Validity of "generalized" equations for body composition analysis in male athletes

Equations by Durnin and Womersley [(D-W), Br. J. Nutr. 32:77, 1974], Jackson and Pollock [(J-P), Br. J. Nutr. 40:497, 1978], and Lohman [(L), Human Biol., 53:181, 1981] for estimating body density (BD) purportedly overcome the problem of specificity by accounting for age and/or the curvilinear relationship between skinfolds (SF) and BD. Their equations were validated on 265 male athletes against percent fat measured by underwater weighing [(UWW); mean +/- SD = 9.2 +/- 4.4%]. Equations by Sloan [(S), J. Appl. Physiol. 23:311, 1967], Katch and McArdle [(K-M), Human. Biol. 45:445, 1973], and Forsyth and Sinning [(F-S), Med. Sci. Sports 5:174, 1973] were included as "linear regression models" to compare to the curvilinear models of J-P, D-W, and L. Differences between UWW and estimated mean values ranged from -1.1 to +5.9%; correlations ranged from 0.58 to 0.85; SEE ranged from +/- 2.41 to +/- 3.61% and total error (E) ranged from 2.38 to 6.97%. The seven D-W equations overestimated mean percent fat by from 3.9 to 5.9%. The K-M, S, and L equations overestimated by 1.3, 0.5, and 1.7%, respectively. The F-S equations overestimated by 2.4 to 3.8%. Of the 21 equations evaluated, only 3 by J-P gave estimates not significantly different from UWW percent fat. Regression analyses of the relationship between UWW (y) and estimated (x) percent fat values from those equations were: y = 1.037x - 0.08 +/- 2.38, E = 2.38, r = 0.84; 0.869x + 1.36 +/- 2.45, E = 2.51, r = 0.83; 1.107x - 1.14 +/- 2.51, E = 2.53, r = 0.82.(ABSTRACT TRUNCATED AT 250 WORDS)