An innovative model for predicting coronary heart disease using triglyceride-glucose index: a machine learning-based...

Area of Science:

• Cardiology
• Metabolic Syndrome
• Machine Learning in Healthcare

Background:

• Existing coronary heart disease (CHD) predictive models lack optimal accuracy.
• These models often overlook insulin resistance and triglyceride (TG) levels, despite their known impact on CHD.
• The triglyceride-glucose index (TyG-index) is a validated indicator of insulin resistance.

Purpose of the Study:

• To enhance CHD predictive models by incorporating the TyG-index.
• To evaluate the TyG-index's efficacy as a CHD predictor compared to diabetes.
• To develop machine learning models for improved CHD risk assessment.

Main Methods:

• A cohort of 2000 Iranian adults (aged 20-74) was followed for a mean of 9.9 years.
• Multivariate Cox proportional hazard models assessed the association between TyG-index and CHD incidence.
• Machine learning models were developed, substituting the TyG-index for diabetes in established CHD risk scores.

Main Results:

• CHD incidence was 14.5% in the study population.
• Higher TyG-index quartiles were significantly associated with increased CHD risk (HR 2.32 for the highest quartile).
• A TyG-index > 8.42 demonstrated high negative predictive value; TyG-index-based SVM models outperformed diabetes-based models.

Conclusions:

• The TyG-index is a more potent predictor of CHD than diabetes.
• The TyG-index emerged as the most significant predictor after age in machine learning models.
• Clinical implementation of the TyG-index is recommended for CHD risk identification and prevention.