (%) Heart failure, n (%) Diabetes, n (%) Active cancer without metastases, n (%) Status post heart valve surgery, n (%) Chronic obstructive pulmonary disease, n (%) History of stroke, n (%) Active cancer with metastases, n (%) Other thromboembolism, n (%) History of cancer, n (%) Dialysis, n (%) Thrombophilia, n (%) *Patients could have had more than 1 ADE. 62 (49-72) 92 (19.9) 246 (53.1) 27.4 (23.8-33.1) 13 (7-23) 243 213 140 126 125 121 (52.5) (46) (30.2) (27.2) (27) (26.1) 106 104 74 68 66 (22.9) (22.5) (16) (14.7) (14.3) 45 42 40 34 30 16 (9.7) (9.1) (8.6) (7.3) (6.5) (3.5) index was 27.4 kg/m2. Patients who suffered anticoagulantassociated ADEs had a median length of stay of 13 days. Patients with anticoagulant-associated ADEs had a high frequency of medical conditions, such as atrial fibrillation, history of deep vein thrombosis or pulmonary embolism, heart failure, ischemic heart disease, chronic kidney disease, and stroke, that increase the risk of thromboembolism. Characteristics of Anticoagulant-associated Adverse Drug Events In 250,725 admissions over the 5-year study period, there were 463 anticoagulant-associated ADEs reported. Of these anticoagulant-associated ADEs, 226 were medication errors (48.8%), 141 were ADRs (30.5%), and 96 (20.7%) involved both a medication error and ADR. Seventy percent of anticoagulant-associated ADEs were potentially preventable (Table 2). Unfractionated heparin (58%) and warfarin (21%) were the most commonly implicated drugs in anticoagulant-associated ADEs. An average of 1.4 anticoagulant-associated ADEs occurred per patient over the 5-year study period. Characteristics of Anticoagulant-associated Medication Errors The most frequent type of anticoagulant-associated medication errors were missed medication doses (25%) and wrong rate or frequency (23%) (Table 3). Transcription errors (48%) were the most frequent root cause of anticoagulantassociated medication errors, followed by memory lapses (16%). Characteristics of Anticoagulant-associated Adverse Drug Reactions The most frequent anticoagulant-associated ADRs were abnormal coagulation studies (72%), any bleeding (25%), and Piazza et al Anticoagulation-associated Adverse Drug Events Table 3 Types and Root Causes of Anticoagulant-associated Medication Errors (n œ≠ 323) Type of medication error* Missed dose, n (%) Wrong rate or frequency, n (%) Medication not discontinued when ordered, n (%) Extra dose, n (%) Wrong dose, n (%) Wrong time of administration, n (%) Failure to act on laboratory result, n (%) Known allergy or contraindication, n (%) Wrong route, n (%) Medication expired, n (%) Wrong patient, n (%) Wrong drug, n (%) Preparation error, n (%) Wrong technique of administration, n (%) Medication administered without an order, n (%) Inadequate monitoring, n (%) Root cause of medication error Transcription error, n (%) Memory lapse, n (%) Infusion or parenteral administration problem, n (%) Rule violation, n (%) Lack of knowledge about drug, n (%) Faulty drug identity checking, n (%) Drug preparation error, n (%) Drug stocking or delivery problem, n (%) Faulty interaction between services, n (%) Lack of information about the patient, n (%) Faulty dose checking, n (%) Known allergy or contraindication, n (%) 79 75 31 26 23 17 17 13 10 8 7 7 6 3 2 1 (24.5) (23.2) (9.6) (8.1) (7.1) (5.3) (5.3) (4) (3.1) (2.5) (2.2) (2.2) (1.9) (0.9) (0.6) (0.3) 155 (48) 52 (16.1) 28 (8.7) 19 18 17 12 9 7 4 1 1 (5.9) (5.6) (5.3) (3.7) (2.8) (2.2) (1.2) (0.3) (0.3) *Patients may have had more than 1 medication error type. thrombocytopenia (18%) (Table 4). Seventy-two percent of all ADRs were associated with at least one occurrence of excessive anticoagulation, such as a super-therapeutic international normalized ratio or activated partial thromboplastin time. Seventeen percent of anticoagulant-associated ADRs culminated in transfusion of at least one unit of packed red blood cells. Undetected predisposing conditions (58%), such as a previously unknown drug allergy, and medication errors (40%) comprised the most common root causes of anticoagulant-associated ADRs. Outcomes of Anticoagulant-associated Adverse Drug Events Death within 30 days of anticoagulant-associated ADEs occurred in 11% of patients. In-hospital death occurred after 5.6% of anticoagulant-associated ADEs, while death after discharge but within 30 days of the ADE was observed in 5% (Table 5). A rehabilitation center stay was frequently required (40%) after an admission in which an anticoagulant-associated ADE occurred. Patients who suffered anticoagulant-associated ADEs had a high rate of re-hospitalization within 30 days of the ADE (17.5%). 1139 Cost Analysis of Anticoagulant-associated Adverse Drug Reactions Most hospitalization expenditures after an anticoagulantassociated ADR were attributable to nursing costs (mean $33,189 per ADR) followed by pharmacy costs (mean $7451 per ADR) (Table 6). ADRs that were determined to add incremental expense were associated with significant increases in total hospitalization cost (mean $118,429 vs $54,858, P œ≠ .02) as well as cost after the ADR (mean Table 4 Characteristics and Root Causes of Adverse Drug Reactions (ADRs) (n œ≠ 238) Characteristic of ADR Abnormal coagulation studies, n (%) 172 (72.3) Occurrences of excessive anticoagulation during 171 (71.9) hospitalization, n (%) Any bleeding event, n (%) 59 (24.8) Occult bleed 21 (8.8) Bleeding event related to surgery 14 (5.9) Hematoma 8 (3.4) Any gastrointestinal bleeding 6 (2.5) Decrease in hematocrit 6 (2.5) Retroperitoneal hemorrhage 3 (1.3) Median length of stay post bleeding event, days 9 (5-20) (interquartile range) Blood transfusion administered, n (%) 40 (16.8) Blood transfusion administered with 48 hours 13 (5.5) after surgery, n (%) Treatment of bleeding ADR, n (%) Vitamin K 10 (4.2) Protamine 2 (0.8) Fresh frozen plasma 11 (4.6) Surgery 6 (2.5) Catheterization laboratory 1 (0.4) GUSTO bleeding classification, n (%) I 19 (8) II 32 (13.5) III 8 (3.4) Thrombocytopenia, n (%) 43 (18.1) Heparin-induced thrombocytopenia, n (%) 31 (13) Any thromboembolic event, n (%) 16 (6.7) Deep vein thrombosis 6 (2.5) Pulmonary embolism 6 (2.5) Myocardial infarction 1 (0.4) Stroke 2 (0.8) Other arterial thromboembolic event 1 (0.4) Thromboembolic ADR requiring treatment, n (%) 8 (3.4) Treatment of thromboembolic ADR, n (%) Catheterization laboratory 3 (1.3) Interventional radiology procedure 1 (0.4) Surgery 2 (0.8) Root cause of ADR Undetected predisposing condition, n (%) 138 (58) Medication error, n (%) 94 (39.5) Drug-drug interaction, n (%) 3 (1.3) Patient medication non-adherence, n (%) 2 (0.8) GUSTO œ≠ Global Use of Strategies to Open Occluded Coronary Arteries study. 1140 Table 5 The American Journal of Medicine, Vol 124, No 12, December 2011 Outcomes of Adverse Drug Events (ADEs) (n œ≠ 463) Outcome In-hospital death, n (%) Death after discharge but within 30 days of ADE, n (%) Death due to thromboembolism, n (%) Death due to bleeding event, n (%) Discharge status, n (%) Home Rehabilitation center Hospice Acute care facility Deceased Readmission within 30 days of ADE, n (%) 26 (5.6) 23 (5) 4 (0.9) 1 (0.2) 236 186 13 2 26 81 (51) (40.2) (2.8) (0.4) (5.6) (17.5) $89,733 vs $23,680, P œ≠ .004) compared with ADRs in which no incremental cost was determined to be incurred (Table 7). ADRs that were determined to add incremental cost to hospitalization were associated with significant increases in costs of nursing, pharmacy, blood products, clinical laboratory, and radiology. DISCUSSION We found that 48.8% of anticoagulant-associated ADEs were medication errors, 30.5% were ADRs, and 20.7% involved both medication errors and ADRs. Seventy percent of all reported anticoagulant-associated ADEs were potentially preventable. Transcription errors were the most common root cause of medication errors, while undetected predisposing conditions and medication errors were the most frequent root causes of ADRs. We noted high 30-day mortality (11%) in patients who experienced an anticoagulantassociated ADE during hospitalization. We observed high post-ADE hospitalization costs, largely attributable to nursing and pharmacy expenditures. We had implemented computerized provider order entry, an electronic medication administration record, ‚Äúsmart‚Äù infusion pumps, and barcode technology before the registry was instituted. Therefore, we were surprised to observe that 70% of anticoagulant-associated ADEs were potentially preventable and that 40% of related ADRs were due to medication errors. Our data corroborate previous reports demonstrating that a substantial proportion of ADEs are due to medication errors and therefore potentially preventable.4,5 Computerized provider order entry,5 improved infusion pump technology,11 and implementing barcode technology8,9 reduce medication errors. However, our root cause analysis suggests that further improvements can be made to reduce anticoagulant-associated medication errors, particularly those due to transcription errors. While computerized provider order entry, ‚Äúsmart‚Äù infusion pump technology, electronic medication administration records, and barcode technology reduce the frequency of some transcription errors, human error may occur at transition points among these clinical tools and result in mistakes at any number of steps in medication transcription. The emergence of novel oral anticoagulants that are administered in fixed doses may further reduce dosing and infusion pump errors.19 We observed high 30-day mortality (11%) in patients suffering an anticoagulant-associated ADE during hospitalization. We speculate that this may be related to a combination of high medical acuity of patients suffering anticoagulant-associated ADEs and complications of the ADEs themselves. Patients with anticoagulant-associated ADEs had a high frequency of medical conditions such as heart failure, ischemic heart disease, chronic kidney disease, and stroke, all of which may increase patient vulnerability to the complications of anticoagulant-associated ADEs. Furthermore, anticoagulant-associated ADEs may be a marker for fragile patients with complicated hospital courses and long lengths of stay. The contribution of medical errors, including those associated with ADEs, to deaths occurring during hospitalization, may be overestimated in studies relying on physician review.20 Regardless of whether anticoagulantassociated ADEs contribute to increased mortality or are Table 6 (ADRs)* Cost Associated with Adverse Drug Reactions Mean ($) œÆ SD‚Ä† Overall and departmental costs Total hospitalization Hospitalization pre-ADR Hospitalization post-ADR Nursing Pharmacy Blood products Clinical laboratory Radiology Surgical Interventional procedure Anesthesia Drug-related costs Argatroban Bivalirudin Thrombolytic therapy Low-molecular weight heparin Unfractionated heparin Lepirudin Warfarin Fondaparinux Aprotinin Vitamin K Desmopressin Protamine Aminocaproic acid 88,842 œÆ 166,708 29,851 œÆ 49,264 58,991 œÆ 146,065 33,189 œÆ 81,236 7451 œÆ 21,878 2318 œÆ 9235 2001 œÆ 5069 1864 œÆ 4175 225 œÆ 1349 210 œÆ 1485 46 œÆ 257 1570 œÆ 7097 1102 œÆ 5879 87 œÆ 579 49 œÆ 154 35 œÆ 64 34 œÆ 327 23 œÆ 45 14 œÆ 76 4 œÆ 51 2œÆ6 2 œÆ 12 1œÆ6 0œÆ1 ADR œ≠ adverse drug reaction. *Patients may have suffered multiple ADRs during a hospital admission. Costs incurred were analyzed after the initial event. ‚Ä†Costs are rounded to the nearest US dollar. Piazza et al Anticoagulation-associated Adverse Drug Events Table 7 1141 Incremental Cost Associated with Adverse Drug Reactions (ADRs)* Overall and departmental costs (Mean œÆ SD)‚Ä† Total hospitalization Hospitalization pre-ADR Hospitalization post-ADR Nursing Pharmacy Blood products Clinical laboratory Radiology Surgical Interventional procedure Anesthesia Drug-related costs (Mean œÆ SD)‚Ä† Argatroban Bivalirudin Thrombolytic therapy Lepirudin Low-molecular weight heparin Unfractionated heparin Warfarin Fondaparinux Aprotinin Vitamin K Desmopressin Protamine Aminocaproic acid No Incremental Cost (n œ≠ 85) Incremental Cost (n œ≠ 74) P-Value 54,858 œÆ 115,659 31,178 œÆ 58,172 23,680 œÆ 65,283 14,008 œÆ 42,457 2830 œÆ 9243 548 œÆ 1703 902 œÆ 2469 798 œÆ 2228 0œÆ0 0œÆ0 0œÆ0 118,429 œÆ 196,840 28,696 œÆ 40,265 89,733 œÆ 185,395 49,888 œÆ 101,194 11,473 œÆ 28,123 3858 œÆ 12,359 2957 œÆ 6406 2792 œÆ 5158 421 œÆ 1828 393 œÆ 2019 87 œÆ 348 .02 .75 .004 .005 .01 .02 .01 .002 .0496 .1 .03 231 œÆ 1441 70 œÆ 366 24 œÆ 205 0œÆ0 53 œÆ 183 27 œÆ 67 15 œÆ 43 0œÆ4 0œÆ0 1œÆ5 0œÆ0 0œÆ0 0œÆ0 2735 œÆ 9486 2001 œÆ 7946 142 œÆ 766 63 œÆ 446 45 œÆ 125 41 œÆ 60 30 œÆ 47 26 œÆ 102 8 œÆ 69 3œÆ7 3 œÆ 17 3œÆ8 0œÆ1 .03 .04 .2 .23 .72 .14 .04 .03 .35 .01 .17 .003 .21 *Patients may have suffered multiple ADRs during a hospital admission. Costs incurred were analyzed after the initial event. ‚Ä†Costs are rounded to the nearest US dollar. markers for higher medical acuity, patients who suffer these events represent a particularly vulnerable population. Accordingly, Quality Improvement initiatives to reduce anticoagulant-associated ADEs are critical. Most hospitalization costs after an anticoagulant-associated ADE were attributable to nursing and pharmacy expenses. ADRs that incurred incremental cost doubled the total hospitalization costs and nearly quadrupled postevent costs compared with ADRs that did not incur any incremental expense. Our findings are similar to other reports that demonstrate the heavy financial burden of ADEs.13,21,22 Based on these data, efforts to reduce anticoagulant-associated ADEs have the potential to reduce hospitalization costs as well as improve patient safety. Our data were obtained from a voluntary patient safety reporting system and therefore might be limited by underreporting and selective reporting, in which only the most severe ADEs or those with the greatest consequences are reported. Because we excluded patient care areas in which medication administration records were not consistently computerized, we may have omitted subgroups of patients who had suffered a higher proportion of anticoagulant-associated ADEs. Because anticoagulant medications are given throughout the hospital but not all patient care areas have electronic medication administration records, it is not possible to calculate the denominator of patients exposed to anticoagulants. Therefore, we could not calculate the incidence of anticoagulantassociated ADEs and could not identify an appropriate comparison population of patients exposed to anticoagulant drugs who did not suffer ADEs. In addition, our analysis did not encompass community-acquired ADEs that resulted in hospitalization. While 5 deaths were clearly attributable to ADE-related bleeding (n œ≠ 1) or thromboembolism (n œ≠ 4), we were not able to quantify the contribution of ADEs to other deaths in this cohort. Finally, because of limitations in our cost accounting software, we were unable to separate costs incurred after the initial ADR from costs associated with subsequent events in a particular patient during the same hospitalization. The methodology utilized in this analysis is consistent with published criteria for evaluating the scientific value of clinical data registries.23 We utilized a large database generated from a tertiary care center that is representative of similar acute care facilities. Our study provides real-world insights into anticoagulant-associated ADEs from a medical center that has implemented an electronic health record that 1142 The American Journal of Medicine, Vol 124, No 12, December 2011 integrates computerized provider order entry, an electronic medication administration record, ‚Äúsmart‚Äù infusion pumps, and barcode technology. We utilized a multidisciplinary group of physicians, pharmacists, and a patient safety officer to provide the highest accuracy for classifying events and determining root causes of ADEs. Finally, we provided outcomes data for patients who suffered anticoagulantassociated ADEs with 100% follow-up. CONCLUSIONS Most anticoagulant-associated ADEs among inpatients result from medication errors and are therefore potentially preventable. Transcription errors are the most common cause of medication errors which, in turn, are a frequent cause of anticoagulant-associated ADRs. We observed elevated 30-day mortality among patients who suffered an anticoagulant-associated ADE and high hospitalization costs following ADRs. Further efforts to reduce anticoagulant-associated medication errors are warranted to improve patient safety and decrease health care expenditures. References 1. Zaidenstein R, Eyal S, Efrati S, et al. Adverse drug events in hospitalized patients treated with cardiovascular drugs and anticoagulants. Pharmacoepidemiol Drug Saf. 2002;11:235-238. 2. Silverman JB, Stapinski CD, Churchill WW, et al. Multifaceted approach to reducing preventable adverse drug events. Am J Health Syst Pharm. 2003;60:582-586. 3. Piazza G, Goldhaber SZ. Computerized decision support for the cardiovascular clinician: applications for venous thromboembolism prevention and beyond. Circulation. 2009;120:1133-1137. 4. Morimoto T, Sakuma M, Matsui K, et al. Incidence of adverse drug events and medication errors in Japan: the JADE Study. J Gen Intern Med. 2010;26:148-153. 5. Hug BL, Witkowski DJ, Sox CM, et al. Adverse drug event rates in six community hospitals and the potential impact of computerized physician order entry for prevention. J Gen Intern Med. 2009;25:31-38. 6. Classen DC, Jaser L, Budnitz DS. Adverse drug events among hospitalized Medicare patients: epidemiology and national estimates from a new approach to surveillance. Jt Comm J Qual Patient Saf. 2010;36: 12-21. 7. Bates DW, Cullen DJ, Laird N, et al. Incidence of adverse drug events and potential adverse drug events. Implications for prevention. ADE Prevention Study Group. JAMA. 1995;274:29-34. 8. Poon EG, Keohane CA, Yoon CS, et al. Effect of bar-code technology on the safety of medication administration. N Engl J Med. 2010;362: 1698-1707. 9. Poon EG, Cina JL, Churchill W, et al. Medication dispensing errors and potential adverse drug events before and after implementing bar code technology in the pharmacy. Ann Intern Med. 2006;145:426-434. 10. Leape LL, Bates DW, Cullen DJ, et al. Systems analysis of adverse drug events. ADE Prevention Study Group. JAMA. 1995;274:35-43. 11. Grissinger MC, Hicks RW, Keroack MA, et al. Harmful medication errors involving unfractionated and low-molecular-weight heparin in three patient safety reporting programs. Jt Comm J Qual Patient Saf. 2010;36:195-202. 12. Fanikos J, Stapinski C, Koo S, et al. Medication errors associated with anticoagulant therapy in the hospital. Am J Cardiol. 2004;94:532-535. 13. Bates DW, Spell N, Cullen DJ, et al. The costs of adverse drug events in hospitalized patients. Adverse Drug Events Prevention Study Group. JAMA. 1997;277:307-311. 14. Fanikos J, Cina JL, Baroletti S, et al. Adverse drug events in hospitalized cardiac patients. Am J Cardiol. 2007;100:1465-1469. 15. Hanlon JT, Pieper CF, Hajjar ER, et al. Incidence and predictors of all and preventable adverse drug reactions in frail elderly persons after hospital stay. J Gerontol A Biol Sci Med Sci. 2006;61:511-515. 16. Hajjar ER, Hanlon JT, Artz MB, et al. Adverse drug reaction risk factors in older outpatients. Am J Geriatr Pharmacother. 2003;1:82-89. 17. Nebeker JR, Barach P, Samore MH. Clarifying adverse drug events: a clinician‚Äôs guide to terminology, documentation, and reporting. Ann Intern Med. 2004;140:795-801. 18. An international randomized trial comparing four thrombolytic strategies for acute myocardial infarction. The GUSTO investigators. N Engl J Med. 1993;329:673-682. 19. Eikelboom JW, Weitz JI. New anticoagulants. Circulation. 2010;121: 1523-1532. 20. Hayward RA, Hofer TP. Estimating hospital deaths due to medical errors: preventability is in the eye of the reviewer. JAMA. 2001;286: 415-420. 21. Jennings HR, Miller EC, Williams TS, et al. Reducing anticoagulant medication adverse vents and avoidable patient harm. Jt Comm J Qual Patient Saf. 2008;34:196-200. 22. Eckman MH, Levine HJ, Pauker SG. Making decisions about antithrombotic therapy in heart disease. Decision analytic and cost-effectiveness issues. Chest. 1995;108:457S-470S. 23. Alpert JS. Are data from clinical registries of any value? Eur Heart J. 2000;21:1399-1401. CLINICAL RESEARCH STUDY Antidepressant Use and Cognitive Decline: The Health and Retirement Study Jane S. Saczynski, PhD,a,b,c Allison B. Rosen, MD,b,c Ryan J. McCammon, AB,d Kara Zivin, PhD,e,f,g,h Susan E. Andrade, PhD,b Kenneth M. Langa, MD, PhD,d,f,h Sandeep Vijan, MD,d,f Paul A. Pirraglia, MD,i Becky A. Briesacher, PhDj a Department of Medicine, University of Massachusetts Medical School, Worcester; bMeyers Primary Care Institute, Worcester, Mass; Department of Quantitative Health Sciences, University of Massachusetts Medical School, Worcester; dDivision of General Medicine, Department of Medicine, University of Michigan Medical School, Ann Arbor; eDepartment of Veterans Affairs, National Serious Mental Illness Treatment Resource and Evaluation Center, Ann Arbor, Mich; fDepartment of Veterans Affairs, Center for Clinical Management Research, Ann Arbor, Mich; gDepartment of Psychiatry, University of Michigan Medical School, Ann Arbor; hInstitute for Social Research, University of Michigan, Ann Arbor; iDepartment of Medicine, Alpert School of Medicine at Brown University, Providence, RI; jDepartment of Pharmacy and Health Systems Sciences, Northeastern University, Boston, Mass. c ABSTRACT BACKGROUND: Depression is associated with cognitive impairment and dementia, but whether treatment for depression with antidepressants reduces the risk for cognitive decline is unclear. We assessed the association between antidepressant use and cognitive decline over 6 years. METHODS: Participants were 3714 adults aged 50 years or more who were enrolled in the nationally representative Health and Retirement Study and had self-reported antidepressant use. Depressive symptoms were assessed using the 8-item Center for Epidemiologic Studies Depression Scale. Cognitive function was assessed at 4 time points (2004, 2006, 2008, 2010) using a validated 27-point scale. Change in cognitive function over the 6-year follow-up period was examined using linear growth models, adjusted for demographics, depressive symptoms, comorbidities, functional limitations, and antidepressant anticholinergic activity load. RESULTS: At baseline, cognitive function did not differ signiÔ¨Åcantly between the 445 (12.1%) participants taking antidepressants and those not taking antidepressants (mean, 14.9%; 95% conÔ¨Ådence interval, 14.315.4 vs mean, 15.1%; 95% conÔ¨Ådence interval, 14.9-15.3). During the 6-year follow up period, cognition declined in both users and nonusers of antidepressants, ranging from √Ä1.4 change in mean score in those with high depressive symptoms and taking antidepressants to √Ä0.5 change in mean score in those with high depressive symptoms and not taking antidepressants. In adjusted models, cognition declined in people taking antidepressants at the same rate as those not taking antidepressants. Results remained consistent across different levels of baseline cognitive function, age, and duration of antidepressant use (prolonged vs short-term). CONCLUSIONS: Antidepressant use did not modify the course of 6-year cognitive change in this nationally representative sample. √ì 2015 Elsevier Inc. All rights reserved. The American Journal of Medicine (2015) 128, 739-746 KEYWORDS: Antidepressants; Cognition; Depression; Epidemiology Funding: See last page of article. ConÔ¨Çict of Interest: See last page of article. Authorship: See last page of article. Requests for reprints should be addressed to Jane S. Saczynski, PhD, Division of Geriatric Medicine, Department of Medicine, 377 Plantation St, Suite 315, Worcester, MA 01605. E-mail address: Jane.saczynski@umassmed.edu 0002-9343/$ -see front matter √ì 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.amjmed.2015.01.007 Depression is associated consistently with cognitive impairment and an increased risk for dementia in clinical and epidemiologic studies of older adults.1-8 A meta-analysis of case-control and prospective studies concluded that a history of depression approximately doubled the risk for dementia.1 However, whether treatment for depression can reduce the rate of cognitive impairment and dementia is unclear.9-17 740 The American Journal of Medicine, Vol 128, No 7, July 2015 Research to date on the effects of antidepressants on the that includes assessments of depressive symptoms and rate of decline in cognitive function has been limited by cognitive function. The HRS began in 1992, and particismall sample sizes (<100 patients), short (<12 months) pants are re-interviewed every 2 years with high follow-up follow-up, lack of a comparison group, and selected subrates (90%-95%).22 The PDS was a mail survey distribgroups of patients, such as those who responded to therapy uted to a subsample of the HRS, drawn from respondents to (ie, those whose depressive symptoms decreased after the 2004 HRS wave, designed to track changes in pretreatment).11-17 A number of scription drug use among beneÔ¨Åciaries as Medicare Part D was studies have also examined CLINICAL SIGNIFICANCE phased in. The PDS sample inwhether antidepressant treatment cludes HRS respondents born in can slow the rate of cognitive Cognitive function did not differ at 1942 or earlier (age !65 years in decline or progression of dementia baseline between antidepressant users 2007) or those who were already among patients with preexisting and nonusers. covered by Medicare or Medicaid cognitive impairments or demen Rate of decline in cognitive function between 2002 and 2004. This tia.15,18-20 From these studies, it analysis starts with the 2004 data has been difÔ¨Åcult to disentangle over 6 years did not differ between anto correspond with the Ô¨Årst wave the association between antidetidepressant users and nonusers. of the PDS. This study was pressant treatment and long-term Treatment with antidepressant medicaapproved by the University of cognitive outcomes in a broadly tion use does not seem to modify the Michigan Institutional Review representative population of inwell-established association between Board; participants provided dividuals with a wide range of depression and cognitive decline. informed consent at enrollment. cognitive function, from nonThe current study included all impaired to moderately impaired, PDS respondents born before when depression is assessed. 1943 who were community dwelling, self-respondents Little work has been done on the relationship between providing cognitive function and depressive symptom asantidepressant use and cognition in population-based samsessments in the 2004 HRS interview. Participants‚Äô cogniples. One study of 595 patients found that antidepressant use tive function was assessed at each wave through the 2010 was associated with an increased risk of cognitive decline survey, with up to 4 total assessments. Our primary analyses over 4.5 years among depressed patients without cognitive focus on change in cognitive function in patients who were, impairment.10 Within the Women‚Äôs Health Initiative Memcompared with those who were not, taking antidepressants ory Study, antidepressant use was associated with a 70% according to the PDS. increased risk of incident mild cognitive impairment over 7.5 years.9 Of note, although the Women‚Äôs Health Initiative Memory Study collected information on the type of antiAssessment of Cognitive Function depressant taken, neither study assessed the anticholinergic activity of antidepressants. Medications with anticholinergic At each HRS wave, cognitive function was assessed using a effects, including amitriptyline, doxepin, paroxetine, and previously described and validated 27-point scale based on a nortriptyline among others, can block muscarinic receptors battery of tests that included tests of memory, serial 7 causing impairment in various cognitive functions, subtractions, and naming.23 This battery is a subset of an including memory, executive function, and processing expanded battery (range, 0-35) administered to participants speed.21 aged !65 years in the HRS; the expanded battery includes measures of orientation.24 Participants requiring proxy inWe examined whether antidepressant use was associated with cognitive decline over a 6-year period using data from terviews (because of cognitive or physical impairments that the nationally representative Health and Retirement Study limited their ability to self-respond) were excluded from this (HRS) and the HRS Prescription Drug Study (PDS), which analysis. Cut points for cognitive function were based on include serial assessments of cognitive function, depressive prior studies with the HRS data,25,26 as well as methods symptoms, and antidepressant treatment. We hypothesized used for the Aging, Demographics, and Memory Study, a that treatment with antidepressant medications would be supplemental study of dementia in the HRS.27 These cut associated with slower rates of cognitive decline. points deÔ¨Åned a level of cognitive function that was generally consistent with normal function (12-27 points), cognitive impairment but no dementia (7-11 points), and dementia (0-6 points).28 MATERIALS AND METHODS Data Source Study data were drawn from the 2004, 2006, 2008, and 2010 waves of the HRS, and from the 2005 and 2007 waves of the PDS. The HRS is a longitudinal, nationally representative survey of US residents aged 51 years and older Assessment of Depression Baseline depression status was based on the 2004 wave of the HRS using an 8-item version of the Center for Epidemiologic Studies Depression Scale.29,30 This version of the Saczynski et al Antidepressant Use and Cognitive Decline 741 Center for Epidemiologic Studies Depression Scale included the following questions with ‚Äúyes/no‚Äù response options: Much of the time during the last week, I felt depressed. I felt everything I did was an effort. My sleep was restless. I was happy (reverse coded). I felt lonely. I enjoyed life (reverse coded). I felt sad. I could not ‚Äúget going.