Evaluating the impact of metabolic indicators and scores on cardiovascular events using machine learning

Abstract

Cardiovascular diseases such as coronary artery disease, myocardial infarction, and heart failure impact millions of people annually globally and are a major cause of disease and death. This study explores the predictive capabilities of novel metabolic indices (TyG, HOMA-IR, TG/HDL-C, and VAI) for major adverse cardiovascular events (MACE) and analyzes their relationships with diabetes and cardiovascular risks. Using data from the National Health and Nutrition Examination Survey (NHANES) spanning from 2003 to 2018, we applied multiple machine learning algorithms to evaluate nine metabolic indicators including cholesterol levels, triglycerides, insulin, and waist circumference. Through cross-validation to validate model performance, the XGBoost algorithm demonstrated the most accurate performance in predicting cardiovascular outcomes, particularly for diseases like angina and heart failure. Additionally, SHAP value analysis confirmed the critical roles of waist circumference and METS-IR in predicting adverse cardiovascular events. Furthermore, we employed 100 machine learning algorithms to calculate the AUC values of metabolic indicators in predicting AP, CHD, HF, and MI, showing that METS-IR had the greatest contribution in these aspects. This research highlights the significance of metabolic indices in stratifying cardiovascular risks and presents potential avenues for targeted preventive strategies.

Keywords: Angina pectoris; Coronary disease; Heart failure; Metabolism; Myocardial infarction.

Fig. 1

Machine learning model evaluation of various metabolic indicators and scores in the angina pectoris population. A AUC results graph of the training set. B AUC results graph of the testing set. C Forest plot of AUC scores. D Summary of the shap model of the XGBoost algorithm. E Importance ranking results of the shap model of the XGBoost algorithm

Fig. 2

Machine learning model evaluation of various metabolic indicators and scores in the coronary disease population. A AUC results graph of the training set. B AUC results graph of the testing set. C Forest plot of AUC scores. D Summary of the shap model of the XGBoost algorithm. E Importance ranking results of the shap model of the XGBoost algorithm

Fig. 3

Machine learning model evaluation of various metabolic indicators and scores in the myocardial infarction population. A AUC results graph of the training set. B AUC results graph of the testing set. C Forest plot of AUC scores. D Summary of the shap model of the XGBoost algorithm. E Importance ranking results of the shap model of the XGBoost algorithm

Fig. 4

Machine learning model evaluation of various metabolic indicators and scores in the heart failure population. A AUC results graph of the training set. B AUC results graph of the testing set. C Forest plot of AUC scores. D Summary of the shap model of the XGBoost algorithm. E Importance ranking results of the shap model of the XGBoost algorithm

Fig. 5

Evaluation of AUC values for AP, CHD, HF, MI prediction using 100 machine learning algorithms. A presents the AUC values for AP prediction using 100 machine learning algorithms. B shows the AUC values for CHD prediction using 100 machine learning algorithms. C displays the AUC value of SVM + GBM algorithm on the training set. D illustrates the AUC value of SVM + GBM algorithm on the testing set. E demonstrates the AUC value of Enet[alpha = 0.3] + GBM algorithm on the training set. F depicts the AUC value of Enet[alpha = 0.3] + GBM algorithm on the testing set. G exhibits the AUC value of Enet[alpha = 0.8] + GBM algorithm on the training set. H showcases the AUC value of Enet[alpha = 0.8] + GBM algorithm on the testing set. I shows the AUC value of GBM algorithm on the training set. J reveals the AUC value of GBM algorithm on the testing set. K represents the AUC value for HF prediction using 100 machine learning algorithms. L indicates the AUC value for MI prediction using 100 machine learning algorithms