Asymmetry of the total body water prediction bias using the impedance index

Our purpose was to prove on a geometric basis that the bias of total body water (TBW) prediction equations based on the impedance index is far greater in fluid overloading than in dehydration. We used formal evaluation of conventional bioimpedance regression equations in both normal and abnormal body fluid status. We plotted the hyperboloid function generated from a standard prediction equation for the TBW over the resistance-reactance (RXc) plane containing the bivariate tolerance intervals (ellipses) of the healthy population. The equation estimated 35 L TBW for the average man (both sexes) of 170 cm height. Leaving the center of the tolerance ellipses, over which the function was relatively flat, the predicted TBW rapidly increased to absurd values for the shorter vectors, indicating fluid overloading (e.g., > 100 L for R < 170 ohm). Migration of the longer impedance vectors beyond the upper pole of 95% tolerance ellipse, which is in the dehydration region, produced less biased estimates of TBW (e.g., < 22 L for the extreme R values > 850 ohm). Different formulas produced TBW prediction bias of the same order. Due to the hyperbolic shape, functions of the impedance index are critically dependent on the region of the RXc plane where they are calculated and they produce misleading results in patients with fluid overload. (C) Elsevier Science Inc. 1997.