Study population and design
The participants in the present study were community-dwelling older adults aged 60 years and over from Shandong, China (Retrospectively Registered: ChiCTR-EOC-17013598) [36–38]. Baseline data were collected between April 2007 and November 2011. A total of 13,251 older individuals were eligible and enrolled based on the following exclusion criteria: congestive heart failure, myocardial infarction, stroke, thyroid and parathyroid disease, Cushing’s syndrome, dementia, psychosis, liver dysfunction, renal dysfunction, dialysis treatment, acute and chronic infectious diseases, malignancy, autoimmune diseases, connective tissue diseases, alcohol and drug abuse, contraindications to bioelectrical impedance analyzer, and unwillingness to provide informed consent. Individuals who planned to leave the study area within five years and failed to complete the baseline evaluation were also excluded from the study.
The present research was conducted in compliance with the Declaration of Helsinki and the study protocol was approved by the Research Ethics Committee of the Institute of Basic Medicine, Shandong Academy of Medical Sciences, Jinan, China. Written informed consent was obtained from all participants.
Glucose metabolic status assessment and classification
Blood samples from each participant were collected by experienced nurses in the morning after overnight fasting. Plasma and serum were separated and stored at -80℃ for subsequent analyses. The concentration of fasting plasma glucose (FPG) was measured using a routine enzymatic method with a Hitachi 7600 (Hitachi, Ltd., Tokyo, Japan) automated biochemical analyzer. Hemoglobin A1c (HbA1c) was measured using ion-exchange high-performance liquid chromatography.
Glucose metabolic status was divided into the following categories: euglycemia (FPG < 100 mg/dL, HbA1c < 5.7%, no diagnosed diabetes, no use of antihyperglycemic medication), prediabetes (FPG 100–125 mg/dL and/or HbA1c 5.7–6.4%, no diagnosed diabetes, and no use of antihyperglycemic medication), uncontrolled diabetes (FPG ≥ 126 mg/dL and/or HbA1c ≥ 6.5%, diagnosed diabetes, and use of antihyperglycemic medication), and well-controlled diabetes (FPG < 126 mg/dL and HbA1c < 6.5%, diagnosed diabetes, and use of antihyperglycemic medication) [39, 40].
Anthropometric and body composition assessments
Body weight, height, WC, hip circumference, and visceral fat area (VFA) were assessed in a quiet, bright, and warm room (22–24℃) by trained nurses who were blinded to the clinical data for participants. Participants were asked to fast for 3 h, empty their bladder, and wear a minimal amount of clothing and no shoes. Body weight was measured to the nearest 0.1 kg using a segmental multifrequency bioelectrical impedance analyzer (BIA, InBody720, Biospace Co., Seoul, Korea). Height was determined to the nearest 0.1 cm using a wall-mounted stadiometer. BMI was calculated as body weight (kg)/body height (m) squared. WC and hip circumference were measured using inelastic tape to the nearest 0.1 cm at the midpoint between the iliac crest and the lower rib margins and at the level of the great trochanter. WHR was calculated as WC (cm)/hip circumference (cm). VFA (cm2) was evaluated using a BIA (InBody720, Biospace Co., Seoul, Korea) according to the manufacturer’s protocol. VFA (cm2) of 30 male and 30 female randomly selected patients measured using magnetic resonance imaging (MRI; 3.0-Tesla scanner, Signa Horizon LX, GE Medical Systems, Pittsburgh, PA, USA) according to a turbo spin echo imaging protocol was used to correct the BIA device’s measurements. There was a good consistency between the VFAs determined using MRI and using BIA with correlation coefficient of 0.81.
The correlation coefficient was 0.81 between the measures using MRI and bioelectrical impedance analyzer with good consistency.
In the present study, participants were classified into non-general vs. general obesity groups depending on their BMI (< 30 kg/m2 vs. ≥30 kg/m2), non-central vs. central obesity groups depending on their WHR (≤ 0.8 vs. >0.8 for female participants and ≤ 0.9 vs. >0.9 for male participants), and low vs. high VFA groups depending on the VFA quintile (sex-specific three low quintiles vs. two high quintiles).
Plasma lipid, serum creatinine, and hemoglobin assessment
Plasma concentrations of triglyceride (TG), total cholesterol, low-density lipoprotein cholesterol, and high-density lipoprotein cholesterol and serum concentration of creatinine were determined using a routine enzymatic method with a Hitachi 7600 (Hitachi, Ltd., Tokyo, Japan) automated biochemical analyzer. Hemoglobin was detected using a hemoglobin colorimetric assay kit (Beyotime, Shanghai, China).
