Predicting the Risk of Fracture at Any Site in the Skeleton: Are All Bone Mineral Density Measurement Sites Equally Effective?
- First Online:
- Cite this article as:
- Blake, G.M., Knapp, K.M., Spector, T.D. et al. Calcif Tissue Int (2006) 78: 9. doi:10.1007/s00223-005-0127-3
- 71 Downloads
The ability to assess a patient’s risk of fracture is fundamental to the clinical role of bone densitometry. Fracture discrimination is quantified by the relative risk (RR), defined as the increased risk of fracture for a 1 standard deviation decrease in bone mineral density (BMD). The larger the value of RR, the more effective measurements are at identifying patients at risk of fracture. Epidemiological studies show that RR values for predicting the risk of any fracture are approximately the same for all BMD measurement sites. In this study, we show theoretically that this interesting observation is predictable and a consequence of two related observations: (1) that fracture prediction by BMD measurement sites distant from the fracture site is quantitatively explained by the correlation of BMD measurements and (2) that all correlation coefficients between distant BMD sites are comparable, with values in the range r = 0.55–0.65. The first of these conditions (referred to as the correlation hypothesis) is important because it sets a lower limit on the RR values at distant BMD sites on the assumption that measurements at these sites contain no independent information about fracture risk over and above that provided by their correlation with the fracture site BMD. If the correlation hypothesis is true, the present study points to the importance of the correlation coefficient between BMD sites as a key index that is indicative of the ability of different types of measurement to predict fracture risk. If, on the contrary, the correlation hypothesis is not valid, there is scope to improve bone densitometry by further studies to better identify those measurements that do provide independent information about fracture risk and how best to integrate this information with existing techniques to improve decision making.