International Journal of Obesity (2014) 38, 1110–1114 & 2014 Macmillan Publishers Limited All rights reserved 0307-0565/14 www.nature.com/ijo ORIGINAL ARTICLE The obesity paradox in heart failure patients with preserved versus reduced ejection fraction: a meta-analysis of individual patient data R Padwal1, FA McAlister1, JJV McMurray2, MR Cowie3, M Rich4, S Pocock5, K Swedberg6, A Maggioni7, G Gamble8, C Ariti5, N Earle8, G Whalley9, KK Poppe8, RN Doughty8 and A Bayes-Genis10 for the Meta-analysis Global Group in Chronic Heart Failure (MAGGIC) Investigators11 BACKGROUND: In heart failure (HF), obesity, defined as body mass index (BMI) X30 kg m À 2, is paradoxically associated with higher survival rates compared with normal-weight patients (the ‘obesity paradox’). We sought to determine if the obesity paradox differed by HF subtype (reduced ejection fraction (HF-REF) versus preserved ejection fraction (HF-PEF)). PATIENTS AND METHODS: A sub-analysis of the MAGGIC meta-analysis of patient-level data from 14 HF studies was performed. Subjects were divided into five BMI groups: o22.5, 22.5–24.9 (referent), 25–29.9, 30–34.9 and X35 kg m À 2. Cox proportional hazards models adjusted for age, sex, aetiology (ischaemic or non-ischaemic), hypertension, diabetes and baseline blood pressure, stratified by study, were used to examine the independent association between BMI and 3-year total mortality. Analyses were conducted for the overall group and within HF-REF and HF-PEF groups. RESULTS: BMI data were available for 23 967 subjects (mean age, 66.8 years; 32% women; 46% NYHA Class II; 50% Class III) and 5609 (23%) died by 3 years. Obese patients were younger, more likely to receive cardiovascular (CV) drug treatment, and had higher comorbidity burdens. Compared with BMI levels between 22.5 and 24.9 kg m À 2, the adjusted relative hazards for 3-year mortality in subjects with HF-REF were: hazard ratios (HR) ¼ 1.31 (95% confidence interval ¼ 1.15–1.50) for BMI o22.5, 0.85 (0.76–0.96) for BMI 25.0–29.9, 0.64 (0.55–0.74) for BMI 30.0–34.9 and 0.95 (0.78–1.15) for BMI X35. Corresponding adjusted HRs for those with HF-PEF were: 1.12 (95% confidence interval ¼ 0.80–1.57) for BMI o22.5, 0.74 (0.56–0.97) for BMI 25.0–29.9, 0.64 (0.46–0.88) for BMI 30.0–34.9 and 0.71 (0.49–1.05) for BMI X35. CONCLUSIONS: In patients with chronic HF, the obesity paradox was present in both those with reduced and preserved ventricular systolic function. Mortality in both HF subtypes was U-shaped, with a nadir at 30.0–34.9 kg m À 2. International Journal of Obesity (2014) 38, 1110–1114; doi:10.1038/ijo.2013.203 Keywords: body mass index; heart failure; prognosis; obesity paradox; ejection fraction INTRODUCTION were largely performed in patients with HF-REF.2 In addition, Obesity, defined as a body mass index (BMI) X30 kg m À 2,isan discrepant findings have been reported. For example, independent risk factor for CV morbidity and mortality and an observational study of 4700 subjects hospitalized with HF doubles the risk of heart failure (HF).1 In the presence of reported that obese subjects with HF-PEF had lower mortality established HF, obesity has been paradoxically associated with a compared with normal weight subjects (hazard ratio (HR) ¼ 0.77 reduced risk of mortality, a finding that has been termed ‘the (0.70–0.86)), whereas obese subjects with HF-REF exhibited obesity paradox’.2,3 Several explanations for the obesity paradox increased mortality risk than leaner patients with HF-REF have been proposed and it is currently unclear if it is a true (HR ¼ 1.21 (1.01–1.45)).9 Thus, there is a need to clarify whether phenomenon or a consequence of methodological bias.3–7 the paradox is specific to one HF subtype or is found to an Two major subtypes of HF exist—HF with reduced ejection equivalent degree in both HF-PEF and HF-REF. fraction (HF-REF) and HF with preserved ejection fraction (HF-PEF). We used the Meta-analysis Global Group in Chronic Heart Although patients with HF-REF exhibit higher mortality rates, both Failure (MAGGIC) individual patient data meta-analysis to explore subtypes confer a high absolute mortality risk (B120–140 deaths the relationship between BMI and survival and whether the per 1000 patient years).8 Studies describing the obesity paradox association was similar in those with HF-REF and HF-PEF. 1Department of Medicine, University of Alberta, Edmonton, Canada; 2BHF Cardiovascular Research Centre, University of Glasgow, Glasgow, UK; 3Faculty of Medicine, National Heart and Lung Institute, Imperial College London (Royal Brompton Hospital), London, UK; 4Washington University School of Medicine, St Louis, MO, USA; 5Department of Medical Statistics, London School of Hygiene and Tropical Medicine, London, UK; 6Department of Emergency and