Commentary: Smoking during pregnancy—genes and environment weigh in

In their article, ‘Familial factors confound the association between maternal smoking during pregnancy and offspring overweight’, Iliadou et al.¹ report three key messages: (i) maternal smoking during pregnancy (SDP) is associated with increased risk for offspring overweight in young adulthood; (ii) this association is partly confounded by familial factors; and (iii) specific familial factors common to both SDP and offspring overweight need to be identified. These messages offer important contributions to the literature on the role of SDP and later offspring outcome. First, this research demonstrates additional support for the value of using genetically informative designs to address this complex question. Secondly, we are reminded of the complexity surrounding the human condition and to tread carefully when making inferences of causation. Rarely, in the world of human research, are we considering scientific inquiries that are simple questions of nature vs nurture, or whether genes plus environment equal outcome. Rather, we are dealing with multifaceted questions that, perhaps unsurprisingly, have multifaceted answers.

When one sets out to study the effects of SDP in human populations, the question of whether this pre-natal exposure predicts later outcome becomes considerably messy. Gone is the control of the laboratory environment or the controlled delivery of pre-natal nicotine exposure, as in animal studies. Gone also is the control of genetic background (i.e. inbred animal strains). When one sets out to examine SDP in humans, one is doing so in a highly variable population—variable on the genetic level, the environmental level and the level of the independent and dependent variables. Moreover, in humans, SDP results in fetal exposure not only to nicotine but also to a large amount of other toxic components, such as carbon monoxide, ammonia, nitrogen oxide and lead and other metals.² Thus, one should not limit the effects of SDP in humans to nicotine alone. Further, if a woman smokes during her pregnancy, is it not also likely that she smoked after pregnancy providing not only pre-natal, but also post-natal, exposure?

Up until a few years ago, this question of whether DSP predicted later outcome involved the comparison of two independent samples of women: those who smoked during pregnancy and those who did not, with some kind of statistical control for maternal and/or paternal characteristics. Unsurprisingly, there are inherent differences between two such samples of women, which begged the question: can we really compare them? For example, mothers who smoke during pregnancy tend to be substance dependent,³ have other comorbid psychopathology,³ have a history of criminal conviction,⁴ have substance-dependent partners⁵ and are more likely to have less education.⁴ Consider this: if mothers with ADHD more commonly smoke during pregnancy, and also confer increased child risk of ADHD via genetic transmission, the observed correlation between SDP and childhood ADHD would be largely spurious, with limited aetiological relevance. Recent efforts have been made to try to address this issue within the context of SDP across a variety of outcomes—by using quasi-experimental genetically informative approaches.^4,6 One example of such an approach, as exemplified in the current report,¹ is to look at sibling pairs who are discordant for SDP.^1,4,6–9 That is, mothers have smoked during one pregnancy but not during another. Such a quasi-experimental genetically informative design can begin to overcome the fact that the vast majority of existing studies provide only limited control for the fact that pre-natal exposures may be correlated with parental behaviours that could act as more proximal risk factors that are in turn transmitted to their offspring. It is the prior failure to control for such (possibly heritable) confounding factors that may account for a large part of the suggested associations between SDP and offspring outcomes that have been reported to date. This is, in fact, what has been found for many behavioural outcomes such as ADHD,^7,10 conduct disorder,⁷ criminality⁹ and academic achievement.^9,11 In other words, once genetic and shared environmental influences are accounted for, the effect of SDP on outcome is attenuated and often no longer significant. Iliadou et al.¹ report similar results for offspring overweight.

This research is to be commended. Iliadou and colleagues approach a reported and important public health problem—that is, an association between SDP and later obesity—from a new viewpoint and with a very large sample incorporating data from several linked Swedish national registers (N = 124 203 males including 8441 sibling pairs). They are to be applauded in their attempt to gain a deeper understanding of this reported relationship. However, while the Swedish registries are an outstanding resource, there are limitations to utilizing this type of data. The authors outline some of these: (i) SDP was assessed only in the first trimester; (ii) possible misclassification of smoking exposure between pregnancies; (iii) the inability to distinguish between effects of pre-natal and post-natal smoking exposure; and (iv) the outcome measure (body mass index, BMI) was only available in males precluding the evaluation of gender differences. One additional limitation not discussed includes the fact that these national registries do not have focused, extensive personal assessments of behaviour (both maternal and offspring). Thus, only ‘broad brushstroke’ research questions can be addressed. These findings by Iliadou et al. provide further justification for increased attention to the issue of SDP and offspring outcomes, even though they do not establish, nor can they given the data limitations, what aspects of the specific familial environment are involved.

Important to their message, the authors note that the association between SDP and offspring overweight diminished in both full and half siblings, ‘indicating that regardless of how many genes siblings have in common, it is the shared environment that accounts for this association.’ They also suggest using caution when interpreting this as the lack of association in half siblings could have been due to random error. Full and half siblings share 50% and 25%, respectively, of their genetic material on average; thus, the sibling comparison alone does not indicate the extent to which genetic variation within sibling pairs might contribute. Thus, in addition to the inability to determine specific environmental mechanisms involved in the association between SDP and offspring overweight, the data also do not allow for further research on specific genetic pathways involved. Moreover, differences between full and half siblings might also be attributable to paternal characteristics,⁴ which are unmeasured and untested in these data.

Despite these key limitations, in this important work, Iliadou et al. provide support to the growing argument that familial (both genetic and environmental) confounds may explain all, or part, of the association between SDP and later offspring outcome. That is, SDP might not be causally related to later outcomes, but rather genetic and/or environmental factors shared within families could account, in part, for these reported relationships. More work is needed before conclusions of specific familial mechanisms that are integral to the relationship between SDP and offspring overweight can be made. A combination of analytical designs and strategies will likely be necessary to provide answers to the complicated nature of this question.

Nevertheless, given that a central goal of the ‘epidemiological method’ is to attempt to derive a relatively unbiased estimate of the magnitude of the association between exposure and outcome, it will only be through increasingly rigorous methods, including quasi-experimental genetically informative designs, that such an estimate will be realized. The reality is that, in humans, we do not understand how much of the association between SDP and offspring outcomes can be attributed to either nicotine or other smoking by-products.⁶ By putting more realistic boundaries on the impact of SDP on behavioural as well as poor pregnancy outcomes¹² and not continuing to ignore confounding factors, we open the door for other avenues of intervention and prevention—opportunities that have heretofore been missed.

Funding

This work was supported, in part, by a grant from the National Institute on Drug Abuse (DA023134).

Conflict of interest: None declared.