Limited utility of clinical indices for the prediction of symptomatic fracture risk in postmenopausal women

Bone mineral density (BMD) is the primary predictor of fracture, and is utilised in the definition of osteoporosis. Mass screening for osteoporosis is, however, currently not recommended. The primary objective of this study was to develop, validate and assess a simple, non-invasive scoring system to identify women at high risk of fracture. Using baseline data of the Dubbo Osteoporosis Epidemiology Study, a sample of 1256 women aged 60 or above was randomly divided into a development cohort (n=846) and a validation cohort (n=410). Low BMD was evaluated by DXA, with respect to 2.0 or 2.5 SD below the mean for young normal women at the femoral neck and lumbar spine. A logistic regression model was used to derive a predictive score, “DOEScore”, in the development cohort, and the performance of this score was then assessed in the validation cohort. Incident fractures over 9395 person-years (median of follow-up duration: 8.4 years) were identified by X-ray records. Approximately 57% and 40% of women (in both cohorts) had T-scores less than −2.0 and greater than −2.5, respectively. Only age, body weight, and previous fracture were significantly related to BMD at both the femoral neck and lumbar spine. These three variables were used in the development of the DOEScore. When applied to the validation cohort, the sensitivity and specificity of DOEScore were 0.82 and 0.52, respectively, for selecting women with T-scores less than −2.5; the area under the receiver operating characteristic (ROC) curve was 0.75. These goodness-of-fit indices were comparable to, or better than, those obtained by the FOSTA, SOFSURF and ORAI score systems. However, neither the DOEScore nor other score systems reliably identify women with incident fractures; for DOEScore, the sensitivity and specificity were 0.52 and 0.49, respectively, with an area under the ROC curve of 0.48. Clinical risk scores can be used to identify women likely to have low BMD (albeit with low specificity), but they are not a useful tool to identify women who will have a fracture.