Endpoint-specific associations of the cholesterol, high-density lipoprotein, and glucose (CHG) index and its adiposity- and inflammation-related derivatives with incident cardiovascular disease and mortality: evidence from CHARLS and NHANES

Abstract

The cholesterol-high-density lipoprotein–glucose (CHG) index combines lipid and glycemic measures, but whether its adiposity- and inflammation-related derivatives add information beyond their components and established risk factors is uncertain. We compared nine CHG-related indices with incident cardiovascular disease (CVD) and mortality. We included 7516 adults aged ≥ 45 years without baseline CVD from the China Health and Retirement Longitudinal Study (CHARLS). Associations across four follow-up intervals were estimated using multivariable discrete-time complementary log–log models. False discovery rate (FDR) correction was applied within prespecified families. Component-separated analyses and internally cross-validated prediction models benchmarked derivatives against components and conventional risk models. Complementary mortality analyses used survey-weighted Cox models in the 1999–2018 National Health and Nutrition Examination Survey (NHANES); complete-case analyses included 8658 participants for eight core indices and 7059 for CHG-CRP. Over 9 years, 1849 CHARLS participants developed CVD. In fully adjusted Model 3, all nine indices were associated with incident CVD per 1-SD increment (HRs 1.077–1.213; all FDR-adjusted P ≤ 0.002), with the largest estimate for CHG-CVAI (HR 1.213, 95% CI 1.154–1.275). All indices remained associated with stroke (HRs 1.234–1.332), whereas five adiposity-related derivatives remained associated with heart disease (HRs 1.116–1.164). No multiplicative product term remained significant after FDR correction in component-separated models. Adding CHG-BMI, CHG-WC, or CHG-CVAI to a core conventional model increased 9-year AUC by 0.009–0.011, but none improved AUC or Brier score after model extension, and no derivative outperformed its component. Associations were broadly preserved in sensitivity analyses. In NHANES, none of the eight core indices was associated with mortality after FDR correction. CHG-CRP remained associated with all-cause mortality (HR 1.14, 95% CI 1.06–1.23; FDR-adjusted P = 0.006), whereas its cardiovascular mortality association did not survive correction (HR 1.21, 95% CI 1.04–1.40; FDR-adjusted P = 0.123). CHG-related indices were associated with incident CVD, more consistently with stroke, but their apparent advantages largely overlapped with component variables and conventional risk factors. Mortality associations differed and were exploratory. No single index showed consistent superiority or sufficient incremental performance to support standalone clinical use.