Reassessment-stage metabolic indices and subsequent cardiorenal outcomes in adults with diabetes: CHARLS and an independent hospital validation cohort

Background Simple metabolic indices may capture residual cardiorenal risk beyond conventional glycemic markers, but their reassessment-stage value and external reproducibility remain unclear. We evaluated the triglyceride-glucose (TyG) index, metabolic score for insulin resistance (METS-IR), triglyceride total cholesterol body weight index (TCBI), and cholesterol-high-density-lipoprotein-glucose (CHG) index in relation to subsequent cardiorenal outcomes. Methods The CHARLS cohort comprised participants with diabetes identified in 2011 who attended the 2015 survey, remained free of overt cardiorenal outcomes, and had post-2015 follow-up. The primary outcome was the first cardiovascular or kidney event. Analyses included survey-wave-based Cox and discrete-time models, Kaplan-Meier curves, restricted cubic splines, selection weighting, proportional-hazards diagnostics, 10-fold cross-validated calibration and decision curves, and replication in an independent hospital validation cohort. Results The eligible CHARLS reassessment cohort included 1,104 participants and 311 composite cardiorenal outcomes; glycemia-adjusted complete-case Cox analyses included 787–800 participants. Each 1-SD increase in TyG, TCBI, and METS-IR was associated with higher risk, with HRs (95% CIs) of 1.26 (1.06-1.49), 1.19 (1.03-1.37), and 1.20 (1.04-1.38), respectively, whereas CHG was not significant. Selection-weighted and discrete-time estimates were similar, and proportional-hazards tests showed no violations. In the wave-approximated 4-year prediction sample (N = 749; 173 events), apparent AUC gains were small and not statistically significant; 10-fold cross-validated AUC increased only from 0.665 to 0.669-0.671, without a consistent decision-curve advantage. In the independent hospital validation cohort (N = 1,578; 296 events), with post-landmark follow-up administratively censored at 4 years, positive composite-outcome associations were reproduced for TyG, TCBI, and METS-IR, whereas CHG showed a weaker and graphically inconsistent pattern. Conclusions Among adults with diabetes identified in 2011 who returned for the 2015 CHARLS reassessment and remained free of overt cardiorenal outcomes, higher TyG, TCBI, and METS-IR were associated with subsequent composite cardiorenal outcomes. Robustness analyses supported these associations, whereas predictive improvement was limited. These indices may serve as exploratory, low-burden adjunctive markers at reassessment, with supportive independent clinical validation; clinical utility requires further study.