Related Experiment Video
Updated: Sep 25, 2026
Method and Instrumented Fixture for Femoral Fracture Testing in a Sideways Fall-on-the-Hip Position
Published on: August 17, 2017
Evaluation of a simplified hip structure analysis method for the prediction of incident hip fracture events
B C C Khoo1,2,3, J R Lewis4,5, K Brown6
1Medical Technology and Physics, Sir Charles Gairdner Hospital, Nedlands, Western Australia, Australia. ben.khoo@health.wa.gov.au.
Unlabelled:
Many attempts have been made to improve the predictive ability of areal bone mineral density (aBMD) which integrates bone mass and area. The addition of an extra variable derived from the hip dual-energy X-ray (DXA) image TR_σ, which describes distribution of mass within the scanned area of the trochanter, improved prediction of 15-year hip fracture probability in elderly women.
Introduction:
Two-dimensional DXA imaging of the proximal femur to produce an aBMD is a clinically useful predictor of future fracture risk. Further analysis of the DXA image to produce an eight-variable hip structure analysis (Beck HSA) has been developed to improve understanding of structural factors determining hip bone strength at each of three proximal femur sites, the narrow femoral neck (NN), intertrochanter (TR) and shaft (S). Recently, data on four measurements derived from the currently used eight Beck HSA variables were used to capture population variation in bone structure at each site. These include two previously used variables, the localised aBMD and the sub-periosteal width (W) applying to 5-mm sections (at each sites), and two new variables, standard deviation of normalised mineral-mass projection profile distribution (σ), and displacement between centre-of-mineral mass and geometric centre-of-mineral mass of projection profile (δ).
Methods:
Using a cohort of 1159 women, mean baseline age 75, who sustained 139 hip fractures over 15 years, we determined whether these measures significantly improved 15-year hip fracture prediction compared to current approach utilising age and total hip aBMD. To describe the most parsimonious model for hip fracture risk prediction, the 12 base measures (4 from each site), total hip aBMD and age were evaluated in stepwise logistic regression models.
Results:
The final model included TR_σ, total hip aBMD and age and provided improved utility for hip fracture prediction compared to total hip aBMD and age alone (C-statistic 0.73 vs. 0.69, P = 0.009 and net reclassification improvement 0.164, P < 0.001, respectively).
Conclusions:
Addition of TR_σ to total hip aBMD and age substantially improved prediction of 15-year hip fracture risk in this cohort of elderly women.
