Uric Acid to HDL-C Ratio as a Novel Biomarker for Sarcopenia: A National Study with Machine Learning Insights

Background Sarcopenia is a major age-related health burden. Although the uric acid to high-density lipoprotein cholesterol ratio (UHR) has been linked to sarcopenia, the underlying pathways and its clinical utility for risk stratification remain unclear. Methods We analyzed 9,853 middle-aged and older adults from the China Health and Retirement Longitudinal Study (CHARLS). Sarcopenia was defined by Asian Working Group for Sarcopenia 2019 Consensus (AWGS 2019) criteria. We assessed the UHR-sarcopenia relationship using multivariable logistic regression and restricted cubic splines. Mediation analyses examined roles of insulin resistance (TyG index) and renal function (eGFR). Five machine learning algorithms—logistic regression, decision tree (DT), random forest (RF), support vector machine (SVM), and extreme gradient boosting (XGBoost)—were developed. The dataset was randomly divided into training (70%) and testing (30%) sets. Within the training set, 5-fold cross-validation was applied for hyperparameter tuning. Model performance in the test set was primarily evaluated using the area under the receiver operating characteristic curve (ROC-AUC), accuracy, sensitivity, and specificity. Feature selection was conducted using Boruta and least absolute shrinkage and selection operator (LASSO) methods, and model interpretability was assessed using SHapley Additive exPlanations (SHAP). Results In this nationally representative cohort, higher UHR was independently associated with lower odds of sarcopenia (adjusted odds ratio (OR) per- standard deviation (SD) increase = 0.75, 95% CI: 0.69–0.81, P < 0.001), showing a nonlinear dose-response. UHR correlated with better handgrip strength and chair-stand performance. The TyG index and eGFR mediated 16.44% and 2.03% of this association, respectively. In machine learning analyses, model performance was broadly comparable across algorithms. In the test set, the area under the receiver operating characteristic curve (ROC-AUC) ranged from 0.73 to 0.77. Among the evaluated models, XGBoost demonstrated the best performance (ROC-AUC = 0.77, accuracy = 0.70, sensitivity = 0.70, specificity = 0.71). SHAP analysis based on the XGBoost model further identified UHR as the second most important predictor of sarcopenia, following age. Conclusions Our study provides robust evidence that a higher UHR is a protective factor against sarcopenia in Chinese middle‑aged and older adults, partially mediated by metabolic and renal pathways. UHR demonstrates strong discriminatory value in risk profiling, supporting its potential as a simple, integrative biomarker for sarcopenia screening in this population.