Clinical application of techniques for assessing bone mineral density (BMD) requires accurate and precise measurements that can be related to clearly defined normal ranges. In this study we investigated the clinical interpretation of BMD values in a group of individuals measured on the same day with two different dual-energy x-ray densitometers (Lunar DPX and Hologic QDR 1000). The BMD results were analyzed as absolute values in g/cm2 and with respect to young and age-specific normals as defined by each manufacturer. Absolute BMD values measured by the two instruments were highly correlated (lumbar spine r = 0.98, femoral neck r = 0.95; p less than 0.0001). In the lumbar spine, the two instruments assigned almost identical values when expressed as a percentage of age-matched values and as a percentage of young normals, despite a small but systematic difference between the values assigned for the latter index. In the femoral neck, however, there were significant differences in assignments between instruments, expressed both as a percentage of young normal (mean difference 6.2%) and with respect to age-matched values (mean difference 3.3%). In particular, in premenopausal subjects femoral neck values with the Hologic instrument were assigned significantly lower values. This study shows effective comparability between these two instruments for absolute and relative values for the lumbar spine, as well as for absolute values at the femoral neck, but important differences for normality assignments at the femoral neck. These latter differences may produce bias in the "diagnosis" of femoral neck osteoporosis and may have important implications for clinical decision making.(ABSTRACT TRUNCATED AT 250 WORDS)