A Causally Guided Imaging-Informatics Framework for Ranking DXA-Derived Hip Imaging Phenotypes in Hip Fracture Risk Prediction

Abstract Dual-energy X-ray absorptiometry (DXA) provides multiple hip measurements, but their value for fracture risk stratification is rarely evaluated within an imaging-informatics framework. We developed a causally guided framework to rank DXA-derived hip imaging phenotypes and assess whether ranked phenotypes improve hip fracture prediction. We analyzed 21,098 UK Biobank participants with linked health records, DXA-derived hip measures, and prespecified covariates. Sixteen phenotypes spanning bone mineral content (BMC), bone mineral density (BMD), and T-score across hip-related regions were evaluated. Confounder selection was guided by a prespecified directed acyclic graph (DAG). Backdoor-adjusted average treatment effects (ATEs) were estimated on the absolute risk-difference scale per standard deviation (SD) increase. Effect heterogeneity was evaluated for total femur BMD, and prediction was assessed using clinical variables combined with ATE-ranked phenotypes. Among 21,098 participants, 115 had hip fractures. All 16 phenotypes showed negative backdoor-adjusted ATEs per SD increase. The largest ATEs were observed for total femur BMC and BMD, each with a risk difference of − 0.0047, corresponding to 4.7 fewer hip fractures per 1000 participants per SD higher phenotype value. Conditional effects of total femur BMD were stronger among older participants and those with lower BMI. Clinical variables plus the top 11 ATE-ranked phenotypes achieved higher AUC than FRAX with femoral neck BMD (0.842 vs. 0.709), with higher sensitivity (0.748 vs. 0.443) and similar specificity (0.793 vs. 0.777). DXA-derived hip phenotypes differed in backdoor-adjusted ATEs. This framework may support interpretable phenotype selection for fracture risk stratification.