Machine learning-based evaluation of lipid biomarkers for cardiovascular risk prediction in chronic kidney disease: A cross-sectional study.

Chronic kidney disease (CKD) is an important public health issue globally, greatly increasing the prevalence and mortality of cardiovascular disease (CVD). Dyslipidemia is a prevalent metabolic disease in people with CKD that is linked to atherosclerosis and cardiovascular events. Nevertheless, the association between various lipid markers and CVD is still uncertain. This study aimed to analyze the associations between various lipid markers and CVD in patients with CKD using machine learning methods and to identify the optimal lipid biomarkers for risk prediction. This study analyzed 2696 CKD participants from the National Health and Nutrition Examination Survey from 2005 to 2018 through multivariate logistic regression to examine the relationship between various lipid markers and CVD, used restricted cubic splines to evaluate linear and nonlinear relationships between variables and outcomes, and evaluated the predictive value of various lipid markers for cardiovascular risk using machine learning models. Total cholesterol (odds ratio [OR]: 0.679 [0.614-0.751], P < .001), low-density lipoprotein cholesterol (LDL-C; OR: 0.629 [0.560-0.704], P < .001), and apolipoprotein B (OR: 0.986 [0.982-0.991], P < .001) were independently associated with CVD after multivariable adjustment in patients with CKD, among which total cholesterol, LDL-C, and apolipoprotein B showed significant L-shaped associations with CVD, while remnant cholesterol and triglycerides exhibited a U-shaped relationship. High-density lipoprotein cholesterol showed an almost linear relationship. Among the lipid markers, LDL-C demonstrated the strongest discriminative performance for CVD. In patients with CKD, lipid markers were significantly associated with CVD. Incorporating these variables into predictive models improved model discrimination and may enhance cardiovascular risk stratification in this population. Further validation in external CKD cohorts is warranted.