Validity and intertester error of anthropometric estimations of body density

The purposes of this study were to evaluate the intertester error in anthropometry and its effect on estimations of actual body density (ABD), as well as to assess the validity of predicted body density (PBD) from selected anthropometric equation. Sixteen adult males X +/- SD = 21.98 +/- 1.65 years) volunteered to be measured by three experienced and reliable testers at eight circumference, two diameter, and eight skinfold sites as well as by underwater weighing. Eighteen commonly used anthropometric equations which predict body density were selected for evaluation. Repeated measures ANOVA with Tukey post-hoc comparisons were used to determine differences between testers for each anthropometric measure and the PBD values, as well as between the PBD values and ABD. There were significant (p less than 0.01) differences between testers for all anthropometric measures except bi-iliac diameter and hips circumference. However, the magnitude of the differences between testers were considered to be quite small and of no practical concern. There were significant differences among the PBD values and ABD for all equations except the base 10 logarithmic transformations of Durnin and Womersley. Seven of the equations, however, resulted in a constant error of less than or equal to 0.0085 g.ml-1. Overall, there was no single anthropometric equation which satisfied all of the cross-validation criteria suggested by previous investigators, however the linear equation of Forsyth and Sinning which included the scapular and abdominal skinfolds, as well as the bitrochanteric diameter resulted in an extremely small constant error (0.0024 g.ml-1) and an appropriate standard deviation of the PBD values (0.0127 g.ml-1).