Weight centile crossing in infancy: correlations between successive months show evidence of growth feedback and an infant-child growth transition(1)

Background: Early rapid weight gain is associated with later overweight, which implies that weight centile crossing tracks over time. Objective: Centile crossing is defined in terms of the change or deviation in weight z score during 1 mo, and the correlations between successive deviations are explored at different ages. Design: Two Cambridge (United Kingdom) growth cohorts were used: Widdowson (1094 infants born during 1959–1965) and the Cambridge Infant Growth Study (CIGS; 255 infants born during 1984–1987), each with weights measured monthly in the first year. Weights were converted to WHO age- and sex-adjusted z scores, deviations were calculated as the change in z score between adjacent measurement occasions, and the correlations between deviations were studied. Results: In both cohorts, the correlations between successive monthly deviations were positive in the first 6 mo and highest at ages 3–4 mo (r = 0.3, P < 0.0001), whereas after 6 mo they were negative and were lowest at ages 10–11 mo (r = –0.3, P < 0.0001), with the correlation decreasing linearly with age between these extremes. Thus, during the first 6 mo of age, infants crossing centiles in 1 mo tended to continue crossing centiles in the same direction the following month, whereas after 6 mo they tended to cross back again. This represents positive and negative feedback, respectively. At age 12 mo, the correlation was close to zero, which suggests an infant-child transition in growth. Conclusions: The results confirm that weight centile crossing tracks over time, with the correlations between successive periods that change with age suggesting a complex feedback mechanism underlying infant growth. This may throw light on the link between early rapid weight gain and later overweight. Clinically, the correlations indicate that when predicting future weight from current weight, recent centile crossing affects the prediction in an age-dependent manner.