‚Äù The total number of ‚Äúyes‚Äù responses were summed to calculate an overall depression score (range, 0-8); participants with scores !4 were considered to have high depressive symptomatology.30 We conducted several sensitivity analyses: one in which we stratiÔ¨Åed the analysis on the basis of cognitive function at baseline (no impairment, moderate impairment, dementia) and a second one in which we stratiÔ¨Åed the analysis on the basis of age at baseline (61-74 vs !75 years) to examine whether the association between antidepressant use and change in cognitive function varied according to level of cognitive function or age. In additional sensitivity analyses, we restricted the sample to those taking antidepressants at baseline (n ¬º 445) and examined the effect of self-reported antidepressant use >1 year at baseline compared with shorter-term use and the effect of continuous antidepressant use over the entire follow-up period compared with discontinuation. All analyses were conducted using SAS version 9.3 (SAS Institute Inc, Cary NC) and MPlus version 7.1. Assessment of Antidepressant Use Antidepressant use was assessed using the 2005 and 2007 PDS. Respondents were asked to list all medications prescribed and directed to provide the information on drug name and dosage directly from the prescription bottle label. Medications were then matched to the American Hospital Formulary System, a database of all prescription drugs. Respondents were considered to be taking an antidepressant if any of their medications were matched to therapeutic class code 28:16.04, which includes selective serotonin reuptake inhibitors, serotonin norepinephrine reuptake inhibitors, other novel antidepressants (eg, bupropion), tricyclic antidepressants, and monoamine oxidase inhibitors. Other Study Variables The following sociodemographic variables from the 2004 HRS survey were included as covariates: age, sex, education (less than high school vs high school or more), race (nonHispanic-white, non-Hispanic-black, Hispanic, and other), chronic conditions (count of 7 self-reported conditions, including high blood pressure, diabetes, heart disease, stroke, lung disease, cancer, and arthritis), self-reported visual or hearing impairment, dependence in Activities of Daily Living,31 and self-reported duration of antidepressant use. We also included a measure of baseline anticholinergic burden (based on the Anticholinergic Drug Scale),32 because anticholinergic activity is high in some antidepressant medications and is associated with cognitive decline.33 Analyses Baseline (2004) characteristics were compared for participants who were taking antidepressants as assessed by the 2005 PDS compared with those who were not taking antidepressants using the chi-square test for categoric variables and analysis of variance analyses for continuous variables. A linear latent growth model was used to examine change on the 27-point cognitive battery over the 6-year follow-up. In this model, time is measured as years since baseline (2004), and random effects are used to capture individual differences in baseline levels (intercepts) and rates of change (slopes). We adjusted for depression, age, sex, education, race, hearing/visual impairment, comorbidity burden, and Activities of Daily Living impairment. We also adjusted for anticholinergic activity load of medications reported in the 2005 PDS. RESULTS Medication data from the 2005 PDS were available for 4320 individuals. Participants born after 1942 (n ¬º 185), those who required a proxy respondent (n ¬º 373), and those with incomplete data on the Center for Epidemiologic Studies Depression Scale (n ¬º 48) were excluded from the study sample, resulting in an analytic sample of 3714 participants. Of the 3714 eligible respondents, 445 (12%) were taking antidepressants in 2005. Participants taking antidepressants were more likely to be female, to drink alcohol daily, and to be white in contrast to participants not taking antidepressants (Table 1). Those taking antidepressants had a higher burden of comorbidity; they were more likely to need assistance with Activities of Daily Living and Instrumental Activities of Daily Living and more likely to have hearing and visual impairments. As expected, participants taking antidepressants were more likely to report a higher level of depressive symptoms than participants not taking antidepressants. Age, education, marital status, and smoking status did not differ according to antidepressant use (Table 1). Antidepressant Use and Cognitive Function Table 2 shows unadjusted cognitive performance by baseline antidepressant use and depressive symptoms. At baseline, no signiÔ¨Åcant differences were detected in average cognition scores between the 445 (12.1%) participants taking antidepressants and those not taking antidepressants (mean 14.9, 95% conÔ¨Ådence interval [CI], 14.3-15.4 vs mean 15.1%; 95% CI, 14.9-15.3). During the 6-year follow-up period, both users and nonusers of antidepressants experienced cognitive decline ranging from √Ä1.4 change in mean score (95% CI, √Ä1.1 to √Ä1.7) in people with high depressive symptoms and taking antidepressants to √Ä0.5 change in mean score (95% CI, √Ä0.6 to √Ä0.3) in people with high depressive symptoms and not taking antidepressants. In our initial linear growth model, adjusted for age and depressive symptoms, there was a nonsigniÔ¨Åcant negative association between antidepressant use and baseline cognitive 742 Table 1 The American Journal of Medicine, Vol 128, No 7, July 2015 Sample Characteristics by 2005 Antidepressant Use Antidepressant Use Measure Baseline age, mean, y (95% CI) Sex, % female Education less than high school, % White Black Hispanic/other Married, % Current smoker, % Intake of 1√æ drinks per day, % CES-D, mean (95% CI) CES-D depression* % No. of chronic conditions, % 0 1 2 3√æ Chronic conditions, mean (95% CI) Visual impairment Hearing impairment 2005 AA load 0 1 or 2 3√æ 2005 AA load, mean (95% CI) ADL, % any dependence IADL, % any dependence Duration antidepressant use, % !1 y No (n ¬º 3269) % (95% CI) Yes (n ¬º 445) % (95% CI) 72.0 (71.7-72.4) 56.7 23.6 85.0 8.1 6.9 57.4 10.1 12.3 1.3 (1.2-1.4) 12.0 71.9 (70.4-72.8) 69.3 20.7 93.9 2.3 3.8 52.3 13.5 10.8 2.3 (2.0-2.6) 26.6 13.4 26.3 29.8 30.5 1.9 (1.9-2.0) 19.2 23.2 8.3 20.8 30.1 40.8 2.3 (2.1-2.5) 25.8 25.7 64.8 26.1 9.1 0.8 (0.7-0.8) 15.2 11.9 ‚Äî 14.3 52.8 33.0 2.1 (2.0-2.3) 28.9 23.0 89.8 P Value .72 <.001 .91 <.001 .09 .042 .41 <.001 <.001 <.001 <.001 .006 .34 <.001 <.001 <.001 <.001 AA ¬º anticholinergic activity; ADL ¬º Activities of Daily Living; CES-D ¬º Center for Epidemiologic Studies Depression Scale; CI ¬º conÔ¨Ådence interval; IADL ¬º Instrumental Activities of Daily Living. *CES-D !4. function (Intercept; Table 3, Model 1), but a signiÔ¨Åcant decline in cognitive function over the 6-year study period in antidepressant users (Slope; Table 3, Model 1). Depression and age were both negatively associated with baseline level of function, but only age was associated with change (decline) in cognitive function over time. In subsequent models, we found a nonsigniÔ¨Åcant negative association between antidepressant use and change in cognitive function when we further adjusted for sex, education, race, Activities of Daily Living impairments, comorbidity burden (Table 3, Model 2), and anticholinergic load (Table 3, Model 3). Figure 1 presents unadjusted model-based estimates by baseline depressive symptom status (high/low) and antidepressant use corresponding to Model 1 (Table 3) for ages 72 years (2004) to 78 years (2010). Subgroup and Sensitivity Analyses We conducted several sensitivity analyses. First, we examined whether the association between antidepressant use and change in cognitive function varied by baseline level of cognitive function or by age. In linear growth models adjusted for age and depression, antidepressant use was associated with a signiÔ¨Åcant decline in cognitive function among participants with normal cognitive function at baseline (n ¬º 2817; Slope ¬º √Ä0.094 [standard error ¬º 0.04], P < .05) but nonsigniÔ¨Åcant decline in cognitive function among participants with cognitive impairment but no dementia (n ¬º 721; Slope ¬º √Ä0.095 [0.10]) and those who were demented (n ¬º 176; Slope ¬º √Ä0.133 [0.17]). Of note, although the association between antidepressant use and cognitive decline was signiÔ¨Åcant in the participants with normal cognitive function at baseline and not in the 2 cognitively impaired groups, we observed a negative association between antidepressant use and cognitive change across levels of baseline cognitive function. In additional models and similar to Ô¨Åndings in the overall cohort, there was a nonsigniÔ¨Åcant association between antidepressant use and cognitive decline in younger participants (age 61-74 years, n ¬º 2398; Slope ¬º √Ä0.065 [standard error ¬º 0.05]) and older participants (age 75√æ years, n ¬º 1316; Slope ¬º √Ä0.133 [standard error ¬º 0.08]). We also conducted sensitivity analyses among participants taking antidepressants at baseline (n ¬º 414) to examine Saczynski et al Table 2 Antidepressant Use and Cognitive Decline 743 Changes in Cognitive Functioning Over 6-Year Period by Baseline Antidepressant Use Baseline (95% CI) Time 1 (95% CI) Time 2 (95% CI) Time 3 (95% CI) Low depressive symptoms and not taking antidepressants Normal, % Borderline, % Impaired, % Mean score n 81.3 (79.6-83.0) 15.6 (13.8-17.4) 3.0 (2.4-3.6) 15.3 (15.1-15.6) 2832 79.8 (77.6-81.9) 15.0 (13.3-16.7) 5.3 (4.4-6.1) 15.1 (14.9-15.4) 2673 78.2 (75.9-80.5) 16.3 (14.4-18.3) 5.5 (4.4-6.5) 15.0 (14.7-15.2) 2424 73.2 (70.9-75.4) 17.9 (16.2-19.5) 9.0 (7.7-10.2) 14.5 (14.3-14.8) 2071 Low depressive symptoms and taking antidepressants Normal, % Borderline, % Impaired, % Mean score n 80.6 14.8 4.6 15.1 324 (75.0-86.3) (9.8-19.7) (1.8-7.4) (14.4-15.8) 74.4 16.6 9.0 14.6 302 (69.7-79.0) (12.5-20.7) (5.4-12.6) (14.0-15.2) 72.4 15.8 11.8 14.5 262 (65.8-79.0) (10.4-21.3) (7.8-15.8) (13.8-15.3) 68.7 20.6 10.6 14.4 219 (61.8-75.7) (15.4-25.8) (6.3-14.9) (13.8-15.1) High depressive symptoms and not taking antidepressants Normal, % Borderline, % Impaired, % Mean score n 64.0 29.3 6.7 13.1 437 (57.9-70.1) (24.3-34.4) (3.5-9.8) (12.6-13.6) 62.6 25.8 11.7 13.1 401 (56.9-68.2) (20.8-30.7) (7.7-15.6) (12.5-13.6) 60.1 28.1 11.8 13.2 341 (52.5-67.7) (21.8-34.5) (8.2-15.3) (12.5-13.8) 58.0 25.1 16.9 12.6 257 (51.8-64.3) (19.7-30.5) (11.6-22.1) (11.9-13.3) High depressive symptoms and taking antidepressants Normal, % Borderline, % Impaired, % Mean score n 74.1 19.5 6.4 14.2 121 (65.2-83.0) (11.9-27.1) (1.7-11.1) (13.3-15.1) 67.4 23.1 9.4 13.6 114 (56.7-78.2) (14.0-32.2) (3.7-15.2) (12.4-14.7) 69.8 19.7 10.4 13.7 93 (58.1-81.5) (10.3-29.2) (3.6-17.2) (12.5-14.8) 54.2 29.9 15.9 12.8 71 (40.6-67.9) (19.3-40.6) (7.0-24.7) (11.6-14.0) CI ¬º conÔ¨Ådence interval. the effect of duration and discontinuation of antidepressant use. In linear growth models adjusted for depressive symptoms and age, duration of antidepressant use at baseline and continuity of antidepressant use over the follow-up period were nonsigniÔ¨Åcantly associated with cognitive change (Slope for >1 year use ¬º 0.103 [standard error ¬º 0.10]) (Slope for continuous use ¬º 0.026 [standard error ¬º 0.09]). DISCUSSION In a nationally representative cohort of older adults, we found that antidepressant use was not associated with changes in cognitive function over a 6-year period, adjusting for depressive symptoms, comorbidity burden, and anticholinergic load. Findings were consistent across levels of baseline cognitive function (normal function, mild cognitive impairment, and dementia). The results suggest that pharmacologic treatment for depression does not modify the negative association observed between depression and cognitive decline in clinical and epidemiologic studies. Many clinical trials (mostly small) have been conducted to assess the association between antidepressant treatment and cognitive decline, and Ô¨Åndings have been mixed. When responders and nonresponders are examined separately, responders to treatment tend to show improvements in cognitive function,11-13 whereas nonresponders show no improvement12 or decline.13 Other studies have shown only modest improvements in cognition after treatment with antidepressants, equal to the practice effects observed in cognitive performance in control groups.14-17 One study found that treatment with antidepressants decreased the risk of dementia, but only among patients with mild cognitive impairment at baseline.15 Little work has been published on treatment with antidepressants and cognitive trajectories in population-based studies. However, the Ô¨Åndings suggest that, among nonimpaired participants, antidepressant use is associated with an increased risk of incident mild cognitive impairment in up to 7.5 years of follow-up.9,10 We found a nonsigniÔ¨Åcant decline in cognitive function over a 6-year follow-up period among participants taking antidepressants at baseline. Our sample was not restricted to cognitively healthy participants at baseline, and results were similar across strata of baseline cognitive function (normal, mild cognitive impairment, and dementia). To our knowledge, the effect of anticholinergic activity load has not been controlled or accounted for in previous studies of the longitudinal association between antidepressant use and cognitive decline. High anticholinergic activity 744 The American Journal of Medicine, Vol 128, No 7, July 2015 Table 3 Linear Latent Growth Models of 2004-2008 Health and Retirement Study 27-Point Cognition Measure (n ¬º 3714) Model Intercept Antidepressant use CES-D 4√æ Age Female <12 y of school Black Hispanic Impaired vision Impaired hearing Chronic conditions (range, 0-7) 1 ADL impairment 2√æ ADL impairments AA load of 1 or 2 AA load of 3√æ Mean Residual variance Slope Antidepressant use CES-D 4√æ Age Female <12 y of school Black Hispanic Impaired vision Impaired hearing Chronic conditions 1 ADL impairment 2√æ ADL impairments AA load of 1 or 2 AA load of 3√æ Mean Residual variance 1 CoefÔ¨Åcient (SE) 2 CoefÔ¨Åcient (SE) 3 CoefÔ¨Åcient (SE) √Ä0.11 (0.25) √Ä1.72‚Ä° (0.23) √Ä0.18‚Ä° (0.01) √Ä0.39 √Ä0.54‚Ä° √Ä0.16‚Ä° 0.42‚Ä† √Ä2.72‚Ä° √Ä2.84‚Ä° √Ä2.44‚Ä° √Ä0.28 √Ä0.68‚Ä° √Ä0.11 √Ä0.65‚Ä† √Ä1.04‚Ä† √Ä0.36 √Ä0.54‚Ä° 0.16‚Ä° √Ä0.43‚Ä† √Ä2.72‚Ä° √Ä2.85‚Ä° √Ä2.45‚Ä° √Ä0.27 √Ä0.67‚Ä° √Ä0.10 √Ä0.64‚Ä† √Ä1.03‚Ä† √Ä0.03 √Ä0.13 17.40‚Ä° (0.16) 11.33‚Ä° (0.42) 18.39‚Ä° (0.16) 8.10‚Ä° (0.37) 18.40‚Ä° (0.16) 8.10‚Ä° (0.37) √Ä0.09* (0.04) 0.002 (0.04) √Ä0.02‚Ä° (<0.01) √Ä0.8 √Ä0.01 √Ä0.02‚Ä° √Ä0.02 √Ä0.01 0.02 0.16‚Ä† √Ä0.02 0.04 0.02 0.01 0.01 √Ä0.08 √Ä0.01 √Ä0.02‚Ä° √Ä0.02 √Ä0.01 0.02 0.16‚Ä† √Ä0.02 0.04 0.02 <0.01 0.01 √Ä0.01 √Ä0.02 √Ä0.07‚Ä† (0.02) 0.07‚Ä° (0.01) √Ä0.12‚Ä° (0.03) 0.06‚Ä° (0.01) (0.24) (0.20) (0.01) (0.14) (0.16) (0.27) (0.26) (0.19) (0.16) (0.07) (0.21) (0.32) (0.04) (0.05) (<0.01) (0.03) (0.03) (0.04) (0.05) (0.03) (0.03) (0.01) (0.04) (0.05) (0.26) (0.20) (0.01) (0.14) (0.16) (0.27) (0.27) (0.18) (0.16) (0.06) (0.21) (0.32) (0.16) (0.20) (0.05) (0.05) (<0.01) (0.03) (0.03) (0.04) (0.05) (0.03) (0.03) (0.01) (0.04) (0.06) (0.03) (0.04) √Ä0.10‚Ä° (0.03) 0.06‚Ä° (0.01) AA ¬º anticholinergic activity; ADL ¬º Activities of Daily Living; CES-D ¬º Center for Epidemiologic Studies Depression; SE ¬º standard error. *P < .05. ‚Ä†P < .01. ‚Ä°P < .001. load is recognized increasingly for its association with negative cognitive effects, and anticholinergic activity is present, often in high doses, in many antidepressants. We were surprised by the small effect of anticholinergic activity load on baseline cognitive function and cognitive decline in our study. A number of population-based studies have examined anticholinergic activity load and cognitive decline, with signiÔ¨Åcant associations observed in some,33-36 but not all,37-39 studies. The association between anticholinergic activity and cognitive decline may be related to level of cognitive function at baseline. For example, in a study of more than 13,000 patients, the strongest association between anticholinergic activity load and decline on the Mini Mental State Exam was in participants with scores in the ‚Äúnormal‚Äù range (26-30).33 We found that antidepressant use was associated with a decline in cognitive function across all levels of baseline cognitive function (normal, cognitive impairment no dementia, and dementia) and that the association was strongest among participants with normal cognitive function at baseline and before and after adjustment for anticholinergic activity load of antidepressants. Study Strengths and Limitations The present study has a number of strengths. The HRS is a nationally representative cohort, and our analysis was not Saczynski et al Antidepressant Use and Cognitive Decline 745 results were consistent across baseline levels of cognitive function and age, and did not change when we controlled for the anticholinergic activity load of antidepressants. Because depression is a well-known risk factor for cognitive decline and dementia, consideration of nonpharmacologic approaches to treatment for depression that also are related to cognitive function (eg, social engagement, physical activity, and diet) may be used in combination with pharmacologic approaches to address cognitive function and depressive symptoms. References Figure 1 Linear growth model of antidepressant use, depression, and cognition. Estimates are for participant who is aged 72 years at baseline (2004). CESD ¬º Center for Epidemiologic Studies Depression Scale. restricted to select subgroups of patients, such as those who responded to therapy or those without preexisting cognitive impairment. The PDS medication data, although selfreported, collects detailed information on medications used, so therapeutic class code can be determined by matching with the American Hospital Formulary System. As a result, we were able to code medications for anticholinergic activity load and examine the effect of high anticholinergic activity load on cognitive trajectories that has been missing from previous studies. There are several potential limitations to our study. We did not look at speciÔ¨Åc classes of antidepressants and do not know whether patients were taking antidepressants for depression or other indications, such as weight loss. Our study was not designed to examine responsiveness to antidepressant treatment; however, we controlled for depressive symptoms in our analyses. Duration of antidepressant use was self-reported and thus could have resulted in misclassiÔ¨Åcation, particularly among cognitively impaired participants. The battery of cognitive tests allowed us to examine global cognitive function but not individual cognitive domains, and it did not provide a clinical diagnosis of dementia. However, we used cut points that have been validated against clinical diagnosis of dementia.26 Likewise, the Center for Epidemiologic Studies Depression Scale assesses depressive symptoms rather than providing a clinical diagnosis of depression; however, we used a wellestablished cut point that has been validated against clinical diagnosis of depression.30 CONCLUSIONS In this large nationally representative cohort of older adults, we found that antidepressant use was not signiÔ¨Åcantly associated with 6-year change in cognitive function. These 1. Jorm AF. History of depression as a risk factor for dementia: an updated review. Aust N Z J Psychiatry. 2001;35:776-781. 2. Barnes DE, Alexopoulos GS, Lopez OL, Williamson JD, Yaffe K. Depressive symptoms, vascular disease, and mild cognitive impairment: Ô¨Åndings from the Cardiovascular Health Study. Arch Gen Psychiatry. 2006;63:273-279. 3. Jorm AF, van Duijn CM, Chandra V, et al. Psychiatric history and related exposures as risk factors for Alzheimer‚Äôs disease: a collaborative re-analysis of case-control studies. EURODEM Risk Factors Research Group. Int J Epidemiol. 1991;20:S43-S47. 4. Speck CE, Kukull WA, Brenner DE, et al. History of depression as a risk factor for Alzheimer‚Äôs disease. Epidemiology. 1995;6:366-369. 5. Green RC, Cupples LA, Kurz A, et al. Depression as a risk factor for Alzheimer disease: the MIRAGE Study. Arch Neurol. 2003;60: 753-759. 6. Saczynski JS, Beiser A, Seshadri S, Auerbach S, Wolf PA, Au R. Depressive symptoms and risk of dementia: the Framingham Heart Study. Neurology. 2010;75:35-41. 7. Dotson VM, Beydoun MA, Zonderman AB. Recurrent depressive symptoms and the incidence of dementia and mild cognitive impairment. Neurology. 2010;75:27-34. 8. Goveas JS, Espeland MA, Woods NF, Wassertheil-Smoller S, Kotchen JM. Depressive symptoms and incidence of mild cognitive impairment and probable dementia in elderly women: the Women‚Äôs Health Initiative Memory Study. J Am Geriatr Soc. 2012;59:57-66. 9. Goveas JS, Hogan PE, Kotchen JM, et al. Depressive symptoms, antidepressant use, and future cognitive health in postmenopausal women: the Women‚Äôs Health Initiative Memory Study. Int Psychogeriatr. 2012;24:1252-1264. 10. Ravaglia G, Forti P, Lucicesare A, et al. Prevalent depressive symptoms as a risk factor for conversion to mild cognitive impairment in an elderly Italian cohort. Am J Geriatr Psychiatry. 2008;16:834-843. 11. Butters MA, Becker JT, Nebes RD, et al. Changes in cognitive functioning following treatment of late-life depression. Am J Psychiatry. 2000;157:1949-1954. 12. Mandelli L, Serretti A, Colombo C, et al. Improvement of cognitive functioning in mood disorder patients with depressive symptomatic recovery during treatment: an exploratory analysis. Psychiatry Clin Neurosci. 2006;60:598-604. 13. Culang ME, Sneed JR, Keilp JG, et al. Change in cognitive functioning following acute antidepressant treatment in late-life depression. Am J Geriatr Psychiatry. 2009;17:881-888. 14. Nebes RD, Pollock BG, Houck PR, et al. Persistence of cognitive impairment in geriatric patients following antidepressant treatment: a randomized, double-blind clinical trial with nortriptyline and paroxetine. J Psychiatr Res. 2003;37:99-108. 15. Reynolds CF 3rd, Butters MA, Lopez O, et al. Maintenance treatment of depression in old age: a randomized, double-blind, placebocontrolled evaluation of the efÔ¨Åcacy and safety of donepezil combined with antidepressant pharmacotherapy. Arch Gen Psychiatry. 2011;68: 51-60. 16. Herrera-Guzman I, Herrera-Abarca JE, Gudayol-Ferre E, et al. Effects of selective serotonin reuptake and dual serotonergic-noradrenergic 746 17. 18. 19. 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. reuptake treatments on attention and executive functions in patients with major depressive disorder. Psychiatry Res. 2010;177:323-329. Gallassi R, Di Sarro R, Morreale A, Amore M. Memory impairment in patients with late-onset major depression: the effect of antidepressant therapy. J Affect Disord. 2006;91:243-250. Rosenberg PB, Mielke MM, Han D, et al. The association of psychotropic medication use with the cognitive, functional, and neuropsychiatric trajectory of Alzheimer‚Äôs disease. Int J Geriatr Psychiatry. 2012;27:1248-1257. Devanand DP, Pelton GH, Marston K, et al. Sertraline treatment of elderly patients with depression and cognitive impairment. Int J Geriatr Psychiatry. 2003;18:123-130. Lopez OL, Wisniewski SR, Becker JT, Boller F, DeKosky ST. Psychiatric medication and abnormal behavior as predictors of progression in probable Alzheimer disease. Arch Neurol. 1999;56:1266-1272. Mulsant BH, Pollock BG, Kirshner M, Shen C, Dodge H, Ganguli M. Serum anticholinergic activity in a community-based sample of older adults: relationship with cognitive performance. Arch Gen Psychiatry. 2003;60:198-203. Health and Retirement Study. Sample Sizes and Response Rates. Available at: http://hrsonline.isr.umich.edu/sitedocs/sampleresponse. pdf. Accessed August 16, 2012. Ofstedal M, Fisher GG, Herzog AR. Documentation of Cognitive Functioning Measures in the Health and Retirement Study. Ann Arbor, MI: University of Michigan; 2005. Herzog AR, Wallace RB. Measures of cognitive functioning in the AHEAD Study. J Gerontol B Psychol Sci Soc Sci. 1997;52B:37-48. Langa KM, Chernew ME, Kabeto MU, et al. National estimates of the quantity and cost of informal caregiving for the elderly with dementia. J Gen Intern Med. 2001;16:770-778. Langa KM, Larson EB, Karlawish JH, et al. Trends in the prevalence and mortality of cognitive impairment in the United States: is there evidence of a compression of cognitive morbidity? Alzheimers Dement. Mar 2008;4:134-144. Langa KM, Plassman BL, Wallace RB, et al. The Aging, Demographics, and Memory Study: study design and methods. Neuroepidemiology. 2005;25:181-191. Crimmins EM, Kim JK, Langa KM, Weir DR. Assessment of cognition using surveys and neuropsychological assessment: the Health and Retirement Study and the Aging, Demographics, and Memory Study. J Gerontol B Psychol Sci Soc Sci. 2011;66(Suppl 1):i162-i171. Radloff L. The CES-D Scale: a self-report depression scale for research in the general population. Appl Psychol Methods. 1977;1:385-401. The American Journal of Medicine, Vol 128, No 7, July 2015 30. StefÔ¨Åck D. Documentation of Affective Functioning Measures in the Health and Retirement Study. Ann Arbor, MI: University of Michigan; 2000. 31. Katz S, Downs TD, Cash HR, Grotz RC. Progress in development of the index of ADL. Gerontologist. 1970;10:20-30. 32. Carnahan RM, Lund BC, Perry PJ, Pollock BG, Culp KR. The anticholinergic drug scale as a measure of drug-related anticholinergic burden: associations with serum anticholinergic activity. J Clin Pharmacol. 2006;46:1481-1486. 33. Fox C, Richardson K, Maidment ID, et al. Anticholinergic medication use and cognitive impairment in the older population: the Medical Research Council Cognitive Function and Ageing study. J Am Geriatr Soc. 2011;59:1477-1483. 34. Campbell NL, Boustani MA, Lane KA, et al. Use of anticholinergics and the risk of cognitive impairment in an African American population. Neurology. 2010;75:152-159. 35. Carriere I, Fourrier-Reglat A, Dartigues JF, et al. Drugs with anticholinergic properties, cognitive decline, and dementia in an elderly general population: the 3-city study. Arch Intern Med. 2009;169: 1317-1324. 36. Ancelin ML, Artero S, Portet F, Dupuy AM, Touchon J, Ritchie K. Nondegenerative mild cognitive impairment in elderly people and use of anticholinergic drugs: longitudinal cohort study. BMJ. 2006;332:455-459. 37. Fox C, Livingston G, Maidment ID, et al. The impact of anticholinergic burden in Alzheimer‚Äôs Dementia-the Laser-AD study. Age Ageing. 2011;40:730-735. 38. Low L-F, Anstey KJ, Sachdev P. Use of medications with anticholinergic properties and cognitive function in a young-old community sample. Int J Geriatr Psychiatry. 2009;24:578-584. 39. Bottiggi KA, Salazar JC, Yu L, et al. Long-term cognitive impact of anticholinergic medications in older adults. Am J Geriatr Psychiatry. 