Structural determinants of hip fracture in elderly women: re-analysis of the data from the EPIDOS study

Hip fracture is the most disastrous osteoporotic fracture, characterized by high mortality, morbidity and institutionalization for the patient and by high economic costs for the health care system. The morphology of the upper part of the femur can influence the risk of hip fracture, e.g., a longer femoral neck is associated with a higher risk of cervical fractures, but not trochanteric ones. In this study, we evaluated the prediction of hip fracture risk by morphological parameters estimated from DXA measurements, and we compared their predictive value for cervical and trochanteric fractures in elderly women by reanalyzing previously published data (Duboeuf et al. J Bone Miner Res 1997 12 1895). This nested case-control study was performed in 232 elderly community-dwelling women from the EPIDOS cohort, including 65 women who sustained a hip fracture. After adjustment for confounding variables, women who sustained a cervical fracture had lower areal bone mineral density (aBMD), lower cortical thickness and a higher average buckling ratio (P < 0.005 for all) as well as longer femoral neck (P < 0.01) than controls. Women who sustained a trochanteric fracture had lower aBMD, lower cortical thickness and higher buckling ratio than controls (P < 0.0001) and than women who sustained a cervical fracture (P < 0.05). Their bending resistance (cross-sectional moment of inertia-CSMI, section modulus) was significantly lower in comparison with controls (P < 0.05-0.001). A decrease in aBMD, cortical thickness, CSMI and section modulus as well as an increase in buckling ratio were predictive of all hip fractures (OR -1.42-2.46 per 1 SD, P < 0.05-0.0001), but the ORs for all structural parameters were markedly higher for trochanteric than for cervical fractures. CSMI and section modulus were predictive of trochanteric, but not cervical fractures. However, aBMD was strongly correlated with the CSA, cortical thickness and buckling ratio ( r(2) > 0.74), which suggests that they convey the same information. CSMI and section modulus correlated with aBMD more weakly, but their OR lost statistical significance after adjustment for aBMD. In conclusion, low femoral neck aBMD, CSA and cortical thickness as well as a high buckling ratio are associated with the higher risk of hip fracture, especially trochanteric ones. These indices are highly correlated with aBMD and convey the same message. The calculated CSMI and section modulus predict trochanteric fractures, but not cervical fractures, and their statistical significance is lost after adjustment for aBMD, indicating that they reflect mainly aBMD, not mechanical properties. Thus, the independent contribution of the external diameter of the femoral neck to the risk of hip fracture cannot be reliably estimated by this technique.