Adiposity-associated metabolic activity assessment
In the present study, homeostatic model assessment of insulin resistance (HOMA-IR), adiponectin-to-leptin ratio (ALR), and triglyceride-glucose index (TyG) were used to assess adiposity-associated metabolic activities. HOMA-IR was calculated using the following formula: HOMA-IR = FPG (mmol/L) × insulin (mU/L)/22.5 [41, 42]. A landmark for investigating adipose-mediated tissue crosstalk, ALR [43, 44] was calculated by dividing adiponectin by leptin. TyG was determined using the following formula: TyG = ln [TG (mg/dL) × FPG (mg/dL)/2] [45, 46].
Plasma concentrations of insulin, leptin, and adiponectin were detected using enzyme-linked immunosorbent assay kits (Beyotime, Shanghai, China) following the manufacturer's instructions. The minimum detectable concentrations were 1.54 mU/L for insulin, 26.6 pg/mL for leptin, and 56 pg/mL for adiponectin. Intra- and inter-assay coefficients of variations were both < 10% for insulin, leptin, and adiponectin.
Covariates
Age, smoking status, alcohol consumption, exercise, history of hypertension, history of dyslipidemia, history of atrial fibrillation, use of antihypertension, lipid-lowering, anticoagulation, and anti-atrial fibrillation medications, systolic blood pressure (SBP), diastolic blood pressure (DBP), heartbeat, and plasma lipid levels were included as covariates. The plasma concentration of N-terminal pro-B-type natriuretic peptide (NT-proBNP) was detected using obtained plasma with a recombinant human NT-pro-BNP kit (Beyotime, Shanghai, China). In addition, since renal dysfunction and anemia are significantly associated with HF [10, 47], estimated glomerular filtration rate (eGFR) and hemoglobin level were also included as covariates in the present study. The eGFR was estimated according to the formula provided by the Chronic Kidney Disease Epidemiology Collaboration using serum creatinine level on the basis of age, sex, and ethnicity [38, 48].
Outcomes of interest
First HF occurrence at hospitalization and specialized outpatient visit during the follow-up period were the primary outcomes of interest in the present study. HF was determined according to the International Classification of Diseases 10th Revision code I50 and adjudicated by a committee based on the review of medical records, including clinical signs and symptoms, echocardiogram recording, NT-proBNP level, and medication use.
Statistical analysis
Continuous data were represented as the mean with standard deviation (SD) or median with interquartile range (IQR, 25th to 75th percentiles), depending on the normality determined by the Kolmogorov-Smirnov test. Categorical data were represented as frequencies with percentages. Cumulative incidence values for HF were assessed among participants stratified by glucose metabolic status (euglycemia vs. prediabetes, uncontrolled diabetes, and well-controlled diabetes) and adiposity (non-general vs. general obesity, non-central vs. central obesity, and low vs. high VFA) using Fine-Gray models accounting for competing risk of death. Multiplicative interactions were included in the models to evaluate the effect of adiposity modulation on the associations between incident HF and glucose metabolic status [each continuous and categorical variable of adiposity (BMI, WHR, and VFA and general obesity, central obesity, and high VFA)*glucose metabolic status]. The confounders included in separate models were as follows: Model 1 for none; Model 2 for age, smoking status, alcohol consumption, regular physical activity, history of hypertension and dyslipidemia, antihypertensive and lipid-lowering medication use, SBP, DBP, heart rate, plasma lipid level, eGFR level, NT-proBNP level, and hemoglobin level; Model 3 for confounders in Model 2 plus BMI; Model 4 for confounders in Model 2 plus WHR; and Model 5 for confounders in Model 2 plus VFA. Restricted cubic splines method was used to assess the associations of BMI, WHR, VFA, HOMA-IR, TyG, and ALR with incident HF and to adjust for confounders in Model 2.
The population attributable risk percentage (PARp) for incident HF across the categories of adiposity was assessed using the following formula: PARp = 1-[(1-S0(t))/(1-S(t))], where S0(t) is the counterfactual survival function and S(t) is the factual survival function [35].
Multiple linear regression models were generated to assess the associations of HOMA-IR, ALR, and TyG with BMI, WHR, and VFA adjusted for confounders in Model 2. Mediating effect analysis was carried out to assess the roles of HOMA-IR, ALR, and TyG in the association between adiposity and HF adjusted for confounders in Model 2.
For sensitivity analyses, diabetes patients were divided into uncontrolled and well-controlled diabetes groups to assess the association between diabetes and HF. The adiposity characteristics were used as categorical and continuous variables to evaluate the association with HF.
R (version 4.3.2, R Foundation for Statistical Computing) and SPSS (version 26.0, IBM Corp., Armonk, NY, USA) software packages were used to perform all statistical analyses. GraphPad Prism (version 9.1.0, GraphPad Prism Software Inc., San Diego, CA, USA) and R (version 3.6.3, R Foundation for Statistical Computing) software packages were used to generate graphs. Two-tailed P-value of < 0.05 was considered statistically significant.