Association of Birth Weight Centiles and Gestational Age With Cognitive Performance at Age 5 Years

IMPORTANCE Birth weight percentiles (BWPs) are often dichotomized at the 10th percentile and show statistically significant association with later cognitive performance, for both preterm and term-born children. However, research testing nonlinear associations between BWPs and cognitive performance is scarce. OBJECTIVE To investigate culturally invariant, nonlinear associations of BWPs and gestational age with later cognitive performance. DESIGN, SETTING, AND PARTICIPANTS In this cohort study, participants with valid neonatal and cognitive data were combined from 4 observational cohorts, including the Millennium Cohort Study, the National Longitudinal Survey of Youth 1979 Child and Young Adult cohort, Growing Up in Ireland, and the Longitudinal Study of Australian Children, with children born between 2000 and 2002, 1980 and 2010, 2007 and 2008, and 2003 and 2004, respectively. Neonatal data were parent reported before age 1 year. At approximately 5 years of age, multiple cognitive tests were performed. Follow-up at 5 years of age was the predominant focus. Data were analyzed July 17, 2023. EXPOSURE The parent-reported neonatal data were used to calculate BWPs according to the Fenton growth chart. MAIN OUTCOME AND MEASURE Scores for IQ were created from multiple measures of cognition, which were z standardized separately within each cohort. RESULTS Of 30 643 participants (50.8% male), 7.5% were born preterm (before 37 weeks gestation) and 92.5% were term born (between 37 and 42weeks gestation). In the pooled data using multivariate adaptive regression splines, IQ linearly increased by 4.2 points as BWPs increased from the first to the 69th percentile before completely plateauing. For gestational age, IQ linearly increased by 1.3 points per week up until 32 weeks, with the association reducing to 0.3 points per week after 32 weeks. The association of BWP with IQ was not moderated by gestational age. For term-born infants, the estimated IQ score was only clinically meaningfully lower than average when birth weight was below the third percentile. Consistent results were found when instead using multivariable regression where gestational age and BWPs were categorized into groups. CONCLUSIONS AND RELEVANCE In this cohort study, lower BWPs and gestational age were independently associated with lower IQ. For term-born infants, a cutoff of the third percentile would be more appropriate than the traditionally used 10th percentile when the aim is estimating meaningful cognitive differences.