Tobacco and psychosis: Not quite a smoking gun

In the wake of the embarrassing and deadly delay in establishing that smoking causes cancer, the Royal Society of Medicine posed the question, ‘In what circumstances can we pass from … [an] observed association to a verdict of causation? Upon what basis should we proceed to do so?’ (Hill, 1965). In response, Austin Bradford Hill set out the following criteria by which we can judge whether an association is likely to be causative: the strength of the association between the environmental agent and the disease, the consistency of the association in different settings, whether the exposure to the environmental agent preceded the illness temporally, the specificity of the exposure to the disease, the presence of a biological gradient whereby greater exposure leads to increased risk, the biological plausibility of a link between the exposure and the disease, whether epidemiological and biological findings are coherent, if experiment (either natural or designed) supports the hypothesis and when analogous casual associations have been demonstrated between other exposures and diseases. Hill urged that these criteria be used cautiously and he did not believe that it was possible to ‘lay down some hard-and-fast rules of evidence that must be obeyed before we accept cause’, nor did he believe that all or any number of his criteria must be met before causation can be inferred. Rather, he argued that these criteria should be used to ‘help us to make up our minds on the fundamental question—is there any other way of explaining the set of facts before us, is there any other answer equally, or more likely, than cause and effect?’

Applying Hill’s criteria to the association between cannabis and schizophrenia, we find a modestly strong association, consistently demonstrated in a number of studies (Moore et al., 2007). Cannabis use almost always predates psychosis, is associated with an earlier age at onset (Donoghue et al., 2014; Myles et al., 2012b) and heavier and more potent cannabis users have an increased risk of developing psychosis (Di Forti et al., 2009) and do so at an earlier age than people smoking milder cannabis (Di Forti et al., 2014). There is evidence that the endocannabinoid system effects brain functions that are relevant to schizophrenia making the relationship plausibly causal (Bossong et al., 2014). Further, there is experimental evidence that cannabis can precipitate psychotic symptoms in healthy individuals (Martin-Santos et al., 2012). While none of these findings alone is conclusive, the weight of evidence is enough to convince most researchers that cannabis plays a causal role in some cases of schizophrenia.

It has long been noted that patients with schizophrenia have very high rates of cigarette smoking, and it is increasingly recognized that this is a major contributor to the mortality gap experienced by people with mental illness (Callaghan et al., 2014). However, the notion that smoking might have a causative role, increasing risk of psychosis, is only now being seriously considered. Previously, the association was assumed to reflect patients’ self-medication for distress, boredom or the side effects of antipsychotic drugs (Poirier et al., 2002). In fact, the self-medication hypothesis has received little support (Hahn et al., 2013) and other explanations for the association must be considered. One possibility is that tobacco use is causally associated with psychosis. Researchers have often overlooked the obvious – that almost all cannabis users also smoke tobacco – so could tobacco use be the real cause of the reported association between cannabis and psychosis? Or could both cannabis and tobacco smoking independently or interactively contribute to the risk of psychosis?

In this issue of the journal, McGrath and associates examined whether age at first tobacco use was associated with diagnosed non-affective psychosis, hallucinations or delusion-like experiences among 3752 individuals born at the Mater Hospital in Brisbane, Australia, between 1981 and 1984 (McGrath et al., 2015). They found that those who commenced tobacco use aged 15 years or younger were significantly more like to have non-affective psychosis, delusion-like experiences and hallucinations. This increased risk persisted after excluding cannabis users, and adjusting for age and sex. This finding suggests that there may be a biological gradient in the association between exposure to tobacco and psychosis risk and is consistent with the biologically plausible idea that exposure during a critical period of brain development is especially associated with psychosis risk. The results of this study are also consistent with three earlier studies. In the first study, cross-sectional data from the Australian National Survey of Mental Health and Wellbeing showed that those who used tobacco daily and particularly heavy tobacco users and those who stated smoking before the age of 15 were more likely to have had delusion-like experiences (Saha et al., 2011). In the second study, cigarette smoking was associated with psychosis-like experiences in 1929 healthy adults in the Netherlands, an association that persisted after adjusting for cannabis use (Van Gastel et al., 2013). In the third study, data from the Avon Longitudinal Study of Parents and Children found that tobacco use before the age of 18 better explained the emergence of psychotic experiences at the age of 18 than did cannabis use at a similar age (Gage et al., 2014).

A recent meta-analysis examined both the strength of the association between tobacco use and psychosis and the temporal relationship between exposure and illness (Myles et al., 2012a). This study examined smoking in first episode psychosis in order to minimize the any confounding effect of self-medication. The results were that worldwide, 59% people with first episode psychosis use are regular tobacco users, that people with first episode psychosis have a six-fold greater probability of being smokers than the general population and that smoking initiation occurs on average 5.3 years before the onset of psychosis. Consistent with these epidemiological findings are clinical findings that among patients with existing schizophrenia, a greater degree of nicotine dependence is associated with more severe psychotic symptoms (Krishnadas et al., 2012).

Finally, genetic evidence also implicates tobacco use in the biology of schizophrenia. For example, DNA-sequence variations in the CHRNA5-A3-B4 gene cluster are strongly associated with heavy smoking, and an associated variant was recently found by the Psychiatric Genomics Consortium to be associated with schizophrenia (Flint and Munafo, 2014), implying either that such variants increase the risk of both smoking and schizophrenia or that smoking lies on the causal pathway, that is, that the genetic association between CHRNA5-A3-B4 variants and schizophrenia might be mediated by increased smoking.

McGrath and colleagues are rightly cautious about the implications of their findings. However, it can be seen that the association between smoking and psychosis is statistically strong, temporal, has a biological gradient and is consistent and plausible. The recent Australian study adds further weight to the emerging evidence of a causal association between tobacco use and psychosis. While not conclusive, it is this sort of evidence that we have to consider while making up our minds about the nature of this important and worrying association.