Accurate prediction of children's target height from their mid-parental height

BackgroundFor the past 50 years, standard guidelines have recommended the use of sex-adjusted mid-parental height to predict a childs final height. Here, we studied the accuracy of this procedure. MethodsWe used height data in a cohort of 23 very large nuclear families (Mean = 11 adult children per family). We compared the actual final height of the children to their height predicted by the standard procedure, as well as to alternative height predictions that incorporate corrections of mid-parental height for age, sex, and regression to the mean. ResultsStandard mid-parental height explained 36% of the variance in childrens heights, with a heritability of 74%, and children were on average 2.7 cm taller than predicted by their target heights. When we introduced a non-linear correction for the age of the parents, employed a multiplicative (rather than additive) correction for sex, and accounted for regression to the mean, the variance explained increased to 40%, heritability increased to 80%, and prediction bias was reduced from 2.7 cm to 0.14 cm (representing an improvement in prediction by half a standard deviation of the height distribution). We further measured the empirical distribution of heights of adult children around their predicted height. We describe how this distribution can be used to estimate the probability that a childs height is within the normal expected range. Conclusions and RelevanceBased on these observations, we propose an improved method for predicting childrens target heights. Our procedure for determining whether the deviation of a childs projected height from the target height is in the normal range can be used to assess whether the child should be tested further for potential medical abnormalities.