T-score-based diagnosis of osteoporosis might lead to diagnostic misclassification when using multiple-site bone mineral density (BMD) measurements. To compare the diagnostic concordance of T-score-based diagnosis of osteoporosis among different skeletal sites and its correlation to osteoporotic fracture, we studied 1200 postmenopausal women with (441) and without (759) fragility fracture after measuring BMD at the femoral neck, Ward's triangle, trochanter, and spine. Agreement rates of T-score-based diagnosis of osteoporosis were statistically different between pairs of measurements taken at different skeletal sites (McNemar test, p < 0.001). Fragility fractures poorly matched T-score-based diagnosis of osteoporosis (Cohen and Younden indexes <0.4). Technique inaccuracies support these discrepancies as also shown by the large range of T-score values (from -2 to -3) with similar abilities to predict fractures by ROC curve area comparison. Concordance rates between T-score and fragility fracture diagnosis of osteoporosis (marginal homogeneity test, p < 0.001) were also different across the various measurement sites. Our data show that the T-score leads to diagnostic inconsistencies among different skeletal sites and low concordance with fragility fracture based diagnosis of osteoporosis. Integration of the T-score with multiple risk assessment from clinical sources should be tested to better diagnose osteoporosis and related fracture risk.