2006;14:980-984. Funding: The Health and Retirement Study is funded by the National Institute on Aging (U01 AG09740). JSS was supported in part by funding from the National Institute on Aging (K01AG33643) and the National Heart, Lung, and Blood Institute (U01HL105268). This research was supported in part by P01AG031098 from the National Institute on Aging and the Department of Veterans Affairs, Health Services Research and Development (CD2 07-206-1 and VA IIR 10-176-3). ConÔ¨Çict of Interest: None. Authorship: All authors had access to the data and played a role in writing this manuscript. ˇ˛Antidiabetic drugs W stephen Waring Abstract over recent years, there has been rapid expansion of different classes of antihyperglycaemic drugs. each possesses a unique pharmacological mechanism of action and, correspondingly, they have diverse toxicological profiles. antidiabetic drugs have the potential to exert severe toxic effects, and patients normally require urgent medical assessment and treatment. Keywords acarbose; antidiabetic; antihyperglycaemic; DPP-iV inhibitor; GLP-1; insulin; metformin; overdose; sulphonylureas; thiazolidinediones Mechanisms of toxicity Insulin Insulin modifies glucose transporter function via its tyrosine kinase-linked receptor. Parenteral administration may cause pro- found hypoglycaemia and activation of neurohormonal counter- regulatory mechanisms. Overdose may be associated with significantly longer duration of action than therapeutic doses. Inhaled insulin has limited bioavailability, and significant hypogly- caemia is less likely. Insulin is rapidly degraded in the gastrointes- tinal tract, and ingestion is not expected to cause systemic effects. Sulphonylureas Sulphonylureas bind to a specific islet cell receptor and stimulate insulin secretion. Acute ingestion may cause significant hypogly- caemia; onset is normally within 8 hours of ingestion.1 Biguanides Metformin suppresses gluconeogenesis from lactate and pyru- vate, and enhances tissue sensitivity to insulin. Typically, met- formin ingestion does not cause hypoglycaemia, but may cause severe lactic acidosis.2 Thiazolidinediones Thiazolidinediones enhance sensitivity to insulin by stimulating peroxisomal proliferator-activated receptor gamma (PPAR-≥). Ingestion may cause mild hypoglycaemia in a small number of W Stephen Waring FRCPE PhD is a Consultant Toxicologist at the Royal Infirmary of Edinburgh, and honorary Senior Lecturer in Clinical Pharmacology at the University of Edinburgh. He has previously worked within the pharmaceutical industry, and research interests include early drug development, assessment of drug safety, and clinical aspects of poisoning. Conflicts of interest: none declared. patients.3 Hepatic injury is a recognized adverse effect of regular treatment, but risk after single acute ingestion is likely to be low. ±-glucosidase inhibitors These inhibit hydrolysis of complex carbohydrates and impair absorption from the small intestine. Gastrointestinal effects are common, including diarrhoea and abdominal pain. Hypoglycaemia is not a recognized feature. Glucagon-like peptide-1 analogues and dipeptidyl peptidase-IV inhibitors Glucagon-like peptide-1 (GLP-1) enhances insulin responsive- ness, and is degraded by dipeptidyl peptidase-IV (DPP-IV). Both GLP-1 analogues and DPP-IV antagonists are relatively new drug classes, and their toxicity is poorly characterized. Given that their effects are insulin-dependent, ingestion alone would not be expected to cause significant hypoglycaemia. Toxicity and acute features Poisoning by sulphonylureas, biguanides and insulin are asso- ciated with moderate-to-severe poisoning in 4.6%, 12.2%, and 14.7% respectively, and fatal poisoning in 0.9%, 6.1% and 3.6% respectively.4 Insulin and sulphonylurea overdose may cause significant hypoglycaemia. Features are due to both hypoglycaemia and activation of neurohormonal mechanisms, and include agitation, altered behaviour, excess sweating, slurred speech, tachycardia, seizures, and coma (Table 1). Metformin ingestion can be associated with abdominal pain, vomiting, and diarrhoea due to drug accumulation in the Poisonous substances Summary of toxic features associated with antidiabetic drug classes Class Insulin subcutaneous (short and long acting) Sulphonylureas, e.g. gliclazide, glipizide Biguanides, e.g. metformin Thiazolidinediones, e.g. pioglitazone, rosiglitazone ±-glucosidase inhibitors, e.g. acarbose, voglibose GLP-1 analogues, e.g. exenatide, liraglutide DPP-IV inhibitors, e.g. sitagliptin, vildagliptin Risk of hypoglycaemia Profound, may be prolonged for several days Profound, may be prolonged for hours to days negligible Low risk, mild negligible Low risk Low risk Other features Local injection site (parenteral bolus) severe lactic acidosis, gastrointestinal symptoms Possible hepatic dysfunction gastrointestinal tract and, in patients who develop severe lactic acidosis, agitation, confusion, tachypnoea and hypotension. Management Hypoglycaemia should be reversed by administration of intra- venous dextrose or glucose, aiming to achieve a target plasma glucose concentration of e"4 mmol/L (higher target concentra- tions may be needed to avoid symptoms in some patients with poorly controlled diabetes). Prolonged intravenous infusions can be required, and this is associated with the risk of hepatocel- lular glycogen accumulation and reversible liver dysfunction.5 Lactate provides an alternative substrate for brain metabolism, and lessens the risk of neuroglycopenia. Glucagon and octreotide antagonize the effects of insulin, and may reduce the quantity of carbohydrate administration needed to maintain euglycaemia.1,6 Surgical excision should be considered only in cases where very large amounts of insulin have been identified in a single depot and other approaches to management have not proved success- ful. There are insufficient data to establish whether any of these treatments confer a better outcome than systemic administration of dextrose alone. Patients who develop severe lactic acidosis require close observation of haemodynamic status, urinary output, serum electrolytes and acid base balance and, where appropriate, cor- rection of any fluid or electrolyte disturbance. Haemodialysis or filtration may be required for severe acidosis and renal failure in the context of metformin poisoning. Complications Profound hypoglycaemia may lead to persisting neurological deficits, for example cognitive impairment and delayed reaction times.7 Risks are greatest in those exposed to profound hypogly- caemia for a prolonged duration. Therefore, prompt restoration and maintenance of euglycaemia are indicated to minimize the risks of neuropsychiatric sequelae. Anti-HMGCR Autoantibodies in European Patients With Autoimmune Necrotizing Myopathies Inconstant Exposure to Statin Yves Allenbach, MD, PhD, Laurent Drouot, MSc, Aude Rigolet, MD, Jean Luc Charuel, PharmD, Fabienne Jouen, MD, Norma B. Romero, MD, Thierry Maisonobe, MD, Odile Dubourg, MD, Anthony Behin, MD, Pascal Laforet, MD, PhD, Tania Stojkovic, MD, Bruno Eymard, MD, PhD, Nathalie Costedoat-Chalumeau, MD, PhD, Emmanuelle Campana-Salort, MD, Anne Tournadre, MD, Lucile Musset, MD, Brigitte Bader-Meunier, MD, Isabelle Kone-Paut, MD, PhD, Jean Sibilia, MD, PhD, Laurent Servais, MD, PhD, Olivier Fain, MD, Claire Larroche, MD, Elisabeth Diot, MD, PhD, Benjamin Terrier, MD, PhD, Raphael De Paz, MD, Antoine Dossier, MD, Dominique Menard, MD, Chafika Morati, MD, Marielle Roux, MD, Xavier Ferrer, MD, Jeremie Martinet, PharmD, Sophie Besnard, MD, Remi Bellance, MD, Patrice Cacoub, MD, PhD, Laurent Arnaud, MD, PhD, Bernard Grosbois, MD, PhD, Serge Herson, MD, Olivier Boyer, MD, PhD, and Olivier Benveniste, MD, PhD, on behalf of the French Myositis Network Abstract: Necrotizing autoimmune myopathy (NAM) is a group of acquired myopathies characterized by prominent myofiber necrosis with little or no muscle inflammation. Recently, researchers identified autoantibodies (aAb) against 3-hydroxy-3-methylglutaryl-coenzyme A reductase (HMGCR) in patients with NAM, especially in statin-exposed patients. Here we report what is to our knowledge the first European cohort of patients with NAM. The serum of 206 patients with suspicion of NAM was tested for detection of anti-HMGCR aAb using an addressable laser bead immunoassay. Forty-five patients were found to be anti-HMGCR positive. Their mean age was 48.9 ¬± 21.9 years and the group was predominantly female (73.3%). Statin exposure was recorded in 44.4% of patients. Almost all patients had a muscular deficit (97.7%), frequently severe (Medical Research Council [MRC] 5 ‚â§3 in 75.5%). Subacute onset (<6 mo) was noted for most of them (64.4%). Nevertheless, 3 patients (6.6%) had a slowly progressive course over more than 10 years. Except for weight loss (20%), no extramuscular sign was observed. The mean CK level was high (6941 ¬± 8802 IU/L) and correlated with muscle strength evaluated by manual muscle testing (r = ‚àí0.37, p = 0.03). Similarly, anti-HMGCR aAb titers were correlated with muscular strength From the AP-HP, H√¥pital Piti√©-Salp√™tri√®re, Department of Internal Medicine 1 and Inflammation-Immunopathology-Biotherapy Department (I2B), East Paris Neuromuscular Diseases Reference Center, Inserm U974, Universit√© Pierre et Marie Curie, Paris 6, Paris (YA, AR, SH, O. Benveniste); Inserm, U905, Immunology Department, Normandie Univ, IRIB and Rouen University Hospital (LD, FJ, JM, O. Boyer); AP-HP, H√¥pital Piti√©-Salp√™tri√®re, Department of Immunochemistry, Universit√© Pierre et Marie Curie, Paris 6, Paris (JLC, LM); AP-HP, H√¥pital Piti√©-Salp√™tri√®re, Institute of Myology, Unite de Morphologie Musculaire, CNRS-UMR7215, Paris (NBR); APHP, H√¥pital Piti√©-Salp√™tri√®re, Department of Neuropathology, Universit√© Pierre et Marie Curie, Paris 6, Paris (TM, OD); AP-HP, H√¥pital Piti√©Salp√™tri√®re, Department of Neurology, East Paris Neuromuscular Diseases Reference Center, Universit√© Pierre et Marie Curie, Paris 6, Paris (AB, PL, TS, BE); AP-HP, H√¥pital Cochin Centre de R√©f√©rence Maladies AutoImmunes et Syst√©miques Rares, Service de M√©decine Interne P√¥le M√©decine, Universit√© Ren√© Descartes Paris V, Paris (NCC, BT); APHM, H√¥pital la Timone, Centre de R√©f√©rence des Maladies Neuromusculaires, Marseille (ECS); Clermont-Ferrand University Hospital, Department of Rheumatology, Clermont-Ferrand (AT); AP-HP, H√¥pital Necker, Department of Paediatric Rheumatology, Paris (BBM); AP-HP, H√¥pital Bic√™tre, Department of Paediatric Rheumatology, Kremlin-Bic√™tre, Paris (IKP); CHU Strasbourg, Department of Rheumatology, Strasbourg (JS); H√¥pital Piti√©Salp√™tri√®re, Institute of Myology, Paris (LS); AP-HP, H√¥pital Jean Verdier, Department of Internal Medicine, La Seine-Saint-Denis (OF); AP-HP, 150 www.md-journal.com (r = ‚àí0.31; p = 0.03) and CK level (r = 0.45; p = 0.01). Mean duration of treatment was 34.1 ¬± 40.8 months, and by the end of the study no patient had been able to stop treatment. This study confirms the observation and description of anti-HMGCR aAb associated with NAM. The majority of patients were statin naive and needed prolonged treatments. Some patients had a dystrophic-like presentation. Anti-HMGR aAb titers correlated with CK levels and muscle strength, suggesting their pathogenic role. (Medicine 2014;93: 150‚Äì157) Abbreviations: aAb = autoantibodies, ALBIA = addressable laser bead immunoassay, AU = arbitrary units, C5b‚Äì9 = membranolytic attack complex, CK = creatine kinase, DMARD = disease-modifying antirheumatic drugs, HMGCR = 3-hydroxy-3-methylglutaryl-coenzyme A reductase, IVIg = intravenous immunoglobulin, MHC = major histocompatibility complex class I antigen up-regulation, MMT = manual muscular testing, MRC = Medical Research Council, NAM = necrotizing autoimmune myopathy, SRP = signal recognition particle. H√¥pital Avicenne, Department of Internal Medicine, Bobigny (CL); CHRU Tours, Department of Internal Medicine, Tours (ED); Fondation A. de Rothschild, Department of Neurology, Paris (RDP); AP-HP, H√¥pital Bichat, Department of Internal Medicine, Paris (AD); CHU Rennes, Department of Neurology, Rennes (DM); Centre Hospitalier de la Region d‚ÄôAnnecy, Department of Internal Medicine, Annecy (CM); H√¥pital Pierre Oudot, Department of Internal Medicine, Bourgouin (MR); H√¥pital du Haut Lev√™que, Department of Neurology, Bordeaux (XF); CHU Rennes, Department of Internal Medicine, Rennes (SB, BG); CHU Fort de France, Department of Neurology, Fort de France (RM); AP-HP, H√¥pital Piti√©-Salp√™tri√®re, Departement HospitaloUniversitaire I2B, UPMC Univ Paris 06, UMR 7211, INSERM, UMRS 959, Department of Internal Medicine 2, Paris (PC, LA); France. Financial support and conflicts of interest: The authors have no funding or conflicts of interest to disclose. Reprints: Yves Allenbach, MD, PhD, AP-HP, H√¥pital Piti√©-Salp√™tri√®re, Department of Internal Medicine 1 and Inflammation-Immunopathology-Biotherapy Department (I2B), East Paris Neuromuscular Diseases Reference Center, Inserm U974, Universit√© Pierre et Marie Curie, Paris 6, 75013, Paris, France (e‚Äêmail: yvesallenbach@gmail.com). Copyright ¬© 2014 by Lippincott Williams & Wilkins ISSN: 0025-7974 DOI: 10.1097/MD.0000000000000028 Medicine ‚Ä¢ Volume 93, Number 3, May 2014 Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine ‚Ä¢ Volume 93, Number 3, May 2014 INTRODUCTION N ecrotizing myopathies are characterized by predominant muscle fiber necrosis and regeneration with little or no inflammation. Among the inflammatory myopathies, this histologic pattern defined a new subgroup called immune-mediated necrotizing myopathy7 or necrotizing autoimmune myopathy (NAM). The immune-mediated nature of these myopathies was first suggested by their response to immunosuppressive treatments4,8,14 and their frequent association with anti-signal recognition particle (anti-SRP) autoantibodies (aAb).8,14 Nevertheless, anti-SRP aAb are observed in only 16% of necrotizing myopathies.5 Remarkably, Mammen and colleagues5,17 recently identified a new specific autoantibody recognizing 200 and 100 kDa proteins, initially in 16 United States patients (61%) with pathologic features of necrotizing myopathy without any known myositis-specific antibody. This United States cohort was then expanded by 29 patients (n = 45) in whom the 100 kDa protein was identified as the target of statins 3-hydroxy-3methylglutaryl-coenzyme A reductase (HMGCR).9 AntiHMGCR-positive (Anti-HMGCR+) patients were in their fifth decade of life9,11 and were very frequently exposed to statin (72.7% of cases).11 They presented proximal and symmetric muscular deficit with high creatine kinase (CK) levels (mean, 9718 ¬± 7383 IU/L) and responded to immunosuppressants.9 Since anti-HMGCR aAb were not observed in a large cohort of statin-exposed patients, including patients with self-limited statin intolerance,10 anti-HMGCR aAb are considered specific for NAM.10 To validate this concept in a different cohort of patients by using an alternative independent testing assay, we report herein a second cohort, equivalent in size, of anti-HMGCR+ patients with their clinico-pathologic features. METHODS Anti-HMGCR Detection and Quantitative Titration For anti-HMGCR aAb detection and quantification of serial titers, we recently developed a new addressable laser bead immunoassay (ALBIA), using the same strategy as that previously used for anti-SRP aAb quantification.2 The technical description and diagnostic value of this assay are detailed elsewhere.5a Briefly, the catalytic domain of recombinant human HMGCR protein fused to a glutathione S transferase tag was obtained from Sigma (Saint Louis, MO) and coupled to Bio-PlexR COOH beads with the Bio-PlexR amine coupling kit according to the manufacturer‚Äôs protocol (Bio-Rad, Hercules, CA); 1250 HMGCR-coated beads were added to 100 ŒºL of patient or control serum for 2 hours under shaking. Biotinylated mouse anti-human IgG-specific secondary Ab (Southern Biotech, Birmingham, AL) was added, and, after washing, beads were further incubated with streptavidin-R-phycoerythrin (Qiagen). Blank (no serum, secondary antibody only), negative controls (anti-HMGCR Ab-negative serum), and positive controls (highly positive human anti-HMGCR Ab serum) were included in every assay. Mean fluorescence intensity (MFI) was determined on a Bio-Plex apparatus using the Bio-Plex Manager Software 4.0 (Bio-Rad Laboratories, Hercules, CA). Anti-HMGCR Ab titers were determined at a 1/D dilution using the following formula: [MFI serum/MFI calibrator] √Ç [titer of calibrator] √Ç D/500. The calibrator is a highly positive human anti-HMGCR Ab serum (the same throughout the study) whose titer was arbitrarily set to 100 arbitrary units (AU/mL). ¬© 2014 Lippincott Williams & Wilkins Anti-HMGCR Autoimmune Necrotizing Myopathies We tested the first 37 sera that were found positive with this assay, and all 37 sera also scored positive for anti-HMGCR aAb using the recently developed Myositis Euroline 7 line immunoassay from Euroimmun (L√ºbeck, Germany). The assay was always negative when testing control sera from healthy blood donors (n = 100) or from patients with different inflammatory/ autoimmune diseases (n = 142), according to established classification criteria: American College of Rheumatology revised criteria for systemic lupus erythematosus13 with anti-dsDNA aAb, American Rheumatism Association criteria for rheumatoid arthritis1 with anti-CCP antibodies and/or rheumatoid factor, revised European criteria for primary Sj√∂gren syndrome16 with anti-SSA and/or anti-SSB aAb, Bohan and Peter criteria3 for dermatomyositis, Troyanov criteria for overlap myositis15 with anti-tRNA-synthetase Ab, and Griggs criteria for inclusionbody myositis.6 Patients and Data Collection We tested first the sera of 23 patients having the diagnosis of NAM based on clinical and histologic criteria. These patients were recruited by clinicians (YA, AR, AB, PL, TS, BE, SH, and OB) from the neuromuscular diseases reference center of East Paris. The positive first results were communicated to the French Myositis Network. Sera of 183 patients (including 8 pediatric patients) selected by clinicians of the network were sent to what was until June 2013 the sole French immunology laboratory able to detect anti-HMGCR. These 183 patients had a suspicion of NAM based on clinical and/or pathologic criteria. All these patients were also selected because of their negativity for myositis-specific aAb anti-Jo-1 and anti-SRP using a commercial kit (dTek or Euroimmun). Furthermore, all patients diagnosed as anti-HMGCR aAb positive were also tested for the presence of the following myositis-specific aAb: anti-PL-7, anti-PL-12, anti-SRP, anti-PMSCl, and anti-Mi2 (using a commercial kit, either dTek or Euroimmun). We controlled negativity for anti-SRP aAb detection using analytical sensitivity and specificity of ALBIA‚ÄìSRP.2 All patients determined to be anti-HMGCR aAb positive by ALBIA-HMGCR from July 2012 to June 2013 were recorded for this study. For all patients, the following parameters were recorded: age, sex, past medical history, statin exposure, characteristics of myopathy including date of diagnosis, clinical manifestation with muscular deficit evaluation (by muscle manual testing [MMT] using the Medical Research Council [MRC] scale on 5 points), CK level, and reports of muscle biopsy. Biopsies of 20 patients were performed following standardized procedures and were analyzed by the same pathologists exclusively specialized in muscle pathology (OD and TM) permitting semiquantitative analyses and standardized immunohistochemistry analyses. In those cases, muscle tissue was frozen and stored at ‚àí80¬∞C. On 8 Œºm-thick cryosections, morphologic analysis (hematin and eosin counter staining) and immunohistochemistry analyses were performed. Features such as the presence of necrosis/regeneration fibers, inflammation, major histocompatibility complex (MHC) class I antigen upregulation, membranolytic attack complex (C5b‚Äì9) deposits and vacuoles were recorded. First-line treatments including corticosteroid and diseasemodifying anti-rheumatic drugs (DMARD) were recorded as well as treatment intensification during the follow-up. Treatment intensification was defined as an introduction or an increase of more than 50% of the dose of corticosteroids and/or the introduction of DMARD. www.md-journal.com Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 151 Medicine ‚Ä¢ Volume 93, Number 3, May 2014 Allenbach et al TABLE 1. Characteristics of 45 Anti-HMGCR+ Patients Characteristic Age, yr; mean Sex ratio (M:F) Statin exposure Muscular involvement Muscular deficit Subacute onset Progressive onset Severe deficit (‚â§3) Myalgia Dysphagia CK level Extraskeletal muscular involvement Weight loss Interstitial lung disease Cardiac insufficiency Arthralgia Raynaud phenomenon 48.9 ¬± 21.9 0.36 44.4% (n = 20) 97.7% (n = 44) 64.4% (n = 29) 33.3% (n = 15) 75.5% (n = 34) 53.3% (n = 24) 26.7% (n = 12) 6941 ¬± 8802 IU/L 20% (n 2.2% (n 2.2% (n 11.1% (n 11.1% (n = 9) = 1) = 1) = 5) = 5) For this medical research where we used identifiable human sera and data, we obtained agreement from the French Ministry of Research (AC 2008-87) for the collection, analysis, storage, and reuse. All samples were obtained for diagnostic purposes. In this situation of retrospective study, patients‚Äô consents were impossible or impractical to obtain. We conducted the research after consideration and approval of the research ethics committee of Piti√©-Salp√™tri√®re Hospital. Statistical Analysis Categorical variables are reported as numbers and/or percentages and were compared using a chi-square or, when appropriate, Fisher exact test. Quantitative variables are reported as mean (¬± standard deviation) and compared using a nonparametric test. For all statistical analyses, p < 0.05 was considered significant. Correlation analysis was performed using the Spearman test. Statistical analyses were conducted using Prism software. RESULTS CK level. Four other young girls were initially diagnosed as having limb girdle muscular dystrophy (LGMD) because of a slowly progressive muscle deficit. Autoimmune diseases were noted in the past medical history of 5 patients (thyroiditis, n = 4, and type I diabetes, n = 1). Statin exposure was present in 20 (44.4%) patients. These patients received mainly atorvastatin (n = 10/20) and rosuvastatin (n = 4/20). Statin-exposed patients were older than statin-naive patients (mean age, 64.4 ¬± 6.8 yr vs. 36.6 ¬± 21.7 yr, respectively; p = 0.001). Muscular Deficit and CK Level in Anti-HMGCR+ Patients The circumstances of discovery of NAM were mainly signs related to muscle weakness (n = 37). Otherwise, 3 patients declared having isolated myalgia, and 4 consulted for increased CK level without any other symptoms. For 1 patient, antiHMGCR presence was suggested by a cytoplasmic fluorescence pattern on antinuclear screening test for the etiology of deep venous thrombosis. During the course of the disease, with a mean follow-up of 49 ¬± 22 months, 53.3% of patients (n = 24/45) suffered from myalgia, and 97.7% of patients (n = 44/45) had a muscle weakness. Subacute onset (<6 mo) was noted for most of them (see Table 1); nevertheless, 33.3% of patients (n = 15/45) had a slow progressive muscular deficit ranging from 10 months to years. Of note, 3 cases had been considered to be muscular dystrophy for more than 10 years prior to the detection of antiHMGCR aAb. In most of the cases (n = 34/45, 75.5%), MMT showed a severe proximal deficit (MRC score ‚â§3/5 for the weakest muscular group), and 5 patients were bedridden at diagnosis (see Table 1). Axial weakness and dysphagia were found in 35.5% (n = 16/45) and 26.7% (n = 12/45) of patients, respectively. No facial deficit was observed. Marked muscle atrophy was noted for 22.2% of patients (n = 10/45), and 2 patients also had scapular winging. All patients had increased CK levels (mean, 6941 ¬± 8802 IU/L). Furthermore, CK level correlated with muscular strength (r = ‚àí0.37, p = 0.03) according to the weakest muscle groups (Figure 1). Except for the onset age, no significant difference was observed in statin-exposed patients versus nonexposed patients Characteristics and Statin Exposure in Anti-HMGCR+ Patients Forty-five anti-HMGCR+ patients, coming from 17 different French hospitals in the myositis network, were included in the study. This represented 21.8% of the total of tested patients (n = 45/206), 26% of the first series of patients (n = 6/23), and 21.3% of the second series of patients (n = 39/183). Most patients were white, except 3 African and 1 Asian patients. Patients had a mean age of 48.9 ¬± 21.9 years at the time of the first signs of the disease. They were predominantly female (n = 33/45; 73%) (Table 1). One patient had onset of the disease at 3 months of pregnancy, and a second at immediate childbirth. Eight pediatric cases (age range, 4-16 yr) were observed. Of note, only 1 patient among the 8 was tested and found to be anti-HMGCR+ before the age of 16 years. For the 7 others, the diagnosis was performed in adulthood. Four girls were initially diagnosed as having an inflammatory myopathy (polymyositis, n = 1; dermatomyositis, n = 2; NAM, n = 1) because of rapid progressive scapular and pelvic deficit with high 152 www.md-journal.com FIGURE 1. Correlation between CK level and muscular deficit in patients with NAM. For each patient the muscular deficit is given as the Medical Research Council score of the weakest muscular group using manual muscular testing (MMT MRC). ¬© 2014 Lippincott Williams & Wilkins Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine ‚Ä¢ Volume 93, Number 3, May 2014 Anti-HMGCR Autoimmune Necrotizing Myopathies concerning percentage of myalgia, severity of muscular deficit, percentage of dysphagia, and mean CK level (data not shown). Histologic Necrotizing Myopathy Pattern All but 2 patients had a muscle biopsy (n = 43) performed for the diagnosis of the myopathy. Morphologic analysis showed for 42 patients (97.6%) a necrotizing myopathy pattern defined by the presence of fibers with necrosis and/or regeneration and no or few inflammatory infiltrates. Results of 20 muscular biopsies performed in our reference center for neuromuscular disorders following standardized procedures and analyzed by the same myo-pathologists are represented in Table 2. Necrosis and/or regenerative phenomena varied from 1 patient to another, ranging from few to intense muscular necrosis (see Table 2 and Figure 2A and B). This pattern was associated with irregularity of the size of fibers in 90% of cases (n = 18/20) due to presence of atrophic fibers with nonspecific repartition, notably in the perimysium (see Table 2 and Figure 2C). Most patients (60%, n = 12/20) did not have muscular infiltrates. Six patients had small muscle infiltrates, and only 2 had mild (but obvious) inflammatory infiltrates (see Figure 2C). C5b9 deposit was frequently observed on necrotic fibers but also at the membrane of few normal fibers (65%, n = 13/20) (see Table 2 and Figure 2D). Deposits of C5b9 decorating a few muscle capillaries were occasionally observed (25%; n = 5/20) but not as dramatically as is typically seen in dermatomyositis. Inflammation, when present, was usually in perivascular areas. Overexpression of MHC class I by fibers was mostly observed on regenerative or necrotic fibers (Figure 2E). Nevertheless, sometimes (n = 8) a few normal fibers, in focal area, had MHC class I overexpression. Only 2 patients (including the 1 with the most obvious inflammatory infiltrates) had diffuse and intense MHC class I overexpression (Figure 2F), as can be observed in polymyositis or inclusion body myositis. Extramuscular Involvement Some patients (20%; n = 9/45) had fatigue associated with weight loss (range, 2‚Äì18 kg). Twenty-one percent of patients (n = 6/28) had increased C-reactive protein serum levels (range, 6‚Äì79 mg/mL). Among the 27 patients who had a computed thoracic scan, none had an interstitial lung disease, except 1 patient in whom an idiopathic interstitial lung disease was diagnosed by a lung biopsy 8 years before the onset of the myopathy. Among the 27 patients who had pulmonary function tests, 7 (25.9%) TABLE 2. Histologic Muscle Analysis* Patient Fiber Necrosis and Disparity of Muscular Regeneration Fiber Size Infiltrates Topography of Infiltrates C5b9 Deposits ‚àí Capillaries and fibers Fibers 1 ++ Yes ‚àí 2 ++ Yes + 3 4 5 6 7 8 + +++ + + ++ No Yes No Yes Yes Yes ‚àí + ‚àí ‚àí ‚àí ‚àí Pericapillary and perinecrotic ‚àí Pericapillary ‚àí ‚àí ‚àí ‚àí 9 +++ Yes + Pericapillary 10 11 12 ++ ++ +++ Yes Yes Yes ‚àí + ++ 13 ++ Yes ++ 14 15 16 17 ++ + ++ +++ Yes Yes Yes Yes ‚àí + ‚àí + ‚àí Pericapillary Endomysial and pericapillary Endomysial and pericapillary ‚àí Pericapillary ‚àí Endomysial 18 19 +++ +++ Yes Yes ‚àí ‚àí ‚àí ‚àí 20 +++ Yes ‚àí ‚àí ‚àí ‚àí Fibers Diffuse Fiber MHCI Endomysial Rimmed Overexpression Fibrosis Vacuoles No No No No No No No No No No No No No Yes No No No No No No No No No No ‚àí ‚àí Capillaries and fibers Capillaries and fibers ‚àí ‚àí Fibers No No No No No Yes Yes No Yes No No Yes Fibers No No No No No No No Yes No No No No No No No Yes No No No Yes No No No No Fibers ‚àí Fibers Capillaries and fibers Fibers Capillaries and fibers Fibers *Table shows the results of 20 patients with muscular biopsy following the same standardized procedures performed in 1 center. Semiquantitative analysis of muscular necrosis and regeneration were performed (+ corresponds to <10 necrosis and/or regenerative fibers, ++ corresponds to 10‚Äì30, and +++ corresponds to >30 per biopsy). C5b9 deposits were considered only on non-necrotic fibers or capillaries. ¬© 2014 Lippincott Williams & Wilkins www.md-journal.com Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 153 Medicine ‚Ä¢ Volume 93, Number 3, May 2014 Allenbach et al FIGURE 2. Histologic and immunohistologic analysis of muscle biopsies from anti-HMGCR+ patients. A, Muscle biopsy from a NAM patient showing rare necrotizing fiber (arrow) with some atrophic fibers (arrowheads) (hematin eosin stain). B, Muscle biopsy from another patient showing a high number of necrotized fibers (arrow) with important numbers of regenerative fibers (arrowhead). C, Muscular infiltrates are sometimes present (arrow) and in low intensity as represented here (arrow) for 1 patient. D, Immunohistologic analysis showing C5b9 deposits decorating some necrotic and non-necrotic fibers (arrows). E, Immunohistologic analysis of MHC class I expression showing a representative case of MHC overexpression on 2 fibers (arrows). F, A few patients may have diffuse and intense MHC overexpression not only on necrotic fibers (arrows) but also on non-necrotic fibers (arrowheads). had a restrictive syndrome with a decrease of forced vital capacity (range, 33%‚Äì42% of predicted volumes). On electrocardiogram, 4 patients (12.9%; n = 4/31) had conduction abnormalities (right branch block, n = 2; left hemi branch block, n = 2). One patient had an auricular fibrillation, and 1 an atrial flutter. For these 2 patients, 1 (aged 72 yr) was diagnosed 3 years before the onset of the myopathy and the other in the context of venous thromboembolism. Among 25 patients who had an echocardiogram, only 1 showed a decrease of ejection fraction (measured at 20% of normal); this patient was known to have an idiopathic dilated cardiomyopathy for more than 20 years before the onset of the skeletal myopathy. Seven patients had Raynaud phenomenon (n = 5/45) or acrosyndrome (n = 2/45), and 5 patients had arthralgia without synovitis. No patient had any association with another connective tissue disease, nor presented other specific aAb. Five patients had cancer during their medical history. The cancers were diagnosed 19 years (breast adenocarcinoma), 4 years (breast adenocarcinoma), 12 months (melanoma), 6 months 154 www.md-journal.com (renal adenocarcinoma), and 2 months (pulmonary adenocarcinoma) after the first signs of the NAM. Treatment and Evolution Thirty-nine of the 45 patients (86.6%) received immunosuppressive drugs and/or immune-modulatory treatment. Six patients (statin-exposed, n = 4) were not treated. One non-statinexposed patient without muscular deficit was not treated, but follow-up time by the end of the study was less than 6 months. Two patients refused the treatment. Three remaining statinexposed patients did not receive any treatment. They slowly improved after statin withdrawal without any other intervention. One had complete remission (normal strength and CK level) 8 months after statin withdrawal, and because she had severe familial hypercholesterolemia, statins were reintroduced without any muscular symptom or CK increase. Another patient improved in strength (with a MRC score at 4 vs. 3 concerning the weakest muscular group) and decreased CK level (500 IU/L vs. ¬© 2014 Lippincott Williams & Wilkins Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine ‚Ä¢ Volume 93, Number 3, May 2014 Anti-HMGCR Autoimmune Necrotizing Myopathies TABLE 3. Treatments Used for NAM Patients Patient Number First-Line Treatment 2 3 4 6 7 8 9 11 12 13 14 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 CT + IVIg + MTX CT + MTX MTX + CT CT + MTX + PE + IVIg CT + MTX CT + MTX CT + MTX + PE + IVIg CT CT + IVIg CT + MTX CT + IVIg + MTX CT + IVIg CT + MTX CT + MTX CT + IVIg + MTX CT + EDX CT + MTX CT CT CT CT CT + MTX CT CT + IVIg + MTX CT CT + IVIg CT IVIg IVIg CT + MTX CT CT CT + MTX + IVIg CT + MTX CT CT CT CT + MTX CT + MTX Number of Treatment Intensification Disease-Modifying Antirheumatic Drugs 0 0 0 1 0 2 0 5 4 4 2 4 1 0 0 0 3 5 6 1 10 1 1 2 2 1 1 0 0 3 3 3 0 1 1 1 1 0 1 none none none CT/MTX/RTX none CT/MMF/TACRO/PE/MTX/IVIg none CT/AZA/MTX/CYC/MMF/IVIg CT/AZA/MTX/RTX/PE/IVIg CT/IVIg/RTX CT/AZA/IVIg/RTX CT/MTX/IVIg/MMF/RTX CT/MTX none none none CT/AZA CT/IVIg/AZA/MTX/RTX CT/IVIg/MTX/AZA/MMF CT/AZA CT/IVIg/AZA/CYC/TACRO/RTX IVIg CT/IVIg/PE/MTX IVIg MTX/PE/IVIg/RTX CT/EDX IVIg none none CT/MTX/IVIg/MMF/RTX/AZA CT/MMF CT/EDX/IVIg CT/MTX/IVIg CT/MTX/IVIg IVIg CT/AZA MTX none CT/MTX Treatment Duration (Months) 3 1 3 3 3 24 18 114 41 42 40 62 36 4 1 3 80 180 90 34 120 18 5 8 11 23 12 7 7 42 89 36 4 3 50 6 84 11 12 Abbreviations: AZA = azathioprine, CT = corticosteroid, CYC = cyclosporine, EDX = cyclophosphamide, IVIg = intravenous immunoglobulin, MMF = mycophenolate mofetil, MTX = methotrexate, PE = plasmapheresis, RTX = rituximab, TACRO = tacrolimus. 