In humans, it is unclear how different estimates of energy balance (EB) compare with each other and whether the resulting changes in body weight (bw) and body composition (BC) are predictable and reproducible.
This is a secondary data analysis of effects of sequential 7d over- (OF), 21d under- (UF) and 14d refeeding (RF) on EB. Energy intake (EI) was controlled at +/− 50% of energy needs in a 32 normal weight men (see Am J Clin Nutr. 2015; 102:807–819). EB was calculated (i) directly from the difference between EI and energy expenditure (EE) and (ii) indirectly from changes in BC. Changes in fat mass (FM) were compared with predicted changes according to Hall et al. (Lancet 2011; 378:826–37). Finally, within-subject reproducibility of changes in bw and BC was tested in a subgroup.
There were interindividual and day-by-day variabilities in changes in bw and BC. During OF and RF, the two estimates of EB were similar while with UF the indirect approach underestimated the direct estimate by 10593 ± 7506 kcal/21d (p < 0.001). Considerable differences became evident between measured and predicted changes in FM. Adjusting measured for predicted values did not reduce their interindividual variance. During UF, changes in bw and BC were reproducible, while corresponding changes during OF were not. Conclusion During hypercaloric nutrition the direct estimate of EB corresponded to the indirect estimate whereas this was not true during UF. Changes in bw and BC in response to OF were not reproducible while they were during UF.