The relative accuracy of skinfolds compared to four-compartment estimates of body composition

Background: Body composition estimates using skinfold thickness are common in field settings and can provide a reasonably accurate measure when more advanced technology is unavailable. However, the observed error between skinfolds and criterion body composition measures may be influenced by the methodology used to derive the criterion measure. Aims: The aim of this study was to examine the relative accuracy of body composition estimates derived from measures of skinfold thickness when compared to four-compartment (4C) models that utilize underwater weighing (UWW) and dual energy X-ray absorptiometry (DXA)-derived body volume (BV). Methods: The sample consisted of adult males (n = 96) and females (n = 91) who were apparently healthy (age = 21.85 +/- 4.82 years; BMI = 24.45 +/- 4.62 kg/m(2)). %Fat was assessed via skinfold using three common equations. BV assessed via UWW and DXA were used to estimate %Fat derived as part of a 4C model. Between group differences were assessed using a repeated measures analysis of variance. Results: %Fat(4c) ranged from 4.7 to 39.7 %Fat (21.9 +/- 8.1 %Fat). Estimated %Fat(SF) using the SF7 Jackson, SF7 Evans, and SF4 Peterson were significantly lower than %Fat(4c) as measured via UWW (-4.8 +/- 3.5 % Fat, -3.3 +/- 3.9 %Fat,-3.1 +/- 6.9 %Fat, respectively, all p <.001). The estimated %Fat(SF) error was lowest when compared to the %Fat(4c) that used the Smith-Ryan DXA-derived BV equation (-0.2 +/- 7.3 to -1.9 +/- 4.6 % Fat) and highest for the Wilson DXA-derived BV equation (-6.5 +/- 7.1 to -8.3 +/- 3.3 %Fat). Conclusions: Skinfold prediction methods can provide reasonable accuracy when estimating %Fat in field settings when more advanced methods are unavailable or undesirable due to increased participant burden. In addition, clinicians and researchers should use caution when selecting a method of estimating body volume via DXA, as the methodology and equation used to derive body volume as part of the 4C model can introduce differences in error. (C) 2019 Elsevier Ltd and European Society for Clinical Nutrition and Metabolism. All rights reserved.