Experimental validation of DXA-based finite element models for prediction of femoral strength

Osteoporotic fractures are a major clinical problem and current diagnostic tools have an accuracy of only 50%. The aim of this study was to validate dual energy X-rays absorptiometry (DXA)-based finite element (FE) models to predict femoral strength in two loading configurations.Thirty-six pairs of fresh frozen human proximal femora were scanned with DXA and quantitative computed tomography (QCT). For each pair one femur was tested until failure in a one-legged standing configuration (STANCE) and one by replicating the position of the femur in a fall onto the greater trochanter (SIDE). Subject-specific 2D DXA-based linear FE models and 3D QCT-based nonlinear FE models were generated for each specimen and used to predict the measured femoral strength. The outcomes of the models were compared to standard DXA-based areal bone mineral density (aBMD) measurements.For the STANCE configuration the DXA-based FE models (R²=0.74, SEE=1473N) outperformed the best densitometric predictor (Neck_aBMD, R²=0.66, SEE=1687N) but not the QCT-based FE models (R²=0.80, SEE=1314N). For the SIDE configuration both QCT-based FE models (R²=0.85, SEE=455N) and DXA neck aBMD (R²=0.80, SEE=502N) outperformed DXA-based FE models (R²=0.77, SEE=529N). In both configurations the DXA-based FE model provided a good 1:1 agreement with the experimental data (CC=0.87 for SIDE and CC=0.86 for STANCE), with proper optimization of the failure criteria.In conclusion we found that the DXA-based FE models are a good predictor of femoral strength as compared with experimental data ex vivo. However, it remains to be investigated whether this novel approach can provide good predictions of the risk of fracture in vivo.