Early-life stature, preschool cognitive development, schooling attainment, and cognitive functioning in adulthood: a prospective study in four birth cohorts

Abstract

Background: Nutrition is important for growth and brain development and therefore cognitive ability. Growth faltering in early childhood, an important indicator of early adversity, is associated with poorer developmental outcomes, some into adulthood, but this association probably reflects early-life deprivation. We aimed to investigate the associations between early-life stature, child IQ, and adult IQ.

Methods: In this cohort study, we used prospective longitudinal data collected in four birth cohorts from Brazil (born in 1993), Guatemala (born in 1969-77), the Philippines (born in 1983-84), and South Africa (born in 1990). Using multivariable linear models, we estimated the relative contributions of early-life stature, child IQ, and schooling (highest school year completed) to adult IQ, including interaction effects among the early-childhood measures and schooling.

Findings: We included 2614 individuals in the analysis. Early-life stature was associated with adult IQ (range across eight site-by-sex groups -0·14 to 3·17 IQ points) and schooling (-0·05 to 0·77 years) per height-for-age Z-score. These associations were attenuated when controlling for child IQ (-0·86 to 1·72 for adult IQ and -0·5 to 0·60 for schooling). The association of early-life stature with adult IQ was further attenuated when controlling for schooling (-1·86 to 1·21). Child IQ was associated with adult IQ (range 3·91 to 10·02 points) and schooling (0·25 to 1·30 years) per SD of child IQ in all groups; these associations were unattenuated by the addition of early-life stature to the models. The interaction between schooling and child IQ, but not that between schooling and early-life stature, was positively associated with adult IQ across groups.

Interpretation: The observed associations of early-life stature with adult IQ and schooling varied across cohorts and sexes and explained little variance in adult IQ beyond that explained by child IQ. These findings suggest that interventions targeted at growth for health and early development are important. Our results are consistent with the inference that improving long-term cognitive outcomes might require interventions that more specifically target early cognitive ability.

Funding: Bill & Melinda Gates Foundation.

Figure 1

Conceptual model of the relationship between early life determinant and adult cognitive outcomes We propose that early-life stature and child IQ are correlated and have both shared and specific determinants. Associations between child stature, schooling, and adult IQ are at least partly due to this correlation. Pathways from the early-life determinants to schooling and adult IQ that do not go through early-life stature or child IQ are not shown.

Figure 2

Associations between early-life stature and child IQ with adult IQ and schooling Plotted on the y axis are the unadjusted and adjusted β coefficients with their 95% confidence intervals for each regression model. Unadjusted models show the β coefficients for a regression model including either child IQ or early-life stature and the covariates. Adjusted models include both early-life stature and child IQ and therefore adjustment refers to the inclusion of the other main variable of interest in the regression model. All data including p values are shown in the appendix (p 2).

Figure 3

Adult IQ according to categories of early-life stature (A) and child IQ (B) and schooling in four birth cohorts, by sex Bars indicate mean adult IQ by site and sex, within categories of HAZ at around age 2 years and schooling attainment. HAZ is categorised as less than –3, –3 to less than –2, –2 to –1, –1 or greater. No individuals from Guatemala had a HAZ of 1 or greater and no individuals from Brazil had a HAZ below –3. Child IQ was measured at age 4·5 years (Brazil), 4–7 years (Guatemala), 8·5 years (the Philippines), and 4·5 years (South Africa). Child IQ was categorized as less than –1, –1 to +1, or greater than +1 SD of the site-specific distribution. Schooling was categorised as low (did not compete primary [Guatemala] or secondary school [Brazil, the Philippines, and South Africa]) versus high (completed primary [Guatemala] or secondary school [Brazil, the Philippines, and South Africa]). HAZ=height-for-age Z score.