Assessment of Ten Insulin Resistance Surrogate Indexes Predicts New-Onset Cardiovascular Disease Incidence in Patients with Prediabetes or Diabetes: Insights from CHARLS Data with Machine Learning Analysis.

Among 3,532 participants with prediabetes or diabetes followed longitudinally, 874 (24.7%) developed new-onset CVD by Wave 4 follow-up.

• Study used data from the China Health and Retirement Longitudinal Study (CHARLS), a longitudinal cohort of middle-aged and elderly participants.
• Baseline data collected at Wave 1; incident CVD events assessed at Wave 4.
• Sample size: 3,532 participants with prediabetes or diabetes.

Each standard deviation increase in eGDR was associated with a statistically significant reduction in CVD risk.

• OR = 0.822, 95% CI: 0.696–0.969 per standard deviation increase in eGDR.
• eGDR (estimated glucose disposal rate) is an insulin resistance surrogate index.
• Association was assessed using multivariate logistic regression adjusted for confounders.

Each standard deviation increase in CVAI was associated with a statistically significant increase in CVD risk.

• OR = 1.124, 95% CI: 1.028–1.229 per standard deviation increase in CVAI.
• CVAI (Chinese visceral adiposity index) is an insulin resistance surrogate index.
• Association was assessed using multivariate logistic regression adjusted for confounders.

Participants in the highest eGDR quartile had a 47.3% lower CVD risk compared to those in the lowest quartile.

• OR = 0.527, 95% CI: 0.353–0.789, P = 0.0018 for highest vs. lowest eGDR quartile.
• This represented a 47.3% reduction in CVD risk.

Participants in the highest CVAI quartile had a 33.1% higher CVD risk compared to those in the lowest quartile.

• OR = 1.331, 95% CI: 1.038–1.709, P = 0.0243 for highest vs. lowest CVAI quartile.
• This represented a 33.1% increase in CVD risk.

Incorporating eGDR and CVAI into a K-Nearest Neighbors (KNN) machine learning model achieved the highest discriminative performance among the nine algorithms tested.

• KNN model AUC = 0.936, 95% CI: 0.928–0.943.
• Nine machine learning algorithms were employed to develop predictive models.
• Model performance was evaluated via ROC curves, calibration curves, and decision curve analysis.

Ten insulin resistance surrogate indexes were evaluated for their predictive performance for new-onset CVD in patients with prediabetes or diabetes.

• The ten indexes evaluated were: TyG, TyG-BMI, TyG-WC, TyG-WHtR, METS-IR, AIP, TyHGB, CTI, eGDR, and CVAI.
• Non-linear relationships between IR indexes and CVD were explored using restricted cubic spline analyses.
• Multivariate logistic regression was used to assess associations between indexes and CVD, adjusted for confounders.