Adiposity and insulin resistance indices and progression to advanced cardiovascular-kidney-metabolic syndrome in Chinese adults

Abstract Metabolic and adiposity-related indices may help identify individuals at increased risk of cardiovascular-kidney-metabolic (CKM) progression; however, comparative prospective evidence remains limited. We examined the associations of seven metabolic and adiposity-related indices with progression to advanced CKM stages and assessed their incremental discrimination and exploratory internal machine-learning performance. This prospective analysis included 2,871 participants from the China Health and Retirement Longitudinal Study with CKM stages 0–2 at baseline in 2011. CKM progression was defined as progression to stages 3–4 at the 2015 follow-up. Associations of the triglyceride-glucose index (TyG), TyG-waist circumference (TyG-WC), TyG-waist-to-height ratio (TyG-WHtR), lipid accumulation product (LAP), visceral adiposity index (VAI), waist-to-height ratio (WHtR), and body roundness index (BRI) with CKM progression were examined using multivariable Cox regression after multiple imputation. Dose–response, subgroup, sensitivity, and incremental discrimination analyses were performed. Separate LAP-, TyG-WC-, and TyG-WHtR-based machine-learning frameworks were evaluated using a stratified 70:30 training/test split. During follow-up, 595 participants (20.7%) progressed to CKM stages 3–4. In fully adjusted analyses, the highest versus lowest quartile of each index was associated with a higher risk of CKM progression, with HRs ranging from 1.52 for TyG (95% CI, 1.20–1.94) to 2.03 for TyG-WC (95% CI, 1.60–2.57; all P for trend < 0.001). Corresponding HRs were 2.01 (95% CI, 1.57–2.57) for TyG-WHtR and 1.98 (95% CI, 1.55–2.54) for LAP. Six indices showed nonlinear dose–response associations. Subgroup analyses showed significant interactions by sex for TyG-WHtR, LAP, WHtR, BRI, and VAI; by smoking status for TyG-WHtR, LAP, and VAI; and by age for WHtR (all P for interaction ≤ 0.049). Addition of individual indices to the covariate model yielded C-indices of 0.702–0.715, compared with 0.701 for the basic model; TyG-WC achieved the highest C-index. In the held-out internal test set, logistic regression achieved the highest ROC-AUC in the LAP (0.721), TyG-WC (0.729), and TyG-WHtR (0.726) frameworks. SHAP analyses of the corresponding logistic regression models identified age as the predictor with the largest mean absolute contribution across all three frameworks. Elevated levels of all seven indices were associated with CKM progression, particularly TyG-WC, TyG-WHtR, and LAP. Given the structural overlap between several index components and the CKM staging framework, the incremental discrimination and machine-learning findings should be considered exploratory and require external validation.