Methods

The Taiwan Society of Cardiology-Heart Failure with Reduced Ejection Fraction (TSOC-HFrEF) registry was a multicenter study aimed to prospectively investigate the prognosis of patients with systolic HF in Taiwan. The registry was conducted in 22 medical centers and enrolled patients hospitalized for acute worsening HF, whose echocardiographic left ventricular ejection fraction were <45%. BMI was defined as body weight (kg)/square of height (m ² ). Hypertension was defined as systolic blood pressure ≥140 mm Hg, or diastolic blood pressure ≥90 mm Hg, or the use of antihypertensive medications. Diabetes mellitus was defined as fasting glucose ≥126 mg/dl or the use of hypoglycemic medications. Dyslipidemia was defined as fasting low-density lipoprotein cholesterol ≥130 mg/dl, or high-density lipoprotein cholesterol <40 mg/dl in men or <50 mg/dl in women, or the use of lipid-lowering medications. The baseline data were obtained on admission and/or at discharge including demographic parameters, cause of HF, co-morbidities, clinical status, electrocardiographic and echocardiographic findings, laboratory data, interventional, and medical treatments. The participants were divided to 4 groups by BMI categories at discharge: <20.5 kg/m ² (underweight), 20.5 to 24.0 kg/m ² (normal weight), 24.1 to 27.4 kg/m ² (overweight), and ≥27.5 kg/m ² (obesity), according to the guidelines of the Department of Health in Taiwan and the results from the Eastern Taiwan integrated health care delivery system of Coronary Heart Disease (ET-CHD) registry.

Outcomes of interest were (1) all-cause mortality; (2) cardiovascular mortality, defined as death of sudden cardiac death, ischemic heart disease, stroke, refractory HF, and lethal arrhythmia; and (3) noncardiovascular mortality. Patients were followed up at outpatient department for evaluating their clinical condition, laboratory, and imaging examinations every 6 months. The origin and the issue date of mortality were verified according to the medical record. The study complies with the Declaration of Helsinki that the jointed ethics committee has approved the protocol and that informed consent has been obtained from the subjects.

Characteristics of patients in each BMI group were compared using either analysis of variance or chi-square analysis and reported as mean ± SD or percent for continuous and categorical variables, respectively. The analysis used the time for follow-up at the patients’ first enrollment (October 2013 to October 2014) with censoring at the occurrence of mortality or end of follow-up (February 1, 2015). Kaplan–Meier analysis was used to assess the association of BMI categories with mortality. Curves were compared using the log-rank test. Cox proportional hazard regression analyses were used to assess the multivariable association of BMI (as a continuous or categorical variable, respectively) with mortality, adjusting for potential confounders. In model 1, age, gender, and cause of HF were adjusted, and in model 2, demographic variables, co-morbidities, severity of HF, and medications at discharge were additionally adjusted. Stratified analyses exploring the association between BMI and all-cause mortality within age <65 and ≥65 years were performed. p Value <0.05 was used for as the criterion for variables to remain in the model 2. All statistical analyses were performed with SAS, version 9.2 (SAS Institute, Cary, North Carolina).

Subsequently, we sought previous studies evaluating the effect of BMI on the all-cause mortality among East Asian patients with HF from MEDLINE. The East Asian populations included those living in the territory of China, Japan, Mongolia, North Korea, South Korea, and Taiwan. We used keywords of body mass index, BMI, mortality, or the term of obesity paradox (or reverse epidemiology), and heart failure to search for the eligible studies written in English from 2000 to 2015. Two reviewers (GML and YHL) independently reviewed these papers and performed data extraction. Disagreements were resolved through consensus. Studies were required to report the association of BMI, treated as either a continuous variable or categorical variable, with mortality in patients with HF. A priori, we knew that not all studies, especially conducted in East Asian countries, would use the World Health Organization BMI classification system of <18.5, 18.5 to 24.9, 25.0 to 29.9, and ≥30.0 kg/m ² , respectively for underweight, normal, overweight, and obesity. To avoid eliminating studies with important data, we considered BMI levels within 2.5 kg/m ² of standard levels to be acceptable. Studies comparing obese and nonobese were excluded unless outcomes in the normal BMI population alone were available. The primary outcome of interest was all-cause mortality.

HRs and 95% CIs were risk factors adjusted and retrieved from the studies for meta-analysis. If BMI was treated as a continuous variable for all-cause mortality, all the retrieved HRs and their corresponding CIs were converted to HRs associated with 5-kg/m ² increase in BMI. When BMI levels were categorical, all the retrieved HRs and their CIs were converted into HRs using BMI: 18.5 to 23.9 kg/m ² as reference group. RevMan (version 5.2) (Cochrane Author Support Tool; website: http://tech.cochrane.org ) was used to perform random-effect meta-analysis of the converted HRs and their corresponding CIs by the inverse variance method for 5-kg/m ² increase in BMI and BMI category, respectively. Heterogeneity was examined using the Higgins I ² test. Roughly, Higgins I ² values of 25, 50, and 75% were interpreted as indicating low, moderate, and high heterogeneity.