5000 IU/L) at month 4. The remaining patient was diagnosed 18 months after statin withdrawal. At this period, muscle strength was 3 for the weakest muscular group and CK level was 500 IU/mL, whereas it was 2500 IU/mL 18 months before. Four months later, the patient‚Äôs strength spontaneously increased from 3 to 4, myalgia disappeared, and CK level decreased to 240 IU/L. For the other statin-exposed patients (n = 16), statin was removed and immunosuppressive drugs were initiated. Of note for 2 patients, statin was reintroduced without significant flare. Mean duration of treatment for 39 treated patients was 34.1 ¬± 40.8 months (range, 1‚Äì180 mo). By the end of the study, no ¬© 2014 Lippincott Williams & Wilkins patient had been able to stop the treatment for a prolonged time over 1 year because of relapse. For the first line of treatment all patients but 2 were treated with corticosteroids, frequently associated with immunosuppressive drugs (methotrexate, n = 20; cyclophosphamide, n = 1) or intravenous immunoglobulin (IVIg, n = 10, Table 3). One bedridden patient was also treated by plasmapheresis as firstline treatment. All patients but 1 who were treated for more than 1 year required intensification by DMARD (see Table 3). Because of flares or insufficient control of the NAM, the mean number of treatment intensifications was 1.8 ¬± 2.1 (range, 1‚Äì10). www.md-journal.com Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 155 Medicine ‚Ä¢ Volume 93, Number 3, May 2014 Allenbach et al FIGURE 3. Correlation between muscular deficit and CK level with anti-HMGCR aAb titer. For each patient, we analyzed the correlation between the anti-HMGCR aAb titer and muscular deficit given as the MRC score of the weakest muscular group using manual muscular testing (MMT MRC) (A) or CK level (B). DMARD were methotrexate (n = 15), azathioprine (n = 10), mycophenolate mofetil (n = 6), cyclosporine (n = 2), tacrolimus (n = 2), cyclophosphamide (n = 2), rituximab (n = 9), plasmapheresis (n = 3), IVIg (n = 17). The mean treatment duration and the number of treatment intensifications was not significantly different between statin-exposed and statin-naive patients (32.2 ¬± 39.4 mo vs. 30.8 ¬± 32.3, respectively; p = 0.83; and 1.9 ¬± 1.8 vs. 1.1 ¬± 1.5, respectively; p = 0.31). By the end of the study only 1 patient had died, due to aspiration pneumonia; this patient was bedridden. Titer of Anti-HMGCR Finally, in an attempt to indirectly investigate a possible physiopathogenic role of anti-HMGCR, we looked for correlation between aAb titer and the muscular parameters: strength and CK level. We observed a significant correlation (r = ‚àí0.31; p = 0.03) between the anti-HMGCR aAb titer and the MRC score of the weakest muscle groups (Figure 3A). Similarly, we observed a significant correlation (r = 0.45; p = 0.01) between the aAb titers and CK level tested the same day (Figure 3B). Of note, no significant difference was observed between HMGCR titer in statin-exposed and statin-naive patients (data not shown). Differentiating statin-exposed and statin-naive patients, we did not observe a significant correlation between CK levels and aAb titers (r = 0.6; p = 0.06 and r = 0.38; p = 0.09, respectively). But a significant correlation between muscle strength and aAb titers was observed in the statin-naive group (r = 0.56; p = 0.01), which we did not observe in the statin-exposed group (r = 0.06; p = 0.6). DISCUSSION In this study describing the second case series of antiHMGCR NAM patients (n = 45), we found that both children and adults may be affected by a severe muscular deficit with an acute or a more slowly progressive onset, without extramuscular involvement in most cases. Most patients required several lines of treatment over a prolonged period. The high CK level correlates with muscular strength, and the titer of antiHMGCR aAb correlates with muscular strength and CK level. We observed a marked female predominance (73%) that the previous study did not observe (57.8%).9 Of note, the onset of disease in 2 patients (5%) was during or just after pregnancy, which is a rare situation in other forms of autoimmune myopathies.11 The 2 newborns did not present any sign of myopathy. 156 www.md-journal.com Eight of the 45 cases involved patients aged younger than 16 years. This observation and the 2 pediatric cases reported by Mammen et al9 suggest that anti-HMGCR NAM should be considered in the differential diagnosis in children with myopathy, as was previously shown for anti-SRP NAM.12 Dermatomyositis is not the sole inflammatory myopathy occurring during childhood. We observed that the onset of the muscular deficit in antiHMGCR+ patients may be slow, and 3 patients had a disease progression for more than 10 years before the diagnosis. Typically, these patients were initially considered to have a limb girdle muscular dystrophy with no molecular diagnosis. Indeed, muscle histologic analysis may show a dystrophic pattern with necrosis/regeneration, associated with irregular size of fibers and endomysial fibrosis. This pitfall was also described in antiSRP NAM.2 It led us and others12 to recommend testing for anti-SRP and anti-HMGCR aAb in patients presenting clinical and pathologic features compatible with limb girdle muscular dystrophy with no molecular diagnosis. Of note, muscular inflammation may also be observed on muscle biopsy. Yet, our cohort of anti-HMGCR+ subjects did not include patients with prominent inflammatory cell infiltrates, since anti-HMGCR aAb were only looked for in patients suspected of having NAM and, presumably, patients with significant degrees of inflammation were not screened for anti-HMGCR antibodies. Contrary to the case of other overlap syndromes associated with myositis-specific antibodies, such as antisynthetases, we did not observe clear extramuscular involvement in antiHMGCR+ patients. For example, no patient had interstitial lung disease. This is in line with the observations of Mammen et al.9 Only 44% of patients were statin exposed in the current study, compared to the 72.7% frequency reported in the study by Mammen et al.17 While we also observed that statin-naive patients were younger, we did not find any other difference between the statin-naive and statin-exposed groups, whereas Mammen et al9 found that mean CK levels were lower in statin-exposed patients. The difference in the prevalence of statin exposure in our study compared to Mammen‚Äôs study may be explained by the fact that we studied a different population. For example, our study is a collaborative work involving 2 pediatric and 2 dystrophic centers. Thus, our cohort is younger on average (aged 44 ¬± 19 vs. 52 ¬± 16 yr) and includes subjects who were initially considered to have an inherited myopathy. Another difference is that Mammen and colleagues9 screened their entire cohort of subjects suffering from ¬© 2014 Lippincott Williams & Wilkins Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine ‚Ä¢ Volume 93, Number 3, May 2014 inflammatory myopathies for anti-HMGCR antibodies, whereas we screened mainly those suspected of having NAM. Three statin-exposed patients improved within months after statin withdrawal without any other intervention. Such cases were not previously described. It may suggest that some muscular statin-intolerant patients are in fact real NAM cases. Mammen et al10 tried to address this point and showed that none of 51 patients with self-limited statin intolerance was antiHMGCR aAb positive. Nevertheless, anti-HMGCR+ NAM diagnosis cannot be ruled out in patients diagnosed as ‚Äústatinintolerant‚Äù especially if a high CK level is observed. This difference may be due to the technology used for the detection of anti-HMGCR aAb and/or patient recruitment, since statinintolerant had a mild increase in CK level (131.44 ¬± 71.07 IU/L) as compared to our 3 patients with CK level above 1000 IU/L. Anti-HMGCR treated patients had long treatment duration, in line with those previously reported.17 At the study‚Äôs end, all patients continued their treatment and almost everyone required DMARD (ranging from 1 to 10) for intensification, suggesting that anti-HMGCR NAM is a severe disease justifying prolonged treatment. The correlation between CK level and muscular strength suggests that CK level monitoring may be a good surrogate biomarker of disease activity. Nevertheless, the correlation coefficient was not high. This is not surprising since other parameters may be involved in muscular deficit, such as uncontrolled disease duration and/or muscular adipose involution. We also confirmed what Mammen and colleagues17 first reported, that anti-HMGCR titers correlated with muscle strength and CK levels. Together, these results argue for a possible pathogenic role of anti-HMGCR in the pathophysiology of this condition, and mimic what we also observed in anti-SRP+ patients.2 To conclude, to our knowledge this is the first study outside the United States to confirm the observation and description of anti-HMGCR+ NAM. We showed a marked female predominance, the existence of pediatric cases, and that disease onset may be insidious, leading to an erroneous diagnosis of muscle dystrophy. The majority of our patients were not exposed to statin. The CK level may be a good biomarker for the follow-up of patients who most often need prolonged treatments because of the risk of disease flare, in a disease where antiHMGCR aAb may have a role in the physiopathology. ACKNOWLEDGMENT The authors are indebted to Andrew Mammen for critical reading of the manuscript and helpful discussions. REFERENCES 1. Arnett FC, Edworthy SM, Bloch DA, McShane DJ, Fries JF, Cooper NS, Healey LA, Kaplan SR, Liang MH, Luthra HS, Medsger TA Jr, Mitchell DM, Neustadt DH, Pinals RS, Schaller JG, Sharp JT, Wilder RL, Hunder GG. The American Rheumatism Association 1987 revised criteria for the classification of rheumatoid arthritis. Arthritis Rheum. 1988;31:315‚Äì324. 2. Benveniste O, Drouot L, Jouen F, Charuel JL, Bloch-Queyrat C, Behin A, Amoura Z, Marie I, Guiguet M, Eymard B, Gilbert D, Tron F, Herson S, Musset L, Boyer O. Correlation of anti-signal recognition particle autoantibody levels with creatine kinase activity in patients with necrotizing myopathy. Arthritis Rheum. 2011;63:1961‚Äì1971. 3. Bohan A, Peter JB. Polymyositis and dermatomyositis (first of two parts). N Engl J Med. 1975;292:344‚Äì347. 4. Bronner IM, Hoogendijk JE, Wintzen AR, van der Meulen MF, 9Linssen WH, Wokke JH, de Visser M. Necrotising myopathy, an ¬© 2014 Lippincott Williams & Wilkins Anti-HMGCR Autoimmune Necrotizing Myopathies unusual presentation of a steroid-responsive myopathy. J Neurol. 2003;250:480‚Äì485. 5. Christopher-Stine L, Casciola-Rosen LA, Hong G, Chung T, Corse AM, Mammen AL. A novel autoantibody recognizing 200-kd and 100-kd proteins is associated with an immune-mediated necrotizing myopathy. Arthritis Rheum. 2010;62:2757‚Äì2766. 5a. Drouot L, Allenbach Y, Jouen F, Charuel JL, Martinet J, Meyer A, Hinschberger O, Bader-Meunier B, Kone-Paut I, Campana-Salort E, Eymard B, Tournadre A, Musset L, Sibilia J, Marie I, Benveniste O, Boyer O. Exploring necrotizing autoimmune myopathies with a novel immunoassay for anti-3-hydroxy-3-methyl-glutaryl-CoA reductase autoantibodies. Arthritis Res Ther. 2014;16:R39. Epub ahead of print. 6. Griggs RC, Askanas V, DiMauro S, Engel A, Karpati G, Mendell JR, Rowland LP. Inclusion body myositis and myopathies. Ann Neurol. 1995;38:705‚Äì713. 7. Hoogendijk JE, Amato AA, Lecky BR, Choy EH, Lundberg IE, Rose MR, Vencovsky J, de Visser M, Hughes RA. 119th ENMC international workshop: trial design in adult idiopathic inflammatory myopathies, with the exception of inclusion body myositis, 10‚Äì12 October 2003, Naarden, The Netherlands. Neuromuscul Disord. 2004;14:337‚Äì345. 8. Kao AH, Lacomis D, Lucas M, Fertig N, Oddis CV. Anti-signal recognition particle autoantibody in patients with and patients without idiopathic inflammatory myopathy. Arthritis Rheum. 2004;50:209‚Äì215. 9. Mammen AL, Chung T, Christopher-Stine L, Rosen P, Rosen A, Doering KR, Casciola-Rosen LA. Autoantibodies against 3-hydroxy-3-methylglutaryl-coenzyme A reductase in patients with statin-associated autoimmune myopathy. Arthritis Rheum. 2011;63:713‚Äì721. 10. Mammen AL, Pak K, Williams EK, Brisson D, Coresh J, Selvin E, Gaudet D. Rarity of anti-3-hydroxy-3-methylglutaryl-coenzyme A reductase antibodies in statin users, including those with self-limited musculoskeletal side effects. Arthritis Care Res (Hoboken). 2012;64:269‚Äì272. 11. Silva CA, Sultan SM, Isenberg DA. Pregnancy outcome in adult-onset idiopathic inflammatory myopathy. Rheumatology (Oxford). 2003;42:1168‚Äì1172. 12. Suzuki S, Satoh T, Sato S, Otomo M, Hirayama Y, Sato H, Kawai M, Ishihara T, Suzuki N, Kuwana M. Clinical utility of anti-signal recognition particle antibody in the differential diagnosis of myopathies. Rheumatology (Oxford). 2008;47:1539‚Äì1542. 13. Tan EM, Cohen AS, Fries JF, Masi AT, McShane DJ, Rothfield NF, Schaller JG, Talal N, Winchester RJ. The 1982 revised criteria for the classification of systemic lupus erythematosus. Arthritis Rheum. 1982;25:1271‚Äì1277. 14. Targoff IN, Johnson AE, Miller FW. Antibody to signal recognition particle in polymyositis. Arthritis Rheum. 1990;33:1361‚Äì1370. 15. Troyanov Y, Targoff IN, Tremblay JL, Goulet JR, Raymond Y, Senecal JL. Novel classification of idiopathic inflammatory myopathies based on overlap syndrome features and autoantibodies: analysis of 100 French Canadian patients. Medicine (Baltimore). 2005;84:231‚Äì249. 16. Vitali C, Bombardieri S, Jonsson R, Moutsopoulos HM, Alexander EL, Carsons SE, Daniels TE, Fox PC, Fox RI, Kassan SS, Pillemer SR, Talal N, Weisman MH; European Study Group on Classification Criteria for Sjogren‚Äôs Syndrome. Classification criteria for Sjogren‚Äôs syndrome: a revised version of the European criteria proposed by the American-European Consensus Group. Ann Rheum Dis. 2002;61:554‚Äì558. 17. Werner JL, Christopher-Stine L, Ghazarian SR, Pak KS, Kus JE, Daya NR, Lloyd TE, Mammen AL. Antibody levels correlate with creatine kinase levels and strength in anti-3-hydroxy-3-methylglutaryl-coenzyme A reductase-associated autoimmune myopathy. Arthritis Rheum. 2012;64:4087‚Äì4093. www.md-journal.com Copyright ¬© 2014 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 157 ˇ˛Articles Lancet 2010; 37: 664 72 Published Online February 4, 2010 DOI:10.1016/S0140- 6736(09)61962-0 See Comment page 618 Department of Internal Medicine (W A Nienhuis MD, Y Stienstra MD), Department of Epidemiology (J P Schouten MSc), and Infectious Diseases Service and Tuberculosis Unit, Department of Internal Medicine and Department of Pulmonary Diseases and Tuberculosis (Prof T S van der Werf MD), University Medical Centre Groningen, University of Groningen, Netherlands; Agogo Presbyterian Hospital, Agogo,Ghana (W A Thompson MD, K M Abass); Nkawie-Toase Government Hospital, Nkawie, Ghana (P C Awuah MD, W Tuah); Kumasi Centre for Collaborative Research in Tropical Medicine, Kwame Nkrumah University of Science and Technology, Kumasi, Ghana (N Y Awua-Boateng, Prof O Adjei PhD); National Buruli Ulcer Program, Accra, Ghana (E O Ampadu MD); and Department of Infectious Diseases and Tropical Medicine, Ludwig-Maximilians University of Munich, Munich, Germany (V Siegmund PhD, G Bretzel MD) Correspondence to: Prof Tjip S van der Werf, University Medical Centre Groningen, PO Box 30 001, 9700 RB Groningen, Netherlands t.s.van.der.werf@int.umcg.nl Antimicrobial treatment for early, limited Mycobacterium ulcerans infection: a randomised controlled trial Willemien A Nienhuis, Ymkje Stienstra, William A Thompson, Peter C Awuah, K Mohammed Abass, Wilson Tuah, Nana Yaa Awua-Boateng, Edwin O Ampadu, Vera Siegmund, Jan P Schouten, Ohene Adjei, Gisela Bretzel, Tjip S van der Werf Summary Background Surgical debridement was the standard treatment for Mycobacterium ulcerans infection (Buruli ulcer disease) until WHO issued provisional guidelines in 2004 recommending treatment with antimicrobial drugs (streptomycin and rifampicin) in addition to surgery. These recommendations were based on observational studies and a small pilot study with microbiological endpoints. We investigated the efficacy of two regimens of antimicrobial treatment in early-stage M ulcerans infection. Methods In this parallel, open-label, randomised trial undertaken in two sites in Ghana, patients were eligible for enrolment if they were aged 5 years or older and had early (duration <6 months), limited (cross-sectional diameter <10 cm), M ulcerans infection confirmed by dry-reagent-based PCR. Eligible patients were randomly assigned to receive intramuscular streptomycin (15 mg/kg once daily) and oral rifampicin (10 mg/kg once daily) for 8 weeks (8-week streptomycin group; n=76) or streptomycin and rifampicin for 4 weeks followed by rifampicin and clarithromycin (7∑5 mg/kg once daily), both orally, for 4 weeks (4-week streptomycin plus 4-week clarithromycin group; n=75). Randomisation was done by computer-generated minimisation for study site and type of lesion (ulceration or no ulceration). The randomly assigned allocation was sent from a central site by cell-phone text message to the study coordinator. The primary endpoint was lesion healing at 1 year after the start of treatment without lesion recurrence or extensive surgical debridement. Analysis was by intention-to-treat. This trial is registered with ClinicalTrials.gov, number NCT00321178. Findings Four patients were lost to follow-up (8-week streptomycin, one; 4-week streptomycin plus 4-week clarithromycin, three). Since these four participants had healed lesions at their last assessment, they were included in the analysis for the primary endpoint. 73 (96%) participants in the 8-week streptomycin group and 68 (91%) in the 4-week streptomycin plus 4-week clarithromycin group had healed lesions at 1 year (odds ratio 2∑49, 95% CI 0∑66 to infinity; p=0∑16, one-sided Fisher s exact test). No participants had lesion recurrence at 1 year. Three participants had vestibulotoxic events (8-week streptomycin, one; 4-week streptomycin plus 4-week clarithromycin, two). One participant developed an injection abscess and two participants developed an abscess close to the initial lesion, which was incised and drained (all three participants were in the 4-week streptomycin plus 4-week clarithromycin group). Interpretation Antimycobacterial treatment for M ulcerans infection is effective in early, limited disease. 4 weeks of streptomycin and rifampicin followed by 4 weeks of rifampicin and clarithromycin has similar efficacy to 8 weeks of streptomycin and rifampicin; however, the number of injections of streptomycin can be reduced by switching to oral clarithromycin after 4 weeks. Funding European Union (EU FP6 2003-INCO-Dev2-015476) and Buruli Ulcer Groningen Foundation. 664 www.thelancet.com Vol375 February20,2010 Introduction Buruli ulcer is a necrotising infection of subcutaneous tissue caused by Mycobacterium ulcerans.1 The name Buruli ulcer comes from a region near the Nile River delta in Uganda, named Buruli County, where the disease was highly endemic in the 1960s.2 Today, the disease is emerging in west African countries with thousands of cases every year, mainly in children.3,4 A plasmid of M ulcerans encodes the production of mycolactone,4,5 an immunomodulatory macrolide toxin that causes tissue necrosis.6 M ulcerans is acquired near slow-flowing and stagnant water in tropical and subtropical environments. The natural reservoir and mode of transmission of the infection remain largely obscure and might differ between endemic foci around the world.7,8 However, skin injury9 and insect bites10 have been proposed as modes of transmission. M ulcerans infection usually starts as a nodule, papule, plaque, or oedema. When left alone, the lesion breaks open and a typical painless ulcer with undermined edges appears, which can progress to a large necrotic lesion. WHO has defined lesions with a cross-sectional diameter of less than 5 cm as category I, 5 15 cm as category II, and more than 15 cm, lesions on important sites (eye, breast, and genitalia), or multiple lesions as category III. M ulcerans infection can be self-limiting, but scar tissue and contractures in joints leave patients Articles 250 patients screened 70 ineligible 18 lesion too large 16 too young 7 pregnant 2 hearing impairment 7 recurrent disease 20 doubtful diagnosis and PCR and AFB negative 26 with clinical disease but without PCR confirmation* 154 enrolled and started on treatment 151 randomised 143 with infection confirmed by PCR 5 with infection confirmed by other method! 3 without confirmation of infection 3 withdrew/lost before randomisation 1 accidentally operated 1 died 1 non-compliant 75 assigned to streptomycin and rifampicin for 4 weeks followed by rifampicin and clarithromycin for 4 weeks 1 died (in week 16) ß 2 migrated (in week 32 and week 36)ß 75 assessed for primary endpoint 76 assigned to streptomycin and rifampicin for 8 weeks 1 withdrew (in week 6)ß 76 assessed for primary endpoint with functional limitations and can result in social stigma.11,12 The diagnosis can be made clinically but culture is the gold standard. However, this method is difficult and has low sensitivity.1,3,4,13 Since the develop- ment of PCR targeting insertion sequence 2404 (IS2404) a repetitive oligonucleotide unit with more than 200 copies in the genome of M ulcerans14 diagnostic confirmation has improved substantially.13,15,16 Buruli ulcer is one of 19 neglected tropical diseases addressed by WHO in its Global plan to combat neglected tropical diseases 2008 2015.17 In this plan, the organisation describes Buruli ulcer as a disease for which there are no cost-effective control methods. Since the disease s first description in 1948,18 different treatments have been investigated. Extensive surgical debridement, with or without subsequent skin grafting, is standard treatment. However, surgery cannot completely remove all bacilli19 and recurrence is common, with reported rates varying between 6% and 47%.20 22 Although larger excisions might be more effective, they can increase chances of residual functional limitations. In the first of two randomised controlled trials for M ulcerans infection, clofazimine did not show a significant benefit compared with placebo.23 In individuals with small (<5 cm), non-ulcerated lesions, recurrence-free healing without surgery was reported in five of eight participants who were treated with clofazimine compared with five of 17 who were treated with placebo. In ten patients with larger and ulcerated lesions, all except one (in the placebo group) needed surgery. A second study compared the effect of dapsone plus rifampicin with placebo. Of 41 randomised patients, 30 completed the 2-month trial. Rate of healing did not differ between groups. Uneven baseline characteristics might partly explain why patients assigned to active treatment had a larger reduction in lesion size than did patients assigned to placebo.24 Many antimycobacterial agents show activity against M ulcerans in vitro, and experiments in animals, such as the mouse footpad model, show that streptomycin in combination with rifampicin is highly bactericidal. In a pilot study sponsored by WHO, 31 patients clinically diagnosed with pre-ulcerative M ulcerans infection were treatedwithstreptomycinandrifampicinfor0,2,4,8,or 12 weeks.25 All lesions were excised; M ulcerans infection was confirmed by PCR in 21 cases. In ten patients who were treated for 2 weeks or less, viable bacilli could be isolated from excised tissues, whereas M ulcerans could not be cultured from tissue taken from 11 patients who were treated for 4 weeks or longer. Lesions either reduced or stabilised in size in all patients.25 On the basis of these findings, preliminary guidelines were issued by WHO recommending streptomycin in combination with rifampicin as standard treatment for M ulcerans infection,26 with or without additional surgical debridement or skin grafting.27 When our study was designed, clarithromycin was believed to have only bacteriostatic activity in vivo.3,28 We assessed the efficacy of antibiotic therapy with oral Figure 1: Trial profile AFB=acid-fast bacilli.*Patients not enrolled but given 8 weeks of treatment with streptomycin and rifampicin. Participant died of cause unrelated to M ulcerans infection. ! See text for details. ßHealed at time of last assessment, included in the final analysis. rifampicin and intramuscular streptomycin given for 8 weeks for treatment of early M ulcerans infection in patients from Ghana. This regimen was compared with rifampicin and streptomycin given for 4 weeks, followed by an oral combination of clarithromycin and rifampicin for 4 weeks. Our aim was to identify an effective alternative treatment to extensive surgical debridement, and to explore possibilities to keep the use of injectable antimicrobial treatment to a minimum. Methods Participants The study design was partly based on discussions within a WHO expert group on Buruli ulcer that took place between 2001, and 2003. Between April, 2006, and January, 2008, patients were recruited at two sites (Nkawie- Toase Government Hospital, Nkawie, and Agogo Presbyterian Hospital, Agogo) in Ghana. Patients clinically diagnosed with M ulcerans disease were recruited by active case finding. Patients were eligible for enrolment www.thelancet.com Vol375 February20,2010 665 Articles Sex (male) Body-mass index (kg/m2) Tribe Akan Other Lesion surface area (cm2) Type of lesion No ulceration Ulceration HIV infection 8-week streptomycin group (n=76) 27 (36%) 16∑6 (14∑9 19∑4) 20 (26%) 56 (74%) 29 (9 55) 49 (64%) 27 (36%) 0 4-week streptomycin plus 4-week clarithromycin group (n=75) 19 (25%) 17∑2 (15∑4 19∑4) 18 (24%) 57 (76%) 26 (10 46) 43 (57%) 32 (43%) 3 (4%) Age (years) 12 (8 18) 12 (9 22) Study site Agogo 54 (71%) 53 (71%) Nkawie 22 (29%) 22 (29%) Duration of disease (weeks) 4 (2 6) 3 (2 4) Category of lesion I 29 (38%) 29 (39%) II or III 47 (62%) 46 (61%) Lesion distribution (side of body) Left 27 (36%) 28 (37%) Right 49 (64%) 47 (63%) Data are n (%) or median (IQR). Patients in the 8-week streptomycin group were assigned to receive intramuscular streptomycin and oral rifampicin for 8 weeks. Patients in the 4-week streptomycin plus 4-week clarithromycin group were assigned to receive streptomycin and rifampicin for 4 weeks followed by rifampicin and clarithromycin, both orally, for 4 weeks. Table 1: Patient baseline characteristics 666 www.thelancet.com Vol375 February20,2010 if they were aged 5 years or older, had a reported disease duration of less than 6 months, and had lesions with a cross-sectional diameter (indurated area) of 10 cm or less. M ulcerans infection was confirmed by IS2404 dry-reagent- based PCR.29 Exclusion criteria were pregnancy, drug in- tolerance, and renal, hepatic, and acoustic impairment. The protocol and consent forms were approved by the Committee on Human Research, Publication, and Ethics of the School of Medical Science, Kwame Nkrumah University of Science and Technology, Kumasi, and the Komfo Anokye Teaching Hospital, Kumasi (CHRPE/07/01/05), and by the Ethical Review Committee of Ghana Health Services (GHS-ERC-01/01/06). The Medical Ethics Review Committee of the University Medical Centre Groningen, Netherlands, reviewed the protocol before ethics clearance in Ghana. Written and verbalinformedconsentwasobtainedfromallparticipants aged 12 years or older, and from parents, carers, or legal representatives of participants aged 18 years or younger. Procedures After participants had given informed consent, we obtained demographic and clinical information and took blood samples. We undertook pregnancy tests in female participants aged 10 years or older, and hearing tests in all participants (AS208 portable equipment; Inter- acoustics, Assens, Denmark) to obtain baseline audio- metric data. HIV antibody testing was done with cold-stored sera after completion of the study. Lesions were photographed and traced onto acetate sheets. Three 3 mm punch biopsy samples were taken under local anaesthesia; two swabs of ulcerated lesions were also taken. All samples were transported to the Kumasi Centre for Collaborative Research in Tropical Medicine laboratory in Kumasi, Ghana, for IS2404 dry- reagent-based PCR and Ziehl-Neelsen staining to detect acid-fast bacilli; mycobacterial culture was done on Lowenstein-Jensen slopes at 32oC.13 One punch biopsy was reserved for histopathological examination. Participants started streptomycin (15 mg/kg once daily intramuscularly) and rifampicin (10 mg/kg once daily orally) after the diagnostic procedures. After assess- ments and start of treatment at the hospital, most parti- cipants were treated as outpatients. Once a week, participants were given study drugs to take to the nearest health facility to receive directly observed treatment (DOT) for the subsequent days, with daily wound care. Only participants that had extensive oedema or lesions at difficult sites (joints, eye, or genitalia), or lesions with suspected secondary infection were admitted to hospital; participants who could not receive DOT or wound care at home were also admitted to hospital. DOT was recorded on forms by the health-care worker or helper who was observing the treatment. Participants were followed up at weekly intervals during the first 8 weeks. At these visits, clinical assessments and digital photographs were taken, DOT forms were checked, and participants were invited to report any adverse events. Once every 2 weeks, the size of the lesion was traced onto an acetate sheet and blood cell counts were taken; we also undertook liver and kidney function tests and hearing tests in all participants, and pregnancy tests in female participants aged 10 years or older. Randomisation and masking Before the end of week 4, participants with M ulcerans infection confirmed by PCR were randomly assigned to receive streptomycin intramuscularly and rifampicin orally for 4 more weeks (8-week streptomycin group) or rifampicin and clarithromycin (7∑5 mg/kg once daily), both orally, for another 4 weeks (4-week streptomycin plus 4-week clarithromycin group). Randomisation was done with minimisation for study site and type of lesion (ulceration or no ulceration). The study coordinator (WAN) forwarded the information of every enrolled participant by cell-phone text messaging to a statistician (JPS) at the Department of Epidemiology, University Medical Centre Groningen, Netherlands. There, a computer-generated randomisation program was used, and the randomly assigned allocation was then sent by text message to the study coordinator. Individuals who were clinically diagnosed with M ulcerans disease but Articles Study Treatment Sex Age Category of Stage Size of lesion Additional Timepoint Additional diagnostics Diagnostic results site group (years) lesion (mm) information (weeks*) (timepoint, weeks*) Participants with treatment failure recorded before week 52 1 Agogo 8-week streptomycin Male 6 II Ulcer 120◊98 Extensive debridement 6 ∑∑ ∑∑ 2 Agogo 8-week streptomycin Male 10 III Ulcer and oedema, critical site 126◊79 Extensive debridement 20 ∑∑ ∑∑ 3 Agogo 4-week streptomycin plus 4-week clarithromycin Female 12 II Plaque 73◊60 Lesion progression, extensive debridement 8 Surgical resected tissue PCR negative, ZN negative, culture positive 4 Nkawie 4-week streptomycin plus 4-week clarithromycin Female 11 I Nodule 30◊24 Ulceration of lesion; 4 additional weeks of streptomycin and rifampicin 12 Punch biopsy and swab PCR negative, ZN negative, culture negative 5 Nkawie 4-week streptomycin plus 4-week clarithromycin