These findings suggest that indicators of environmental risk thought to influence the diagnosis of ADHD may also have independent effects on the phenotype manifestation of ADHD within a clinical sample, something which has not been explored in previous research. Severity of hyperactive-impulsive symptoms was independently predicted by both lower social class and maternal smoking during pregnancy, with maternal smoking during pregnancy being revealed as the most important factor. Indeed, when we further analysed the data to look at quantity of cigarettes smoked, we found a dose-dependent effect, whereby the children of mothers who reported smoking more than 10 cigarettes a day during their pregnancy had more hyperactive-impulsive symptoms than those whose mothers smoked fewer than 10 cigarettes a day (r2
= 0.02, Beta = 0.15, t = 2.80, p = 0.006). In contrast there was a lack of association observed between any of the indicators of environmental risk and inattentive symptom severity. This finding is in line with previous twin studies which indicate a greater influence of common environmental factors (which social class and maternal smoking during pregnancy would be) for hyperactive-impulsive symptoms, in comparison to inattentive symptoms [44
]. This supports the distinction of ADHD subtypes and suggests that environmental as well as genetic factors contribute to the phenotypic heterogeneity of ADHD characterised by the type of symptoms and their severity. The results also suggest that smoking during pregnancy and environmental risk as indexed by social class may influence the aetiology of the hyperactive-impulsive dimension (or its severity) specifically, rather than all symptoms of ADHD (in a clinical sample).
Maternal smoking during pregnancy also significantly predicted total number of ODD symptoms but not a diagnosis of the disorder, possibly because such a large proportion (47%) of the sample met diagnostic criteria for ODD.
An independent, significant association was also found between maternal smoking during pregnancy and Conduct Disorder symptoms (r2 = 0.04, Beta = 0.18, t = 3.34, p = 0.001), even when other significant covariates were considered. Similarly, a diagnosis of Conduct Disorder was independently associated with maternal smoking during pregnancy. Conduct disorder (but not symptoms scores) was also found to be associated with social class. These results indicate that there is an increase in the risk of a comorbid diagnosis of CD for those ADHD children whose mothers smoke during pregnancy with an odds ratio of 3.14. Furthermore, those ADHD children from a low social class family whose mothers smoke during pregnancy are at an even greater risk of being diagnosed with Conduct Disorder (more than a five-fold increase in risk). Considering the additional impairment and risks in later life for those individuals with both ADHD and Conduct Disorder, these are potentially important findings. Environmental indicators of risk which can be easily assessed, even if not causal, can help pick up children at higher risk of adverse outcomes so that risk reduction strategies and higher levels of follow up and service involvement are targeted at those most at risk.
Inevitably, this study has a number of limitations. First, we were not able to test for main effects of the environmental factors on a diagnosis of ADHD (all our subjects had a diagnosis). However, we specifically chose indicators of environmental risk which have previously been associated with ADHD and the purpose of this study was to investigate phenotypic variation within a clinical ADHD sample.
Second, it is possible that, rather than showing an association between environmental factors and comorbid CD symptoms, the results of our analysis simply revealed main effects of these factors on antisocial behaviour in general. This is certainly plausible as associations between maternal smoking during pregnancy, birth weight, social class and antisocial behaviour have previously been reported. [45
] Unfortunately, this hypothesis can not be tested in this sample as neither an unaffected or conduct disorder alone control group were available. However, the aim of this investigation was not to find main effects of environmental measures, but rather effects within the phenotype.
Third, the failure to detect significant associations between phenotypic and indicators of environmental risk may be due to the reduced variance in ADHD symptoms in this clinical sample. However, we were interested in how the environmental factors have additional effects within the ADHD phenotype. It may even be argued that any independently significant findings are even more interesting as the power to detect them is small, although we recognise that small sample sizes can also lead to false positive findings [40
Fourth, our indicators of environmental risk were collected retrospectively from self report. It is possible that these measures are inaccurate. Furthermore, our question assessed maternal smoking during pregnancy generally, rather than separating it into different trimesters. We therefore cannot determine whether or not there are differing effects of maternal smoking during pregnancy between those who give up at some point during the pregnancy. However, retrospective self report of maternal smoking during pregnancy has been shown to be relatively accurate, corresponding with both contemporary reports [13
] and blood cotinine levels which identify tobacco products in the bloodstream. [47
] Similarly, research using our measure has illustrated high levels of agreement between parental recall of birth weight and smoking during pregnancy with data from medical records. [38
Fifth, social class is unlikely to be a proximal risk factor and is certainly only an indicator of risk. However, it is an easily obtainable indicator of risk that is readily available in clinical practice. Given these findings, there is clear justification for measuring and examining potential proximal risk factors that are indexed by social class and that may have effects on the clinical presentation of ADHD.
Sixth, as discussed previously, these indicators of environmental risk (and indeed the phenotypic measures) show covariation. It is therefore possible that our findings are the result of a separate, unmeasured factor, associated with both the environmental and phenotypic measures. In this analysis, we control for a number of possible covariates by including every environmental indicator measure in the multivariate analysis, whilst also controlling for a range of demographic and clinical variables found to be associated with the phenotypic variable.
The environmental indicator measures chosen represent easily obtainable indicators of risk, but are not necessarily causal themselves. Other designs are needed to test whether the associations are due to causal risk effects [48
]. Nevertheless ascertaining such indicators of risk may be advantageous when identifying high risk subgroups of ADHD children, as well as providing a starting point for research attempting to find more proximal, causal risk factors. The majority of our sample (90%) were male and so these findings may not translate to females (although re-analysis our data with the exclusion of females did not alter our findings). Similarly, this is a clinical sample and so findings may not be generalised to individuals within the community who would meet diagnostic criteria for ADHD.
Finally, there may be a genetic link between our indicators of environmental risk and our phenotypic measures. For example, maternal smoking during pregnancy and child ADHD or conduct disorder behaviour; antisocial or ADHD mothers may be more likely to smoke during pregnancy (indeed there is extensive literature linking substance use and conduct problems e.g. [49
]) and pass genetic risk for such behaviours to their child, regardless of their smoking behaviour. Unfortunately, we do not have data on the mothers of these children (for example psychiatric diagnoses, personality measures or IQ) and so the design of this study cannot distinguish between genetically and environmentally medicated risks. However, twin studies have indicated that maternal smoking during pregnancy has a significant environmental effect on ADHD symptoms, over and above any genetic effects [51
], whilst in our previous work, we have not found any specific gene variants to be associated with both maternal smoking during pregnancy and childhood ADHD [52
]. Similarly, there is some evidence to suggest that social class is associated with genes which influence Verbal IQ [53
]. Although verbal IQ was included as a covariate in our analyses where relevant, the possibility that genetic factors are influencing the role of social class on hyperactive-impulsive and conduct disorder symptoms cannot be ruled out.