Comparison of two software versions for assessment of body-composition analysis by DXA

Objective: To compare two software versions provided by Lunar Co. for assessment of body composition analysis by DXA. Research Methods and Procedures: Soft-tissue phantoms for lean tissue (water) and fat tissue (methanol) were repeatedly scanned using DXA machines (DPX-L; Lunar Co., Madison, WI) and analyzed using software version 1.33 and the updated year 2000-compatible version 1.35. For the intersoftware comparison, the phantoms were scanned 10 times (each scan was analyzed once) with both software versions using all three scanning modes (slow, medium, and fast) for a total of 60 scans and analyses. For the intermachine comparison, the same phantoms were scanned three times (each scan was analyzed once) with a second machine from the same manufacturer using all three scanning modes and version 1.35 only. Percentage of fat was the variable of interest. Results: For version 1.33, fat was 9.9 +/- 0.4%, 10.0 +/- 0.5%, and 11.0 +/- 0.5% (mean +/- SD) for the lean-tissue phantom and 50.8 +/- 0.3%, 50.9 +/- 0.5%, and 51.1 +/- 0.6% for the fat-tissue phantom using the slow, medium, and fast scanning modes, respectively. For version 1.35, the respective fat values were 9.8 +/- 0.7%, 9.9 +/- 0.4%, and 10.3 +/- 0.7%, and 50.6 +/- 0.5%, 50.9 +/- 0.6%, and 50.8 +/- 0.8%, respectively. For the lean-tissue phantom, the estimation of percentage of fat was significantly (p < 0.05) affected by scanning mode but not by software version. For the fat-tissue phantom, the estimation of percentage of fat was not affected by either scanning mode or software version. The use of version 1.35 did not effect intermachine variability. Discussion: Versions 1.33 and 1.35 of the Lunar body composition software appear to be comparable. Soft-tissue phantoms, such as the ones described in this paper, may be useful in monitoring the reproducibility of body composition analyses within and between DXA machines, particularly in longitudinal studies.