Female 5 I Plaque 48◊47 New ulceration after initial closure 34 Swab and surgical resected tissue PCR negative, ZN negative, culture negative Participants who were not healed at time of primary endpoint (week 52) 6 Agogo 8-week streptomycin Male 22 II (with oedema leading to III) Ulcer 112◊80 ∑∑ ∑∑ ∑∑ ∑∑ 7 Agogo 4-week streptomycin plus 4-week clarithromycin Female 20 III Ulcer and multiple nodules ∑∑ Ulcer healed; multiple nodules with ulceration before healing ∑∑ Swab (32), swab (52) 32 weeks: PCR positive, ZN negative, culture postive; 52 weeks: PCR positive, ZN positive, culture negative 8 Nkawie 4-week streptomycin plus 4-week clarithromycin Male 12 II Ulcer 95◊95 Inadequate wound care ∑∑ Swab (29), swab (72), surgically resected tissue (80) 29 weeks: PCR negative, ZN negative, culture negative; 72 weeks: PCR negative, ZN negative, culture positive; 80 weeks: PCR negative, ZN negative, culture negative 9 Nkawie 4-week streptomycin plus 4-week clarithromycin Female 7 II Ulcer 122◊100 ∑∑ ∑∑ ∑∑ ∑∑ 10 Agogo 4-week streptomycin plus 4-week clarithromycin Female 27 III Plaque and two small ulcers 113◊86 HIV positive ∑∑ ∑∑ ∑∑ Treatment failure was recorded if a participant s lesion had not healed by week 52, lesion recurrence occurred within 1 year, or lesion size increased to 150% or more at any timepoint compared with baseline with surgical debridement undertaken as deemed necessary by the attending doctor in the hospital. *Weeks after start of treatment. Results for insertion sequence 2404 dry-reagent-based PCR, Ziehl-Neelsen (ZN) staining to identify acid-fast bacilli, and M ulcerans culture. Patients in the 8-week streptomycin group were assigned to receive intramuscular streptomycin and oral rifampicin for 8 weeks. Patients in the 4-week streptomycin plus 4-week clarithromycin group were assigned to receive streptomycin and rifampicin for 4 weeks followed by rifampicin and clarithromycin, both orally, for 4 weeks. Table 2: Characteristics of ten participants with treatment failure who did not have confirmation by PCR continued treatment with streptomycin plus rifampicin and were not randomised; these individuals were followed up and analysed separately. This was an open-label trial. Follow-up and study outcomes After 8 weeks of antimicrobial treatment, missed doses were not supplemented. Participants were followed up at week 10 and week 12 after start of treatment, and then monthly to week 36, and bimonthly to week 52. Study visits included clinical assessment with reporting of adverse effects, measurement of lesion size (if not healed) by tracing onto an acetate sheet, and photography of the lesion. Participants travel costs werereimbursedandsmallmonthlyincentives(sugar, condensed milk, and cocoa powder) were offered for time spent in the study. Treatment failure was recorded if a participant s lesion had not healed by week 52, lesion recurrence occurred within 1 year, or lesion size increased to 150% or more at any timepoint compared with baseline with surgical debridement undertaken as deemed necessary by the attending doctor in the hospital. Neither the investigators who took measurements of the lesions, nor the attending doctor in the hospital making the final decision for extensive surgical debridement were masked to treatment assignment. Removal of necrosis and slough is part of normal wound care and skin grafting speeds up healing www.thelancet.com Vol375 February20,2010 667 Articles 8-week streptomycin group 4-week streptomycin plus 4-week clarithromycin group 95% CI 0∑0067 0∑10 0∑046 0∑19 0∑084 0∑25 0∑16 0∑36 0∑34 0∑57 0∑51 0∑73 0∑78 0∑93 0∑88 0∑99 Week 1 Week 4 Week 6 Week 8 Weeks 10 12 Weeks 16 20 Weeks 24 28 Weeks 32 36 Weeks 44 52 95% CI 0∑0019 0∑090 0∑028 0∑15 0∑036 0∑17 0∑054 0∑20 0∑094 0∑26 0∑28 0∑49 0∑52 0∑74 35 0∑82 0∑96 21 0∑94 1∑00 7 Total Healed (n) (n) 76 1 72 1 70 0 69 2 66 3 59 12 37 10 18 11 3 2 Treatment failure (n) 0 0 1 0 0 1 0 0 1 Cumulative proportion healed 0∑013 0∑066 0∑076 0∑11 0∑16 0∑37 0∑63 0∑90 0∑99 Total Healed Treatment Cumulative (n) (n) failure (n) proportion healed 75 0 0 74 1 0 0∑027 72 4 0 0∑093 66 2 1 0∑15 60 5 0 0∑24 50 10 0 0∑45 Week 2 75 3 0 0∑053 0∑020 0∑13 75 1 0 0∑013 0∑019 0∑091 Week 5 71 1 0 0∑076 0∑036 0∑17 73 1 0 0∑040 0∑013 0∑12 Week 7 70 0 0 0∑076 0∑036 0∑17 68 2 0 0∑12 0∑064 0∑22 Weeks 8 10 67 1 0 0∑12 0∑064 0∑22 63 2 1 0∑17 0∑11 0∑28 Weeks 12 16 63 4 0 0∑21 0∑14 0∑32 55 5 0 0∑32 0∑22 0∑43 Weeks 20 24 46 9 0 0∑50 0∑39 0∑61 40 5 0 0∑52 0∑41 0∑64 7 0 0∑61 11 1 0∑87 3 4 0∑96 Weeks 28 32 27 9 0 0∑76 0∑65 0∑85 28 7 0 0∑71 0∑60 0∑81 Weeks 36 44 7 4 0 0∑96 0∑90 0∑99 9 2 0 0∑90 0∑81 0∑95 Patients in the 8-week streptomycin group were assigned to receive intramuscular streptomycin and oral rifampicin for 8 weeks. Patients in the 4-week streptomycin plus 4-week clarithromycin group were assigned to receive streptomycin and rifampicin for 4 weeks followed by rifampicin and clarithromycin, both orally, for 4 weeks. Table 3: Actuarial life table for cumulative proportion of healing for both treatment groups, by time interval 668 www.thelancet.com Vol375 February20,2010 but does not affect bacterial load. These interventions were therefore not regarded as evidence of treatment failure. The primary clinical endpoint was lesion healing (complete re-epithelialisation) at 1 year after the start of treatment without recurrence or extensive surgical debridement. Secondary outcomes were time to wound healing and time to complete wound coverage by a crust. Daily sterile dressings were only applied at the health facility if lesions were open and discharging. Before final healing occurs, lesions might turn dry with a crust. At this stage, participants could cover the lesions for protection at home, without visiting the health facility to receive wound care and sterile dressings. Since participants reported this stage of wound healing as beneficial, we also measured time to complete wound coverage by a crust without complete re-epithelialisation as a secondary endpoint. The safety outcome measure was occurrence of adverse events. Statistical analysis When the study was designed, there was no information available about healing rates for the proposed regimens; therefore, we assumed a healing rate of 80% in the 8-week streptomycin group. We calculated that a sample size of 148 randomised and fully assessable participants (74 in each group) would be needed to detect a difference in healing rate of 20% or more (<60% in the 4-week streptomycin plus 4-week clarithromycin group) with a one-sided alpha of 0∑05 and a power of 80%. We calculated an odds ratio for the primary clinical endpoint by use of Fisher s exact test. Because secondary outcome data were interval-censored, we analysed the cumulative incidence of healing by use of actuarial life table analysis and weighted log-rank tests for interval- censored data, in particular the group proportional hazards model30 and a generalised Wilcoxon-Mann- Whitney test,31 which emphasises early events. We calculated the exact permutation p value for the scores of the group proportional hazards model and Wilcoxon- Mann-Whitney tests and the non-parametric maximum likelihood estimate of the survival distribution function.32 Other secondary outcome measures were assessed by actuarial life table analysis. All analyses were by intention to treat. Statistical analysis was done with SPSS version 16.0, R version 2.9.2, and Stata version 10.1. An independent data safety monitoring board reviewed the data for safety purposes after inclusion of 57 and 115 participants. Interim reports were discussed at the annual WHO meeting on Buruli ulcer in Geneva in 2007 and 2008, and presented at the 2008 combined ICAAC/IDSA Annual Meeting in Washington, DC, USA.33 After the trial had been completed, two independent wound experts from University Medical Centre Groningen, who were masked to treatment assignment, assessed the primary study endpoint (healing at 1 year) using the digital photographs of the lesions taken during the trial. This trial is registered with ClinicalTrials.gov, number NCT00321178. Role of the funding source The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, writing Articles 1∑0 0∑8 0∑6 0∑4 0∑2 0 Number at risk 8-week streptomycin group 4-week streptomycin plus 4-week clarithromycin group Group proportional hazards model p=0∑26 (99% CI 0∑22 0∑29) Generalised Wilcoxon-Mann-Whitney test p=0∑60 (99% CI 0∑56 0∑64) 0 4 8 12 16 20 24 28 32 36 40 44 48 52 Weeks 8-weekstreptomycingroup 76 72 71 69 67 66 63 59 46 37 27 18 7 3 4-weekstreptomycinplus 75 74 73 68 63 60 55 50 40 35 28 21 9 7 4-week clarithromycin group Figure 2: Non-parametric maximum likelihood estimates for time to healing The non-parametric maximum likelihood estimates for each treatment group are plotted with shaded rectangles denoting the indeterminate rises in the proportion healed during each time interval. Linear interpolation lines of healing within these indeterminate regions are also shown. Patients in the 8-week streptomycin group were assigned to intramuscular streptomycin and oral rifampicin for 8 weeks. Patients in the 4-week streptomycin plus 4-week clarithromycin group were assigned to receive streptomycin and rifampicin for 4 weeks followed by rifampicin and clarithromycin, both orally, for 4 weeks. ofthereport,orindecisionsaboutsubmissionofresults for publication. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication. Results Figure 1 shows the trial profile. 180 eligible patients started treatment. 26 patients with suspected but unconfirmed M ulcerans infection received streptomycin and rifampicin for 8 weeks. Of 151 participants who were enrolled and randomised, eight had a clinical diagnosis without confirmation of M ulcerans infection by PCR. Five of these eight participants had infection later confirmed by one or more diagnostic tests (Ziehl-Neelsen staining, two; culture, one; histopathology, two). Three randomised participants did not have diagnosis confirmed by any test. Table 1 shows baseline characteristics of study participants. Lesions were more frequently seen on the right side of the body (64%) than on the left side (36%; p<0∑0001). Three (2%) participants were HIV positive; these individuals had initial lesions and clinical presentations that were indiscernible from those of HIV-negative participants. One participant in the 8-week streptomycin group withdrew from the study at week 6. In the 4-week streptomycin plus 4-week clarithromycin group, two participants moved out of the study area and were lost to follow-up (week 32 and 36) and one participant, who later tested positive for HIV infection, died in week 16 of urosepsis. Since these four participants had healed lesions at their last assessment, they were included in the analysis for the primary clinical endpoint and in the analyses for time to healing. Compliance to study treatment was assessed by use of DOT forms, signed by health personnel at the health facilities. Adherence to treatment protocol was 98% in the 8-week streptomycin group and 99% in the 4-week streptomycin plus 4-week clarithromycin group. Treatment failure was recorded in ten participants, three in the 8-week streptomycin group and seven in the 4-week streptomycin plus 4-week clarithromycin group. Table 2 shows the characteristics of these individuals. Five participants were not healed at week 52, all of whom had a substantial decrease in lesion size. One participant had several lesions, and four had large lesions at the start of treatment (one of whom had HIV infection). Of the five participants with treatment failure before week 52, two had large lesions, one had a pre-ulcerative lesion that ulcerated later, one had a progressive lesion, and one had a lesion that www.thelancet.com Vol375 February20,2010 669 Cumulative proportion of participants with healed lesions Articles Patient 1 Patient 3 Patient 5 Study Sex site Agogo Female Nkawie Female Nkawie Male Age Category Stage (years) of lesion 12 II Plaque 5 II Ulcer 12 II Ulcer Indication for culture Lesion progression Pus collection No complete healing at week 52 Timepoint Additional information (weeks)* 11 Extensive surgical debridement 18 Lesion healed without further intervention 72 Inadequate wound care; surgical debridement Diagnostic specimen Surgically resected tissue Swab Swab Patient 2 Patient 4 Nkawie Agogo Female Female 12 20 II III Oedema Ulcer and multiple nodules Development of a second lesion Confirmation of nodule This analysis was not specified by the protocol. All five participants were in the 4-week streptomycin plus 4-week clarithromycin group (streptomycin and rifampicin for 4 weeks followed by rifampicin and clarithromycin for 4 weeks). *Time of tissue specimen collection (weeks after start of treatment). Participant with treatment failure. Table 4: M ulcerans isolated by culture of tissue specimens after 8-week treatment period 14 32 Lesion healed without further intervention Ulcer healed; multiple nodules ulcerated before healing Swab Swab 670 www.thelancet.com Vol375 February20,2010 almost healed, but opened up again. No participants with healed lesions had a recurrence at week 52. 73 (96%) participants in the 8-week streptomycin group and 68 (91%) participants in the 4-week streptomycin plus 4-week clarithromycin group had healed lesions at week 52 (odds ratio for failure in healing for 4-week streptomycin plus 4-week clarithromycin vs 8-week streptomycin 2∑49, 95% CI 0∑66 to infinity, p=0∑16). We obtained consistent findings when the four participants who were not followed up to week 52 were excluded from the analysis and when two wound experts masked to treatment assignment assessed the primary endpoint by use of photographs available at the different timepoints (data not shown). Table 3 shows the actuarial life table for cumulative proportion of healing. The estimated cumulative proportion of patients healed at week 52 was 0∑99 (95% CI 0∑94 1∑00) in the 8-week streptomycin group and 0∑96 (95% CI 0∑88 0∑99) in the 4-week streptomycin plus 4-week clarithromycin group; a difference of 0∑034 (95% CI 0∑024 to 0∑091) between groups. Figure 2 shows the non-parametric maximum likelihood estimates for healing in the intention-to-treat population. Neither the group proportional hazards model (p=0∑26; 99% CI 0∑22 0∑29) nor the generalised Wilcoxon-Mann-Whitney test (p=0∑60; 99% CI 0∑56 0∑64) showed a significant difference in time to healing between groups. The group proportional hazards model suggested a shorter time to healing in the 8-week streptomycin group whereas the Wilcoxon-Mann-Whitney test suggested that time to healing was shorter in the 4-week streptomycin plus 4-week clarithromycin group. Adjustment for study site and type of lesion (ulceration or no ulceration) did not affect the results (data not shown). Five participants received skin grafts, four in the 8-week streptomycin group (at week 16, 24, 24, and 28), and one in the 4-week streptomycin plus 4-week clarithromycin group (at week 20). Time to healing of category I lesions (median 18 weeks, 95% CI 14 22) was significantly shorter than that for category II and III lesions (30 weeks, 95% CI 26 34, p=0∑002; data pooled for the two treatment groups; five participants with skin grafts not included). Time to complete wound coverage by a crust was also significantly shorter for category I lesions than category II and III lesions (14 weeks, 95% CI 11 18, vs 22 weeks, 95% CI 22 26; p=0∑002). Three participants had vestibulotoxic events, one in the 8-week streptomycin group (aged 49 years, starting after 7 weeks of treatment) and two in the 4-week streptomycin plus 4-week clarithromycin group (aged 24 years and 38 years, starting after 4 weeks and 3 weeks of treatment, respectively). Analysis of digital photographs showed that three participants had mild to moderate functional limitations at the end of the study: one had a contracture with substantial decrease in range of movement of the thumb and index finger (4-week streptomycin plus 4-week clarithromycin group); two had ulcers on the back of the hand and wrist that resulted inclaw-hands(oneineachgroup).Noliverorkidney function test abnormalities or audiological deterioration occurred that necessitated termination of streptomycin treatment. One participant developed an injection abscess (4-week streptomycin plus 4-week clarithro- mycin group) and two participants (both in the 4-week streptomycin plus 4-week clarithromycin group) developed an abscess close to the initial lesion which was incised and drained. One participant in the 8-week streptomycin group and two participants in the 4-week streptomycin plus 4-week clarithromycin group reported abdominal discomfort. Some participants had additional diagnostic tests not specified in the protocol. Table 4 shows the characteristics of the five participants in whom M ulcerans was isolated by culture after treatment; all were in the 4-week streptomycin plus 4-week clarithromycin group. Three of these five participants had treatment failure: in two, surgical debridement was done; in the third, multiple nodules ulcerated successively over 52 weeks before final healing. Two participants had lesion healing without further intervention within the study period. Discussion Our study has shown that early, limited M ulcerans infection can be safely and effectively managed by antimicrobial treatment alone, without surgical debridement. The drug regimen proposed by WHO, Articles consisting of 8 weeks of streptomycin and rifampicin, seemed effective and was not associated with deterioration requiring subsequent surgical debridement. Treatment with oral clarithromycin plus rifampicin during the second 4-week period resulted in similar outcomes to continuation of treatment with streptomycin and rifampicin. Our findings are important for patients with M ulcerans infection who live in remote, resource- poor areas in west Africa, where people often need to walk for several hours to reach health-care facilities, skilled personnel are scarce, and patients tend to refrain from treatment because of fear of surgery. Our results also support the use of antimicrobial treatment in individuals who are unable to receive streptomycin eg, pregnant women or those who cannot tolerate aminoglycosides. With few reported side-effects, the treatment regimens used in this trial seemed well tolerated, although vestibulotoxicity remains a concern. The rate of lesion recurrence in our study at 52 weeks was lower than that reported in retrospective studies assessing the effect of surgery, in which rates of between 6% and 47% were reported.20 22 Time to healing was a median of 18 weeks for category I lesions and 30 weeks for larger lesions. The length of this healing period might have obscured the potential of antimicrobial treatment in earlier studies that either looked at healing after 2 months,24 or assigned participants to surgery when early healing was not seen during follow- up.23 HIV was not an important confounder; most case- control studies from west Africa report a low incidence of HIV in patients with M ulcerans infection.34,35 One strength of this study is the large proportion of participants (147 of 151) who were followed up to week 52. Second, most participants (148 of 151) had laboratory- confirmed M ulcerans infection, which contrasts with previous trials that were partly undertaken before PCR- based diagnostic confirmation tests were available.23,24 Finally, the sample size in our study was substantially larger than that in earlier studies. A potential weakness of our study is the open-label design. However, masking would have substantially increased costs, and a trial in which children can be assigned to placebo injections is not justified for safety reasons. Moreover, although only one injection abscess was recorded, intramuscular injections in rural Africa are not the preferred option. Another limitation of our study is that no formal external monitoring was done. However, limited auditing was organised. Additionally, consistent results were obtained when two wound experts who were masked to treatment assignment reviewed all digital photographs available at the different timepoints. We therefore believe that the study was robust. One concern is that healing took a fairly long time. Additionally, we could not address the issue of prevention of disabilities in a formal way, although our analysis of digital photographs combined with clinical assessment showed that only three participants had mild to moderate functional limitations at week 52 (all three involving hand function). Contractures and functional limitations are common in ulcers that are close to joints.11 Future studies should assess prevention of disabilities, include all categories of lesions, and investigate oral drug regimens. Thus, antimicrobial treatment is highly effective for treatment of early, limited M ulcerans infection, and the number of intramuscular injections of streptomycin can be reduced without compromising efficacy. Contributors TSvdW and YS designed and supervised the study. WAN coordinated the study. WAN, WAT, PCA, and EOA were responsible for patient screening and enrolment. KMA, WT, and WAN provided patient care and requested informed consent from participants, participants parents, or legal representatives, and collected the clinical and laboratory data. GB, VS, NYA-B, and OA were responsible for the laboratory confirmation. JPS, WAN, and YS did the statistical analyses. TSvdW, WAN, and YS contributed to the interpretation of the results and the writing and critical review of the report. All authors have seen and approved the final version of the report. ˇ˛European Journal of Internal Medicine 15 (2004) 451 459 Original article Antineutrophil cytoplasmic antibody-associated neutropenia Paul Coppoa,1, David Gheza,1, Vincent Fuentesb, Djaouida Bengoufac, Eric Oksenhendlera, Bruno Triboutd, Jean-Pierre Clauvela, Kaiss Lassouedb,* aService d Immuno-He matologie, Ho∆pital Saint-Louis, Paris bService d Immunologie, CHU d Amiens, France cLaboratoire d Immunopathologie, Ho∆pital Saint-Louis, Paris dService de Pathologie Vasculaire, CHU d Amiens, France Received 14 January 2004; received in revised form 1 July 2004; accepted 31 August 2004 www.elsevier.com/locate/ejim Abstract Background: Antineutrophil cytoplasmic antibodies (ANCA) can be associated with various disorders. However, their association with neutropenia has never been reported. Methods: Nine patients with chronic unexplained neutropenia and ANCA were studied. Clinical charts were extensively analyzed and all patients underwent hematological and immunological investigations. Results: All patients (6 women and 3 men) were Caucasian and had a mean age of 49 years (range 16 67 years). All presented with a neutropenia below 1.5ˇ€‹109/L for more than 6 months. The neutropenia was b0.5ˇ€‹109/L in six cases and moderate in three. There was no evidence of toxic- or drug-related neutropenia or of a hematological malignancy. Autoimmune anemia and/or thrombocytopenia were present in five patients. ANCA, with various specificities, were present in all patients. ANCA were associated with various other autoantibodies in eight patients, including antisurface-neutrophil antibodies in three cases. Four of the six patients with severe neutropenia experienced infections. Five patients were treated with hematopoietic growth factors, steroids, intravenous immunoglobulins, splenectomy, methotrexate and/or cyclophosphamide, allowing the neutrophil count to be restored transiently or permanently. Conclusions: A subset of patients with neutropenia of possible autoimmune origin may develop ANCA. Their detection would provide strong evidence of an autoimmune mechanism. Neutropenia should be added to the list of ANCA-associated diseases. D 2004 Elsevier B.V. All rights reserved. Keywords: Neutropenia; Antineutrophil cytoplasmic antibodies; Autoimmunity 1. Introduction Antineutrophil cytoplasmic antibodies (ANCA) are autoantibodies that are directed against different neutrophil antigens. When detected by immunofluorescence on etha- nol-fixed human neutrophils, ANCA usually displays three major patterns: cytoplasmic (cANCA), perinuclear (pANCA), and atypical (xANCA). The main targets of * Corresponding author. Service d Immunologie, Faculte de Me decine, 3 rue des Louvels, F-80036 Amiens Cedex 1, France. Tel.: +33 3 22 82 79 06; fax: +33 3 22 82 79 07. E-mail address: kaiss.lassoued@sa.u-picardie.fr (K. Lassoued). 1 Contributed equally to this work. 0953-6205/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.ejim.2004.08.009 ANCA are either myeloperoxydase (MPO) or proteinase 3 (PR3) [1 4], but they may also be directed against lactoferrin, elastase, cathepsin, lysozyme [5], bactericidal permeability increasing protein (BPI), and azurocidin [6]. ANCA are observed in a large spectrum of diseases [7 19]. Their specificity can make them a helpful tool in the diagnosis of primary systemic vasculitides [20]. Anti-PR3 antibodies (Abs) are strongly associated with Wegener s disease and, to a lesser extent, with microscopic polyarteritis and necrotizing crescentic glomerulonephritis [7 9]. Anti- MPO antibodies are often associated with systemic vascu- litis but can also be found in various autoimmune disorders and connective tissue diseases without evidence of vasculitis [10,11,13,14,16 18]. Other specificity is not considered to 452 P. Coppo et al. / European Journal of Internal Medicine 15 (2004) 451 459 be a helpful serological marker of any diseases. We report herein nine patients with chronic peripheral neutropenia associated with ANCA. To our knowledge, this association has not previously been described in the literature. 2. Material and methods 2.1. Patients Nine adult patients were referred to the Hematology Departments of Ho∆pital Saint-Louis (Paris, France) and Ho∆pital Sud (Amiens, France) for chronic neutropenia defined as a neutrophil count below 1.5ˇ€‹109/L for more than 6 months and the absence of a toxic or infectious etiology. All patients had a complete physical examination. Medications were systematically recorded, as well as any toxic usage or exposure. All patients had serologies for acute viral infections [EBV, CMV, parvovirus B19, hepatitis B (HBV) and C (HCV) viruses], as well as for HIV and HTLV- 1. In one patient with CVID, a blood sample was examined for the presence of CMV antigenemia, HBS antigen, and HCV and HIV transcripts using PCR. A blood smear examination was done on all patients. Analysis of bone marrow aspirates and/or biopsy, immunophenotyping of blood and/or bone marrow lymphocytes, and an analysis of blood lymphocyte clonality were performed when warranted. 2.2. ANCA detection and characterization Sera samples were stored at ˇ€‹20 8C until analysis. In patients 8 and 9, ANCA were sought and found to be positive 2 years prior to the onset of neutropenia; in all other patients, ANCA were sought for the first time in the setting of neutropenia. In all patients, ANCA detection was repeatedly performed during the follow-up. ANCA were detected by indirect immunofluorescence (IIF) using healthy human neutrophils (INOVA Diagnostics, San Diego, CA, USA) [21]. To rule out false-positive results, artefact results due to the presence of antinuclear antibodies (ANA) in five patients, freshly isolated granulocytes from healthy donors were fixed in different ways [98% ethanol (15 min, ˇ€‹20 8C), 95% methanol (15 min, ˇ€‹20 8C), 4% formaldehyde (10 min, +20 8C), 45% acetone 4% formaldehyde (15 min, +20 8C)] and tested by IIF for the presence of ANCA [22]. The sera were examined independently for their immunofluor- escence patterns by two investigators. ANCA specificity was determined using an enzyme-linked immunosorbent assay (ELISA; Euroimmune, Lqbeck, Germany) and by Western blot. For Western blot analysis, white blood cells were isolated from normal blood donors by centrifugation in Ficoll-Hypaque gradients, followed by hypotonic lysis [0.2% NaCl in distilled water, 1 mM phenylmethylsulfonyl- fluoride (PMSF), 1 mM EDTA]. Granulocytes were resuspended in 1.6% NaCl solution, washed, and lysed in 1% Nonidet P-40 (NP-40), 150 mM NaCl, 50 mM Tris (pH 7.4), 0.2% sodium azide, 1 mM di-isopropylfluorophos- phate, 10 mM EDTA, 20 mM iodoacetamide, and 1 mM PMSF. Cell lysates were centrifuged (20,000ˇ€‹g, 20 min, +4 8C) and the supernatant resuspended in 2ˇ€‹Laemmli s buffer [23]. Granulocyte cytoplasmic proteins were separated using SDS-10% polyacrylamide gel electrophoresis (PAGE) and electrotransferred onto nitrocellulose filters, as previ- ously described [24]. Immobilized proteins were incubated with patients sera diluted at 1/50, then with goat antihuman antibody coupled to horseradish peroxydase, and revealed using an enhanced chemiluminescence (ECL) kit (Amer- sham Arlington Heights, IL, USA). Purified MPO (Sigma, France) was also used in some Western blot experiments. 