Catch-up growth or regression to the mean? Recovery from stunting revisited

An important question for policy is the extent to which catch-up growth can ease the impact of early stunting. Martorell et al. (1992) showed that stunted Guatemalan infants remain stunted into adulthood, whereas Adair (1999) found appreciable catch-up growth in Filipino children from 2-12 years. Both groups defined catch-up as an inverse correlation between early height and subsequent growth, but Martorell based the correlation on height, whereas Adair used height z scores. The statistical phenomenon of regression to the mean is much like catch-up growth, an inverse correlation between initial height and later height gain. The objective of this study was to reexamine the relationship between stunting and later catch-up growth in the context of regression to the mean. The design was a theoretical analysis showing that catch-up growth is more evident based on height z scores than on height, validated using data on 495 stunted South African children seen at 2 and 5 years of age. The correlation between height at 2 and height change from 2 to 5 was small based on height (-0.11) but large and highly significant based on height z score (-0.58), providing strong evidence of catch-up growth. We argue that catch-up growth should be estimated using height z score not height and that catch-up is present only when the change in z score exceeds that predicted by regression to the mean. This leads to a compact definition of catch-up growth: if z1 and z2 are the initial and final (mean) height z scores, and r is the correlation between them, then catch-up growth for groups or individuals is given by (z2 - rz1).