A machine learning-based approach for low-density lipoprotein cholesterol calculation using age, and lipid...

Area of Science:

• Biochemistry
• Cardiovascular Medicine
• Machine Learning in Healthcare

Background:

• Low-density lipoprotein cholesterol (LDL-C) is a key cardiovascular disease biomarker.
• Current methods for LDL-C estimation lack consensus in Chinese laboratories.
• Accurate LDL-C measurement is crucial for cardiovascular risk assessment.

Purpose of the Study:

• To develop and validate a machine learning (ML) based method for estimating LDL-C in the Chinese population.
• To compare the performance of ML-derived equations against established LDL-C calculation methods.
• To address the need for a reliable LDL-C estimation technique in Chinese clinical settings.

Main Methods:

• Utilized an extensive dataset of 111,448 samples, randomized into five subsets for cross-validation.
• Developed ML-based equations incorporating age, sex, and lipid parameters.
• Externally validated ML equations across three independent datasets and compared performance with Friedewald, Martin/Hopkins, and Sampson equations.

Main Results:

• ML equations demonstrated reduced bias compared to traditional methods, including for patients with high triglycerides (TG ≥ 400 mg/dL) or low LDL-C (< 40 mg/dL).
• The ML method's performance showed less susceptibility to variations in patient age.
• External validation confirmed the generalizability and robustness of the developed ML equations across different datasets.

Conclusions:

• Machine learning models integrating sex, age, and lipid parameters offer a more robust and reliable method for LDL-C calculation.
• The developed ML approach shows significant potential for improving cardiovascular risk assessment in China.
• This study provides a foundation for adopting advanced computational methods in clinical lipid diagnostics.