2.3. Other biological investigations The patients were evaluated with routine laboratory tests, including blood cell count, bone marrow aspirates and/or bone marrow biopsy. All patients had extensive immuno- logical investigations. Detection of ANA and identification of their specificities were performed as reported elsewhere [25]. Antisurface neutrophil autoantibodies were detected with granulocyte immunofluorescence (GIFT) and with a granulocyte agglu- tination test (GAT) and antiplatelet antibodies (Abs) by monoclonal Ab immobilization of platelet antigens (MAIPA), as described elsewhere [26 28]. Lymphocyte immunophenotyping and blood lymphocyte clonality analysis were performed as previously described [26]. Control sera were obtained from 8 children with anti- NA1 or anti-NA2 Abs-autoimmune neutropenia, 14 patients with drug-induced neutropenia, 11 patients with isolated immune thrombocytopenia [8] or autoimmune hemolytic anemia [3], 6 patients with Sjfgren s syndrome, 2 patients with Hashimoto s thyroiditis and 8 intravenous immuno- globulin (IVIG)-treated CVID patients with no neutropenia, and 22 healthy individuals. Positive controls for anti-PR3 and MPO Abs were provided with the commercial kits. 3. Results 3.1. Clinical features All nine patients were Caucasians. Three patients were male and six female, and they ranged in age from 16 to 67 years (mean 49 years). In four patients (patients 1, 2, 5, and 6), the past medical history was unremarkable (Table 1). Patient 3 had been diagnosed with Hashimoto s thyroiditis 13 years earlier and had a history of recurrent miscarriages associated with antiphospholipid Abs. Patients 4 and 8 had a long-standing history of non-destructive seronegative poly- arthritis of unknown origin. Patients 8 and 9 had salivary gland infiltration suggestive of Sjfgren s syndrome. In the latter, ANCA were found 2 years before the onset of Table 1 Main clinical and hematological findings P. Coppo et al. / European Journal of Internal Medicine 15 (2004) 451 459 453 Patient Sex/age Organomegaly 1 M/16 Peripheral adenopathies 2 F/46 3 F/56 Splenomegaly 4 M/60 Splenomegaly Hepatomegaly 5 F/37 6 M/44 Peripheral Adenopathies 7 F/67 8 F/62 9 F/54 Infections Anal abcess, folliculitis None Furunculosis None Clonal CD8+gy+ T cell expansion None Pneumonia None Recurrent bronchitis Other conditions Raynaud s phenomenon Livedo Evan s syndrome Hashimoto s thyroiditis Livedo Necrotizing vasculitis Multiple miscarriages Hemolytic anemia Non destructive polyarthritis Hemolytic anemia Thrombocytopenia Myalgias Good s syndrome Evan s syndrome Polyclonal CD8+ T-cell Expansion Neutropenia (ˇ€‹109/L) Lower value b100 1200 b100 b100 1200 1150 b100 Bone marrow Myeloid maturational arrest (myelocyte/ metamyelocyte stage) ND Normal Lymphocyte expansion ND Normal Myeloid maturational arrest (myelocyte/ metamyelocyte stage) Lymphocyte expansion Myeloid maturational arrest (myelocyte/ metamyelocyte stage) Normal Treatment IV Ig Steroids G-CSF No treatment G-CSF Splenectomy G-CSF Methotrexate Steriods No treatment No treatment IV Ig G-CSF Steroids Cyclophosphamide Steroids Methotrexate Steroids G-CSF Methotrexate Non-destructive 200 polyarthritis Sjfgren s syndrome Polyclonal CD8+ ah+T-cell expansion (LGL) Sjfgren s syndrome 150 Male (M), female (F), large granular lymphocytes (LGL), not done (ND), intravenous immunoglobulins (IVIG), granulocyte colony-stimulating factor (G-CSF). neutropenia. Patient 7 had had recurrent upper respiratory tract infections related to a common variable immunodefi- ciency (CVID) diagnosed 2 years earlier. She was being treated monthly with IVIG (O.4 g/kg). Moderate splenomegaly was present in patients 3 and 4 and was associated with hepatomegaly in patient 3. This patient underwent a splenectomy early in the course of the disease. Peripheral lymph node enlargement was found in two patients (patients 1 and 6). In one (patient 1), a lymph node biopsy showed a non-specific lymphoid hyperplasia. Cutaneous manifestations were seen in two patients (patients 1 and 3) and consisted of a livedo in both cases, associated with a Raynaud s phenomenon in patient 1 and with necrotizing vasculitis in patient 3. Infections occurred in four patients with severe neutro- penia (b0.5ˇ€‹109/L) and consisted of folliculitis with an anal abscess in one case (patient 1), diffuse furunculosis in one case (patient 3), pneumonia in the patient with CVID (patient 7), and recurrent bronchitis in patient 9, who had Sjfgren s syndrome. Aside from these two patients, there was no patient with respiratory manifestations. There was no evidence of sinus involvement in any of the nine patients. Neither drug intake nor exposure to toxic compounds that could induce neutropenia could be recorded. 3.2. Hematological findings (Table 1) All patients had neutropenia of variable severity (Table 1). Non-periodic fluctuations in the neutrophils were observed in patients 3 and 4. Four patients (patients 2, 6, 8, and 9) had neutropenia with no other cytopenia. In three patients (patients 1, 5, and 7), the neutropenia was associated with a peripheral thrombocytopenia that was severe (b10ˇ€‹109/L in patients 1 and 7) or moderate (100ˇ€‹109/L in patient 5). Four patients had hemolytic anemia (patients 1, 3, 4, and 7). In patient 1, Evan s syndrome was diagnosed 5 years after the neutropenia, and in patient 4, hemolytic anemia appeared 7 years later. Thrombocytopenia and/or hemolytic anemia were present at the time of diagnosis in three patients. The monocyte count was normal in all patients. 454 P. Coppo et al. / European Journal of Internal Medicine 15 (2004) 451 459 Blood smear examination showed large granular lym- phocytes (LGL) in patient 8, accounting for 40% of the total lymphocyte count (i.e., 0.8ˇ€‹109/L). This LGL population consisted of polyclonal TCR ah+ CD3+, CD4ˇ€‹, and CD8+ lymphocytes. Patient 4 had persistent clonal TCR gy+ CD3+, CD4ˇ€‹, CD8+ lymphocyte expansion in peripheral blood, accounting for 30% of the blood lymphocyte count, with no LGL. The cytological pattern did not suggest any hematological malignancies. Blood lymphocyte count and immunophenotyping were normal in the remaining patients. Blood lymphocyte clonality was also assessed in patients 3 and 9 and did not reveal any predominant lymphocyte clones. A bone marrow analysis was performed in six patients; in all cases, the bone marrow aspirate showed normal or increased cellularity. Myeloid cell count was normal in three cases (patients 3, 6, and 9) and showed a myeloid maturational arrest at the myelocyte/metamyelocyte stage in three others (patients 1, 7, and 8). Among these, one (patient 7) had a polyclonal TCR ah+ CD3+, CD8+ cell expansion (38%) and another (patient 4) had a TCR gy+ CD3+, CD4ˇ€‹, CD8+ lymphocyte expansion that was also found in blood. 3.3. ANCA detection and characterization Using an IIF assay and ethanol-fixed neutrophils, ANCA were found to be positive in all patients (Table 2). The fluorescence pattern was cytoplasmic (cANCA) in four cases (patients 1, 2, 5, and 6), perinuclear (pANCA) in two cases (patients 3 and 7), and atypical (xANCA) in three (patients 4, 8, and 9; Fig. 1). Sera were further tested by IIF using ethanol-, methanol-, formaldehyde-, acetone-form- aldehyde-fixed neutrophils and found to label neutrophil cytoplasm in all cases. Using ELISA, the ANCA specificity was anti-PR 3 (patient 2), anti-MPO plus anti-elastase (patient 3), anti- MPO (patient 8), and anti-lactoferrin (patient 7). Reac- tivity of patients 3 and 8 with MPO was confirmed by Western blot using the purified enzyme (data not shown). Fig. 1. Fluorescence photomicrograph of neutrophils after indirect immunofluorescence staining using serum from patient 1. Magnification ˇ€‹40. In the remaining patients, the specificity was undetermined by ELISA using MPO, PR3, cathepsin G, elastase, BPI, and lactoferrin as substrates and by Western blot using MPO. ANCA were not detected by IIF or ELISA in the control sera obtained from 22 healthy individuals, from 8 children with anti-NA1 or anti-NA2-associated neutrope- nia, from patients with drug-induced neutropenia or various autoimmune disorders, or from 8 intravenous Ig- treated CVID patients, ruling out the possibility that ANCA were passively transmitted through IVIG prepara- tions in patient 7. Three sera (patients 1, 4, and 5), negative by ELISA, were tested by Western blot using cytoplasmic extracts of freshly isolated polymorphonuclear cells. As shown in Fig. 2, these sera displayed a heterogeneous pattern of reactivity. The serum of patient 1 showed a strong reactivity with two proteins of 60 and 33 kD and a weaker reactivity with two proteins of 45 and 38 kDa (lane 3). A strong reactivity with 61 and 52 kDa proteins and a weaker reactivity with additional proteins (45, 43, and 27 kDa) were obtained with patient 4 serum (lane 4). Patient 5 serum revealed two major bands of 59 and 53 kDa, and one additional faint band of 45 kDa (lane 5). The serum of patient 8 (lane 6), which Table 2 Main immunological findings Immunological data Patients ANCA Titer (at diagnosis)a Fluorescence pattern Specificity Titer 100 400 400 Fluorescence pattern speckled homogeneous homogeneous 123 45 67 80 640 cANCA pANCA n.i. lactoferrin ˇ€‹ˇ€‹ ˇ€‹+ 89 80 160 ANCA xANCA MPO n.i. ˇ€‹ˇ€‹ ++ 160 320 speckled homogeneous ˇ€‹ˇ€‹ ˇ€‹ ND 320 80 cANCA cANCA NI PR3 +ˇ€‹ˇ€‹ ++ Antisurface neutrophil Abs AntinuclearAbs ˇ€‹+ˇ€‹ ˇ€‹+ Coombs test +ˇ€‹+ +ˇ€‹ Antiplatelet Abs ˇ€‹ ˇ€‹ ˇ€‹ ˇ€‹ + Antibodies (Abs), not identified (n.i.), not done (ND). a Positive if N20. ˇ€‹+ ND+ 80 5120 80 pANCA xANCA cANCA MPO/elastase n.i. n.i. P. Coppo et al. / European Journal of Internal Medicine 15 (2004) 451 459 455 3.5. Other findings Fig. 2. Neutrophil cytoplasmic proteins were electrophoretically separated by SDS 10% PAGE, transferred onto nitrocellulose, and probed with normal serum (lane 1), control serum with anti-MPO Abs (lane 2), and sera from patients 1 (lane 3), 4 (lane 4), 5 (lane 5), and 8 (lane 6). Molecular weights are indicated on the right. contained anti-MPO Abs, recognized the 58 kDa heavy chain of MPO, as did the anti-MPO control serum (lane 2), each of which reacted with additional proteins. 3.4. Other biological findings (Table 2) Three patients had antisurface-neutrophil Abs detected by GIFT, the specificity of which could not be identified (patients 1, 4, and 5); anti-NA1, NA2, NB1, and NB2 specificities were ruled out in all three patients. ANA were detected in five patients, displaying a homogeneous pattern or a speckled pattern. No specificity could be identified in any of the cases; there were neither antidouble-stranded DNA nor anti-ENA autoantibodies. Rheumatoid factor with decreased C3, C4, CH 50 was found in two patients associated with a type III (patient 3) or a type II (patient 8) mixed cryoglobulinemia. Antikeratin and antifilaggrin Abs were not detected in the two patients with non-destructive arthritis (patients 4 and 8). Patients 1, 3, 4, and 7 had a positive Coombs test associated with hemolytic anemia. Cold agglutinins (polyclonal IgM with an anti-I specificity) were also present in patient 3 at a titer of 125. Antiplatelet glycoprotein IIb/IIIa Abs was found in the two patients who had peripheral thrombocytopenia. The gamma globulin level was normal in all patients and none of them had a deficiency in serum IgA or IgG subclasses, except for the patient with the CVID. Serum creatinine level was normal and the search for proteinuria and hematuria was consistently negative. Serol- ogies for acute viral infections (EBV, CMV, parvovirus B19), as well as for HIV and HTLV-1, were negative. CMV antigenemia, HBS antigen, and HCV and HIV-1 transcripts were absent in the CVID patient. Computerized abdominal and chest tomography was performed in six patients (1, 3, 4, 7, 8, and 9). It was normal in patients 1, 8, and 9 and confirmed the hepatomegaly in patient 4 and the spleno- megaly in patients 3 and 4. In patient 7, it revealed a thymoma, consistent with the diagnosis of Good s syn- drome, since this patient had a CVID. 3.6. Treatment and follow-up In this retrospective analysis of patients from two separate hospitals with various underlying conditions and different degrees of severity of neutropenia, treatment varied. Patients 2, 5, and 6, who had neutropenia between 1.15ˇ€‹109/L and 1.5ˇ€‹109/L, were not treated. The neutro- penia remained stable and none of them experienced any infections within a median follow-up of 2 years (range 1 7 years). They have remained in good health without any evolution to a characterized disease. Patients 1, 3, 4, and 9 were variably treated with IVIG, steroids, G-CSF, splenectomy, and/or oral methotrexate, resulting in a complete recovery in all cases (long-term in patients 1 and 9). Patients 3 and 4 required intermittent therapy with G-CSF in order to maintain neutrophil counts above 1ˇ€‹109/L. When neutropenia appeared in patient 7, who had a CVID, she was initially treated with G-CSF (250 Ag/day) for 8 days with little effect. She was then given prednisolone (0.5 mg/kg per day) for 5 days and two courses of cyclophosphamide, at a dose of 800 mg every month, allowing for an increase in the neutrophil count. Yet, she subsequently died of pneumonia. Patient 8 was treated with steroids (1 mg/kg per day for 1 month) with no effect; neutrophil counts have remained below 0.5ˇ€‹109. Because he never experienced infectious complications after 3 years of follow-up, he did not receive further treatment. The evolution under treatment of the neutrophil count in patients 3, 4, and 7 is shown in Fig. 3. None of the patients, in particular patient 2, who had anti-PR3 Abs, developed manifestations suggestive of Wegener s disease, microscopic polyangiitis, or crescentic glomerulonephritis after a follow-up of 7 years. During follow-up, ANCA remained present with fluctuating titers in the sera of all patients, and there was no clear correlation between ANCA titers and neutrophil counts. None of the patients, in particular patient 2, who had anti-PR3 Abs, developed manifestations suggestive of 456 P. Coppo et al. / European Journal of Internal Medicine 15 (2004) 451 459 Fig. 3. Evolution of the neutrophil count in three patients. Cyclo- phosphamide (CPM), methotrexate (MTX), granulocyte colony-stimulating factor (G-CSF). Wegener s disease, microscopic polyangiitis, or crescentic glomerulonephritis after a follow-up of 7 years. During follow-up, ANCA remained present with fluctuating titers in the sera of all patients, and there was no clear correlation between ANCA titers and neutrophil counts. 4. Discussion ANCA is considered to be important serological markers for primary systemic vasculitides. When detected by IIF assays using the patient s serum and ethanol-fixed human neutrophils, ANCA display three different fluorescent staining patterns: a cytoplasmic (cANCA) pattern, a perinuclear (pANCA) pattern and, more rarely, an atypical (x-ANCA) pattern. cANCA are usually directed to the 29- kDa neutral serine protease PR3, whereas pANCA are directed to MPO or, occasionally, to other components of azurophilic granules, such as lysozyme, elastase, cathepsin G, lactoferrin, azurocidin, or BPI. ANCA specificity can be routinely assessed with Western blot or ELISA using specific antigens. Anti-PR3 serum reactivity is present in up to 90% of patients with Wegener s granulomatosis [9,29,30], but it can also be detected in patients with microscopic polyarteritis and necrotizing crescentic glomer- ulonephritis [7,8,12]. The detection of ANCA is, however, very rare in the general population, as opposed to ANA or rheumatoid factor [31], and false-positive results have been chiefly described in patients with polyclonal hypergamma- globulinemia and HIV infection [32]. We report herein nine patients who share both non- cyclical chronic neutropenia and a positive ANCA test. ANCA specificity could be identified in only four patients and was different from one patient to the next (MPO, elastase, and lactoferrin). Sera from the five other patients did not exhibit any reactivity with PR3, MPO, elastase, cathepsin G, BPI, or lactoferrin according to the ELISA. With Western blot, three of these five ELISA-negative sera showed reactivity to several granulocyte cytoplasmic proteins whose precise identification will require further investigation. However, we cannot rule out the possibility that some other antigenic targets may not have been detected with Western blot. In keeping with this hypothesis, ANCA was often reported to react poorly by immunoblot- ting with either crude neutrophil extracts or azurophilic granule extracts [33]. There was no evidence of either a toxic or immunoaller- gic mechanism. Hypersplenism does not seem probable since only two patients had moderate splenomegaly and the neutropenia relapsed in the one who underwent a splenec- tomy. There was no paroxysmal nocturnal hemoglobinuria clone detected in our patients (data not shown). Bone marrow examination showed a myeloid maturational arrest at the myelocyte/metamyelocyte stage with no evidence of myelodysplasia or blast excess. Chronic T-cell proliferation was observed in the periph- eral blood and/or in bone marrow in three patients. One was a polyclonal T CD8+ bone marrow cell proliferation associated with Good s syndrome. This case is reminiscent of those of two other patients previously reported who had an IgG2 subclass deficiency associated with a polyclonal T- lymphocyte bone marrow infiltration and severe neutropenia [34]. It is important to mention that, although agranulocy- tosis has been described in Good s syndrome, it is not usually associated with bone marrow lymphoid infiltration [35]. Two patients had chronic T-cell expansion that consisted of clonal T gy+ lymphocytes both in peripheral blood and in bone marrow, or polyclonal CD3+ CD8+ ah+ LGL in peripheral blood only, associated with or without splenomegaly. Since both patients suffered from a sero- negative non-destructive polyarthritis, Felty s syndrome or a Felty s-like syndrome might have been envisaged. More- over, ANCA have been reported at very early stages of rheumatoid arthritis, albeit without any association with neutropenia, and seem to be predictive of rapid radiographic joint destruction [36]. That our patients never developed any joint destruction or serological markers of rheumatoid arthritis pleads strongly against this diagnosis. The mech- anism of neutropenia in patients with T-cell proliferation is unknown. While LGL are known to have multiple effects on the hematopoietic system, in vitro studies have generally failed to demonstrate any suppressive effect of LGL on granulopoiesis [37,38]. In our six other patients, an autoimmune mechanism was strongly suggested by the association with manifestations of autoimmunity such as Hashimoto s thyroiditis, Sjfgren s syndrome, autoimmune hemolytic anemia or thrombocyto- penia, and the presence of a large variety of autoantibodies. Furthermore, the presence of circulating autoantibodies directed against surface neutrophil antigens in three cases provides additional evidence for the autoimmune origin of the neutropenia. The absence of detectable surface neutro- phil autoantibodies in the remaining six patients may be explained by the nonexpression of the target antigen by the neutrophils used in the assays, the very low titer of the circulating autoantibodies, and/or a very high affinity, causing the Abs to be essentially bound to their targets. The pathogenic role of antineutrophil autoantibodies is still not completely clear, apart from the autoimmune neutropenia of children, which generally combines an isolated granulocytopenia and circulating autoantibodies usually directed against isoforms of CD16 such as NA1, NA2, NB1, and NB2 [39]. In adults, the neutropenia is often associated with other autoimmune manifestations, particu- larly autoimmune thrombocytopenia and hemolytic anemia [40,41]. However, the clinical significance of antineutrophil autoantibodies is far from unequivocal as they can often be found in various autoimmune disorders such as SLE or Sjfgren s syndrome in patients with a normal neutrophil count [41,42]. In diseases commonly associated with ANCA, in particular Wegener s granulomatosis, there is often a marked hyperleukocytosis [43]. Therefore, the presence of ANCA in neutropenic patients is unexpected and raises the question of their possible role in the pathogenesis of the neutropenia. One hypothesis is that the usually intracytoplasmic targets of ANCA might be redistributed on the patients neutrophil surface, thereby enabling their recognition by ANCA and neutrophil destruction, by either a complement-mediated or antibody-dependent cytotoxicity. Interestingly, it has been established that antigens like PR3 and MPO can be surface- expressed upon neutrophil activation [44 47]. Therefore, ANCA-associated neutropenia might represent an uncom- mon subset of autoimmune neutropenia with particular antineutrophil autoantibodies directed against targets that are normally located in the cytoplasm but could be induced or aberrantly expressed on the neutrophil surface of these patients. This would provide an explanation for the negativity of assays like GATT and GIFT in most of our cases since they use pooled neutrophils from healthy donors. In keeping with our hypothesis, sera from four of our patients transiently stained both the cytoplasm and the surface of all-trans retinoic acid-stimulated myelomonocytic HL60 cells (personal data). Experiments are currently being conducted to identify the target antigens recognized by these Abs. As has been suggested for other autoantibodies [48], ANCA have recently been shown to enter the PMN and cause cell damage, including apoptosis [49]. However, we cannot exclude the possibility that T lymphocytes may play a major role. They could either be self-reactive cytotoxic T lymphocytes, producing granulopoiesis inhibitory cyto- kines, or induce apoptosis of polymorphonuclear cells via Fas/Fas ligand interaction. In this latter setting, one hypothesis is that ANCA might be induced by a massive release of target antigens by apoptotic neutrophils [48] that could stimulate self-reactive T and B lymphocytes as a secondary event. This may explain the absence of a correlation between ANCA titers and neutrophil counts, as well as the large number of antigen targets detected by Western blot and ELISA. To summarize, we report a series of nine patients with a heterogeneous spectrum of underlying diseases but who all share the association of ANCA and neutropenia of unknown origin in which an autoimmune mechanism is strongly suggested. The responsibility of ANCA in the pathogenesis of the neutropenia cannot be affirmed. A systematic analysis of a larger series of patients is required to better define the clinical significance of neutropenia associated with ANCA. We believe that ANCA should now be sought in every neutropenia of undetermined origin. This would provide a strong argument for an autoimmune mechanism. Until a pathogenic role can clearly be determined, ANCA should at least be considered to be an interesting marker of auto- immune neutropenia. Treatment should be considered individually since ANCA-associated neutropenia may be of variable severity and can be associated with various other conditions. Acknowledgements The authors thank Muriel Bargis-Touchard, Sylvie Duflot,andBe ne dictePecquet for their expert secretarial assistance. ˇ˛European Journal of Internal Medicine 20 (2009) 403 406 Contents lists available at ScienceDirect European Journal of Internal Medicine journal homepage: www.elsevier.com/locate/ejim Original article Antioxidant enzyme levels in cases with gastrointesinal cancer Yalcin Kekec, Semra Paydas N , Abdullah Tuli, Suzan Zorludemir, Gurhan Sakman, Gulsah Seydaoglu Cukurova University Faculty of Medicine Departments of General Surgery, Medical oncology, Biochemistry, Pathology and Biostatistics, Turkey article info Article history: Received 26 August 2008 Received in revised form 20 November 2008 Accepted 18 December 2008 Available online 26 January 2009 Keywords: Gastrointestinal cancer Antioxidant system SOD G6PD GR 1. Introduction Oxygen free radicals, including highly reactive superoxide radicals termed as ROS have been of increasing interest in cancer researches in the past two decades. The effects of ROS can be harmful and beneficial [1,2]. At low levels, ROS can induce a mitogenic effect and participates intracellular signallng pathways [3,4]. The induction of ROS induces DNA damage and has been shown to be involved in the etiology of several diseases including cancer [2,4]. These harmful effects of ROS are protected by both enzymatic and non-enzymatic antioxidants [1]. ROS are generated during normal aerobic metabolism and increased levels are present during oxidative stress. It has been proposed that ROS is necessary for life and essential for the regulation of essential physiologic functions. However, at high concentrations, ROS are cytotoxic. ROS are important in cell differentiation, apoptosis, and cell proliferation. These functions are regulated by redox-sensitive signal transduction pathways. The amount of antioxidants in the cells is high and so cells prevent or repair itself from damages caused by ROS. ROS-induced damage can result in cell death, mutations, chromosomal aberrations and also carcinogenesis [5]. For this reason there is great interest to ROS in cancer in the last years. These N Corresponding author. Cukurova University Faculty of Medicine Department of Medical Oncology, Adana/Turkey. E-mail address: sepay@cu.edu.tr (S. Paydas). abstract The aim is to evaluate the antioxidant enzyme levels in tumoral tissues and accompanying normal tissues in gastrointestinal cancer; and compare the colorectal cancer (CRC) with gastric cancer (GC). Method: Antioxidant enzymes including glutathione reductase (GR), glutathione peroxidase (GPX), super- oxide dismutase (SOD), malondialdehyde (MDA) and glucose 6 phosphate dehyrogenase (G6PD) which are important for anti-oxidant functions were evaluated in fresh tumor tissues and adjacent normal tissues obtained from a total of 58 patients. Results: All the enzyme levels were higher in tumoral tissues compared to normal tissue from non-cancerous disease. There was not a significant difference for enzyme levels between CRC and GC groups except GPx. GPx activity tended to be higher in cases without serosal involvement (SI), and this activity was higher in cases without lymph node (LN) involvement in normal tissue (p = 0.012). MDA activity was higher in cases without serosal involvement compared to with SI groups in tumor tissue (p=0.050). G6PD activity in normal tissue was higher in cases with serosal involvement and LN involvement (p=0.064, 0.046, respectively). GR activity was higher in signet ring cell cancer (SRC) than adeno cancer. In GC, G6PD activity in tumor was tended to be higher in undifferentiated cancer (p=0.071). Conclusion: The antioxidant enzymes activities such as GPX, SOD, G6PD, MDA and GR were found to be related with malignant phenotype in gastrointestinal cancers. We need further studies to understand the biologic and clinical importance of these enzymes in GI cancers. © 2009 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. antioxidants are antioxidant enzymes and some vitamins. There are three major types of antioxidant enzymes in mammalian cells: superoxide dismutase (SOD), catalase (CAT), and peroxidase, of which glutathione peroxidase (GPx) is the most important component of these [6,7]. MDA is a naturally occurring product of lipid peroxidation and prostaglandin synthesis. Lipid hydroperoxides are formed in vivo through free radical pathways from the action of ROS on the polyunsaturated fatty acids. MDA is formed during decomposition of lipid hydropeoxides. The major adduct to DNA is a pyrimidopurinone called M1G (M1G is a secondary DNA damage product arising from primary reactive oxygen species (ROS) damage to membrane lipids or deoxyribose) and this appears to be a major endogenous DNA adduct in human tissues. This may contribute to cancer development [8,9]. In this study, antioxidant enzymes such as GR, GPX, SOD, MDA and G6PD activities were studied in tumoral tissues and normal tissue from non-cancerous disease in 58 cases with gastrointestinal cancer and compared them according to some clinical and histo-pathological characteristics of these tumors. 1.1. Method Antioxidant enzyme (AOE) system including glutathione reductase (GR), glutathione peroxidase (GPX), superoxide dismutase (SOD), malondialdehyde (MDA) and glucose 6 phosphate dehyrogenase 0953-6205/$ see front matter © 2009 European Federation of Internal Medicine. Published by Elsevier B.V. All rights reserved. doi:10.1016/j.ejim.2008.12.003 404 Y. Kekec et al. / European Journal of Internal Medicine 20 (2009) 403 406 (G6PD) was evaluated both in fresh tumor tissues and adjacent normal tissues obtained from a total of 58 patients with gastrointestinal cancer (33 had CRC and 25 were had GC). 2. Patients and methods 58 cases were included in this study; 25 of them had GC and 33 had CRC. Fresh tumor tissues and normal tissue from non-cancerous disease were used as study material. 2.1. Sample preparation Fresh tumor tissues and adjacent normal tissues obtained from patients with colorectal and gastric cancer immediately after surgery was placed in cold 0.9% NaCl solution. The tissues were blotted on filter paper, weighed and in 0.25 M sucrose and 0.15 M NaCl with addition of 6 ºl 250 mM butylated hydroxytoluene in ethanol, to prevent formation of new peroxides during the assay. The tissues were homogenized under standardized conditions and centrifuged at 10.000 ◊g for 10 min at 4 ∞C. The supernatant was kept in an ice- cold condition until assayed. 3. Biochemical assays Superoxide dismutase (SOD) enzyme was assayed according to the methods of McCord et al. [10]. The enzyme specific activity was expressed in units per milligram protein. Glutathione reductase (GR) activity was determined by the method of Mize and Langdon by monitoring the oxidation of nicotinamide adenine dinucleotide phosphate (NADPH) at 340 nm [11]. The activity was expressed in millimoles NADPH per minute per milligram protein. Glutathione peroxidase (GSH-Px) activity was measured spectro- photometrically using a technique based on the method of Paglia and Valentine [12]. The activity was expressed in universal units. One unit of the enzyme activity is defined as the amount of enzyme catalyzing the oxidation of ºmol NADPH/min/mg protein. Glucose 6-phosphate dehydrogenase enzyme activity was assayed by the method of Beutler and Aksoy et al. [13,14]. The enzyme activity was expressed in micromoles NADPH reduced per minute per mg protein. The lipid peroxide level in tissue was assayed in terms of malondialdehyde level (pmol/mg protein), the end product of lipid peroxidation, by using method described by Ohkawa et al. [15]. Protein was determined in diluted aliquots of the tissue homogenates by the method of Lowry et al. with bovine serum albumin as the standard [16]. 3.1. Statistical methods For each continuous variable, normality was checked. Student t test was chosen for normally distributed data. Appropriate non- parametric tests were chosen (Mann Whitney U and Kruskal Wallis test) for the data was not normally distributed. Since analysis of variance was significant, comparisons were applied using the Mann Whitney U test. Bonferroni's correction was applied (p b 0.05/n; where n=number of comparisons) when multiple comparison were made. Results were presented as mean±SD and median. 