With a novel natural experiment design, we were able to separate prenatal and inherited effects in offspring of maternal smokers. We found, in line with previous research, smoking in pregnancy to be associated with offspring ADHD symptoms in the related mother-offspring pairs and that the association was not accounted for by measured confounds including parent ADHD. The magnitude of association was significantly greater in the related mother-offspring pairs than in the unrelated pairs, which suggests that the well-established link between maternal smoking in pregnancy and offspring ADHD symptoms might represent an inherited rather than a true environmental risk effect. However, our results illustrate that the link between maternal smoking in pregnancy and the comparison outcome variable of offspring birth weight, where the magnitude of association was the same in both groups, is attributable to prenatal risk effects.
It has been widely assumed that it is possible to deal with the contribution of inherited factors and other confounds by including variables such as parent ADHD in analyses and statistically controlling for confounding factors in large observational studies. Our results suggest that this might not be a safe assumption. Using a genetically sensitive “natural experiment” design provides a different way of testing the prenatal environmental mediation hypothesis, because it allowed us experimentally to remove inherited confounds and not rely on statistical controls that in this instance, for ADHD, are not adequate.
Community-based clinical studies and pooled analysis (where mothers and offspring are related) have consistently shown that exposure to smoking in pregnancy is associated with ADHD (1,3,4
). Many studies show a dose–response relationship with the number of cigarettes smoked during pregnancy, and there are biologically plausible mechanisms that could explain this link (11,23
). Thus, smoking in pregnancy has come to be viewed as an important environmental risk factor for ADHD. However, when the results of our study are considered together with other emerging findings, there are several reasons to be cautious in assuming smoking plays a causal role in ADHD.
First, in contrast to studies of birth weight, animal studies on cigarette and nicotine exposure in utero have not consistently demonstrated motor and cognitive changes similar to ADHD (11,12,14
). Second, studies examining the children of twins reveal environmentally mediated effects of maternal smoking in pregnancy on birth weight (24
) but not ADHD, at least in the offspring of alcoholics (25
). Finally, studies examining siblings discordant for exposure to smoking in pregnancy suggest that maternal smoking in pregnancy might be indexing unmeasured familial risk that is not tapped by measurement of the usual confounders. Siblings not exposed to smoking in pregnancy but from families where mother smoked in another pregnancy show increased attentional, behavioral, and scholastic problems (26,27
). Thus, when results from all these designs are considered together, a consistent pattern of findings emerges that suggest that the association between maternal smoking in pregnancy and offspring ADHD, in contrast to offspring birth weight, might not be causal (28
). The shared inherited liability between maternal smoking in pregnancy and offspring ADHD symptoms also raises the possibility that there might be genetic risk variants that confer susceptibility to both nicotine dependence and ADHD.
How important are natural experimental designs such as ours when much larger observational studies are available? Epidemiological studies suggest that measured confounders do not account for the link between smoking in pregnancy and offspring ADHD (3,4
), and in accordance with this, association in our related group remained when including covariates. That is, the inherited effect does not seem to be indexed by measured confounders and so would not be picked up by large observational studies.
Interestingly mother's current smoking and the father's smoking in pregnancy were also associated with offspring ADHD in the maternally related and paternally related groups, respectively, and the degree of association for current smoking and paternal smoking in pregnancy was similar to that for maternal smoking in pregnancy. These observations also further indicate that this maternal (and paternal) behavior is indexing inherited liability for ADHD. If paternal smoking was important in terms of effects through passive smoking effects on the fetus, we would have expected to also observe association in the paternally unrelated sperm donation and embryo donation groups but not the related surrogacy group, because surrogates do not live with father; but we do not observe this (results available from first author).
In terms of potential clinical implications, our findings together with those from other studies suggest that it is important for mothers to quit smoking in pregnancy and that smoking cessation could result in increased offspring birth weight but that this intervention might not be a useful public health intervention for reducing ADHD symptoms.
As with all research designs, there are limitations. First, although with this sample size we were able to detect association between maternal smoking in pregnancy and birth weight in the unrelated group, we cannot rule out that association was not detected for ADHD, given the small number of smokers in our sample and in the unrelated group in particular. As a result, our power to detect significant prenatal environmental influences was low, and future replications are needed. Thus, we cannot be certain that there are no prenatal effects, only that there is significant evidence of inherited effects. Notwithstanding this observation, our findings, at the very least, suggest that a proportion of the association between smoking in pregnancy and ADHD observed in previous studies can be attributed to inherited factors. Second, we did not examine ADHD diagnoses; however, all the research to date suggests etiologic continuity between normal variation in ADHD symptoms and extreme scores (29,30
). Third, families that have used Assisted Reproductive Technologies might be unusual. However, comparison with representative community samples showed that family income and measures of psychological adjustment are similar (17
). The income distribution in this sample was expected, given that fertility treatment is freely available in the United Kingdom under the National Health Service.
Unsurprisingly, the rate of maternal smoking in this sample was lower than currently reported UK rates. It is likely therefore that the mothers who smoked in our sample are the most highly dependent on nicotine, which could potentially increase the shared genetic liability with ADHD. However, the magnitude of association between maternal smoking in pregnancy and ADHD in the related group is similar to that reported in general population samples (19
). Another potential issue is that there are some differences between the related and unrelated groups. Given that every design has its own set of strengths and weaknesses, different designs are needed, and there can be greater confidence when there are similar findings across a variety of designs, as is the case here (28
Our findings highlight the need to test causal hypotheses with genetically sensitive designs (7
). Results from traditional observational designs do not necessarily pick up inherited confounds and could therefore be misleading. Our results suggest that the previously observed association between maternal smoking in pregnancy and ADHD might represent an inherited confound.