Triglyceride-glucose-sarcopenia index and risk of cardiovascular diseases in multinational prospective cohorts

Abstract

Abstract Background Muscle tissue dictates insulin-driven glucose uptake; muscle dysfunction paired with insulin resistance can trigger a compound cardiometabolic failure. However, whether this dual pathology confers incremental long-term cardiovascular risk stratification is yet to be established. Methods We evaluated the triglyceride-glucose-sarcopenia [TyG-Sarcopenia (ALM/BMI)] index, calculated as the triglyceride-glucose (TyG) index × [1/(appendicular lean mass [ALM]/body mass index [BMI])], in two national cohorts. Two additional TyG-Sarcopenia indices based on handgrip strength and gait speed were also evaluated. The primary analysis included 14,314 adults from the National Health and Nutrition Examination Survey. Survey-weighted Cox models assessed associations with cardiovascular and all-cause mortality, while time-dependent area under the curve (AUC), concordance index (C-index), net reclassification improvement (NRI), and integrated discrimination improvement (IDI) assessed predictive performance. A machine-learning framework was used for internal validation. External analysis included 7,818 participants from the China Health and Retirement Longitudinal Study (CHARLS) followed for incident cardiovascular disease. Results Over a median follow-up of 133.7 months, 509 cardiovascular deaths and 1,023 all-cause deaths were documented. The highest TyG-Sarcopenia tertile was associated with significantly higher risks of cardiovascular mortality (HR, 1.984; 95% CI, 1.194–3.298) and all-cause mortality (HR, 1.706; 95% CI, 1.343–2.167), with nonlinear positive dose-response patterns. Supplementary analyses showed positive associations of TyG-Sarcopenia (gait speed) with cardiovascular and all-cause mortality (HRs, 2.887 and 2.154; 95% CIs, 1.629–5.117 and 1.642–2.826, respectively; both P < 0.001). TyG-Sarcopenia (ALM/BMI) showed a higher 10-year time-dependent AUC than TyG for all-cause mortality (0.625 vs. 0.609; P = 0.012); the corresponding AUCs for cardiovascular mortality were 0.655 and 0.639 ( P = 0.144). When added to a traditional clinical model, TyG-Sarcopenia changed the C-index from 0.887 to 0.888 for cardiovascular mortality, while the C-index remained 0.857 for all-cause mortality ( P = 0.537 and P = 0.163, respectively); continuous NRI values were 0.1246 and 0.1158, respectively (both P < 0.001), whereas IDI was not significant. Machine learning achieved an ROC-AUC of 0.885, with SHAP analyses confirming the index’s contributory role alongside conventional predictors. In CHARLS, TyG-Sarcopenia showed higher 5-year predictive performance than TyG for incident cardiovascular disease (C-index, 0.551 vs. 0.539; P = 0.036; continuous NRI, 0.1109; IDI, 0.0033; both P < 0.001). Conclusions The TyG-Sarcopenia index was independently associated with cardiovascular and all-cause mortality and may support cardiovascular risk stratification. Further prospective studies are warranted.