Osteoporosis is a musculoskeletal disorder characterised by a loss in bone strength leading to incidence of fragility fractures1. The current clinical gold standard in osteoporosis diagnosis and treatment prognosis is based on the areal bone mineral density (aBMD) measured using dual energy X-ray absorptiometry (DXA) scan as a surrogate for bone strength. However, aBMD alone fails to identify up to 50% of patients at high risk of fracture2. Because of the 2D nature of DXA scan, DXA-based analysis methods lack the volumetric density distribution, crucial to estimating bone strength and improving fracture prevention.

3D patient-specific finite element (FE) models based on quantitative computed tomography (QCT), have been shown to better estimate bone strength3, 4, 5, 6 and fracture risk7, 8, 9, 10, 11, 12, 13, compared to DXA-based aBMD. However, most of these studies have been carried out on small clinical cohorts. This is mostly due to a need for QCT, which results in increased X-rays radiation dose to the patients, high cost of CT imaging and long processing time for QCT based FE modelling pipeline. These factors prevent QCT based FE from being a clinically cost-effective tool in osteoporosis diagnosis and management. Thus, there is a need for a robust, clinically cost-effective tool for improving osteoporosis diagnosis and management.

To overcome the respective limitations of DXA and QCT, 3D modelling methods were proposed, to estimate the 3D shape and density distribution of bones from a limited number of DXA scans14, 15, 16, 17. Briefly, 3D-DXA methods use a 3D statistical shape and density model of the proximal femur built from a database of QCT scans. The statistical model is registered onto the DXA scan to obtain a patient-specific 3D model of the proximal femur (3D geometry and bone density distribution). 3D-DXA methods were validated against QCT for measuring cortical and trabecular volumetric bone mineral density (vBMD) and geometrical parameters14. 3D-DXA cortical and trabecular vBMD measurements showed superior efficacy over DXA measurements in monitoring different treatment strategies18 and estimating fracture risk19.

Recently, 3D-DXA methods have been combined with FE analyses to estimate the risk of fragility fracture. Studies used ex vivo femurs to compare 3D-DXA against QCT-based FE models20,21. Steiner et al22 used either single or two orthogonal simulated DXA projections to develop 3D FE models that were validated against experimental results. The two orthogonal projections overperformed the single projection method, in predicting femur strength22. However, two orthogonal DXA scans of femur are not routinely available in the clinical setup, where the standard protocol consists in an anteroposterior (AP) projection. Ruiz Wills et al23 used DXA-based patient-specific FE models and simulations, and they provided biomechanical descriptors able to discriminate fracture cases better than BMD features, in a case-control study. However, they did not assess femur strength. As a crucial step in this direction, the accuracy of femur strength based on routine AP DXA scans and 3D-DXA methods, still needs to be explored against QCT, by using in vivo data.

Accordingly, this study aimed to develop and evaluate 3D-DXA-FE: a patient-specific FE modelling pipeline to estimate femur strength using 3D-DXA methods and a routine AP DXA scan. The femur strength values based on the 3D-DXA-FE pipeline were compared against the gold standard technique for non-invasive femur strength estimation, i.e., QCT-based FE analysis.