Familial Confounding of the Association Between Maternal Smoking in Pregnancy and Autism in Offspring

Background:  Maternal smoking in pregnancy is a preventable but common exposure thought to lead to neurodevelopmental problems in offspring such as attention deficit hyperactivity disorder (ADHD). Epidemiological studies have evaluated its role in autism with inconsistent findings. Resolving links between maternal smoking and autism is especially difficult because of strong confounding by social class and genetic factors that influence smoking behavior and autism occurrence.

Objectives:  To examine whether maternal smoking in pregnancy increases the risk of autism spectrum disorder (ASD), with a focus on disentangling individual and familial effects to account for confounding.

Methods:  The study cohort comprised all children born in Denmark in 1991 through 2012 with complete parent and maternal smoking information. The cohort (1,294,906 persons, including 993,301 siblings nested within 728,271 families) was followed from 1 year of age until an ASD diagnosis (13,547), death, emigration, or December 31, 2012. Diagnoses of ASD and the presence of co-occurring ADHD (3319 ASD+ADHD and 10,228 ASD-ADHD) were based on ICD-8 and ICD-10 codes from psychiatric and patient register databases. Maternal smoking data were obtained from the birth register as reported by the mother to her midwife at the first prenatal care visit. We estimated adjusted hazard ratios (aHR) and 95% confidence intervals (95% CI) between any maternal smoking and ASD using Cox survival models and robust standard errors to account for familial clustering, adjusting for child sex, parity, parental age, education, income, psychiatric case history, and calendar year. To separate familial from individual effects, we constructed similar models that included a term for the mean level of maternal smoking in the family and the individual’s deviance from the family mean (maternal propensity model). A third modeling approach stratified on maternal siblings and considered siblings discordant for maternal smoking in pregnancy to account for unmeasured confounding at the family level.

Results:  In the cohort, 21.5% of individuals had a report of maternal smoking in pregnancy. In standard adjusted models, the risk for ASD was slightly elevated for maternal smoking: aHR 1.17 (95% 1.13-1.22). The risk for ASD from smoking was attenuated, however, based on either the maternal propensity model, aHR 0.98 (95% CI 0.88-1.09) or the discordant sibling analysis, aHR 0.86 (95% CI 0.64-1.15). In standard adjusted models of case subgroups, the risk for ASD+ADHD from smoking was higher: aHR 1.54 (95% CI 1.40-1.68) than for ASD-ADHD: aHR 1.09 (95% CI 1.03-1.16), and both were attenuated in the maternal propensity model: ASD+ADHD aHR 1.15 (95% CI, 0.89-1.40) and ASD-ADHD aHR 0.93 (95% CI, 0.82-1.06).

Conclusions:  An association between maternal smoking in pregnancy and ASD diagnosis was observed after adjusting for a range of confounding variables, but was attenuated following designs that more fully accounted for unmeasured factors at the family-level, with similar effects observed for ASD with and without co-occurring ADHD. These results suggest that maternal smoking in pregnancy is not a risk factor for ASD and that elevated associations reported in previous studies may arise in part from not fully accounting for confounding influences at the family level.