In this study we report first the concordance and variation in diagnostic osteoporosis classification using multiple skeletal site measurements compared with the lumbar spine only; and secondly, at the lumbar spine, the variation and diagnostic osteoporosis reclassification using the lowest individual vertebra T-score compared with the L1-L4 mean T-score. One hundred and fifty early postmenopausal women were evaluated as part of the recruitment for a multicenter osteoporosis prevention study. Bone mineral density (BMD) was restricted such that no more than 10% of the subjects had a lumbar spine BMD below 0.8 g/cm2. Forty-seven per cent of the subjects were classified as having low bone mass (T-score < or = -1.0) at the lumbar spine, 63% at the mid-forearm, 39% at the distal forearm and 50% at the hip (p < 0.05). The greatest proportion of subjects were categorized as osteoporotic at the lumbar spine, followed by the forearm and then the hip. Correlation between sites ranged from 0.57 to 0.60 (p < 0.01). Eighty-one percent of the subjects had a significant difference between their highest and lowest individual lumbar vertebra T-score (defined as a difference outside the 90% confidence interval coefficient of variation T-score value). Using the lowest individual lumbar T-score, recategorized 33% of the subjects classified as osteopenic (based on the mean L1-L4 T-score) as osteoporotic, and 23% of those classified as normal as osteopenic (p < 0.05). Of all four vertebrae, L2 had the highest T-score in 37.7% of the subjects (mean -0.3) and L4 the lowest in 61% (mean -1.5) (mean difference 1.2 units, 95% CI 0.7 to 1.7). The classification of osteoporosis varies according to skeletal site, with pronounced differences in the early menopausal population. T-scores are useful for characterizing subjects with the highest risk of osteoporosis but BMD and fracture risk must be recognized in a continuum. Individual T-scores of the lumbar vertebrae show wide variation in the absence of degenerative spinal disease or vertebral collapse and the use of the lowest, significantly different, individual lumbar vertebra T-score reclassified over half of the subjects in this study. This poses a great therapeutic dilemma in clinical practice, particularly if these fractures are at higher risk of future collapse.