Association of 14 triglyceride-glucose (TyG)-related indices with new-onset cardiovascular disease in middle-aged and older adults: evidence from the CHARLS and ELSA aging cohorts

Background: Cardiovascular disease (CVD) continues to be a predominant contributor to global illness and death rates.Insulin resistance and adiposity are closely linked to cardiometabolic risk, but the comparative value of single-time and cumulative average triglyceride-glucose (TyG)-based indices for predicting new-onset CVD remains uncertain. This study aimed to compare multiple TyG-based indicators, evaluate the contribution of cumulative exposure, and develop a clinically accessible risk-stratification tool. Methods: This longitudinal cohort study incorporated a total of 5,028 middle-aged and older adult participants, drawn from two major national aging surveys: the China Health and Retirement Longitudinal Study (CHARLS) and the English Longitudinal Study of Ageing (ELSA). Cumulative average TyG-based indicators were constructed using two pre-follow-up measurement waves, with follow-up beginning in 2015 for CHARLS and 2008 for ELSA. Seven single-time TyG-based indicators and seven cumulative average indices were evaluated. Cox regression analysis, restricted cubic splines(RCS), random-effects meta-analysis, subgroup analyses, mediation analyses, and predictive performance assessments were performed. A nomogram and web-based prediction tool were developed for 5-year CVD risk estimation. Results: During a median follow-up of 5.0 years in CHARLS and 13.5 years in ELSA, 1,067 incident CVD events occurred. In the pooled cohort, all 14 TyG-based indicators were significantly associated with incident CVD. Composite indices integrating TyG with obesity-related measures generally showed stronger associations and better predictive performance than the TyG index alone. Cumulative average indices provided only modest improvement over their single-time counterparts, with heterogeneity across cohorts. Restricted cubic spline analyses showed nonlinear associations for TyG-BMI and cumTyG-BMI, whereas other indices showed approximately linear associations. CumTyG-BRI showed relatively robust cross-cohort performance, with an HR of 1.54 (95%CI, 1.27-1.86) for Q4 versus Q1 and low between-cohort heterogeneity (I² = 6.6%). Mediation analysis suggested that hypertension partly mediated the association between CumTyG-BRI and CVD risk, accounting for 20.6% of the total effect. A TyG-BRI-based nomogram showed moderate discrimination, with a C-index of 0.679, and was deployed as an online tool for preliminary 5-year CVD risk estimation. Conclusions: CumTyG-BRI may be a relatively robust TyG-related indicator for predicting new-onset CVD across cohorts, with low heterogeneity and an approximately linear association. Hypertension may partly mediate this association. The online tool may support preliminary CVD risk stratification, but further validation is needed before routine clinical use.