Construction and internal validation of a risk prediction model for rapid bone loss during aromatase inhibitor therapy.

PURPOSE: To establish and validate a risk prediction model for identifying hormone receptor-positive breast cancer patients at high risk of aromatase inhibitor-associated rapid bone loss. METHODS: This single-center retrospective study enrolled 247 patients receiving aromatase inhibitor (AI) therapy between 2016 and 2025. Rapid bone loss, defined as an annual bone mineral density (BMD) loss rate greater than 4% at either the lumbar spine or total hip, was the primary outcome. Predictor selection was conducted using the least absolute shrinkage and selection operator (LASSO) regression. A multivariable logistic regression model was constructed to establish the predictive nomogram. Model performance was evaluated comprehensively using the area under the receiver operating characteristic curve (AUC) for discrimination, bootstrap calibration plots (1,000 resamples) with the Hosmer-Lemeshow (H-L) test for calibration, and decision curve analysis (DCA) for clinical utility. RESULTS: Seven predictive indicators were incorporated into the final model: baseline hip BMD, body mass index (BMI), history of diabetes, high platelet-to-lymphocyte ratio (PLR), AI exposure before baseline DXA, red blood cell (RBC) count and follow-up interval. The nomogram shows acceptable discrimination after internal validation, with an optimism-corrected AUC of 0.709 (bootstrap, 1,000 resamples); the apparent AUC was 0.735 (95% CI: 0.672-0.798; sensitivity: 71.2%; specificity: 65.4%). Calibration was acceptable, with a bias-corrected calibration slope of 0.835 and a bias-corrected Brier score of 0.219; the apparent Hosmer-Lemeshow test was non-significant (P = 0.833). The DCA indicates potential clinical benefit within a wide range of threshold probabilities. CONCLUSION: This single-center study developed an exploratory prediction model that, after optimism correction, showed modest discrimination (corrected AUC 0.709) and may serve as a preliminary screening framework to identify postmenopausal patients at high risk of rapid bone loss related to early AI therapy. However, extensive prospective, multi-center external validation is strictly required before this model can be implemented to guide clinical decision-making.