Development of a CHD risk prediction model using novel lipid markers based on lipidomics and VAP technology

This study aimed to develop a novel risk prediction model for coronary heart disease (CHD) based on lipid subfraction and lipidomics data.

Lipid subfraction analysis was conducted on 624 subjects (312 CHD, 312 controls) using Vertical Auto Profile (VAP). The participants were randomly divided into a training cohort (n = 422, 211 CHD, 211 controls) and a validation cohort (n = 202, 101 CHD, 101 controls) at a 2:1 ratio, stratified by CHD status. Untargeted lipidomics was performed on 16 subjects (8 CHD, 8 controls) randomly selected from the total enrolled population using UHPLC-Q-TOF/MS. Lasso regression with ten-fold cross-validation was applied to screen key predictive factors, followed by multivariate logistic regression to construct a risk prediction model. The false discovery rate (FDR) was controlled using the Benjamini-Hochberg procedure for multiple comparisons. Model discriminative performance was evaluated using ROC curve analysis.

A total of 2,335 lipid molecules across eight classes were identified. After FDR correction, the differential lipids remained statistically significant. CHD patients exhibited significant upregulation of RLP-C and sdLDL-C-related lipid components, while HDL2B-related lipids were markedly downregulated. KEGG pathway enrichment analysis revealed that differential lipids were mainly enriched in glycerophospholipid metabolism, choline metabolism in cancer, and fat digestion and absorption. Multivariate logistic regression identified Lp(a) (OR = 1.288), RLP-C (OR = 3.848), sdLDL-C (OR = 5.317), age (OR = 1.053), systolic blood pressure (OR = 1.075), and fasting plasma glucose (OR = 3.903) as independent risk factors for CHD, whereas HDL2B (OR = 1.416 × 10^-6) was a protective factor. In the validation cohort, the model achieved an AUC of 0.867 (95% CI: 0.8087-0.9252). Using an optimal cutoff of 0.4726 (Youden index = 0.628), the model demonstrated a sensitivity of 81.4% and a specificity of 81.4%.

This study successfully constructed a multivariate logistic regression risk prediction model integrating novel lipid subfractions (sdLDL-C, RLP-C, HDL2B, Lp(a)) and routine clinical variables (age, SBP, FPG). The model exhibits acceptable predictive performance and may provide a complementary tool for CHD risk assessment, pending external validation in larger prospective cohorts.