There are uncertainties concerning the optimal threshold value of bone mineral density (BMD) for the diagnosis of osteoporosis using dual-energy X-ray absorptiometry (DXA). The most widely accepted clinical criteria are based on T- and Z-scores, derived statistically from an age-dependent BMD reference range that includes healthy young adults. This study has compared different analyses of BMD data of healthy, nonfractured, adult white females aged 18-90 years, consisting of 953 studied with DXA in the L1-L4 spine and 1018 measured at the hip. Four regression curves were applied to each reference data set; mean, peak BMD (PBMDR) and the age-independent standard deviation (SDR) were calculated. These data, plus derived T-score thresholds, were compared with equivalent values (PBMDYN, SDYN) determined from BMDs of a subgroup of young normals. Choice of regression curve caused a variation of 0.044 g/cm(2) in spinal PBMDR and 0.015-0.022 g/cm(2) in the hip. The variation in calculated SDR between the regressions was minimal (maximum variation, 0.03 SD unit in the spine). A split-linear model was chosen as the best overall fit for both the spine and the hip on the basis of goodness-of-fit and biological plausibility. PBMDR and PBMDYN varied by less than 1% at the spine or hip. Although the difference between SDR and SDYN was less than or equal to7% for the hip sites, SDR was 0.022 g/cm(2) > SDYN for the spine (17% difference). At each hip site, BMD T-score thresholds, derived from regression analysis of the entire reference range (T-R) or from young normals only (T-YN), were not significantly different; however, equivalent spinal values differed by 0.062 g/cm 2 for T = -2.5. The percentage of the Australian female population (AFP) aged 50 yr or older and classified as osteoporotic, using population census data and the split-line-based T-score threshold of T-R = -2.5, was 23% for spinal BMD and 17% for the total hip. For T-YN = -2.5, the values were 36% (spine) and 17% (total hip). The hip BMDs of this study were greater than those of NHANES III in the third, seventh and eighth decades by as much as 0.6 SD. This produced a difference of 0.049-0.077 g/cm(2) for T-YN = -2.5 between the two reference ranges. However, this study is in broad agreement with an Eastern Australian and the U.S. DXA manufacturer's reference range.