Body composition changes by DXA, BIA and skinfolds during exercise training in women

Purpose: Few studies have examined responsiveness of bioimpedance (BIA) to detect changes over time in body composition using a longitudinal design. Accuracy of BIA and skinfold thickness in estimating body composition among 39-64 year-old women was investigated using dual-energy X-ray absorptiometry (DXA) as a criterion method both cross-sectionally and during a training intervention. Methods: 97 women had percentage of fat assessed using DXA, skinfolds and eight-polar BIA using multi-frequency current. Fat mass and lean mass were estimated by DXA and BIA. Measurements were performed before and after the 21-week training intervention. Results: At baseline relative to DXA, BIA under predicted percentage of fat (-6.50 %) and fat mass (-3.42 kg) and overestimated lean mass (3.18 kg) considerably. Also skinfold measurement under predicted percentage of fat compared to DXA, but the difference was smaller (-1.69 % units). Skinfold measurement overestimated percentage of fat at low values and underestimated at high values (r (2) = 0.535). A significant bias was detected between DXA and BIA’s estimate of change in percentage of fat, fat mass and lean mass. Compared to DXA, BIA and skinfolds underestimated the training-induced positive changes in body composition. Conclusions: BIA and skinfold methods compared to DXA are not interchangeable to quantify the percentage of fat, fat mass and lean mass at the cross-sectional design in middle-aged women. Moreover, exercise training-induced small changes in body composition cannot be detected with BIA or skinfold method, even though DXA was able to measure statistically significant within-group changes in body composition after training.