Statistical significance of body impedance measurements in estimating body composition

Abstract

This study considers the statistical significance and predictive value of measuring the electrical impedance of the body as an indicator of fat-free mass (FFM). Anthropometric and impedance measurements of the body were compared to a reference measurement of hydrodensitometrically determined FFM. A sample of 64 healthy adult male volunteers aged from 17.3 to 54.6 years were used. Multiple linear stepwise regression analysis, with densitometrically determined FFM as the dependent variable, showed weight, age, trunk circumference, some skin-fold thickness measurements and total body impedance to have a strong predictive ability. The regression model developed explained 85% of the total variance when predicting FFM (r = 0.9145, SEE = 0.03). Partial correlation coefficients were calculated to asses the specific relationship between individual variables measured and FFM. The significant partial correlation coefficients ranged from r<inf>part</inf> = 0.80 to r<inf>part</inf> = 0.39, explaining 15% to 64% of the residual variance accounted for by the individual variables. A comparative analysis of different predictive regression equations showed anthropometry to be strongly predictive and bio-electrical impedance analysis (BIA) techniques to be less so. Previously published regression equations were also found to be less related to these data. In conclusion, anthropometry is a valid predictor of FFM, with electrical impedance not adding significant statistical strength. Group and sample specificities are seen, and caution is advised when using BIA to predict body composition. Application of the BIA technique and further validation studies are strongly discouraged until a clearer understanding of the importance of the electrical properties of the body is reached.