Association between red cell distribution width-albumin ratio and osteoarthritis in middle-aged and older adults: Analysis of NHANES data (1999-2018).

Abstract

This study aimed to investigate the association between the Red blood cell distribution width-to-albumin ratio (RAR) and osteoarthritis (OA) in middle-aged and elderly populations, and to assess its potential value as a predictive indicator for OA risk. Based on data from the National Health and Nutrition Examination Survey (NHANES) 1999-2018, this study included 19,967 participants aged 40 years and older. Weighted logistic regression models were used to analyze the association between RAR and OA, with robustness verified through stepwise adjustment for confounding factors. Restricted cubic splines (RCS) and threshold effect analyses were further employed to explore the dose-response relationship between the two, and subgroup analysis was conducted to evaluate potential interactions among variables. Additionally, LASSO regression was utilized to screen key predictive variables to construct an RAR-based OA risk prediction model. The model's discriminatory ability and clinical utility were evaluated using receiver operating characteristic (ROC) curves and decision curve analysis (DCA). Meanwhile, this study compared the predictive performance of five machine learning algorithms and employed five-fold cross-validation to assess model robustness. Weighted logistic regression results showed that RAR was significantly positively associated with OA in the middle-aged and elderly population, and this association remained stable across three progressively adjusted models. RCS analysis indicated a nonlinear relationship between RAR and OA, with a critical inflection point at RAR = 3.42. The prediction model constructed based on LASSO regression was presented in the form of a nomogram, with an AUC of 0.723 (95% CI: 0.714-0.733). DCA results demonstrated that the model had favorable clinical net benefit. Among the five machine learning models, the Random Forest model showed consistently excellent predictive performance in both training and validation sets. Five-fold cross-validation further supported the good robustness of this model. RAR is significantly positively associated with OA risk in the middle-aged and elderly population, suggesting its potential value as a predictive biomarker for OA. However, limited by the cross-sectional study design and data from a single population, the clinical application of this association and the prediction model requires further validation in prospective studies and independent cohorts.