Prenatal smoking, alcohol and caffeine exposure and maternal-reported attention deficit hyperactivity disorder symptoms in childhood: triangulation of evidence using negative control and polygenic risk score analyses

Abstract

Background and aims: Studies have indicated that maternal prenatal substance use may be associated with offspring attention deficit hyperactivity disorder (ADHD) via intrauterine effects. We measured associations between prenatal smoking, alcohol and caffeine consumption with childhood ADHD symptoms accounting for shared familial factors.

Design: First, we used a negative control design comparing maternal and paternal substance use. Three models were used for negative control analyses: unadjusted (without confounders), adjusted (including confounders) and mutually adjusted (including confounders and partner's substance use). The results were meta-analysed across the cohorts. Secondly, we used polygenic risk scores (PRS) as proxies for exposures. Maternal PRS for smoking, alcohol and coffee consumption were regressed against ADHD symptoms. We triangulated the results across the two approaches to infer causality.

Setting: We used data from three longitudinal pregnancy cohorts: Avon Longitudinal Study of Parents and Children (ALSPAC) in the United Kingdom, Generation R study (GenR) in the Netherlands and Norwegian Mother, Father and Child Cohort study (MoBa) in Norway.

Participants: Phenotype data available for children were: N_ALSPAC = 5455-7751; N_GENR = 1537-3119; N_MOBA = 28 053-42 206. Genotype data available for mothers was: N_ALSPAC = 7074; N_MOBA = 14 583.

Measurements: A measure of offspring ADHD symptoms at age 7-8 years was derived by dichotomizing scores from questionnaires and parental self-reported prenatal substance use was measured at the second pregnancy trimester.

Findings: The pooled estimate for maternal prenatal substance use showed an association with total ADHD symptoms [odds ratio (OR)_SMOKING = 1.11, 95% confidence interval (CI) = 1.00-1.23; OR_ALCOHOL = 1.27, 95% CI = 1.08-1.49; OR_CAFFEINE = 1.05, 95% CI = 1.00-1.11], while not for fathers (OR_SMOKING = 1.03, 95% CI = 0.95-1.13; OR_ALCOHOL = 0.83, 95% CI = 0.47-1.48; OR_CAFFEINE = 1.02, 95% CI = 0.97-1.07). However, maternal associations did not persist in sensitivity analyses (substance use before pregnancy, adjustment for maternal ADHD symptoms in MoBa). The PRS analyses were inconclusive for an association in ALSPAC or MoBa.

Conclusions: There appears to be no causal intrauterine effect of maternal prenatal substance use on offspring attention deficit hyperactivity disorder symptoms.

Keywords: ALSPAC; Alcohol; GenR; MoBa; caffeine; childhood ADHD; intrauterine effects; negative control; polygenic risk score; smoking.

Figure 1. Study design

Note: a) The dashed arrow represents the negative control analysis. Assumption includes: the same confounders influence maternal and paternal prenatal substance use and offspring ADHD symptoms, a causal prenatal (intrauterine) effect only exists for maternal prenatal substance use. b) Polygenic risk score analysis was conducted with maternal genetic variants as proxies for prenatal smoking, alcohol and caffeine consumption (3 separate analyses, with polygenic risk scores specific to the substance used).

Figure 2. Meta-analysis of maternal and paternal prenatal smoking, alcohol and caffeine consumption across the cohorts

Note: Meta-analysis of smoking (a) and alcohol consumption (b) are based on mutually adjusted model. Metaanalysis of caffeine consumption (c) is based on adjusted model, because paternal caffeine consumption was not assessed in GenR. Heterogeneity between the cohorts is shown by computing I² (see Methods and Supplementary Table S1 for more details). The statistical difference between maternal and paternal associations for smoking exposure in ALSPAC was (P_ADHD=0.90; Phyp=0.91; Pina=0.34), in MoBa (P_ADHD=0.14; Phyp=0.04; Pina=0.22) and in GenR (P_ADHD=0.07; P_HYP=0.79; P_INA=0.10). The statistical difference between maternal and paternal associations for alcohol exposure in ALSPAC was (P_ADHD=0.001; P_HYP=<0.001; P_INA=0.006), in MoBa (P_ADHD=0.001; P_HYP=0.005; P_INA=0.002) and in GenR (P_ADHD=0.75; P_HYP=0.10; P_INA=0.63) and the statistical difference between maternal and paternal associations for caffeine exposure in ALSPAC was (P_ADHD=0.88; P_HYP=0.99; P_INA=0.68), in MoBa (P_ADHD=0.05; P_HYP=<0.001; p_INA=⁰.⁴⁷⁾.