Body fat percentage via dual energy X-ray absorptiometry following multiple different approaches compared to a laboratory-based 3-compartment model

Abstract

The purpose of this study was to examine the accuracy of using dual energy X-ray absorptiometry (DXA) derived body volume (BV) equations in a multi-compartment model for estimates of body fat percentage (BF%) in comparison to traditional under water weighing (UWW) measures of BV. BF% was estimated using three-compartment (3C), two-compartment (2C) models, and the DXA. The 3C-Siri equation with UWW for BV and deuterium oxide (D2O) for total body water (TBW) was used as the criterion. One hundred twenty-nine adults (55 men and 74 women) volunteered to participate (age = 30 ± 13 years). DXA-derived BV was determined with the recent equations from Smith-Ryan et al. and Wilson et al. and then incorporated into multi-compartment models (i.e., 3CSiri-SR and 3CSiri-W). The 3CSiri-SR and 3CSiri-W DXA-derived BV values were highly correlated with UWW measured BV (74.24L, 74.30L, and 71.97L, respectively, and r=.999 for both). However, the mean BF% was overestimated in all multi-compartment models regardless of which DXA-derived BV equation was used. These results were consistent for the total sample and when stratified by sex, with the observed error ranging from 4.92% to 17.75% (effect size [ES] = .61 to 1.96, all p<.001). The correlation between the DXA-derived BV and 3C-criterion BF% was strongest for both Smith-Ryan et al. and Wilson et al. when utilized in the Siri model (i.e., 3CSiri-SR and 3CSiri-W) in the total sample, (r= .979 and .964, respectively) for men, (r= .974 and .971, respectively) and for women (r= .981 and .973, respectively). The 3CSiri-SR yielded the best accuracy in the total sample, as well as when stratified by men, and women as indicated by the smallest SEE of all methods (1.91%, 1.83%, and 1.76% respectively), although it overestimated BF% by 6% in both sex-specific subgroups. These data indicate that both DXA-derived BV equations are strongly correlated with UWW, however do not provide an accurate measure when incorporated in a 3-compartment model for estimation of BF%. This is likely due to the higher BV values produced by the DXA-derived equations (roughly 2 liters), which overestimated BF% by roughly 5%.