Fracture risk as determined by prospective and retrospective study designs

Both retrospectively and prospectively designed studies consistently show low bone mass and/or bone mineral content (BMC) to be risk factor for low-trauma fractures in postmenopausal women. Along with the reports of such studies there has been concern expressed that BMC measurements overlap between fracture groups, i.e., some women with high BMC develop fractures and some women with low BMC do not. In these commonly used epidemiologic study designs, BMC does not discriminate between those who have and have not experienced the untoward event at some level of the exposure factor. The ability to discriminate is more properly determined by the sensitivity and specificity of the measured value. To contrast the concepts of risk and sensitivity, a nested case-control study was conducted within a 24-year cohort study of women at risk for osteoporosis. We found that for each 1.0 decrement of BMC z-scores, the adjusted relative risk for the prospective study design was 1.67, while the odds ratio obtained from the most recent BMC z-score measurements was 1.87. A receiver operating characteristic (ROC) curve, calculated from the nested case-control study data, showed that BMC z-scores, measured after low-trauma fracture, have both low sensitivity and low specificity to detect existing fracture status.