Key result
Adding LDL-1 and LDL-2 subfractions to neural network models improves ASCVD discrimination by ~0.02 AUROC.
Why the study?
The association between LDL subfractions and established ASCVD, as well as the incremental value of incorporating LDL subfraction features into machine-learning models to identify individuals with ASCVD, required evaluation.
Does incorporating LDL subfraction measurements into machine-learning models improve the identification of individuals with established ASCVD compared to conventional clinical variables alone?
Case-Control (n=374)
No
Does incorporating LDL subfraction measurements into machine-learning models improve the identification of individuals with established ASCVD compared to conventional clinical variables alone?
Answers are grounded in this paper and cite their sources.
Adding LDL-1 and LDL-2 subfraction measurements to machine-learning models improves the discrimination of patients with established ASCVD beyond conventional LDL-C and clinical variables.
Background Atherosclerotic cardiovascular disease (ASCVD) is one of the leading causes of mortality and disability worldwide. This study aimed to evaluate the association between low-density lipoprotein (LDL) subfractions and established ASCVD and to construct a machine-learning model based on these features to improve the identification of individuals with ASCVD. Methods A total of 374 participants, including 185 patients with ASCVD and 189 non-ASCVD controls, were enrolled. Baseline clinical characteristics, biochemical parameters, and lipid profiles were collected. LDL subfractions were quantified using the Lipoprint polyacrylamide gel electrophoresis system. Machine-learning models were developed using conventional clinical variables alone or in combination with LDL subfraction measurements to explore their incremental value in discriminating participants with established ASCVD from non-ASCVD controls. Results Compared with controls, patients with ASCVD exhibited significantly higher levels of LDL-1 and LDL-2, whereas LDL-C levels were comparable between groups. LDL-C showed significant correlations with LDL subclasses. Incorporating LDL-1 and LDL-2 resulted in higher apparent discrimination for some machine-learning models, with the ANN model showing the highest apparent AUROC (0.816). Conclusions LDL subclass analysis provided complementary information beyond conventional LDL-C measurement. Incorporating LDL-1 and LDL-2 into machine-learning models enhanced the apparent discrimination between participants with established ASCVD and non-ASCVD controls. These findings highlight the potential value of LDL subclass profiling as an adjunctive biomarker for characterizing established ASCVD and provide a basis for its further clinical evaluation.
No takes yet. Share an insight, caveat, or question.
Yu et al. (2026) conducted a case-control in Atherosclerotic cardiovascular disease (ASCVD) (n=374). Low-density lipoprotein subfractions (LDL-1 and LDL-2) vs. Conventional clinical variables alone / Non-ASCVD controls was evaluated on Discrimination of established ASCVD using artificial neural network (ANN) AUROC (95% CI 0.745-0.888, p=0.034). Incorporating LDL-1 and LDL-2 subfractions into an artificial neural network model improved ASCVD discrimination compared to clinical variables alone.
Synapse has enriched 10 related papers. Consider them for comparative context: