Explaining the Association Between Fetal Growth and Childhood ADHD Symptoms: Cross-cohort Replication

Well established risk factors for childhood attention-deficit hyperactivity disorder (ADHD) include genetic/familial susceptibility to ADHD, male sex, and restricted fetal growth (Thapar et al., 2013). As a partially environmentally-determined characteristic, fetal growth may be the most amenable to intervention. However, it is not currently understood which fetal growth determinant best explains the association between restricted fetal growth and ADHD symptoms.

Birth weight is a crude approximation of the quality of fetal growth. If an infant’s weight is appropriate for their gestational age, we could assume that their fetal growth rate was typical. The association between lower birth weight and attention problems in childhood has been replicated many times, in many populations (meta-analyses: Franz et al., 2018; Momany et al., 2018). Lower birth weights, even among those born at full term, are linked with an elevated risk of ADHD and academic difficulties (Class et al., 2014; Groen-Blokhuis et al., 2011; Hultman et al., 2007; Momany et al., 2018). Within-twin and within-sibling studies suggest that the association may be independent from genetics and other familial factors (Class et al., 2014; Ficks et al., 2013; Groen-Blokhuis et al., 2011; Hultman et al., 2007; Lim et al., 2018; Pettersson et al., 2015). Experimentally-induced fetal growth restriction in animals has also been shown to influence neurodevelopment and behavior (Lauritz et al., 2012; Meyer et al., 2014). It has therefore been proposed that poor fetal growth is an independent causal factor in the development of ADHD. The replicable association between birth weight and ADHD symptoms in twins also shows that, despite experiencing the same prenatal exposures (e.g., maternal smoking), twin differences in fetal growth can occur and it is this that predicts differences in childhood ADHD symptoms. Large-scale studies of singleton births are needed to understand which prenatal factors contribute to restricted fetal growth, and which account for the association between fetal growth and ADHD symptoms on a population level.

The fetal origins hypothesis (and more recently the Developmental Origins of Health and Disease hypothesis) posits that conditions in utero can permanently program certain aspects of the child’s physiology, explaining why lower birth weights may be causally related to neurodevelopmental issues. However, fetal growth rate and birth weight are the products of many prenatal and pre-pregnancy factors such as maternal malnutrition, smoking, drug-use, certain medications, parental age and pregnancy/gestational complications (Budree et al., 2017; Heaman et al., 2013; McCowan & Horgan, 2009), several of which are also directly associated with subsequent mental health of the child (He et al., 2020; Smith et al., 2016). Proposed mechanisms to explain how prenatal risk factors such as maternal smoking and gestational infection directly impact neurodevelopment have included: lack of oxygen and blood to the fetus leading to altered gene expression (Smith et al., 2016), neurotransmitter function (Laplante et al., 2012), and activation of immune and inflammatory processes (Meyer et al., 2009; Mirza et al., 2015). It may be that birth weight is a convenient proxy for such direct effects of fetal adversity on neurodevelopment. In this study, we quantify the extent to which the association between fetal growth and ADHD symptoms are explained by prenatal determinants of fetal growth. We are particularly interested in the contribution of modifiable prenatal factors such as maternal substance-use in pregnancy. Understanding the extent to which the association is modifiable will help establish the viability of prenatal interventions.

Familial background is a major source of confounding in the theoretically causal pathway between prenatal adversity, lower birth weight and childhood mental health problems. For instance young maternal age is linked with both low birth weights (Aras, 2013) and increased ADHD disruptive behaviours in the child (Tearne, 2015). Household socioeconomics, often exemplified by income and education level, has also shown strong links with both reduced birth weight (Madden, 2014; Martinson & Reichman, 2016) and ADHD risk (Russell et al., 2016). That said, the association between birth weight and ADHD symptoms typically survives correction for familial socioeconomic factors (Abel et al., 2010; Class et al., 2014; Lærum et al., 2017; Pettersson et al., 2019). Low birth weight and ADHD are also not randomly distributed across racial, ethnic or migrant groups. Black women born in the US are more likely to have a low birth weight/preterm child than any other race/ethnicity even after controlling for income and educational attainment, though other unmeasured inequalities cannot be ruled out (Catov et al., 2015; Giscombé & Lobel, 2005). Family psychiatric history is also relevant to both low birth weight and ADHD via several plausible pathways. It captures genetic susceptibility to ADHD (Thapar et al., 2013), but psychiatric issues in the mother may also affect fetal growth via increased stress (Mongan et al., 2019; Wadhwa et al., 2011), and problematic behaviors and coping styles may also be socially learned. Demographic/socioeconomic factors and family psychiatric history should therefore be controlled for when assessing the effect of fetal growth on ADHD symptoms.

No study to our knowledge has assessed the extent to which the association between fetal growth and ADHD symptoms is explained by preceding prenatal factors. This is important information because fetal growth, estimated from weight and age at birth, is not in itself a modifiable factor. Understanding which prenatal factors and to what extent they contribute to the association between fetal growth and ADHD symptoms may help identify prenatal targets for preventative interventions. Similarly, but not equivalently, studies have used prenatal factors to directly predict ADHD diagnosis, where the goal is predictive accuracy rather than explanation (Huhdanpaa et al., 2021; Schwenke et al., 2018; Sciberras et al., 2011; Silva et al., 2014; Willoughby et al., 2020). Other research has assessed the effect of birth weight on ADHD symptoms after adjustment for prenatal factors, however the focus of those studies is generally on the final, fully adjusted, birth weight effect, and not on how it changes due to the addition of each type of explanatory factor (Murray et al., 2016; Silva et al., 2014; Wiles et al., 2006). This study attempts to disentangle potentially confounding factors such as familial income and psychiatric history, and prenatal factors, which temporally precede fetal growth and may explain its effects.

Several other aspects of our study design differentiate it from the existing literature. First, we use a continuous parent-reported scale as our outcome, rather than binary ADHD diagnosis, or a cut-off corresponding to clinical risk, which have typically been the outcomes of choice when assessing prenatal predictors. While diagnosis can be clinically informative, subjective symptom scales can offer greater sensitivity to smaller effects. This may be particularly important given that low birth weight has been associated with ADHD symptom shifts within the sub-clinical range (Milberger et al., 1997; Murray et al., 2016, 2021). Second, unlike studies which assess one risk factor in isolation (e.g., maternal smoking), our data includes an array of pregnancy complications and substances consumed by the mother, allowing us to assess independence of their effects. Finally, our study uses two independent population-based cohorts of children (both aged 9–10), one from the United States (Adolescent Brain Cognitive Development Study; ABCD; N = 8,358) and one from Ireland (Growing Up in Ireland study; GUI; N = 7,724) to test replicability of findings. Meta-analytic findings suggested the association between birth weight and ADHD symptoms was significantly moderated by geographic region suggesting that context may be important when considering this particular association (Momany et al., 2018). Parallel analysis in two nationally-representative samples may help disentangle generalizable from context-specific explanations for the association between fetal growth and ADHD symptoms.

Results from two independent cohorts suggest the association between fetal growth and ADHD symptoms at age 9 can be explained in part by sociodemographic confounds and prenatal events. Across both cohorts, over 25% of the association was attributable to familial confounds (socioeconomics, demographics, family psychiatric history) and the significance of the association survived adjustment for familial confounds but not prenatal factors. However, results differed by cohort regarding the relative contributions of socioeconomic/demographic factors (ABCD > GUI) and prenatal factors (GUI > ABCD) to the effect of fetal growth on ADHD symptoms (Fig. 2; Fig. S4). Fetal growth, as estimated from weight and age at birth, was a relatively context-independent predictor of ADHD symptoms while the explanatory or driving factors of the association were somewhat context-dependent.

Averaging cohort results (Table 2), 23.4% of the association between fetal growth and ADHD symptoms was captured by socioeconomic and demographic factors such as household income, race/ethnicity, parental age, education and single parenthood. This is consistent with other studies which have found the fetal growth (or small-for-gestational-age) effect on ADHD outcomes attenuates after controlling for factors such as maternal age, household income, parental education or single-parenthood (Murray et al., 2016; Pettersson et al., 2019; Wiles et al., 2006). The decrease in fetal growth effect after control of socioeconomic and demographic factors was slightly greater in ABCD (25.4%) compared to GUI (21.3%), however this cohort-discrepancy widened in an ordinal sensitivity analysis (ABCD 31.3% Vs GUI 20.5%; Table S2). This disparity may be explained by national differences in the correlation between socioeconomic factors and fetal growth. Martinson and colleagues (2016) found that income-related inequalities in low birth weight rates were present in the US, the United Kingdom, Canada and Australia, but that the magnitude of this gradient was greatest in the US. The authors suggested that the more generous social support and healthcare systems of the UK, Canada and Australia may play a buffering role. Differences may also be explained by the greater racial and ethnic heterogeneity in ABCD (Table 1) and the greater correlation between certain race/ethnicities and fetal growth in ABCD compared to GUI (Figs. S2-S3).

Pregnancy complications accounted for 5.3% of the fetal growth effect on ADHD symptoms on average, with a larger proportion accounted for in ABCD (7.9%) than GUI (2.7%). Ordinal models approximated this cohort difference (9.9% vs 3.2%). The mean number of pregnancy complications in mothers across cohorts was similar across cohorts (Table 1) however the independent effect of pregnancy complications on ADHD symptoms was stronger in ABCD compared to GUI (Tables S3, S5). This suggests that, despite being similarly prevalent across cohorts, pregnancy complications have more adverse effects on both fetal growth and childhood behavior in ABCD compared to GUI. While we did not assess interactive effects between pregnancy complications and maternal age, the younger age of ABCD mothers compared to those in GUI may explain the difference in results (Table 1). For instance, the association between preeclampsia and small-for-gestational age is stronger for mothers under 25 compared to over 25 (Li et al., 2018). Alternatively, this result may be explained by unmeasured cohort differences such as in maternal weight status, stress access to prenatal care, or quality of care (Bronstein et al., 2018; Fuchs et al., 2022). Rates of obesity, for instance, are higher among women in the US compared to women in comparable countries, and this may interact multiplicatively with certain pregnancy complications to increase the risk of both restricted fetal growth and neurodevelopmental issues (Bronstein et al., 2018; Kong et al., 2020).

Perhaps the most striking aspect of Fig. 2 is the greater proportion of the effect accounted for by maternal substance-use in GUI (22.7%) compared to ABCD (4.8%). This is despite similar reported rates of smoking, and lower rates of alcohol and drug-use among GUI mothers (Table 1). GUI mothers smoked more cigarettes and more persistently than ABCD mothers, smoking ~ 9 cigarettes per day on average throughout pregnancy (Table S8). Maternal smoking was also a significant independent predictor of ADHD symptoms in the GUI, but not ABCD. By contrast, maternal alcohol-use and drug-use during pregnancy were significant independent predictors of symptoms in ABCD, but not in GUI (Tables S3, S5). Our findings are consistent with another analysis of the GUI data which showed strong links between maternal smoking and intrauterine growth restriction in Ireland (Madden, 2014), and support the need for improved smoking cessation programs in Irish maternal hospitals (Reynolds et al., 2017). There has been some evidence that the association between smoking in pregnancy and offspring ADHD is not causal and is fully accounted for by shared genetic factors between mother and child (Rice et al., 2018). Such logic could also be applied to other types of substance-use in pregnancy. However, there is evidence supporting the causal association between maternal smoking in pregnancy and birth weight (Rice et al., 2018). Given reliable associations between fetal growth and ADHD symptoms have been observed in human and animal studies (see introduction), it may be that maternal substance-use in pregnancy, such as smoking, impacts child neurodevelopment via fetal growth restriction (Brannigan et al., 2020).

Our results suggest such prenatal factors can capture up to a quarter of the birth weight effect on ADHD symptoms; however future studies will need to assess, in practice, whether reduction in prenatal risks have any tangible effect on childhood ADHD symptoms at a population level. Effect sizes were small– Table 2 indicated that fetal growth accounted for less than 0.5% of variance in ADHD symptoms across all models (all η_p² < 0.005) and supplementary tables showed that fully-adjusted models explained just 8–10% of variance in outcomes (ABCD R² = 9.6%; GUI R² = 8.3%; Tables S3, S5). A meta-analysis suggested birth weight accounted for 2.25% in the variance in ADHD symptoms (r = -0.15; Momany et al., 2018). While the effect size observed in this study was smaller, this does not necessarily invalidate the clinical relevance of our findings. First, even well-established risk factors for ADHD such as male sex and parental psychopathology had small effect sizes (η_p² < 0.05). This could be explained by our large sample sizes and inclusion of normative symptom scales rather than groups from each end of the spectrum (Kühberger et al., 2014). Second, we know from animal experimentation (Lauritz et al., 2012; Meyer et al., 2014), human meta-analysis (Momany et al., 2018) and multiple twin studies (Ficks et al., 2013; Groen-Blokhuis et al., 2011; Hultman et al., 2007; Pettersson et al., 2015) that fetal growth is a highly replicable predictor of neurodevelopmental problems, even when genetic and social confounds are controlled for. Third, other large cohort studies have showed birth weight accounts for < 1% of the variance in ADHD symptom dimensions (Ficks et al., 2013; van Mil et al., 2015). Finally, the importance of studying prenatal risks to neurodevelopment lies in their temporal precedence over all postnatal risk factors. Small deviations from typical neurodevelopment (reflected by small effect sizes) at an early stage of development have the potential to moderate the effects of all subsequent insults.

The primary strength of this study is its use of two large nationally-representative cohorts. The matched analysis on both cohorts helps determine which findings are generalizable and which may be cohort- or nation-specific. While Ireland and the US are both developed countries, they differ on demographic make-up, healthcare systems, policies and culture. As some of these largely unmeasured factors may confound the association between fetal growth restriction and childhood psychopathology, replicating the association across cohorts can be considered further support for a causal association (Murray et al., 2016). Other strengths include: thorough control of potential confounds, the use of alternative statistical modelling to probe the robustness of findings (ordinal regression), and the specificity of our target population to children aged 9–10 born between 2007 and 2009.

Several aspects of the study design limit interpretation of results. First, comparability of cohort results is limited by differences in outcome scales (CBCL vs SDQ), in definitions of sociodemographic and family psychiatric variables (see Supplementary Material) and in the gestational age range captured (capped at 40 weeks in ABCD). Future studies assessing the generalizability of prenatal contributions to fetal growth and ADHD symptoms should use samples with better matched data, merged into one analysis, to quantitively assess the significance of cohort differences. Second, in ABCD we rely on the retrospective report of gestational and birth events which may be influenced by recall bias (9 years on). In both cohorts, we rely on the mother to report to provide both exposure and outcome data, which may be biased. Third, the study was conducted in singleton-born children thus results are not applicable to twins. Finally, there may be unmeasured sources of confounding such as migrant status, neighborhood poverty, maternal health, weight and stress levels.

We found replicable evidence across Irish and US-based cohorts for the linear association between reduced fetal growth and increased ADHD symptoms by age 9. Background familial confounds (e.g. income, race/ethnicity, family psychiatric history) accounted for over a quarter of that association in both cohorts, however it remained statistically significant. Pregnancy complications and maternal substance-use accounted for different proportions of the fetal growth effect across cohorts, which may reflect societal and cultural differences.