Longitudinal inflammatory-metabolic-adiposity burden and incident heart failure across cardiovascular-kidney-metabolic stages: a prospective cohort study

Figures

• 

  Fig. 1. Flowchart of participant selection. Of 9432 ELSA participants at Wave 2 (2004–2005), 7486 were excluded based on predefined criteria including age below 50 years, missing heart failure or CKM staging data, missing biomarker data (CRP, triglycerides, or glucose), and prevalent heart failure at baseline. The final analytic cohort comprised 1946 participants followed through Wave 8 (2016–2017), among whom 113 developed incident heart failure and 1833 remained free of heart failure

• 

  Fig. 2. Longitudinal distributions of CTGW component biomarkers and heart failure incidence across CKM stages. A ridge density plots of CRP across Waves 2, 4, and 6, with median (vertical line) and mean (diamond) indicated. B ridge density plots of triglycerides across waves. C ridge density plots of fasting glucose across waves. D ridge density plots of WHtR across waves. E violin and box plots of the CTGW composite index across waves displayed on a logarithmic scale, with mean trend line connecting wave-specific means. F bar chart of heart failure incidence rates stratified by baseline CKM stage (Stages 0–4), with event counts and percentages annotated

• 

  Fig. 3. Group-based multi-trajectory modeling of CTGW components. A model selection based on AIC and BIC across K = 2 to 5 trajectory groups, with the optimal solution (K = 3) highlighted. B distribution of average posterior probabilities of group assignment for each trajectory class, with the 0.70 acceptability threshold indicated. C mean CTGW composite index trajectories (± 95% confidence intervals) across Waves 2, 4, and 6 for the three identified groups: low-stable, moderate-metabolic, and high-inflammatory CTGW. D donut chart showing the proportional composition and sample size of each trajectory group. Trajectory groups were named based on their baseline CRP levels and longitudinal biomarker patterns

• 

  Fig. 4. Forest plot of Cox proportional hazards regression for the associations of CTGW trajectory groups and baseline quartiles of CTGW and its components with incident heart failure. Hazard ratios with 95% confidence intervals are presented across three sequential adjustment models: Model 1 (age, sex), Model 2 (Model 1 plus education, marital status, smoking, alcohol consumption), and Model 3 (Model 2 plus hypertension, diabetes mellitus, stroke, and CES-D 8 score). BMI was not included as a covariate to avoid overadjustment with WHtR. For CTGW trajectory analysis, the low-stable group served as the reference. For quartile analyses of baseline CTGW, CRP, TyG index, and WHtR, Q1 served as the reference group, with P for linear trend reported. The dashed vertical line indicates HR = 1.0

• 

  Fig. 5. Restricted cubic spline analysis depicting the dose-response relationship between cumCTGW and heart failure risk. The solid curve represents the adjusted hazard ratio (fully adjusted model) with the shaded area indicating the 95% confidence interval. The reference value was set at the median cumCTGW. The Y-axis is displayed on a logarithmic scale. The background histogram illustrates the frequency distribution of cumCTGW values. The number of knots was selected by minimizing AIC, and P-values for overall association and nonlinearity are annotated. Vertical dotted lines mark threshold values where the hazard ratio crosses 1.0

• 

  Fig. 6. Stratified analysis of CTGW exposure and incident heart failure across baseline CKM stages. Part A presents hazard ratios per one standard deviation increase in cumCTGW within each CKM stratum (Stages 0, 1, 2, and 3–4 combined), with P for interaction. Part B presents hazard ratios for CTGW trajectory groups (with low-stable as reference) within each CKM stratum, with P for interaction. All models were fully adjusted (Model 3 specification)

• 

  Fig. 7. Subgroup analysis of the association between CTGW trajectory groups and incident heart failure. Hazard ratios with 95% confidence intervals for the moderate-stable and high-burden trajectories (versus low-stable reference) are presented across subgroups defined by age (< 65 vs. ≥ 65 years), sex, BMI (< 25 vs. ≥ 25 kg/m2), hypertension status, diabetes status, and CKM risk category. P-values for interaction were derived from likelihood ratio tests. The stratification variable was excluded from covariate adjustment within corresponding subgroups. All models used the full adjustment specification