Machine learning-based prediction model for the efficacy and safety of statins

Area of Science:

• Cardiovascular Medicine
• Biostatistics
• Machine Learning in Healthcare

Background:

• Statin therapy is crucial for reducing atherosclerotic cardiovascular disease (ASCVD) risk.
• Increasing cholesterol levels necessitate rational statin use.
• Statin-associated adverse reactions, like liver enzyme abnormalities and muscle symptoms (SAMS), hinder widespread use.

Purpose of the Study:

• To develop a predictive model for statin efficacy and safety.
• To utilize real-world clinical data and machine learning techniques.
• To create a web-based platform for predicting statin outcomes.

Main Methods:

• Data preprocessing included random forest imputation and Borderline SMOTE oversampling.
• Boruta method for feature selection; 7:3 train-test split.
• Five algorithms (logistic regression, naive Bayes, decision tree, random forest, gradient boosting) were evaluated with cross-validation and bootstrapping.
• SHAP for feature interpretability and development of a web platform.

Main Results:

• Random forest algorithm showed superior performance.
• High AUC and accuracy achieved for predicting LDL-C target attainment (0.883, 0.868), liver enzyme abnormalities (0.964, 0.964), and muscle pain/CK abnormalities (0.981, 0.980).
• Key predictors identified: cerebral infarction, TG, PLT, HDL for efficacy; CRP, CK, oral medications for liver/muscle adverse events.

Conclusions:

• A machine learning-based predictive model for statin efficacy and safety was successfully developed.
• The user-friendly web platform can guide statin therapy decisions and optimize treatment.
• Further research and application are recommended to enhance statin therapy utilization.