Comparison of abilities of various interpretations of bio-electrical impedance to predict reference method body composition assessment

Body composition was assessed in 28 normal men and women using three established reference methods (deuterium dilution for total body water, and densitometry and dual-energy X-ray absorptiometry for body fat and fat-free mass), the four- and three-component models based on these methods, and eight different predictions of the same whole-body bio-electrical impedance measurement. The comparative value of each prediction was assessed using bias and 95% limits of agreement against all reference methods. The magnitude of bias between each interpretation of whole body impedance and each reference method estimate was found to vary considerably, but in general this was consistent over the range studied. However, with few exceptions, magnitudes of the limits of agreement were considerable. For example, a majority of values of 95% limits of agreement in predicting total body water reached bias±3 kg or above, and all values in predicting fat (% body weight) for all subjects reached bias±6%, and for most purposes these could well be unacceptable. Although in absolute terms bias and limits of agreement for fat-free mass were the same as for body fat, estimates of fat-free mass were relatively more acceptable because fat-free mass was (as is the usual case) the larger of these two body components.