Abstract

The general regression equation of Durnin and Womersley for estimating body density from skinfold thicknesses in young men, was examined by comparing the estimated density from this equation, with the measured density of a group of 45 rugby union players of similar age. Body density was measured by hydrostatic weighing with simultaneous measurement of residual volume. Additional measurements included stature, body mass and skinfold thicknesses at the biceps, triceps, subscapular and suprailiac sites. The estimated density was significantly different from the measured density (P < 0.001), equivalent to a mean overestimation of relative fat of approximately 4%. A new set of prediction equations for estimating density was formulated from linear regression using the logarithm of single and sums of skinfold thicknesses. Equations were derived from a validation sample (n = 22) and tested on a crossvalidation sample (n = 23). The standard error of the estimate (s.e.e.) of the equations ranged from 0.0058 to 0.0062 g ml-1. The derived equations were successfully crossvalidated. Differences between measured and estimated densities were not significant (P > 0.05), total errors ranging from 0.0067 to 0.0092 g ml-1. An exploratory assessment was also made of the effect of fatness and aerobic fitness on the prediction equations. The equations should be applied to players of similar age and playing ability, and for the purpose of identifying group characteristics. Application of the equations to individuals may give rise to errors of between -3.9% to +2.5% total body fat in two-thirds of cases.