Endocrine-bone metabolic phenotypes and prevalent osteoporosis in type 2 diabetes: development and internal validation of a risk-classification model

Background Type 2 diabetes mellitus (T2DM) is associated with increased skeletal fragility despite frequently preserved bone mineral density, suggesting substantial metabolic heterogeneity beyond conventional osteoporosis risk factors. Objective To identify endocrine-bone metabolic phenotypes in T2DM and evaluate their associations with prevalent osteoporosis and incremental contribution to an internally validated risk-classification model. Methods This single-center retrospective study included 550 postmenopausal women and men aged ≥50 years with T2DM who underwent dual-energy X-ray absorptiometry. Prevalent osteoporosis was defined by a T-score ≤−2.5 at the lumbar spine, femoral neck, or total hip, or by documented low-trauma hip or vertebral fracture. K-means clustering was performed using glycemic, insulin-resistance, inflammatory, and bone/mineral metabolism variables. Primary phenotype and regression analyses used 537 complete cases. Logistic models, sensitivity analyses, multiple imputation, decision-curve analysis, and 1,000-resample full-pipeline bootstrap validation were performed. Results Three phenotypes were identified: metabolically preserved (n=212), metabolic-inflammatory (n=179), and bone-remodeling imbalance (n=146). Osteoporosis prevalence increased across phenotypes from 20.3% to 43.0% and 71.2% (P for trend<0.001). Compared with the metabolically preserved phenotype, adjusted odds ratios were 1.72 (95% CI 1.03–2.88) and 2.46 (1.34–4.51), respectively. Adding phenotype information to a biomarker-enhanced model increased AUC from 0.847 to 0.884 (DeLong P = 0.018), with continuous NRI 0.32 and IDI 0.11. The final parsimonious model had an optimism-corrected AUC of 0.867. Conclusion Endocrine-bone metabolic phenotypes were strongly associated with prevalent osteoporosis in T2DM and provided incremental predictive information beyond individual biomarkers. The proposed framework is exploratory and requires prospective external validation before clinical implementation.