Estimating total body water in neonates
Editor—The body weight of the preterm babies in the study by Tanget al predicted the total body water better than any other single anthropomorphic measure, and that it accounted for 99% of the variations seen in body water measurements.1
The authors then used a variety of other measurements and anthropomorphic characteristics and combined them in a number of equations and picked the best (involving a constant, body weight, body weight squared, and the length of the foot squared divided by electrical impedance taken under very specific conditions). They claim that by doing this they have produced a potentially useful clinical tool because the equation in their particular group of babies was able to predict not 99%, but 99.5% of the variations in measured body water. This prediction equation has not, so far, been tested in an independent group of babies.
There must be a wide variety of other anthropomorphic characteristics and clinical features—for example, the length of the left eyebrow cubed, divided by the ratio of the lengths of the right palm and index finger—that are both inconvenient to measure and make no effective difference to the total body water prediction based on weight alone. In my view, Tang’s paper suggests that we should assume babies’ total body water to be 0.8 of their body weight, and forget about impedance measurements.
Dr Modi et al respond: We enjoyed reading Dr Coulthard’s amusing letter but fear that his wit might have obscured his understanding. The proportion of body weight that is body water is very variable in newborn infants and changes rapidly after preterm birth, from about 85% at birth in an infant below 32 weeks of gestation, to about 80% at one week, 75% at one month, and 65% at three months.1-2 Dr Coulthard’s suggestion that one simply assumes total body water to be 0.8 of body weight would clearly be unhelpful and indeed nonsensical. He is also mistaken in his belief that the basis of our claim that we have a potentially useful clinical tool is because our regression equation accounts for 99.5% of the variation in measured total body water. Our model incorporating resistance measurements is significantly better than the model using body weight alone (p<0.001) and has a 95% prediction interval of ± 82.5 ml. Although it is self evident that body weight is closely correlated with body water and therefore acceptable as an index of percentage body water in large cross sectional population studies, bioelectrical impedance analysis offers a small but significant improvement to estimates of absolute body water content. This may be of particular clinical value in the assessment of fluctuations in total body water on a day to day basis when body solid mass is also altering rapidly, as in the newborn infant.1-3
