A comparison of three methods to assess body composition.
OBJECTIVE: The aim of this study was to compare the accuracy of measurements of body composition made using dual x-ray absorptiometry (DXA), analysis of computed tomography (CT) scans at the L3 vertebral level, and bioelectrical impedance analysis (BIA). METHODS: DXA, CT, and BIA were performed in 47 patients recruited from two clinical trials investigating metabolic changes associated with major abdominal surgery or neoadjuvant chemotherapy for esophagogastric cancer. DXA was performed the week before surgery and before and after commencement of neoadjuvant chemotherapy. BIA was performed at the same time points and used with standard equations to calculate fat-free mass (FFM). Analysis of CT scans performed within 3 mo of the study was used to estimate FFM and fat mass (FM). RESULTS: There was good correlation between FM on DXA and CT (r2 = 0.6632; P < 0.0001) and FFM on DXA and CT (r2 = 0.7634; P < 0.0001), as well as FFM on DXA and BIA (r2 = 0.6275; P < 0.0001). Correlation between FFM on CT and BIA also was significant (r2 = 0.2742; P < 0.0001). On Bland-Altman analysis, average bias for FM on DXA and CT was 0.2564 with 95% limits of agreement (LOA) of -9.451 to 9.964. For FFM on DXA and CT, average bias was -0.1477, with LOA of -8.621 to 8.325. For FFM on DXA and BIA, average bias was -3.792, with LOA of -15.52 to 7.936. For FFM on CT and BIA, average bias was -2.661, with LOA of -22.71 to 17.39. CONCLUSION: Although a systematic error underestimating FFM was demonstrated with BIA, it may be a useful modality to quantify body composition in the clinical situation.