In clinical practice, precise quantification of insulin resistance (IR) has consistently been challenging. The estimated glucose disposal rate (eGDR) is an innovative metric for assessing IR. IR is vital in the pathogenesis of osteoarthritis (OA). This investigation aims to discover the association of eGDR with OA prevalence and to provide a novel method for identifying individuals with higher probability of OA. The information used for this investigation was sourced from the National Health and Nutrition Examination Survey performed between 1999 and 2018. The relationship between eGDR and OA prevalence was thoroughly investigated employing weighted multivariate regression models, smooth curve fitting (SCF), and subgroup analyses. The acquired data were divided at random into sets of training and validation in a 7:3 ratio. In the training set, LASSO regression, together with multivariate regression analysis, were utilized to identify variables for developing an OA identification model. Subsequently, the identification model was evaluated and validated. This study included 12,989 participants. In each weighted multivariate regression model, a higher eGDR was linked to a decreased prevalence of OA (all P 0.05). Following variable selection, an OA identification model was developed. In the sets of training and validation, this model demonstrated strong discriminating capability, outstanding calibration, and possible net benefits. This investigation identified a negative association between eGDR and OA prevalence. The OA identification model, developed by combining eGDR with other easy-to-obtain variables, demonstrated good discriminating capability, outstanding calibration, and possible net benefits in the sets of training and validation. This model may serve as a simple and practical tool for estimating the probability of OA, informing probability stratification strategies in public health. However, given the cross-sectional design, these findings require confirmation in prospective studies to establish causal relationships and evaluate the model's value for incident OA.
Wang et al. (Wed,) studied this question.