Cross-sectional association between predicted skeletal muscle mass index and osteoporosis in hospitalized patients with type 2 diabetes mellitus

To investigate the cross-sectional association between predicted skeletal muscle mass index (pSMI) and prevalent, clinically recorded osteoporosis in patients with type 2 diabetes mellitus (T2DM), and to evaluate the clinical value of pSMI for identifying osteoporosis in the T2DM population. A total of 12,187 hospitalized patients with type 2 diabetes mellitus (T2DM) were enrolled from Chengdu Integrated TCM & Western Medicine Hospital between 1 January 2018 and 31 December 2025. Osteoporosis was identified through electronic medical records (EMR) or the International Classification of Diseases, Tenth Revision (ICD-10) codes rather than by systematic dual-energy X-ray absorptiometry (DXA) assessment. Participants were divided into three groups according to tertiles of pSMI (T1: 2.76–6.81; T2: 6.82–7.94; T3: 7.95–15.45). Multivariate logistic regression models were used to analyze the association between pSMI and osteoporosis, and restricted cubic spline (RCS) models were applied to evaluate the dose-response relationship. Furthermore, receiver operating characteristic (ROC) curve analysis and SHapley Additive exPlanations (SHAP) analysis were used in exploratory analyses to assess the contribution of pSMI to the osteoporosis prediction model. Among the 12,187 participants included in the analysis, 718 (5.9%) had clinically recorded osteoporosis. Multivariate logistic regression analysis showed that each unit increase in pSMI was associated with lower odds of documented osteoporosis in the crude model (OR = 0.44, 95% CI: 0.40–0.48, P < 0.001), and this inverse association remained significant after full adjustment (Model 3: OR = 0.60, 95% CI: 0.55–0.66, P < 0.001). Compared with the T1 group, the T3 group showed lower odds of documented osteoporosis (OR = 0.23, 95% CI: 0.17–0.32, P < 0.001). RCS analysis revealed a non-linear dose-response relationship between pSMI and documented osteoporosis ( P for non-linearity < 0.001). Subgroup analyses indicated significant effect modification by sex ( P for interaction = 0.002), BMI ( P = 0.003), and hypertension ( P = 0.02). In exploratory ROC curve analysis, adding pSMI to the baseline model was associated with a modest apparent improvement in discrimination for identifying documented osteoporosis in T2DM patients (AUC: 0.7705 to 0.7926; P < 0.001), and the optimal pSMI cut-off value was 7.22 (sensitivity: 80.4%, specificity: 57.1%). SHAP analysis identified pSMI as the second most important predictor after age. In patients with T2DM, higher pSMI is associated with a lower prevalence of clinically recorded osteoporosis. Incorporating pSMI into traditional risk prediction models yielded a modest apparent improvement in discrimination. Not applicable.