Bone mineral content and areal density, but not bone area, predict an incident fracture risk: a comparative study in a UK prospective cohort.
Curtis EM., Harvey NC., D'Angelo S., Cooper CS., Ward KA., Taylor P., Pearson G., Cooper C.
We studied a prospective UK cohort of women aged 20 to 80 years, assessed by dual-energy X-ray absorptiometry (DXA) at baseline. Bone mineral content (BMC) and areal bone mineral density (aBMD), but not bone area (BA), at femoral neck, lumbar spine and the whole body sites were similarly predictive of incident fractures. BACKGROUND: Low aBMD, measured by DXA, is a well-established risk factor for future fracture, but little is known about the performance characteristics of other DXA measures such as BA and BMC in fracture prediction. We therefore investigated the predictive value of BA, BMC and aBMD for incident fracture in a prospective cohort of UK women. METHODS: In this study, 674 women aged 20-80 years, recruited from four GP practices in Southampton, underwent DXA assessment (proximal femur, lumbar spine, total body) between 1991 and 1993. All women were contacted in 1998-1999 with a validated postal questionnaire to collect information on incident fractures and potential confounding factors including medication use. Four hundred forty-three women responded, and all fractures were confirmed by the assessment of images and radiology reports by a research nurse. Cox proportional hazard models were used to explore the risk of incident fracture, and the results are expressed as hazard ratio (HR) per 1 SD decrease in the predictor and 95% CI. Associations were adjusted for age, BMI, alcohol consumption, smoking, HRT, medications and history of fracture. RESULTS: Fifty-five women (12%) reported a fracture. In fully adjusted models, femoral neck BMC and aBMD were similarly predictive of incident fracture. Femoral neck BMC: HR/SD = 1.64 (95%CI: 1.19, 2.26; p = 0.002); femoral neck aBMD: HR/SD = 1.76 (95%CI: 1.19, 2.60; p = 0.005). In contrast, femoral neck BA was not associated with incident fracture, HR/SD = 1.15 (95%CI: 0.88, 1.50; p = 0.32). Similar results were found with bone indices at the lumbar spine and the whole body. CONCLUSIONS: In conclusion, BMC and aBMD appear to predict incident fracture with similar HR/SD, even after adjustment for body size. In contrast, BA only weakly predicted the future fracture. These findings support the use of DXA aBMD in fracture risk assessment, but also suggest that factors which specifically influence BMC will have a relevance to the risk of the incident fracture.