Non-Traditional Lipid Ratios Predict Cardiovascular-Kidney-Metabolic Syndrome: Insights from Machine Learning Model Using NHANES Data

BackgroundCardiovascular-kidney-metabolic (CKM) syndrome is a newly defined multisystem disease continuum characterized by the coexistence of metabolic dysfunction, cardiovascular disease, and chronic kidney disease (CKD). This study aimed to explore the relationship between the non-traditional lipid ratios and CKM syndrome. MethodsA cross-sectional analysis was performed using data from the 2007-2018 National Health and Nutrition Examination Survey (NHANES). CKM stages (0-4) were classified according to the 2023 American Heart Association (AHA) criteria. Non-traditional lipid ratios, including triglyceride to high-density lipoprotein cholesterol (TG/HDL-C), total cholesterol to HDL-C (TC/HDL-C), low-density lipoprotein cholesterol to HDL-C (LDL-C/HDL-C), and non-HDL-C, were assessed. Nonlinear associations were explored using restricted cubic spline models. Variable selection was conducted using the least absolute shrinkage and selection operator (LASSO) regression, and four machine learning models were developed. SHapley Additive exPlanations (SHAP) and partial dependence plots (PDPs) were applied for interpretability. ResultsThe significant associations were observed between non-traditional lipid ratios and CKM risk. Nonlinear relationships were observed for LDL, TG, TC, and HDL. The random forest (RF) algorithm demonstrated superior performance. SHAP analysis identified body mass index (BMI) as the most influential predictor, followed by TG, TG/HDL-C, and glucose. Elevated TG/HDL-C, TC/HDL-C, and non-HDL-C levels were positively associated with increased CKM risk, while LDL-C/HDL-C showed inverse associations. PDPs indicated synergistic interactions among TG/HDL-C, TC/HDL-C, and non-HDL-C. Multivariate logistic regression confirmed TG/HDL-C as an independent predictor (OR = 3.11; 95% CI: 2.49-3.88; p < 0.001). ConclusionTG/HDL-C and non-HDL-C are strong predictors of CKM syndrome. Integration of these markers into risk models may facilitate earlier detection and guide individualized prevention strategies.