Cardiovascular Medicine, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden; 7ANMCO Research Center, Florence, Italy; 8Department of Medicine and National Institute for Health Innovation, University of Auckland, Auckland, New Zealand; 9Faculty of Social and Health Sciences, UNITEC Institute of Technology, Auckland, New Zealand and 10Department of Medicine, Hospital Universitari Germans Trias i Pujol UAB, Barcelona, Spain. Correspondence: Dr R Padwal, General Internal Medicine and Clinical Pharmacology, 2F1.26 WMC, University of Alberta Hospital, 8440 112 Street, Edmonton, Alberta T6G 2B7, Canada. E-mail: [email protected] 11See appendix. Received 27 August 2013; revised 3 October 2013; accepted 15 October 2013; accepted article preview online 31 October 2013; advance online publication, 26 November 2013 Obesity paradox in heart failure R Padwal et al 1111 PATIENTS AND METHODS using the Cochran-Armitage trend test for ordered categorical data. Detailed methods, including details about study selection criteria, included Confidence intervals around person time rates were calculated using studies, and main results of the MAGGIC meta-analysis have been Miettinen’s exact test (http://www.openepi.com, accessed on 11 June described elsewhere.8 In this analysis, we pooled the data from the 14 2013). The Cox proportional hazard of time to all cause death within 3 studies that measured BMI at baseline and included at least 3 years of years from baseline study visit was used to model the hazard of BMI strata, follow-up. Left ventricular ejection fraction (LVEF) was not an entry adjusted for age, sex, LVEF group, ischaemic aetiology, hypertension, criterion for any of the studies. The meta-analysis protocol was approved diabetes, systolic blood pressure and atrial fibrillation. Models were by the University of Auckland Human Subjects Ethics Committee. Data constructed from those individuals with complete data for each model. (including demographics, comorbidities, therapy, symptom status, clinical Imputation of missing data was not performed. Unless otherwise stated, variables, laboratory variables, and outcomes) from the individual studies SAS v9.2 (SAS Institute Inc., Cary, NC, USA) were used for all analyses. were re-coded into a uniform format at the Central Coordinating Centre at All tests were two-tailed and Po0.05 was considered significant. the University of Auckland and incorporated into one database. For this analysis, HF-PEF was defined as a baseline LVEF X50%. Covariates were also defined at baseline. BMI was calculated by taking the weight in kilograms and dividing by the square of the height (in meters). RESULTS Subjects were stratified according to BMI levels (o22.5, 22.5–24.9 BMI data were available for 23 976 patients from 14 studies (mean 2 (referent), 25.0–29.9, 30.0–34.9 and X35.0 kg m À ). Sensitivity analyses age, 66.8 years; 32% women). Most patients had NYHA class II were performed by repeating the analyses within each HF subtype; in the (46%) or III (50%) symptoms. A total of 5609 (23%) patients died World Health Organization BMI strata (o18.5, 18.5–24.9, 25–29.9 over the 914 (interquartile range ¼ 316–1096) day follow-up andX30 kg m À 2); in patients with and without oedema (as a proxy measure for fluid overload); and in subjects in whom estimated glomerular period. filtration rate data were available. The baseline characteristics of patients by BMI group and Data are presented as mean (standard deviation) unless otherwise stratified by HF subtype are shown in Table 1. In both the HF-PEF stated. To compare baseline characteristics, tests for linear trend were and HF-REF subgroups, low-BMI subjects were older and more constructed from orthogonal contrasts for continuous variables and by frequently women. In contrast, hypertension and diabetes were Table 1. Baseline characteristics by BMI strata and HF subtype BMI HF-REF HF-PEF o22.5 22.5–24.9 25–29.9 30–34.9 X35 P-valuea o22.5 22.5–24.9 25–29.9 30–34.9 X35 P-valuea N (14 studies) 3073 3747 7144 2751 987 950 1140 2326 1269 589 Age, years (s.d.) 69 (13) 67 (12) 66 (11) 63 (11) 60 (11) o0.0001 74 (12) 72 (12) 70 (11) 68 (10) 64 (11) o0.0001 Women, % 40 25 22 25 35 o0.0001 63 44 41 49 55 0.0009 Medical history Hypertension 25 31 37 49 59 o0.0001 32 41 48 61 68 o0.0001 Myocardial infarction 44 50 50 48 40 0.964 21 27 29 28 22 0.096 Atrial fibrillation 22 23 22 25 26 0.015 33 28 27 27 26 0.0009 Diabetes 13 17 22 31 39 o0.0001 11 15 19 28 36 o0.0001 Ischaemic aetiology 52 57 59 56 47 0.475 40 45 48 44 35 0.493 Medication ACEi or ARB 68 72 70 68 65 0.207 32 35 36 42 36 0.0006 Beta-blocker 30 38 42 48 49 o0.0001 25 34 40 43 42 o0.0001 Diuretic 86 83 83 84 89 0.654 76 76 74 81 86 o0.0001 Spironolactone 28 25 23 25 25 0.0093 18 17 16 15 20 0.817 Digoxin 59 53