Cross-validation of anthropometric and bioelectrical resistance prediction equations for body composition in older people using the 4-compartment model as a criterion method.

Abstract

To cross-validate existing equations in the literature for their accuracy and precision for estimating body fat in older people from anthropometric measures and height2/resistance, using the 4-compartment model as a criteria method, and to propose new practical equations with improved accuracy and precision. Measurement of body composition in a cross-sectional cohort of healthy men and women and comparison by cross-validation techniques against existing prediction equations. The study was performed on subjects admitted to a General Clinical Research Center. The subjects were 41 healthy women (68.2 +/- 6.6 years; 64.1 +/- 10.2 kg) and 41 healthy men (70.2 +/- 7.0 years; 74.9 +/- 11.0 kg). The criteria method for total body composition was the 4-compartment model based on measurement of total body density by underwater weight, total body water by isotope dilution, and total bone mineral from dual energy X-ray absorptiometry. The other techniques examined for accuracy and validity were body fat estimates derived by skinfolds using the Durnin and Womersley Equations; waist circumference and age using the Lean Equations; and bio-electrical resistance using five published equations, including two derived in the older population. When compared with data derived from the 4-compartment model, the skinfold equation of Durnin and Womersley was cross-validated successfully in women but not in men. The Baumgartner equation was the only bioelectrical resistance equation that met the criteria for successful cross-validation in men and women, although in women the intercept (4.0 +/- 2.1 kg) was close to significantly different from zero (P = .06). Error in the estimates of body fat using the Durnin and Womersley and the Baumgartner equations was significantly and inversely related to fat mass (r = -.39 to r = -.56). In our data, the significant predictors of fat mass were hip circumference, calf skinfold, gender, body weight, height2/resistance, and biceps skinfold, explaining 84% of the variance. The Durnin and Womersley equation is accurate for women and the Baumgartner equation is accurate for both men and women if a correction of +4 kg is made in women; however, for both equations the error in the estimate is inversely related to fat mass. We suggest new anthropometric equations for estimating body fat in older people, which may improve accuracy and precision. The new equations will need to be tested in independent cross-validation studies.