Validation of near-infrared interactance and skinfold methods for estimating body composition of American Indian women.

PURPOSE This study tested the predictive accuracy of the Jackson et al. skinfold (SKF) equations (sigma7SKF and sigma3SKF), a multi-site near-infrared interactance (NIR) prediction equation, and the Futrex-5000 NMR equation in estimating body composition of American Indian women (N = 151, aged 18-60 yr). METHODS Criterion body density (Db) was obtained from hydrodensitometry at residual lung volume. RESULTS Sigma7SKF significantly underestimated Db (P < 0.05). Sigma3SKF and Heyward's NIR equations significantly overestimated Db (P < 0.05). The Futrex-5000 NIR equation significantly underestimated percent of body fat (%BF) (P < 0.05). Prediction errors for SKF and multi-site NIR exceeded 0.0080 g x cc(-1). The SEE for Futrex-5000 was 5.5%BF. Thus, ethnic-specific SKF and NIR equations were developed. For the SKF model, the sigma3SKF (triceps, axilla, and suprailium) and age explained 67.3% of the variance in Db:Db = 1.06198316 -0.00038496(sigma3SKF) -0.00020362(age). Cross-validation analysis yielded r = 0.88, SEE = 0.0068 g x cc(-1), E = 0.0070 g x cc(-1), and no significant difference between predicted and criterion Db. For the NIR model, the hip circumference, sigma2AdeltaOD2 (biceps and chest), FIT index, age, and height explained 73.9% of the variance in Db:Db = 1.0707606 -0.0009865(hip circumference) -0.0369861(sigma2deltaOD2) + 0.0004167(height) + 0.0000866(FIT index) -0.0001894(age). Cross-validation yielded r = 0.85, SEE = 0.0076 g x cc(-1), E = 0.0079 g x cc(-1), and a small, but significant, difference between predicted and criterion Db. CONCLUSIONS We recommend using the ethnic-specific SKF and NIR equations developed in this study to estimate Db of American Indian women.