Oral Presentation Annual Meetings of the Endocrine Society of Australia and Society for Reproductive Biology and Australia and New Zealand Bone and Mineral Society 2016

Predicting hip fracture: the role of structural data from DXA in addition to aBMD in the FRAX prediction model (#147)

Ben CC Khoo 1 2 , MingXiang Yu 3 4 , Joshua Lewis 2 5 6 , Keenan Brown 7 , Richard Prince 2
  1. Sir Charles Gairdner Hospital, Nedlands, WA, Australia
  2. School of Medicine and Pharmacology, University of Western Australia, Perth, WA, Australia
  3. Department of Endocrinology & Metabolism, Zhongshan Hospital, Shanghai, China
  4. Department of Endocrinology & Metabolism, Zhongshan Hospital, Shanghai, China
  5. Centre for Kidney Research, Children's Hospital at Westmead, Sydney, NSW, Australia
  6. School of Public Health, The University of Sydney, Dydney, NSW, Australia
  7. Mindways Software , Austin, Texas, USA


Current FRAX risk fracture prediction algorithms use age, height, weight, alcohol intake, disease, smoking, personal and family fracture histories with or without femoral neck aBMD(FN_aBMD). In Australasian populations these models demonstrate moderate discrimination, suggesting room for improvement. We recently evaluated a new bone mass distribution computed from hip structural analysis variables entitled Sigma_Trochanter(Sigma_TR) that evaluates bone distribution in a trochanter section by demonstrating improved hip fracture prediction independently and additively to age and total hip aBMD(1). Here, we investigated if this model incorporating Age, TH_aBMD and Sigma_TR (clinical model) improves on FRAX in prediction of future fracture in elderly Australian women.



Test cohort was the Longitudinal Study of Aging Women cohort http://www.lsaw.com.au/  consisting of 1,159 elderly women initially enrolled in 5-year RCT of calcium supplementation, mean baseline age 75±3 years, who had ascertainment of hip fracture hospitalisation over 10 years (83 hip fractures) in which Sigma_TR was calculated. FRAX_AU was calculated from FRAX website using baseline LSAW data from 1998/9. All models were adjusted for randomisation to calcium or placebo with ROC analysis used to compare the three models.



FRAX_AU data demonstrated reasonable predictive power for hip fracture which improved with incorporation of FN_aBMD (AUC:0.59 vs 0.64). To assess importance of Sigma_TR in addition to FRAX-FN_aBMD in predicting 10-year probability of hip fracture both were combined and compared to actual data leading to an improvement to c-statistic to +0.07, P=0.016 and a category-based (5 and 10%) net reclassification improvement of 0.38 (15% correctly up and 23% correctly down in risk), P



These findings suggest that fracture risk based on bone structural variables provides improved fracture prediction compared to FRAX models. These findings need to be validated in other cohorts.



Khoo et al. Osteoporos Int. 2016;27(1):241-8.