Construction of a Machine Learning-Based Clopidogrel Resistance Risk Prediction Model

Area of Science:

• Cardiology
• Pharmacogenomics
• Biomedical Informatics

Background:

• Clopidogrel is vital for preventing arterial circulation disorders but exhibits unpredictable antiplatelet efficacy.
• Individual variability in clopidogrel response necessitates predictive tools for optimal patient management.

Purpose of the Study:

• To develop and validate a machine learning model for predicting clopidogrel resistance risk.
• To identify key clinical and laboratory indicators associated with clopidogrel resistance.

Main Methods:

• A cohort of 1592 cardiovascular disease patients treated with clopidogrel was analyzed.
• Lasso and multivariable logistic regression were used for feature selection from clinical, laboratory, and genetic data.
• Logistic Regression, LGBM, Random Forest, and SVC models were evaluated, with Random Forest Classifier selected for the final prediction model.

Main Results:

• Key predictive variables for clopidogrel resistance included white blood cell count, hemoglobin, platelet count, fibrinogen, triglycerides, D-Dimer, mean platelet volume, prothrombin time ratio, uric acid, glycated hemoglobin, and apolipoprotein B.
• The Random Forest Classifier model achieved an AUC of 0.8730 and accuracy of 0.8033.
• A significant year-over-year increase in clopidogrel resistance rates was observed from 2020 to 2022.

Conclusions:

• A robust machine learning model was developed to predict clopidogrel resistance risk.
• The model, utilizing readily available clinical data, can aid clinicians in personalized treatment decisions.
• Improved prediction of clopidogrel resistance can lead to more effective therapeutic strategies for cardiovascular patients.