Associations and potential utility of composite obesity and metabolic indices for osteoporosis risk assessment in Chinese adult men: a cross-sectional study

Abstract

Osteoporosis in men is associated with substantial morbidity and mortality, yet simple tools for preliminary risk assessment remain limited. The aim of this study is to examine the associations and potential utility of composite obesity- and metabolism-related indices for osteoporosis risk assessment in adult men. This cross-sectional study included 702 men aged ≥ 30 years who underwent routine health examinations and dual-energy X-ray absorptiometry-based bone mineral density assessment at a tertiary hospital in China, between July and December 2025. Nine composite obesity- and metabolism-related indices, including triglyceride-glucose body mass index (TyG-BMI), triglyceride-glucose-waist circumference index (TyG-WC), abdominal volume index (AVI), body roundness index (BRI), body adiposity index (BAI), relative fat mass (RFM), waist-to-height ratio (WHtR), fatty liver index (FLI), and Chinese visceral adiposity index (CVAI), were calculated. Multivariate logistic regression, restricted cubic spline (RCS), receiver operating characteristic (ROC) and machine learning analysis were performed to evaluate associations, nonlinear relationships, and predictive performance. The area under the ROC curve (AUC) was used to assess discriminatory ability, and Shapley additive explanations analysis was used to assess variable importance. The prevalence of osteoporosis was 9.7%. After adjustment for potential confounders, TyG-BMI, AVI, BRI, and BAI were positively associated with osteoporosis classification (OR = 1.55, 1.38, 1.33, and 1.32, respectively; all P < 0.05), whereas WHtR showed an inconclusive borderline association ( P = 0.051). Individual indices showed poor discrimination (maximum AUC = 0.591). In the held-out test set, LASSO showed the highest AUC (0.681 [95% CI 0.552–0.806]), followed by ENET (0.644) and LR (0.640); however, the confidence intervals were wide. After training-set out-of-fold Platt calibration, LASSO showed a test-set calibration intercept of -0.12 (95% CI -0.27 to 0.02) and slope of 0.73 (95% CI 0.24 to 1.42). SHAP analysis of the LR model ranked age and TyG-BMI as the leading contributors to model predictions. TyG-BMI, AVI, BRI, and BAI were positively associated with osteoporosis classification after covariate adjustment. Individual indices and internally evaluated multivariable models showed limited-to-moderate discrimination. External validation in larger cohorts using standard axial DXA is required before these measures can be considered for clinical or population screening.