Combining shape and intensity dxa-based statistical approaches for osteoporotic HIP fracture risk assessment

Aiming to improve osteoporotic hip fracture risk detection, factors other than the largely adopted Bone Mineral Density (BMD) have been investigated as potential risk predictors. In particular Hip Structural Analysis (HSA)-derived parameters accounting for femur geometry, extracted from Dual-energy X-ray Absorptiometry (DXA) images, have been largely considered as geometric risk factors. However, HSA-derived parameters represent discrete and cross-correlated quantities, unable to describe proximal femur geometry as a whole and tightly related to BMD. Focusing on a post-menopausal cohort (N = 28), in this study statistical models of bone shape and BMD distribution have been developed to investigate their possible role in fracture risk. Due to unavailable retrospective patient-specific fracture risk information, here a surrogate fracture risk based on 3D computer simulations has been employed for the statistical framework construction. When considered separately, BMD distribution performed better than shape in explaining the surrogate fracture risk variability for the analysed cohort. However, the combination of BMD and femur shape quantities in a unique statistical model yielded better results. In detail, the first shape-intensity combined mode identified using a Partial Least Square (PLS) algorithm was able to explain 70% of the surrogate fracture risk variability, thus suggesting that a more effective patients stratification can be obtained applying a shape-intensity combination approach, compared to T-score. The findings of this study strongly advocate future research on the role of a combined shape-BMD statistical framework in fracture risk determination.