4. Results Thirty-three cases with CRC and 25 cases with GC were included. Mean age was 53.62±1.93 (median: 53 range 22 84) in total cases. Age range was 40 84 and 22 78 for GC and CRC, respectively. Female/ male ratio was 19/14 in CRC group and 10/15 in GC group. Histologically; 28 cases had adeno cancer, 2 had SRC and 3 had undifferentiated cancer in CRC group. In GC group; 15 cases had adeno cancer, 6 had SRC and 4 had undifferentiated cancer. Serosal involvement was detected in 25 cases with CRC and 20 cases with GC. Lymph node involvement was seen in 22 cases with CRC and 20 cases with GC. Table 1 shows the enzyme levels of CRC and GC cases both in tumoral tissues and accompanying normal tissues. All the enzyme levels, were found to be significantly higher in tumoral tissues comparing to normal tissues (pb0.001 for all, not stated in the table). There was not a significant difference for enzyme levels between CRC and GC groups except GPx in normal tissue from non-cancerous disease; GPx activity is significantly higher in CRC group than GC group (p=0.04). There were no significant differences between CRC and GC groups for enzyme levels in tumoral tissues. In Table 2 enzyme levels were evaluated between sex, serosal involvement, lymph node involvement and histologic type groups in non-tumoral and tumoral tissues. Enzyme levels were not different between males and females both in non-tumoral and tumoral tissues. G6PD activity in non-tumoral tissue was found to be higher in cases with lymph node involvement as compared without lymph node involvement (p = 0.046). GPx activity, both in tumor and non-tumoral tissue, tended to be higher in cases without serosal involvemet as compared with serosal involvement group (p = 0.05 and 0.047, respectively) and MDA activity was higher in cases without serosal involvement compared with serosal involvement group in tumor tissue (p = 0.05). According to histological types; GR activity was higher in SRC group than AC (p = 0.011) and GPx activity was higher in AC group than SRC group in tumoral tissue (p = 0.019). These comparisons performed by the subgroups (CRC and GC) of gastrointestinal cancer. However in cases with CRC, G6PD activity in non-tumoral and tumoral tissue was found to be higher in females than males (p = 0.034 and 0.043, respectively). GR activity in normal tissue was found to be higher in females than males (p=0.018). In CRC, GPx activity in normal tissue was higher in cases with without serosal involvement (p = 0.02), GPx activity in tumor tissue was higher in cases Table 1 The enzyme levels of colo-rectal cancer and gastric cancer cases both in tumoral tissues and accompanying normal tissues. Tissues Enzyme Colo-rectal cancer (n=33) Mean ± SD Median Gastric cancer (n=25) p value Mean ± SD Median Non-tumoral GR 0.19±0.04 0.20 ± 0.06 0.18 0.33 ± 0.10 0.31 10.21 ± 6.0 9.6 6.76 ± 4.9 5.0 15.20 ± 10.2 12.0 0.41 ± 0.11 0.38 0.21 ± 0.10 0.19 34.06 ± 10.0 34.2 34.28 ± 15.9 32.0 47.28 ± 26.5 45.0 0.577 0.040 0.455 1.000 0.545 0.132 0.113 0.666 0.666 0.338 Tumor GR 14.0 0.36 ± 0.11 0.37 0.18 GPX 0.38 ±0.12 0.34 SOD 8.95 ± 4.9 9.0 MDA 7.06 ± 5.4 5.0 G6PD 15.15 ± 7.1 GPX 0.23 ± 0.07 0.22 SOD 36.08 ± 13.7 36.0 MDA 35.11 ± 14.2 33.0 G6PD 52.70 ± 22.9 48.0 Superoxide dismutase (SOD): units per milligram protein. Malondialdehyde (MDA): units per milligram protein. Glutathione reductase (GR): millimoles NADPH per minute per milligram protein. Glutathione peroxidase (GSH-Px): ºmol NADPH/min/mg protein. Glucose 6-phosphate dehidrogenase (G6PD): micromoles NADPH reduced per minute per mg protein. Y. Kekec et al. / European Journal of Internal Medicine 20 (2009) 403 406 405 Table 2 The enzyme levels according to the sex, lymph node involvement, serosal invasion and histological subtype. Sex Lymph node involvement Serosal invasion Histological type Mean±SD Mean±SD Mean±SD Mean±SD Female Male No Yes No Yes AC SRC Other (n=29) (n=29) (n=16) (n=42) (n=13) (n=45) (n=43) (n=8) (n=7) Non-tumoral GR 0.20±0.06 0.18±0.03 0.19±0.04 0.19±0.05 0.18±0.03 0.19±0.05 0.19±0.04 0.21±0.10 0.18±0.04 GPX 0.35±0.11 0.36 ± 0.11 0.40 ± 0.13 0.34 ± 0.10 0.42 ± 0.15 0.34 ± 0.09* 0.36 ± 0.11 0.34 ± 0.12 0.35±0.11 SOD 9.2±5.1 9.7±5.7 9.2±3.3 9.6±6.0 8.5±3.2 9.7±5.9 9.5±5.5 11.6±5.9 7.0±3.5 MDA 7.2±5.0 6.6±5.4 5.9±4.7 7.3±5.3 6.5±5.4 7.0±5.2 7.0±5.2 7.6±6.4 5.3±3.4 G6PD 16.3±8.0 13.9±8.9 11.2±6.5 16.6±8.8* Tumor 12.5±5.9 15.9±9.1 14.9±8.5 13.8±7.1 17.8±10.6 GR 0.41±0.12 GPX 0.22±0.08 SOD 36.0±12.8 MDA 34.0±13.3 G6PD 53.2±24.2 0.36±0.10 0.36±0.15 0.22±0.09 0.21±0.07 34.3±11.7 36.6±11.8 35.4±16.4 34.9±17.7 47.4±24.8 45.1±29.2 0.39±0.10 0.22±0.09 34.6±12.4 34.6±13.8 52.3±22.5 0.36±0.11 0.27±0.10 40.1±11.3 42.3±16.7 42.9±23.1 0.39±0.11 0.37±0.12 0.21±0.08* 0.23±0.08 33.8±12.2 35.9±11.7 32.5±13.7* 36.3±15.3 52.5±24.7 49.6±25.6 0.46±0.10 0.17±0.13 33.8±11.2 28.9±13.7 43.2±18.1 0.40±0.06* 0.20±0.07* 32.2±17.2 31.8±12.6 63.1±21.7 *pb0.05 between groups. SRC: signet ring cell cancer. AC: adeno cancer. Superoxide dismutase (SOD): units per milligram protein. Glutathione reductase (GR): millimoles NADPH per minute per milligram protein. Glutathione peroxidase (GSH-Px): ºmol NADPH/min/mg protein. Glucose 6-phosphate dehidrogenase (G6PD): micromoles NADPH reduced per minute per mg protein. without serosal involvement (p = 0.016). In CRC, GPx activity was higher in cases without lymph node involvement (p = 0.012). In GC, G6PD activity in normal tissue was higher in cases with lymph node involvement (p = 0.012). However in CRC, GR activity in normal tissue was higher in AC than others (p = 0.05), GPx and SOD activities in tumor tissue was higher in AC than others (p = 0.004 and 0.029, respectively). In GC, G6PD activity in tumor tended to be higher in undifferentiated cancer (p=0.071). 5. Discussion It is very well known that the enzymatic antioxidant system includes SOD, CAT and GPX. Reactive oxygen metabolites are implicated in the initiation and promotion of cancer [17]. Numerous papers have been published about the altered levels of antioxidant enzymes in cancer cells. Physiologically SODs convert superoxide radical into hydrogen peroxide and molecular oxygen (O2). However the catalases and peroxidases convert hydrogen peroxide into water. In this way, two toxic species, superoxide radical and hydrogen peroxide, are converted to the water. These antioxidant enzymes are thought to be necessary for life in all oxygen metabolizing cells. There is no need to co-factors for SOD and CAT functions, but GPx requires several co-factors and proteins. Two proteins; GR and G-6-PD are considered secondary antioxidant enzymes, since they do not act on ROS, but they are helpers to the GPx to its function [7]. Cancer cells are nearly always low in MnSOD and CAT activity, and usually low in Cu/ZnSOD activity [18]. For this reason MnSOD initially has been proposed as a tumor suppressor gene on the basis of low SOD expression in malignant tumors [19,20]. However several studies have challenged this tumor suppressor concept due to the higher MnSOD expression in neoplasms [17,21,22]. We found higher SOD activity in tumoral tisues as compared with accompanying normal mucosae. However we found SOD activities in tumor tissue was higher in ACthan SRC which is more malignant phenotype. These results suggest and support the variability of this enzyme sytem in GI cancers. It has been shown that in some cancers, reduced expression of MnSOD is due to mutations in the promoter of the gene, while in other types of cancer, reduced levels ofMnSOD are due to abnormal methylation, loss of heterozygosity or mutation in the coding sequence [23 25]. These diferent mechanisms cause the chaotic results about SOD in malignant tumors including GC, CRC and other GI tumors [17,21,22,26 28]. GPx activity is variable [7]. In our study, the only enzyme found to be lower in tumors as compared with normal tissue from non- cancerous disease was GPx and we do not know the exact cause of this. Oxygen radicals react with polyunsaturated fatty acid residues in phospholipids. One of the most abundant carbonyl products of lipid peroxidation is MDA, which reacts with DNA to form adducts. Lipid peroxidation appears to be a major source of endogenous DNA damage in humans that may contribute significantly to the occurrence cancer and other genetic diseases [29]. We found higher MDA levels in tumor tissues as compared with accompanying normal mucosae and also MDA was found to be associated with more malignant phenotype. This is not surprise, because MDA is a product of lipid peroxidation and prostaglandin synthesis which is very important in gastrointest- inal malignant tumor development. The content of AOEs in human neoplastic and nonneoplastic tissues has been reported to be different with regard to the organs and enzyme forms It has not been found an association between SOD levels and clinic-pathological parameters in all studies [17]. The significance of these ROS in human malignant tumors is controversial and requires further studies. 6. Learning points " This study is about the pathogenetic role of the antioxidant system gastrointestinal system. There are many enzyme systems in this pathway and some are harmful for organism. It is important to detect their role and also opposite systems to prevent carcinogenesis. ˇ˛Antiphospholipid syndrome Steven Austin Hannah Cohen Abstract Antiphospholipid syndrome (APS) is an autoimmune disorder in which autoantibody production can lead to a hypercoagulable state, pregnancy failure and/or a multitude of other systemic manifestations. In recent times, research into the antiphospholipid antibodies has increased our understanding of the pathogenic process and encouraged improved detection of antiphospholipid antibodies. However, the precise nature of the pathology of APS remains challenging, as does the management of this disorder. This review outlines the key features of APS, including the laboratory tests and their interpretation, and offers advice regarding the management of patients with APS both in the medical and obstetric settings. Keywords anticardiolipin antibodies; anticoagulation; antiphospholipid syndrome; b2-glycoprotein-I antibodies; lupus anticoagulant; obstetric morbidity; thrombosis Antiphospholipid syndrome (APS) is characterized by throm- bosis (venous and/or arterial) and/or pregnancy failure in association with the persistent production of a group of hetero- geneous autoantibodies known as antiphospholipid antibodies (aPL). The primary targets of aPL are phospholipid-binding proteins, although antibodies directed against phospholipids and other proteins also occur. Additional features, particularly thrombocytopenia, are variably present. In the laboratory, the usual diagnostic tests for aPL are: lupus anticoagulants (LA), which cause prolongation of phospholipid-dependent clotting assays (e.g. activated partial thromboplastin time, dilute Russell s viper venom time [DRVVT]); this effect is abolished by the addition of excess phospholipid (e.g. from platelets) anticardiolipin antibodies (aCL) of immunoglobulin G (IgG) and IgM class, which are determined by enzyme-linked immunosorbent assay (ELISA) b2-glycoprotein-I (b2-GPI) antibodies of IgG and IgM class detected by ELISA The prevalence of aPL in the form of LA or aCL is 1e5% of healthy individuals. The prevalence increases in the elderly and in those with chronic disease. APS has conventionally been divided into primary and secondary forms; the latter is associated Steven Austin MBBS FRACP FRCPA is a Locum Consultant Haematologist at St George s Hospital, London, UK. Competing interests: none declared. Hannah Cohen MD FRCP FRCPath is Consultant Haematologist at University College London Hospitals NHS Foundation Trust, London, UK. Competing interests: none declared. with chronic inflammatory conditions, mainly systemic lupus erythematosus (SLE). However, this distinction was abandoned in the Sydney classification on the basis that it is unknown whether APS and SLE are two diseases coinciding in an indi- vidual, underlying SLE offers a setting for the development of APS, or APS and SLE represent two elements of the same process. Several studies have shown that the prevalence of aPL in SLE patients is variable (15e86%). The frequency of antibody posi- tivity is likely to be around 30%, with the wide variation found in the literature explained by study variations, ethnicity, and extent of autoimmune disease activity. Up to an estimated 40% of patients with SLE and aPL will eventually develop clinical features consistent with APS, whereas under 5% of patients with APS will develop SLE. Pathogenesis Recent progress has furthered our understanding of the mecha- nism behind the development of aPL. However, a precise path- ophysiological mechanism for the clinical features of APS remains elusive. Proposed prothrombotic mechanisms, all of which can result in a prothrombotic state, include: ˇ€‹ acquired resistance to activated protein C ˇ€‹ endothelial cell activation ˇ€‹ up-regulation of tissue factor ˇ€‹ reduced fibrinolysis ˇ€‹ oxidant-mediated endothelial injury ˇ€‹ ADAMTS13 (a disintegrin-like and metalloprotease with thrombospondin type 1 motif, member 13) dysfunction. Platelet and monocyte activation may also promote throm- bosis. Additional mechanisms of displacement of annexin V and complement activation have been linked with direct cellular damage and abnormal placentation, leading to implantation failure and foetal loss. The clinical significance of any one (or more) of these hypotheses remains unclear and reflects the likely multifactorial complex nature of this condition, as is generally the case in acute thrombosis. Animal studies have suggested that the mechanism of molecular mimicry plays an important role in experimental APS. OTHER AUTOIMMUNE DISORDERS What s new? C Antibodies directed against domain I of b2-GPI correlate better with thromboembolic complications than antibodies directed against other domains of b2-GPI C Low titre antiphospholipid antibodies appear to be implicated in recurrent miscarriage C High-intensity anticoagulation is appropriate for some patients e however, substantive evidence is lacking C Low-dose aspirin is ineffective for the prevention of thrombosis in patients with asymptomatic aPL C Plasmapheresis, intravenous immunoglobulin and immuno- suppression are non-evidence-based therapies gaining popu- larity for the management of patients with resistant disease C Statins and ACE inhibitors have been shown to possess anti- inflammatory properties in vitro and may have a role in the armamentarium of APS in the future MEDICINE 38:2 101 O 2009 Elsevier Ltd. All rights reserved. Bacterial peptides homologous to b2-GPI infused into mice have been reported to induce antibodies to b2-GPI along with APS manifestations. The development of aPL is probably only one step towards the development of APS, and it is likely that other factors play a role. Such second hits or triggers may tip haemostasis in favour of a prothrombotic state and include infection, endothelial injury, and other non-immunological pro- coagulant factors. The patient s genetic background (in relation to candidate genes of inflammatory mediators) may also be a critical variable for the development of clinical APS manifes- tations. It is now accepted that anti-b2-GPI are the major path- ological antibodies in APS, although there is no clear consensus on how the occurrence of these antibodies is associated with the various clinical features. Systematic review of published data suggests that LA are a stronger risk factor than aCL for throm- bosis (arterial or venous) and aCL are associated more with arterial than venous thrombosis. Anti-b2-GPI may confer a stronger thrombotic risk than do aCL especially for venous thrombosis. Positivity in more than one assay and immuno- globulin subclass (IgG being more significant for thrombosis) is also important. Low-titre antibodies appear to be implicated in recurrent miscarriage. Clinical features Thrombosis APS is associated with venous and arterial thrombotic events. Among patients with venous thromboembolism, 3e17% have aCL, and 3e14% have LA. Well-established risk factors for thrombosis (e.g. pregnancy and surgery) also increase the risk of thrombosis in APS patients, as do coexistent inherited thrombophilias such as factor V Leiden. The most common venous thrombotic manifestation of APS is lower limb deep venous thrombosis, which occurs in up to 55% of patients, half of whom also have pulmonary embolism. Occasionally, venous thromboses occur in unusual sites (e.g. cerebral venous sinuses), and indeed any part of the venous system can be involved (superficial, mesenteric, portal, intracranial, retinal). Neurological features Ischaemic stroke is the most common neurological feature (>50% of CNS complications), and the commonest type of arterial thrombosis in APS. Recurrent stroke can lead to multi- infarct dementia. More subtle cognitive dysfunction has been reported to be associated with aPL and may represent micro- thrombotic change in the cerebral vasculature or a direct effect of aPL on neuronal tissue. Some studies suggest a high rate of recurrent stroke in aPL-positive patients and a younger age of onset of symptomatic events. aPL have also been linked with Sneddon s syndrome (recurrent stroke and livedo reticularis). Other reported, but unproven, neurological manifestations include migraines, seizures, chorea, transverse myelitis and a multiple sclerosis-like syndrome. Pregnancy morbidity Persistent aPL are seen in about 15% of women with recurrent miscarriage in the first or second trimester. In these women, the fetal loss rate can be up to 90% without pharmacological treatment. Late fetal death or stillbirth can also occur. Other possible obstetric complications of APS include placental insuf- ficiency, HELLP syndrome (haemolysis, elevated liver enzymes and low platelets) and severe early pre-eclampsia. Fetal and neonatal complications include intrauterine growth restriction, prematurity and, rarely, thrombosis. Catastrophic APS (CAPS) Although most patients with APS suffer thrombosis in one area at a time, they occasionally present with life-threatening acute multiple organ failure from extensive microvascular thrombosis ( thrombotic storm ). Laboratory evidence of disseminated intravascular coagulation can occur. Suggested precipitants of CAPS include infection, oral contraceptives, surgery and with- drawal of anticoagulation. The mortality rate is approximately 50%. Seronegative antiphospholipid syndrome A subset of patients has been identified who exhibit clinical manifestations of APS, without any recognized aPL on labo- ratory testing. These individuals are said to have SNAP (seronegative antiphospholipid) syndrome and anticoagulation may be warranted. Subsequent repeat aPL testing can be positive. Other clinical associations of aPL These are listed in Table 1. Diagnosis APS patients present to a wide range of specialties. Two ques- tions are of fundamental importance. OTHER AUTOIMMUNE DISORDERS Clinical associations of antiphospholipid antibodies Conditions associated with production of aPL C Systemic lupus erythematosus, rheumatoid arthritis, systemic sclerosis, Beh 'cet s disease, temporal arteritis, Sjo gren s syndrome C Infections e HIV, varicella, hepatitis C, syphilis, malaria, leprosy C Drugs e phenothiazines, procainamide, phenytoin, quinidine, hydralazine C Lymphoproliferative disease (lymphoma, paraproteinaemia) Clinical manifestations in patients with aPL C Cardiovascular e venous/arterial thromboembolic disease, valvular heart disease, sterile endocarditis with embolism C Obstetric e recurrent miscarriage, intrauterine foetal death (IUFD), stillbirth, early severe pre-eclampsia, HELLP, placental insufficiency, prematurity, intrauterine growth restriction (IUGR) C Neurological e cerebral ischaemic events, chorea, dementia, psychiatric disorders, transverse myelopathy, seizures, GuillaineBarre syndrome, Sneddon s syndrome C Haematological e autoimmune thrombocytopenia, autoimmune haemolytic anaemia C Dermatological e livedo reticularis aPL, antiphospholipid antibodies; HELLP, haemolysis, elevated liver enzymes and low platelets. MEDICINE 38:2 102 O 2009 Elsevier Ltd. All rights reserved. Table 1 What criteria are used to make the diagnosis? Table 2 illustrates the criteria for APS based on the 2006 Sydney update of the 1999 Sapporo classification. Although useful as a clinical tool, these criteria were devised primarily for research purposes, and are tailored to a high diagnostic specificity (in practice such criteria can exclude patients with features consis- tent with APS). It is essential that persistent aPL positivity is confirmed in two separate blood samples; the criteria recom- mend an interval of at least 12 weeks between sampling. Given that thrombotic disease, pregnancy loss and transient aPL posi- tivity are common, it is important to address all other causes of thrombosis and miscarriage during the initial evaluation of sus- pected APS. A careful drug history is also needed: compounds such as phenothiazines, phenytoin and hydralazine are associ- ated with aPL positivity. Who should be tested for APS? Table 3 lists the indications. Testing for aPL should be performed in a specialized haemostasis laboratory. Samples for LA should be taken with minimal venous stasis, rapid draw and immediate anticoagulation. Platelet-poor plasma should be prepared within 1 h of blood collection. Lack of adequate sampling and preparation may result in a false-negative result for LA. Anticoagulation may interfere with laboratory diagnosis of LA; however, in certain specialized laboratories aPL testing can be performed whilst the patient is taking anticoagulant treatment. Strategies include the use of the DRVVT where the international normalized ratio (INR) is only modestly elevated (<3.0) or the use of an alternative coagulation test employing a reagent that is less sensitive to the effects of warfarin (the Taipan snake venom time). Management The Haemostasis and Thrombosis Task Force of the British Society for Haematology (BCSH) has published guidelines on the management of APS and the American College of Chest Table 3 Physicians (ACCP) guidelines include advice on the management of patients with APS and a history of pregnancy morbidity. Use of specialist referral centres is important. Decisions about the commencement and duration of anticoagulant therapy rest on the accurate diagnosis of such events, so appropriate assessment and imaging are vital. The treatment of venous thromboembo- lism must never be delayed whilst awaiting aPL test results. An evidence-based approach to management is difficult because of the limited number of randomized controlled trials involving the APS population. Asymptomatic aPL Prophylactic intervention for the prevention of primary throm- bosis in asymptomatic individuals with persistent aPL is not supported by available evidence. However, it is prudent to observe general thrombotic risk reduction measures (e.g. avoidance of smoking, oestrogen-containing oral contraception and hormone replacement therapy) and give short-term heparin prophylaxis during high-risk periods (surgery, periods of immo- bilization and hospitalization). A recent prospective, randomized clinical trial has shown that prophylactic low-dose aspirin is ineffective for the prevention of thrombosis in individuals with asymptomatic aPL. Whether primary prophylaxis with aspirin is useful for some subsets of aPL patients at particularly high risk of thrombosis, such as those with SLE or with specific patterns of aPL positivity, remains to be established. Venous thromboembolism and stroke Initial management of venous thromboembolism involves stan- dard therapy with low-molecular-weight heparin (LMWH) fol- lowed by oral anticoagulation. The duration of therapy depends on any additional risk factors, the location, severity and conse- quences of previous thrombosis, patient age and the relative risk of further thrombosis versus haemorrhage due to anti- coagulation. In other words, therapeutic decisions must be individualized. Recurrent venous thromboembolic events are managed using long-term warfarin. The optimal intensity of long- term anticoagulation must balance the risk of thrombosis against that of bleeding (1% per year, 0.25% severe haemorrhage). OTHER AUTOIMMUNE DISORDERS Indications for antiphospholipid antibody testing Consider testing in all patients with C Apparently spontaneous venous thromboembolism, especially in the young or in those with thrombosis at unusual sites C Recurrent thrombosis C Stroke, myocardial infarction and peripheral arterial occlusive events presenting at a young age (<40 years), or in selected older patients with no obvious conventional cardiovascular risk factors C Systemic lupus erythematosus or those with autoimmune disease and thrombosis C Recurrent pregnancy loss or pregnancy complications with premature birth C Unexplained thrombocytopenia C Livedo reticularis Diagnostic criteria for antiphospholipid syndrome Clinical criteria C Vascular thrombosis e one or more episodes of arterial, venous or small vessel thrombosis in any tissue or organ (confirmed by imaging or histopathology) C Recurrent pregnancy loss (1 > 10 weeks gestation, or 3 < 10 weeks gestation) or one or more premature births due to preg- nancy complications Laboratory criteria C Lupus anticoagulant in plasma on two occasions at least 12 weeks apart C Anticardiolipin antibodies of IgG and/or IgM isotype on two occasions at least 12 weeks apart C Anti-b2-GPI antibody of IgG or IgM isotype on two occasions at least 12 weeks apart Antiphospholipid syndrome is considered to be definitely present when at least one clinical criterion and one laboratory criterion are met. Ig, immunoglobulin; GPI, glycoprotein-I. Table 2 MEDICINE 38:2 103 O 2009 Elsevier Ltd. All rights reserved. Recent prospective, randomized controlled trials suggest that high-intensity warfarin is not superior to moderate-intensity. However, patients with a high risk of recurrent thrombosis were not included in these trials, and it is likely that for some patients a target INR of 2.5 is insufficient. Although such a group is difficult to define, further events occurring despite therapeutic anticoagulation should alert one to the requirement for more intensive anticoagulation with a target INR of 3.5 (3.0e4.0). In patients with arteriothrombotic stroke associated with aPL, management should involve long-term warfarin. The optimal intensity remains disputed, as prospective studies have not shown a benefit of high-intensity compared with moderate- intensity anticoagulation (but the study population was not representative of an APS population). The addition of aspirin may also be of benefit, but with increased risk of bleeding. There is interest in the use of hydroxychloroquine and statins to reduce thrombotic risk in APS, but their use in this context remains empirical. In view of the association between APS and arterial thrombotic disease, every effort should be made to correct conventional arteriopathic risk factors in all patients with APS. Pregnancy morbidity All pregnant women with APS should be managed within a specialist obstetric unit with ready access to neonatal care. Given the teratogenicity of warfarin (6.4%), which is the highest in the first trimester, warfarinization is contraindicated in early pregnancy. In addition, foetal intracerebral haemorrhage can occur at any gestation. Women on long-term oral anticoagulation must be advised that warfarin should be stopped as early as possible and before 6 weeks gestation. The recommended treatment in women who experience recurrent miscarriage (and no history of thrombosis) associated with aPL is a combination of low-dose aspirin and heparin. However, the evidence for this is driven primarily by the results of a single randomized, controlled trial in which aspirin was used as the control. While some studies suggest aspirin alone or even supportive care only, in vitro data support an adjunctive role for heparin in this situation. At our institution we administer aspirin (75 mg daily) and low-dose LMWH therapy (dalteparin 2500 units daily subcutaneously (SC) to patients with APS associated with recurrent miscarriage, and, on a pragmatic basis, dalteparin 5000 units daily SC to patients with aPL and late obstetric complications). Aspirin should be started as soon as the urine pregnancy test becomes positive (preconceptual aspirin should also be considered), and LMWH when foetal heart activity is established. The ideal duration of such therapy has not been determined, but in women with miscarriage about 25% of treated pregnancies are associated with late obstetric complications. In these patients, a thrombotic basis may be contributory, and it is reasonable to continue therapy to 38 weeks gestation. Thrombotic risk is increased during pregnancy, and women with APS with thrombotic mani- festations require careful anticoagulation, which should be individualized. Catastrophic antiphospholipid syndrome Anecdotal reports suggest that the following agents may be beneficial: anticoagulation, intravenous immunoglobulin, plasma exchange and immunosuppressive therapy, including high-dose corticosteroids, cyclophosphamide and rituximab. Conclusions In recent years there has been significant progress in the under- standing of the aetiology, pathophysiology and diagnosis of APS. However, the precise mechanism of action of aPL remains elusive and more clarification is needed, particularly with regard to the neurological manifestations and late obstetric complications. Furthermore, we lack a solid evidence base to guide optimal management. Current therapy relies on a logical approach based on available evidence. Of paramount importance is recognition of clinical scenarios that warrant investigations for aPL, taking heed of the possibility of transient aPL, and application of an appro- priate protocol for laboratory testing. Interpretation of test results must occur in concert with the clinical history, paying particular attention to any additional thrombotic risk factors. Antith- rombotic agents e aspirin, LMWH and warfarin e are the main- stay of therapy of APS. Further large-scale randomized clinical trials using these agents are needed to increase the evidence base and optimize management of APS. Novel approaches to eliminate the autoantibody, which may offer the potential for cure in the future, are under investigation. Anti-PL-7 (Anti-Threonyl-tRNA Synthetase) Antisynthetase Syndrome Clinical Manifestations in a Series of Patients From a European Multicenter Study (EUMYONET) and Review of the Literature Ane Labirua-Iturburu, MD, Albert Selva-O‚ÄôCallaghan, MD, PhD, Melinda Vincze, MD, Katalin Danko¬¥, MD, PhD, Jiri Vencovsky, MD, PhD, Benjamin Fisher, MD, MRCP, Peter Charles, FRCPath, FIBMS, Maryam Dastmalchi, MD, PhD, and Ingrid E. Lundberg, MD, PhD Abstract: Autoantibodies against several aminoacyl-transfer-RNA synthetases have been described in patients with myositis; anti-threonyltRNA synthetase (anti-PL-7) is one of the rarest. We describe the clinical and laboratory characteristics of a cohort of European anti-PL-7 patients, and compare them with previously reported cases. This multicenter study of patients positive for anti-PL-7, identified between 1984 and 2011, derives from the EUMYONET cohort. Clinical and serologic data were obtained by retrospective laboratory and medical record review, and statistical analyses were performed with chi-squared and Fisher exact tests. Eighteen patients, 15 women, were anti-PL-7 antibody positive. Median follow-up was 5.25 years (interquartile range, 2.8Y10.7 yr), and 4 patients died. All patients had myositis (12 polymyositis, 5 dermatomyositis, and 1 amyopathic dermatomyositis), 10 (55.6%) had interstitial lung disease, and 9 (50%) had pericardial effusion. Occupational exposure to organic/inorganic particles was more frequent in patients with interstitial lung disease than in the remaining patients (5 of 10 vs. 1 of 7; p = 0.152), although the difference was not significant. Concurrent autoantibodies against Ro60 and Ro52 were seen in 8 of 14 (57%) patients studied. In the literature review the most common manifestations of anti-PL-7 antisynthetase syndrome were interstitial lung disease (77%), myositis (75%), and arthritis (56%). As in other subsets of the antisynthetase syndrome, myositis and interstitial lung disease are common features of the anti-PL-7 antisynthetase syndrome. In addition, we can add pericarditis as a possible manifestation related to anti-PL-7 antibodies. (Medicine 2012;91: 206Y211) From the Internal Medicine Department (AL-I, AS-O), Vall d‚ÄôHebro¬¥n General Hospital, Universitat Auto`noma de Barcelona and Vall d‚ÄôHebro¬¥n Research Institute, Barcelona, Spain; 3rd Department of Internal Medicine (MV, KD), Division of Immunology, Medical and Health Science Center, University of Debrecen, Debrecen, Hungary; Institute of Rheumatology (JV), Charles University, Prague, Czech Republic; Kennedy Institute of Rheumatology (BF, PC), Imperial College London, London, UK; Rheumatology Unit (MD, IEL), Department of Medicine, Karolinska University Hospital, Solna, Karolinska Institutet, Stockholm, Sweden. Funding and conflicts of interest: This study was supported in part by the European Union Sixth Framework Programme (project AutoCure; LSH-018661), and the regional agreement on medical training and clinical research (ALF) between Stockholm County Council and Karolinska Institutet. JV is supported by MSM 0021620812 from Ministry of Education, Youth and Sports in the Czech Republic. The authors have no conflicts of interest to disclose. Reprints: Ane Labirua-Iturburu, MD, A1/Urquijo 21, 4 izq, 48008, Bilbao, Spain (e-mail: ane.labirua)gmail.com). Copyright * 2012 by Lippincott Williams & Wilkins ISSN: 0025-7974 DOI: 10.1097/MD.0b013e318260977c 206 www.md-journal.com Abbreviations: ANA = antinuclear antibodies, DLCO = diffusing capacity of the lung for carbon monoxide, DM = dermatomyositis, FEV1 = forced expiratory volume in 1 second, FVC = forced vital capacity, ILD = interstitial lung disease, IQR = interquartile range, PM = polymyositis. INTRODUCTION I diopathic inflammatory myopathiesVpolymyositis (PM) and dermatomyositis (DM)Vare systemic autoimmune diseases characterized by skeletal muscle inflammation, but other organs are frequently involved such as skin in dermatomyositis and lungs and heart in both polymyositis and dermatomyositis. Up to 56% of patients with myositis are positive for various autoantibodies, which can be classified as associated (present in other rheumatic disorders) or specific (positive predominantly in myositis).4 Among the myositis-specific autoantibodies group, antisynthetase antibodies are the most commonly found, and are directed against aminoacyl-transfer-RNA synthetases, a group of cytoplasmic enzymes that catalyze binding of an amino acid to its cognate tRNA, a necessary step in the formation of polypeptides. To our knowledge, 8 autoantibodies against different synthetases have been described to date: anti-Jo-1, antiPL-7, anti-PL-12, anti-EJ, anti-OJ, anti-KS, anti-YRS, and antiZo.1,5,11,14,21,24,28 The clinical profiles associated with these various antisynthetase antibodies, in particular anti-Jo-1, have been investigated previously.7,22 The main associated clinical features are myositis, interstitial lung disease (ILD), arthritis, fever, Raynaud phenomenon, and mechanic‚Äôs hands, comprising the ‚Äò‚Äòantisynthetase syndrome.