A new study published online ahead of print in the Journal of Epidemiology and Community Health concludes that maternal smoking during pregnancy causes criminal behavior later in life (see: Paradis AD, Fitzmaurice GM, Koenen KC, Buka SL. Maternal smoking during pregnancy and criminal offending among adult offspring. J Epidemiol Community Health 2010).
The article's conclusion that exposure to maternal smoking during pregnancy creates criminals later in life has received widespread publicity, leading to headlines such as these:
Smoking in Pregnancy 'Breeds Criminals'
Smoking During Pregnancy 'Increases Risk of Children Turning to Crime'
The study methods are described as follows: "The relationship between MSP [maternal smoking during pregnancy] and adult criminal offending was examined using data from 3766 members of the Providence, Rhode Island, cohort of the Collaborative Perinatal Project. Information on MSP and most potential confounders was collected prospectively throughout pregnancy. In 1999–2000 all offspring had reached 33 years of age and an adult criminal record check was performed. Because previous research has been criticised for not properly accounting for confounding influences, our primary aim was to determine whether the MSP–criminal offending relationship held after efficiently adjusting for a wide range of sociodemographic and family background characteristics using propensity score methods."
The results were as follows: "The association between MSP and adult criminal offending remained after controlling for propensity scores. Offspring of mothers who smoked heavily during pregnancy (≥20 cigarettes per day) had the greatest odds of an adult arrest record (OR 1.31, 95% CI 1.06 to 1.62)."
The study concludes: "Our findings suggest that the relationship between maternal smoking during pregnancy and adult criminal offending is causal."
The Rest of the Story
While it is interesting that there is an association between maternal smoking during pregnancy and criminal behavior among the offspring, there is no way one can draw a causal conclusion from this study. There is a far more likely explanation for this association, rather than the explanation that the smoke exposure caused some sort of brain damage that later led to criminal activity.
The more likely explanation is that mothers who smoke during pregnancy represent a very different population from those who do not, and that factors related to their decision to smoke during pregnancy, rather than the actual smoke exposure, is what predisposes their children to criminal behavior.
In particular, the most likely reason for the association is that mothers who smoke during pregnancy are probably more likely to themselves have a higher incidence of criminal activity. It should come as no surprise that there is a link between criminal activity among parents and their offspring. As the study did not control for parental criminal activity, it cannot possibly conclude that the observed association is a causal one.
Although the paper purports to have controlled for parental anti-social behavior - a critical potential confounding variable - this variable was only available for 720 out of the 3,766 subjects. The study used a multiple imputation procedure to "create" this variable for the subjects with missing data. While multiple imputation is reasonable to use when you have a small proportion of missing data, it is a huge stretch to impute data for the overwhelming majority of the sample. In my view, parental anti-social behavior was clearly not adequately controlled for in this study. Thus, once again, the paper cannot possibly conclude that the observed relationship is a causal one.
Furthermore, the paper didn't control for maternal substance use: the use of alcohol and illicit drugs. It is very likely that maternal substance abuse is related both to smoking during pregnancy and to an increased risk of subsequent criminal activity among the offspring.
Surprisingly, the paper reaches a causal conclusion despite its ready acknowledgment of the possibility that there was unmeasured and residual confounding. The study takes it on faith that this unmeasured and residual confounding is not of a magnitude that would negate the results, even though there is no evidence to support such an assumption.
The paper states: "Due to the potential for unmeasured and residual confounding, our results likely reflect the upper boundary of any true effect of MSP on adult ASB." This is equivalent to stating: "There is strong reason to believe that we did not adequately control for confounding variables. We can make no estimate of the magnitude of this confounding, but we are going to take it on faith that the magnitude of such confounding is not enough to negate our observed results, because then we would have to admit that the relationship is not causal and the paper might not be judged to be publishable."
This is the first time I have ever seen a paper in which it was assumed, on faith, that inadequate control for confounding must not have been of a magnitude that would negate the observed effect, in the absence of any analysis whatsoever of the likely magnitude of that confounding.
Essentially, such reasoning negates the entire point of doing the research in the first place. If you are going to simply assume that any unmeasured confounding does not negate the effect, then why even bother doing the study in the first place? Why not stick with the completely unadjusted estimates and simply make the same assumption?
The conclusion of the paper begins: "While we cannot definitively conclude that MSP (particularly heavy MSP) is a causal risk factor for adult criminal offending, the current findings do support a modest causal relationship." This is a cop out. If the study does not provide enough evidence to (definitively) conclude that maternal smoking during pregnancy causes adult criminal offending, then it cannot draw such a conclusion. What the paper is once again saying is: "We don't have enough evidence to conclude that the relationship is causal, because as we've pointed out, we are quite sure that we've missed important confounding variables. Nevertheless, we want to sensationalize the results as much as possible so we're going to draw a causal conclusion anyway."
The damage caused by the sensationalistic and unsupported conclusion of this study could be substantial. Newspapers are reporting that "exposure to cigarette in the womb may harm developing areas of the brain that affect behaviour, impairing the transmission of chemical signals important for attention and impulse control" and that "smoking in pregnancy can cause harm to the child's developing brain that puts them at greater risk of having a long-term criminal record."
That is quite a stretch for a study that failed to measure the criminal records of the parents themselves, did not adequately control for parental anti-smoking behavior, and readily admitted that "we most likely missed important variables."
Somehow, the scientific rigor of the paper ended just before the drafting of the study conclusion. This is not atypical, however, for modern-day tobacco control research.