Comparative Evaluation of Machine Learning Models and Conventional Formulas for LDL Cholesterol Estimation.

Area of Science:

• Biomedical Informatics
• Clinical Chemistry
• Machine Learning in Healthcare

Background:

• Accurate estimation of low-density lipoprotein cholesterol (LDL-C) is crucial for cardiovascular risk assessment.
• Conventional formulas for LDL-C estimation exhibit limitations, particularly in hypertriglyceridemic individuals.
• Machine learning (ML) presents a potential avenue for improving LDL-C estimation accuracy.

Purpose of the Study:

• To develop and validate ML models for LDL-C estimation.
• To compare the analytical and clinical performance of ML models against conventional formulas.
• To evaluate model performance in individuals with elevated triglyceride (TG) levels.

Main Methods:

• Retrospective analysis of lipid profiles from 11,681 adults.
• Development and evaluation of multiple ML models (linear regression, random forest, support vector regression, XGBoost) using 10-fold cross-validation.
• Performance assessment via error metrics (MAE, RMSE), bias, correlation, Bland-Altman analysis, and clinical classification accuracy, with subgroup analysis for TG levels (TG ≥ 400 mg/dL).

Main Results:

• ML models, particularly XGBoost, demonstrated superior performance with lower error and higher agreement compared to conventional formulas.
• XGBoost achieved MAE of 14.7 mg/dL, RMSE of 20.22 mg/dL, and R² of 0.780.
• ML models maintained stability and higher clinical classification accuracy (up to 66%) across increasing TG levels, outperforming conventional methods in hypertriglyceridemic populations.
• External validation confirmed the stable performance and superior classification accuracy of the XGBoost model.

Conclusions:

• ML-based LDL-C estimation provides a robust alternative to conventional formulas.
• These ML models are particularly beneficial for improving LDL-C estimation accuracy in patients with hypertriglyceridemia.
• The findings support the integration of ML models into clinical practice for more precise lipid management.