Cohort-specific associations of nine cardiometabolic indices with 3-year incident cardiovascular disease in CHARLS and heart-disease mortality in NHANES

To examine cohort-specific adjusted associations of nine cardiometabolic indices with 3-year incident cardiovascular disease (CVD) in CHARLS and heart-disease mortality in NHANES, and to describe exploratory within-cohort prediction performance. We conducted parallel cohort-specific analyses. In CHARLS, 7,608 participants free of CVD at the 2015 baseline and with ascertainable 2018 outcome status were analyzed; 926 developed incident self-reported CVD. Associations per cohort-specific 1-standard-deviation higher index were estimated using fully adjusted logistic regression across 20 index-specific multiple-imputation datasets and combined using Rubin’s rules. In NHANES, 16,579 fasting-subsample participants were analyzed, including 374 heart-disease deaths and 1,195 non-heart-disease deaths. Fully adjusted survey-weighted cause-specific Cox models used pooled fasting-subsample weights and combined strata and primary sampling units. Restricted cubic splines, false-discovery-rate correction, and cohort-specific sensitivity analyses were performed. Prediction-performance analyses were exploratory and evaluated separately within each cohort. In CHARLS, METS-IR, log(CMI), RFM, LAP, CVAI, AIP, TyG-BMI, and TyG-WC were positively associated with incident CVD after false-discovery-rate correction (odds ratios, 1.12–1.25 per 1-standard-deviation higher index); the association for TyG was not statistically significant (odds ratio, 1.07; 95% confidence interval, 0.99–1.15). In NHANES, METS-IR, RFM, LAP, CVAI, TyG-BMI, and TyG-WC were positively associated with heart-disease mortality after false-discovery-rate correction (hazard ratios, 1.09–1.77), whereas TyG, log(CMI), and AIP were not. No index showed statistically significant evidence of nonlinearity in either cohort. Formal paired comparisons of exploratory prediction metrics did not identify any index with multiplicity-adjusted superiority over the conventional base model or other index-extended models. Several cardiometabolic indices were associated with cohort-specific cardiovascular outcomes. However, differences in outcome definitions, follow-up structures, and index composition preclude cross-cohort validation claims, and formal comparisons did not establish a universally superior index or support clinical recommendation of a single index. Not applicable.