Prediction equations €or total body water (TBW) generally use weight and height as predictors, but their ability may be limited because they implicitly assume a constancy of TBW among individuals of similar body size. The objective of this study was to evaluate the relative importance of anthropometric dimensions in predicting TBW and body composition. TBW determined by doubly labeled water (DLW) dilution techniques was used as the frame of reference in 23 healthy Aymara subjects, P 6 5 years, in a rural community of the Bolivian Andes. Predictive performances of anthropometric variables for TBW were examined with multiple regression analyses. The generated equations were tested for cross-validity, using published data for U.S. adults. The resulting errors were compared with those of the published prediction equations of Mellits and Cheek (M&C) and Durnin and Womersley (D&W). The simplified prediction equation using weight and the triceps skinfold (Eq-A1: R2 = 0.989, SEE = 1.041 L) and that using weight and the triceps and subscapular skinfolds (Eq-A2: R2 = 0.990, SEE = 1.020 L) had better R2and smaller SEE than those using any combination of variables, weight, height, age, and sex. In the cross-validation sample, Eq-A1 and Eq-A2 demonstrated higher precision than the D&W and M&C equations. Evaluated by the method of Bland and Altman (mean difference t 2SD), prediction errors for fat mass and fat percent were 0.2 -t 2.8 kg and 0.4 2 5.2% in Eq-AS, 1.1 ? 3.5 kg, and 1.8-C 6.1% in Eq-A1, -2.4 2 3.6 kg and -3.4 2 5.1%) in D&W, and -2.3 ? 7.6 kg and -2.6 ? 10.3% in M&C. Significant underestimation of fat mass and sex differences in the biases were observed with D&W and M&C (P < 0.051, but not with Eq-A2. By including skinfold measurements, a single prediction equation for TBW was valid for males and females across different population samples.

No abstract