Predicting Osteoporotic Fracture in Patients with Early-Stage Diabetic Kidney Disease Using a Radiomic Model: A Longitudinal Cohort Study

OBJECTIVE:There is an urgent need for effective predictive strategies to accurately evaluate the risk of fragility fractures in elderly patients in the early stages of diabetic kidney disease (DKD). METHODS:This longitudinal cohort study included 715 older patients in the early stages of DKD diagnosed between January 2015 and August 2019. Patients were randomly allocated to a training cohort (n = 499) and a validation cohort (n = 216). The least absolute shrinkage and selection operator method was used to select key features for dual-energy X-ray absorptiometry (DXA)-based radiomic analysis. A radiomic model was constructed using Cox proportional hazards regression. The performance of the radiomic model was compared with that of traditional fracture assessment tools through a receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA). RESULTS:Over a mean follow-up period of 4.72 ± 1.60 years, 65 participants (9.09%) experienced incident fragility fractures. Seventeen features were ultimately selected to create the radiomic model. The calibration plots of this model demonstrated satisfactory agreement between the observed and predicted outcomes. Moreover, the radiomic model outperformed traditional fracture assessment tools in both the training and validation cohorts according to the area under the ROC curve and DCA. CONCLUSIONS:The novel radiomic model has demonstrated a more effective prediction of fragility fracture in elderly patients in the early stages of DKD compared to traditional fracture assessment tools.