# Beyond the triglyceride-glucose index, the cholesterol- high-density lipoprotein -glucose index as a superior predictor for diabetes risk in patients with major adverse cardiovascular events: dual evidence from the CHARLS database and real-world data

## TL;DR

This study finds that the cholesterol-high-density lipoprotein-glucose index is a better predictor of diabetes risk in patients with cardiovascular events than the triglyceride-glucose index.

## Contribution

The study introduces the CHG index as a superior metabolic indicator for diabetes risk prediction in cardiovascular patients.

## Key findings

- The CHG index showed higher hazard ratios for diabetes risk in both cohorts compared to the TyG index.
- The CHG index had a slightly higher predictive accuracy and model calibration than the TyG index.

## Abstract

To compare the predictive value of the cholesterol-high-density lipoprotein-glucose (CHG) index versus the triglyceride-glucose (TyG) index for new-onset diabetes in patients with major adverse cardiovascular events (MACE) using two independent cohorts.

This multicenter study enrolled 1,138 patients (median follow-up: 7.99 years) from the China Health and Retirement Longitudinal Study (CHARLS) cohort and 614 patients (median follow-up: 2.61 years) from the Central Hospital of Shaoyang (CHSY) cohort. Multivariable Cox regression, Kaplan-Meier curves, restricted cubic spline (RCS) analysis, subgroup analysis, and sensitivity analysis were employed to assess associations. Predictive performance was compared using receiver operating characteristic (ROC) curves and machine learning.

New-onset diabetes occurred in 181 (15.9%) and 71 (11.6%) patients in the CHARLS and CHSY cohorts, respectively. After adjustment, the hazard ratios (HRs) per 1-standard deviation increase in the CHG index were 1.32 (95% CI: 1.14–1.53) and 1.63 (1.33–2.02) in the two cohorts, respectively. The corresponding HRs for the TyG index were 1.30 (1.12–1.50) and 1.50 (1.22–1.85). Both indices showed a linear positive association (RCS P for nonlinearity >0.05). The CHG index demonstrated a slightly higher area under the curve (AUC) value in the CHARLS study (0.632 vs. 0.626) and greater feature importance in machine learning models.

Both the CHG and TyG indices independently predict the risk of diabetes in patients with MACE. However, the CHG demonstrates a slight advantage in overall predictive accuracy and model calibration, suggesting it may be a better metabolic indicator for assessing the risk of glucose metabolism disorders following cardiovascular events.