It is not clear how bone mineral density (BMD) measurements from several regions of lumbar spine and proximal femur should be utilized in assessing fracture risk. We examined how well the newest ISCD recommendations differentiate subjects with and without prevalent vertebral fractures in 187 postmenopausal women presenting for routine bone densitometry. The association between T-scores from proximal femur and lumbar spine sites and the probability of having a vertebral fracture was modeled via logistic regression with adjustment for age. The lowest T-score of any hip or spine sites (the current ISCD recommendation) and the proximal femur measurements, particularly the femoral neck and total hip, displayed the strongest association with the probability of vertebral fractures.Subjects with a T-score < -2.5 at multiple hip sites had a higher probability of having a vertebral fracture. The sensitivity and specificity associated with particular T-score cutoff values varied greatly depending on the site of measurement.Consequently, T-score values from different sites that had comparable sensitivity/specificity for detecting the presence of vertebral fractures differed by as much as 1.5 T-score units. This finding implies that a single cutoff value, such as -2.5, might not be clinically acceptable when applied to T-scores from different sites.