Clinical risk factor evaluation to defer postmenopausal women from bone mineral density measurement: an Italian study.

To identify clinical risk factors (CRFs) best related to low bone mineral density (BMD) and to assess their abilityin deferring women from dual-energy X-ray absorptiometry (DXA), 1187 Italian postmenopausal women, dividedinto development (709) and validation (478) groups, were studied. CRFs were investigated by questionnaire. DXAwas performed at the spine and femoral neck. A T-score < -2.5 at both measurement sites classified those with lowBMD. In the development group, using the logistic regression, the CRFs best predicting low BMD were years sincemenopause, age at menarche, weight, previous fracture, and muscle weakness. The predicted probability of low BMD(PPL-BMD), calculated by the logistic equation, was used to build receiver-operating characteristic (ROC) curves(area = 0.786, standard error [SE] = 0.017) on diagnosed BMD status. The PPL-BMD cutoff below which to deferwomen from DXA was set corresponding to the ROC curve sensitivity of 99%, 98%, and 97%. The coefficients ofthe logistic regression were then used to calculate the PPL-BMD of the validation group. In the validation group, theoverall ability of CRFs to predict low BMD (ROC area = 0.744, SE = 0.023) was not different from that of the developmentgroup. At the PPL-BMD cutoff of 0.132, the percentage of DXA-deferred cases (14.5% vs 19.0%) and low-BMD-missed cases (0.7% vs 1.7%) was similar in both groups. The rates of DXA-deferred women by CRFs arereproducible and this tool should be useful in clinical practice.