Machine learning evaluation of TyG-based metrics for arteriosclerosis progression

Area of Science:

• Cardiovascular Disease Epidemiology
• Metabolic Syndrome Research
• Biomarker Discovery in Atherosclerosis

Background:

• Insulin resistance (IR) is a key risk factor for arteriosclerosis.
• Triglyceride-glucose (TyG) index and its obesity-related derivatives (TyG-BMI, TyG-WC, TyG-WHtR) are established IR markers.
• Arteriosclerosis can precede diabetes diagnosis, necessitating early detection in non-diabetic individuals.

Purpose of the Study:

• To evaluate the association between cumulative exposure to IR indicators and arteriosclerosis progression in a non-diabetic cohort.
• To assess the predictive value of the TyG index and its derivatives for arteriosclerosis development using machine learning.
• To explore potential nonlinear relationships between IR markers and arteriosclerosis progression.

Main Methods:

• A cohort study of 4,160 participants in southern China (2017-2024) with regular follow-up.
• Multivariable linear regression and restricted cubic splines to analyze associations.
• Machine learning models to predict arteriosclerosis, evaluated using AUC and PR-AUC.

Main Results:

• Cumulative average exposure to the TyG index and TyG-WHtR strongly correlated with arteriosclerosis progression (p < 0.05).
• A potential U-shaped relationship was observed between cumulative TyG-WHtR and pulse wave velocity (baPWV) changes.
• Machine learning models incorporating TyG index or derivatives showed moderate improvement in predicting arteriosclerosis progression (AUC increased to 0.737-0.744).

Conclusions:

• Cumulative exposure to the TyG index and its derivatives are significant predictors of arteriosclerosis progression in non-diabetic populations.
• These readily available, low-cost metrics can aid in the early identification of individuals at high risk for arteriosclerosis.
• Integrating these markers into predictive models enhances early risk stratification for cardiovascular events.