A new equation for estimating low-density lipoprotein cholesterol concentration based on machine learning

Area of Science:

• Cardiovascular Medicine
• Biochemistry
• Data Science

Background:

• Low-density lipoprotein cholesterol (LDL-C) is a key indicator of cardiovascular risk.
• Existing formulas (Friedewald, Martin-Hopkins, Sampson) show inaccuracies in estimating LDL-C for the Chinese population.
• Accurate LDL-C measurement is vital for preventing and managing atherosclerotic diseases.

Purpose of the Study:

• To develop and validate novel machine learning-based equations for precise LDL-C calculation in the Chinese population.
• To compare the performance of the new equations against established methods.
• To provide a more accurate tool for clinical assessment and monitoring of cardiovascular health.

Main Methods:

• Utilized a large dataset of 182,901 patient samples with standard lipid panel measurements.
• Employed polynomial ridge regression to construct a new ternary cubic equation.
• Validated the equation on a separate test set and an additional clinical dataset of 17,285 samples.

Main Results:

• The new ternary cubic equation demonstrated high accuracy (R² = 0.9815, MSE = 37.4250 on testing set).
• It showed significantly smaller differences compared to measured LDL-C than Friedewald, Martin-Hopkins, and Sampson equations.
• The equation performed well in clinical practice (R² = 0.9780, MSE = 24.8482) and across diverse triglyceride and LDL-C levels.

Conclusions:

• The developed machine learning equation provides a highly accurate and user-friendly method for calculating LDL-C concentration.
• It surpasses traditional equations in precision, particularly for the Chinese population.
• This new equation can supplement direct LDL-C measurement, enhancing the prevention and management of atherosclerotic diseases.