The influence of fat free mass on prediction of densitometric body composition by bioelectrical impedance analysis and by anthropometry.

Objective: To examine the effect of varying size of fat free mass (FFM) on the precision and bias of body composition prediction by bioelectrical impedance analysis (BIA) from four equations of Segal et al (BIA-Segal), Gray et al (BIA-Gray, Lukaski et al (BIA-Lukaski) and those from a manufacturer (BIA-EZC), by body mass index (BMI), and by skinfold methods with reference to estimation by densitometry.

Subjects: 73 men and 77 women aged 17-71 years, were measured for height, weight, FFM, bioelectrical impedance and age.

Results: BIA-Segal gave the highest precision (men: R2 = 83%, women: R2 = 75%) and the least bias (men: slope = 0.88; women: slope = 0.81) of all BIA methods. There were poorer precision (R2 < or = 50%) and more bias (slope < 0.70) by BIA-Lukaski and BIA-EZC in both sexes, which were comparable to the simpler BMI method. The skinfold method gave R2 = 83% and slope = 0.84 in men and R2 = 61% and slope = 0.86 in women. Bland and Altman analysis showed BIA-Segal gave prediction of FFM within +/-6 kg of 95% confidence interval limit of agreement of that estimated by UWW in most subjects. Other BIA methods presented unacceptably large underestimates of up to 15-17 kg in FFM.

Conclusions: The BIA-Segal provide the best predictions of the methods tested, but using BMI-or waist-specific equations may be more practical than the original BIA-Segal method, which BIA methods are affected by large FFM, and not better than anthropometric methods in predicting FFM.