‚Äô‚Äô In addition, individual autoantibody specificities may be associated with distinctive clinical features. Non-Jo-1 antisynthetase antibodies seem to be markers of hypomyopathic forms with prominent lung involvement.10,13,15 Few studies have focused on the clinical manifestations of patients with anti-PL-7, an antibody directed against threonyl-tRNA synthetase.12,25,33 We conducted the current study to investigate the clinical and laboratory profiles of patients with anti-PL-7 antibody in a large cohort of European patients, and to compare them with previously reported cases. PATIENTS AND METHODS Patients Patients positive for anti-PL-7 recorded during the period of 1984 to 2011 were identified from the laboratory databases or rheumatology clinics of the following university hospitals: Vall Medicine & Volume 91, Number 4, July 2012 Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine & Volume 91, Number 4, July 2012 d‚ÄôHebro¬¥n General Hospital, Universitat Auto`noma de Barcelona, Barcelona, Spain; Medical and Health Science Center, University of Debrecen, Debrecen, Hungary; Karolinska University Hospital, Stockholm, Sweden; Institute of Rheumatology, Charles University, Prague, Czech Republic; and Imperial College Healthcare NHS Trust, London, United Kingdom. In total 964 myositis patients were identified. Clinical data were obtained by a retrospective review of medical records. Clinical findings included the presence or absence of inflammatory myopathy, ILD, arthritis, Raynaud phenomenon, mechanic‚Äôs hands, fever, skin rash, heart involvement, history of smoking and environmental exposures related to profession, and other relevant clinical features. The diagnosis of dermatomyositis and polymyositis was based on the criteria of Bohan and Peter,2,3 and only patients with definite or probable disease were included. The Sontheimer criteria27 were used to diagnose amyopathic dermatomyositis. The diagnosis of ILD was established when any of the following conditions were present: interstitial infiltrates on chest radiography or high-resolution computed tomography (ground-glass opacities, honeycombing, fibrosis, and/or interstitial thickening) and/or a restrictive pattern on pulmonary function testing (forced vital capacity EFVC^ G80%, forced expiratory volume in 1 second EFEV1^ G70%, diffusing capacity of the lung for carbon monoxide EDLCO^ G75%), and/or positive histology. Pericardial effusion was assessed by transthoracic echocardiography. Autoantibody Analyses Myositis-specific and -associated autoantibodies were identified by line immunoassay (Myositis Profile Euroline, Euroimmun, Lu¬®beck, Germany) or RNA and protein immunoprecipitation assay. Anti-PL-7 antibodies were confirmed by at least 2 of the following techniques: ELISA, line immunoassay (Myositis Profile Euroline), or RNA and protein immunoprecipitation assay. To exclude false-positive cases, we included only patients who repeatedly tested positive for anti-PL-7. Antinuclear antibodies (ANA) and extractable nuclear antigens were assessed by ELISA. Serum samples were obtained after patients provided oral informed consent. Human Leukocyte Antigen Typing Human leukocyte antigen (HLA) class II was detected with a sequence-specific primer and sequence-specific oligonucleotide polymerase chain reaction technique. Ethics The institutional review boards of all the participating centers approved the study. Literature Review We searched the English-language literature in the PubMed database (National Library of Medicine, Bethesda, MD) for related articles published up to March 2011, using the following key words: ‚Äò‚ÄòPL-7 antibody,‚Äô‚Äô ‚Äò‚Äòanti-PL-7,‚Äô‚Äô ‚Äò‚Äòanti-threonyl-tRNA synthetase,‚Äô‚Äô and ‚Äò‚Äòantisynthetase syndrome.‚Äô‚Äô Statistical Analyses Qualitative data are presented as numbers and percentage, and quantitative data as the median and interquartile range (IQR). Association between the presence of anti-PL-7 antibodies and qualitative variables was assessed using the chi-squared and Fisher exact tests. All statistical analyses were performed with SPSS 13.0 software (SPSS, Chicago, IL). Significance was set at a p value of less than 0.05. * 2012 Lippincott Williams & Wilkins Anti-PL-7 Antisynthetase Syndrome RESULTS Clinical Features Eighteen anti-PL-7-positive patients were identified out of 964 patients tested from the contributing centers (1.87%). The median age at diagnosis was 52.5 years (IQR, 40Y58.8 yr), and there was a predominance of women (15 patients, 83%). Eight (44.4%) patients were current or former smokers. PM was diagnosed in 12 (66.7%) patients, DM in 5 (27.8%), and amyopathic DM in 1 (5.6%) patient. Two patients had overlap syndromes: 1 patient with DM met the criteria for rheumatoid arthritis, with positive testing for rheumatoid factor and anticyclic citrullinated peptide, and the other had PM and Sjo¬®gren syndrome. Clinical features of the 18 patients are summarized in Table 1. Twelve (66.7%) patients had arthritis. Ten (55.6%) patients had ILD. Fever attributed to the disease was present in 10 (55.6%) patients, and Raynaud phenomenon in 11 (61.1%). Nine (50%) patients had pericardial effusion with no apparent underlying cause (for example, heart failure). Two of them developed massive effusion with pericardial tamponade: 1 resolved with medical treatment (glucocorticoids), and the other required pericardiocentesis. Mechanic‚Äôs hands were reported in 5 (27.8%) patients. Only 1 patient had a malignant disease, breast cancer, which was diagnosed 4 years after the PM diagnosis. The onset of ILD was before the start of myositis in 2 cases, at the same time as myositis in 7, and after myositis in 1. In all 10 patients with ILD, pulmonary involvement was confirmed by high-resolution computed tomography, which showed various radiologic patterns. Signs of fibrosis were documented in 9 of the 10 patients, interstitial thickening in 5, honeycombing in 4, and ground-glass opacities in 4. Pulmonary function tests were carried out in 5 cases and showed a restrictive pattern with a median FVC at diagnosis of 61% (IQR, 57%Y75.5%) and FEV1 of 69.5% (IQR, 51.4%Y74.3%). Transbronchial biopsy in 1 patient and autopsy findings in another disclosed alveolitis in both cases. One patient required orotracheal intubation due to respiratory failure caused by severe ILD, with later recovery. Six patients in the cohort had been exposed to various organic/inorganic particles at work, including 2 factory workers, a plumber, a farmer, a cleaning worker, and a miner (uranium mines). All but 1 (a factory worker) had underlying ILD, which appeared before or at the same time as myositis. Nonetheless, there was no significant association between working exposure and ILD (5 of 10 with ILD vs. 1 of the 7 remaining patients, p = 0.152). The median follow-up was 5.25 years (IQR, 2.8Y10.7 yr), and 4 patients died. The cause of death was attributed to progression of ILD in 2 patients (1 complicated with Pneumocystis jiroveci infection), acute myocardial infarction in 1 patient, and sepsis secondary to Salmonella species in 1 patient. The following treatment options were used either alone or in combination: glucocorticoids (100%), azathioprine (33%), cyclophosphamide (33%), methotrexate (22%), immunoglobulins (16%), and cyclosporine (5%). Four patients were treated with glucocorticoids alone, 5 patients with glucocorticoids plus 1 immunosuppressive drug, 7 patients with glucocorticoids and 2 immunosuppressive drugs, and 2 patients with 3 or more immunosuppressive drugs (see Table 1). Antibody Profile and HLA Seven patients were ANA-positive, with titers ranging from 1/320 to 1/2560. Anti-Ro/SSA antibody was found in 4 patients, and anti-Ro52 antibody in 5 patients. Anti-La/SSB was found in 2 patients, and anti-PM-Scl, anti-RNP, and anticardiolipin www.md-journal.com Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 207 Medicine Labirua-Iturburu et al & Volume 91, Number 4, July 2012 TABLE 1. Clinical and Serologic Features of 18 European Patients With Antisynthetase (PL-7) Syndrome Patient Age/Sex Smoking (yr) Status Job 1 65/F No Shop worker 2 3 4 5 52/F 56/F 37/F 45/F Yes No No No 6 7 8 9 10 11 12 13 40/F 55/F 32/F 40/F 38/F 55/F 53/F 62/M Yes No No No No No Yes Yes 14 61/M Yes 15 16 46/M 58/F 17 18 69/F 40/F Myositis AR Ray MH Fever ILD PD Ab DM Yes Yes Yes No Administration Housewife Cleaning worker Administration Administration Administration Waiter Toy factory Farmer Car factory Plumber DM ADM DM PM Yes No Yes No No No Yes Yes Yes Yes No No No No Yes Yes PM PM PM PM PM PM PM PM No Yes Yes Yes Yes No Yes No Yes Yes Yes No Yes Yes Yes Yes No No No No Yes No No Yes No No No Yes No No Yes Yes PM No No No Yes Yes Yes Miner (uranium) Song writer Administration No No PM-Scl No No No No RNP No No Ro, La No Yes Ro, La Yes No ANA, ACA Yes Yes Yes Yes ANA, Ro52, CL Yes No Ro52 PM DM No No Yes Yes No No Yes Yes Yes No No NA RF, CCP No Yes Administration Administration PM DM Yes Yes Yes No No No Yes Yes Therapy Yes Yes ANA, RF GC + CP+ MTX+AZA No Yes Ro, Ro52 GC+AZA+IG Yes Yes ANA, Ro52 GC+CP+AZA No Yes ANA, Ro GC+CYS+IG Yes Yes ANA GC+AZA Death Yes; sepsis Yes; MI GC GC GC GC+MTX GC+CP+AZA GC+CP GC+MTX GC+MTX GC GC+AZA+MF GC+AZA+ MTX Yes Yes GC+CP+MTX Yes No ANA, Ro52 GC+CP+AZA+ MTX+IG Yes; ILD Yes; ILD Abbreviations: Ab = antibodies, ADM = amyopathic dermatomyositis, AR = arthritis, AZA = azathioprine, CCP = cyclic citrullinated peptide, CL = cardiolipin, CP = cyclophosphamide, CYS = cyclosporin, GC = glucocorticoids, IG = intravenous immunoglobulin, MF = mycophenolate, MH = mechanic‚Äôs hands, MI = myocardial infarction, MTX = methotrexate, NA = not available, PD = pericardial disease, RF = rheumatoid factor, Ray = Raynaud phenomenon. antibodies were found in 1 patient each. Rheumatoid factor was positive in 2 cases, 1 of whom additionally had anticyclic citrullinated peptide. HLA typing was performed in 13 patients. TABLE 2. Immunogenetic Profile (HLA Class II Alleles) of 18 Patients With Antisynthetase (PL-7) Syndrome Patient HLA-DRB1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 208 HLA-DRB1*03,04 HLA-DRB1*04,15 HLA-DRB1*12 HLA-DRB1*03 HLA-DRB1*07,*07 HLA-DRB1*04,*11 HLA-DRB1*03;*04 HLA-DRB1*11 HLA-DRB1*04,*07 HLA-DRB1*03,*11 HLA-DRB1*01,*11 HLA-DRB1*03,*04 HLA-DRB1*08,*15 www.md-journal.com DRB haplotype is shown in Table 2. The most common haplotypes were DRB1*04 in 6 patients and DRB1*03 in 5 patients. Only 2 of the patients with ILD showed the DRB1*03 allele. Literature Review We identified 54 additional cases of anti-PL-7 antisynthetase syndrome in the literature. The main clinical manifestations in the overall group of 72 patients (present cohort and reported cases) were ILD in 56 (77.8%) patients and myositis in 54 (75%). The prevalence of other signs and symptoms of anti-PL-7 antisynthetase syndrome is summarized in Table 3. DISCUSSION Antisynthetase syndrome, to our knowledge first described in 1990, is recognized by its characteristic clinical manifestations associated with the presence of an anti-tRNA synthetase antibody.20 Anti-Jo-1 is the most frequent, followed by anti-PL-12 and anti-PL-7. It has been suggested that in addition to a common core syndrome, different clinical features may be associated with different antisynthetase antibodies, but the scarcity of published studies, most of which are case reports, precludes any definitive conclusions in this regard.16 In the present study, we analyzed the clinical and laboratory features of a series of anti-PL-7-positive patients from 5 European countries. In common with other reported antisynthetase syndromes, myositis and ILD were the most frequent manifestations of anti-PL-7-associated syndrome, and these common features were confirmed in a review of all cases published * 2012 Lippincott Williams & Wilkins Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine & Volume 91, Number 4, July 2012 Anti-PL-7 Antisynthetase Syndrome TABLE 3. Main Characteristics of Patients With Antisynthetase (PL-7) Syndrome, Previous and Present Reports Anti-PL-7 Report (First Author, Ref.) 12 Hervier Sato25 Fischer9 Yamasaki33 Mathews21 Dugar7 Marguerie20 Targoff 29 Lega17 Troyanov31 Labirua-Iturburu (PR) Total, no. (%) No. of Patients Myositis (No.) ILD (No.) Arth (No.) Ray (No.) MH (No.) Fever (No.) PE (No.) 12 7 7 6 5 4 4 4 3 2 18 72 6 6 3 4 4 3 3 4 1 2 18 54 (75) 12 7 7 6 1 2 3 3 3 2 10 56 (77.8) 4 6 3 3 1 4 4 2 1 12 40 (55.6) 1 4 2 2 2 4 1 1 1 11 29 (40.3) 4 2 1 1 5 13 (18) 8 4 1 10 22 (30.5) 4 1 1 1 9 16 (22.2) Abbreviations: Arth = arthritis/arthralgia, MH = mechanic‚Äôs hands, PE = pericardial effusion, PR = present report, Ray = Raynaud phenomenon. to date in the literature. An unexpected finding in the current cohort was pericardial effusion in half of our cases. Due to the low patient numbers, few studies have focused on the clinical and laboratory characteristics of patients with antiPL-7 antibodies. The most complete are 2 studies performed in Japan, comprising 7 and 6 patients, respectively,25,33 and a 2011 report by Hervier et al12 in France with 12 patients. In these studies, ILD was present in 100% of cases, while myositis appeared in 50%Y86%. In the present series, 6 of 18 patients did not have diagnosed ILD, and there was a much higher prevalence of myositis (all patients had idiopathic inflammatory myopathies except 1, who had amyopathic dermatomyositis). These results differ to some degree from those of previous studies, in which the prevalence of myositis in patients with non-Jo-1 antisynthetase autoantibodies was lower than in patients with anti-Jo1.10,15 However, a pooled analysis of our findings together with those from all published case reports and series in the literature7,9,12,17,20,21,25,29,31,33 (72 anti-PL-7-positive patients) showed that the frequency of ILD and myositis (77.8% and 75%, respectively; see Table 3) does not differ greatly from the clinical findings in patients with anti-Jo-1. One explanation for the discrepancies between individual studies could be the potential selection bias involved in the study of all rare diseases. In small series, some clinical manifestations might be overrepresented; hence to try to overcome these problems, we included a review of all cases published in the literature, in addition to our own cohort, and analyzed the pooled data set. Other possible reasons for the dissimilar results might be the racial/ethnic or genetic background of the patients, a different referral pattern (pulmonary, rheumatology, or internal medicine specialists), and the fact that in many cases, only patients with clinically evident myositis were tested for antisynthetase antibodies. Little is known about the etiopathogenesis of the antisynthetase syndrome, or for that matter, of myopathies in general. According to some authors, an unknown trigger (for example, viral infection) might enter the respiratory tract and lead to a conformational modification of aminoacyl-tRNA synthetase in the pulmonary alveoli. This could result in production of autoantibodies against the synthetase enzyme; this immune response (antibodies and specific T cells) would spread to other internal organs such as muscle and joints.18,19 An interesting factor that should be investigated in well-designed studies is the contribution of organic/inorganic dust inhalation at work, which could be * 2012 Lippincott Williams & Wilkins another etiologic trigger of autoimmunity. There has been some evidence supporting the idea of environmental exposure to antigens as an initial step in the pathogenesis of antisynthetase syndrome.30 In the current series, 6 patients had been occupationally exposed to different organic/inorganic particles and all but 1 had ILD. Nonetheless, statistical significance was not achieved for this factor, probably because of the small number of patients studied. It is noteworthy that 9 of 18 patients in the current series had pericardial effusion. In the study by Hervier et al,12 4 of 12 (33%) patients with anti-PL-7 had pericardial effusion, and 3 other single cases have been reported in different studies.17,21,33 Pericarditis is, compared to systemic lupus erythematosus, infrequent in patients with myositis. The reason patients with anti-PL-7 develop pericardial disease is unknown. It can be speculated that anti-PL-7 antibodies act against certain antigens in pericardial tissue, and thereby lead to pericardial inflammation. Nevertheless, some cases of pericarditis have also been reported in patients with anti-Jo-1.6,23 In a study and a review of literature by Schmidt et al,26 they found a prevalence of 18% of pericarditis in a total of 231 anti-Jo-1 patients, and we cannot exclude that pericarditis is underestimated in myositis patients because of low physical function due to muscle and lung involvement. Other relevant observations can be drawn from this study. First, it seems that clinical heterogeneity is a feature of the antisynthetase syndrome, irrespective of the autoantibody implicated. The results from the pooled analysis indicate that although antisynthetase syndrome can present with a single clinical manifestation (for example, myositis, ILD, Raynaud phenomenon, arthritis, mechanic‚Äôs hands), the most common features are myositis, ILD, and arthritis. The frequency of these symptoms is similar to that seen in patients with anti-Jo-1 antibodies. As for the prognosis of antisynthetase syndrome, which is currently poorly defined, ILD seems to be the major cause of death. It has been suggested that ILD associated with antisynthetase syndrome may have a better prognosis than antibody-negative ILD,8 but this has yet to be confirmed. As was seen in the current study and 1 previously reported study,12 ILD in patients with antisynthetase syndrome can be aggressive and refractory to treatment. It has been reported that coexisting Jo-Ro antibodies may predict more severe ILD,32 but only 1 of our patients who died of ILD was antiRo52-positive. Further study of lung disease in larger antisynthetase syndrome cohorts is needed to elucidate these findings. www.md-journal.com Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. 209 Medicine Labirua-Iturburu et al One limitation of the present study is its retrospective nature and the fact that complete data sets were not available for all patients. Not all patients underwent pulmonary function tests or high-resolution computed tomography; hence, asymptomatic ILD could be more prevalent in our cohort than our results suggest. In addition, a heterogeneous referral pattern could have led to both over- and underdiagnosis of some clinical features. Lastly, the small sample size due to the relative rarity of the syndrome probably contributed to a lack of statistical power in some of the comparisons. In conclusion, our study of a European cohort of anti-PL-7 patients identified myositis as the most frequent clinical manifestation, with ILD being less prevalent than has been previously reported. Nevertheless, when these manifestations were evaluated in all the cases reported to date, including those from the present study, their prevalence did not vary greatly from the rates reported for other antisynthetase antibody syndromes. Occupational exposure should be taken into account in these patients, because environmental antigens could contribute to triggering the immune response and the development of the disease. The unexpected high percentage of patients with pericardial effusion in our cohort is a novel finding and needs to be confirmed in future series. It may be a characteristic of anti-PL-7 antisynthetase syndrome, which could warrant a recommendation of routine transthoracic echocardiography in patients presenting with this serotype. REFERENCES 1. Betteridge Z, Gunawardena H, North J, Slinn J, McHugh N. Anti-synthetase syndrome: a new autoantibody to phenylalanyl transfer RNA synthetase (anti-Zo) associated with polymyositis and interstitial pneumonia. Rheumatology (Oxford). 2007;46:1005Y1008. 2. Bohan A, Peter JB. Polymyositis and dermatomyositis (First of two parts). N Engl J Med. 1975;292:344Y347. 3. Bohan A, Peter JB. Polymyositis and dermatomyositis (Second of two parts). N Engl J Med. 1975;292:403Y407. 4. Brouwer R, Hengstman GJD, Vree Egberts W, Ehrfeld H, Bozic B, Ghirardello A, Gr√∏ndal G, Hietarinta M, Isenberg D, Kalden JR, Lundberg I, Moutsopoulos H, Roux-Lombard P, Vencovsky J, Wikman A, Seelig HP, van Engelen BG, van Venrooij WJ. Autoantibody profiles in the sera of European patients with myositis. Ann Rheum Dis. 2001;60:116Y123. & Volume 91, Number 4, July 2012 11. Hashish L, Trieu EP, Sadanandan P. Identification of autoantibodies to tyrosyl-tRNA synthetase in dermatomyositis with features consistent with antisynthetase syndrome Eabstract^. Arthritis Rheum. 2005;52:S312. 12. Hervier B, Uzunhan Y, Hachulla E, Benveniste O, Nunes H, Delaval P, Musset L, Dubucquoi S, Wallaert B, Hamidou M. Antisynthetase syndrome positive for anti-threonyl-tRNA synthetase (anti-PL7) antibodies. Eur Respir J. 2011;37:714Y717. 13. Hervier B, Wallaert B, Hachulla E, Adoue D, Lauque D, Audrain M, Camara B, Fournie B, Couret B, Hatron PY, Dubucquoi S, Hamidou M. Clinical manifestations of antisynthetase syndrome positive for anti-alanyl-tRNA synthetase (anti-PL12) antibodies: a retrospective study of 17 cases. Rheumatology (Oxford). 2010;49:972Y976. 14. Hirakata M, Suwa A, Nagai S, Kron MA, Tireu EP, Mimori T, Akizuki M, Targoff IN. Anti-KS: identification of autoantibodies to asparaginyl-transfer RNA synthetase associated with interstitial lung disease. J Immunol. 1999;162;2315Y2320. 15. Kalluri M, Sahn SA, Oddis CV, Gharib SL, Christopher-Stine L, Danoff SK, Casciola-Rosen L, Hong G, Dellaripa PF, Highland KB. Clinical profile of anti-PL12 autoantibody: cohort study and review of the literature. Chest. 2009;135:1550Y1556. 16. Labirua A, Lundberg IE. Interstitial lung disease and idiopathic inflammatory myopathies: progress and pitfalls. Curr Opin Rheumatol. 2010;22:633Y638. 17. Lega JC, Cotin V, Fabien N, Thivolet-Bejui F, Cordier JF. Interstitial lung disease associated with anti-PM/Scl or anti-aminoacyl-tRNA synthetase autoantibodies: a similar condition? J Rheumatol. 2010;37:1000Y1009. 18. Levine SM, Raben N, Xie D, Askin FB, Tuder R, Mullins M, Rosen A, Casciola-Rosen LA. A novel conformation of histidyl-transfer RNA synthetase in the lung: the target tissue in Jo-1 autoantibody-associated myositis. Arthritis Rheum. 2007;56:2729Y2739. 19. Lundberg IE, Grundtman C. Developments in the scientific and clinical understanding of inflammatory myopathies. Arthritis Res Ther. 2008;10:220. 20. Marguerie C, Bunn CC, Beynon HL, Bernstein RM, Hughes JM, So AK, Walport MJ. Polymyositis, pulmonary fibrosis and autoantibodies to aminoacyl-tRNA synthetase enzymes. Q J Med. 1990;77:1019Y1038. 21. Mathews MB, Reichlin M, Hughes GRV, Bernstein R. Anti-threonyl-tRNA synthetase, a second myositis-related autoantibody. J Exp Med. 1984;160:420Y434. 5. Bunn CC, Bernstein RM, Mathews AM. Autoantibodies against alanyl-tRNA synthetase and tRNA-ala coexist and are associated with myositis. J Exp Med. 1986;163:1281Y1291. 22. Mileti LM, Strek ME, Niewold TB, Curran JJ, Sweiss NJ. Clinical characteristics of patients with anti-Jo-1 antibodies. J Clin Rheumatol. 2009;15:254Y255. 6. Derk CT, Sandorfi N, Curtis MT. A case of anti-Jo1 myositis with pleural effusions and pericardial tamponade developing after exposure to a fermented Kombucha beverage. Clin Rheumatol. 2004;23:355Y357. 23. Mogulkoc N, Kabasakal Y, Ekren PK, Bishop PW. An unusual presentation of anti-Jo-1 syndrome, mimicking lung metastases, with massive pleural and pericardial effusions. J Clin Rheumatol. 2006;12:90Y92. 7. Dugar M, Cox S, Limaye V, Blumbergs P, Roberts-Thomson PJ. Clinical heterogeneity and prognostic features of South Australian patients with anti-synthetase autoantibodies. Intern Med J. 2011;41:674Y679. Epub 2010 Jan 4. 24. Nishikai M, Reichlin M. Heterogeneity of precipitating antibodies in polymyositis and dermatomyositis. Characterization of the Jo-1 antibody system. Arthritis Rheum. 1980;23:881Y888. 8. Fathi M, Vikgren J, Boijsen M, Tylen U, Jorfeldt L, Tornling G, Lundberg IE. Interstitial lung disease in polymyositis and dermatomyositis: longitudinal evaluation by pulmonary function and radiology. Arthritis Rheum. 2008;59:677Y685. 25. Sato S, Hirakata M, Kuwana M, Nakamura K, Suwa A, Inada S, Mimori T, Ikeda Y. Clinical characteristics of Japanese patients with anti-PL-7 (antithreonyl-tRNA synthetase) autoantibodies. Clin Exp Rheumatol. 2005;23:609Y615. 9. Fischer A, Swigris JJ, du Bois R, Lynch DA, Downey GP, Cosgrove GP, Frankel SK, Fernandez-Perez ER, Gillis JZ, Brown KK. Anti-synthetase syndrome in ANA and anti-Jo-1 negative patients presenting with idiopathic interstitial pneumonia. Respir Med. 2009;103:1719Y1724. 26. Schmidt WA, Wetzel W, Friedlander R, Lange R, Sorensen HF, Lichey HJ, Genth E, Mierau R, Gromnica-Ihle E. Clinical and serological aspects of patients with anti-Jo-1 antibodiesVan evolving spectrum of disease manifestations. Clin Rheumatol. 2000;19: 371Y377. 10. Friedman AW, Targoff IN, Arnett FC. Interstitial lung disease with autoantibodies against aminoacyl-tRNA synthetases in the absence of clinically apparent myositis. Semin Arthritis Rheum. 1996;26:459Y467. 210 www.md-journal.com 27. Sontheimer RD. Cutaneous features of classic dermatomyositis and amyopathic dermatomyositis. Curr Opin Rheumatol. 1999;11:475Y482. * 2012 Lippincott Williams & Wilkins Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited. Medicine & Volume 91, Number 4, July 2012 28. Targoff IN. Autoantibodies to aminoacyl-transfer RNA synthetases for isoleucine and glycine. Two additional synthetases are antigenic in myositis. J Immunol. 1990;144:1737Y1743. 29. Targoff IN, Arnett FC, Reichlin M. Antibody to threonyl-transfer RNA synthetase in myositis sera. Arthritis Rheum. 1988;31:515Y524. 30. Tillie-Leblond I, Wislez M, Valeyre D, Crestani B, Rabbat A, Israel-Biet D, Humbert M, Couderc LJ, Wallaert B, Cadranel J. Interstitial lung disease and anti-Jo-1 antibodies: differences between acute and gradual onset. Thorax. 2008;63:53Y59. 31. Troyanov V, Targoff IN, Tremblay JL, Goulet JR, Raymond Y, Senecal JL. Novel classification of idiopathic inflammatory * 2012 Lippincott Williams & Wilkins Anti-PL-7 Antisynthetase Syndrome myopathies based on overlap syndrome features and autoantibodies: analysis of 100 French Canadian patients. Medicine (Baltimore). 2005;84:231Y249. 32. Vancsa A, Csipo I, Nemeth J, Devenyi K, Gergely L, Danko K. Characteristics of interstitial lung disease in SS-A positive/Jo-1 positive inflammatory myopathy patients. Rheumatol Int. 2009;29:989Y994. 33. Yamasaki Y, Yamada H, Nozaki T, Akaogi J, Nichols C, Lyons R, Loy AC, Chan EK, Reeves WH, Satoh M. Unusually high frequency of autoantibodies to PL-7 associated with milder muscle disease in Japanese patients with polymyositis/dermatomyositis. Arthritis Rheum. 2006;54:2004Y2009. www.md-journal.com Copyright ¬© 2012 Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.