Discrepancies in normative data between Lunar and Hologic DXA systems.

Abstract

Many studies have shown the high correlation between Lunar and Hologic DXA bone mineral density (BMD) measurements despite differences in absolute calibration. However, in clinical practice, raw BMD values (in g/cm2) are not normally used for assessing skeletal status and fracture risk. Instead, the BMD values are expressed in terms of the number of standard deviations above or below the young normal value (commonly referred to as the T-score). If the normative populations of the various systems are consistent, the standard deviation scores should also be consistent. For this reason, the World Health Organization (WHO) recently established diagnostic criteria for osteoporosis based on T-scores and not BMD. However, few studies have compared the instruments in terms of their standard deviation scores. In this study, we used linear regression to compare T-scores in 83 women at L1-4 and 120 women at the femoral neck obtained on a Lunar DPX and a Hologic QDR-1000/W system. patient BMD and T-score measurements were highly correlated between the two systems (r > 0.95). No clinically significant difference in L1-4 T-scores was seen (less than 0.1 SD). However, linear regression analysis confirmed a systematic difference of 0.9 SD between the femoral neck T-scores. This discrepancy is caused by: (1) differences in the normal populations, and (2) differences in statistical models used to determine the young normal mean and standard deviation. In an attempt to correct the discrepancy, the female young normal mean and standard deviation were recalculated for the femoral neck using published epidemiological data from NHANES and existing DXA cross-calibration equations. The Hologic young normal value (mean +/- SD) was redefined as 0.85 +/- 0.11 g/cm2, while the Lunar value was redefined as 1.00 +/- 0.11 g/cm2. When the femoral neck T-scores for the study population were recalculated on the basis of these new values, the results were equivalent between manufactures, effectively eliminating the discrepancy. However, the revised values should be confirmed by additional measurements in young normal adults.