• Cannabis;
  • causation;
  • psychotic symptoms

Macleod & Hickman [1] focus upon the relationships between cannabis and schizophrenia and dismiss research examining the linkages between cannabis and symptom measures on the grounds that the linkage between symptom levels and clinical outcomes is not known. While this focus favours the arguments they develop, it does not represent adequately the literature that they claim to be reviewing. The focus of this literature has not been upon the specific linkages between cannabis and schizophrenia, but rather upon whether the use of cannabis increases rates of psychotic symptoms [2].

Macleod & Hickman propose that any link between cannabis use and psychotic symptoms could be explained by residual confounding or reverse causation. Because of their focus upon schizophrenia, they minimize recent research that has examined these issues using symptom measures. This evidence has been examined in a number of recent reviews [2–8], all of which have concluded that the weight of the evidence favours the view that the associations between cannabis and increased rates of psychotic symptoms are likely to be causal. Perhaps the most comprehensive review of this evidence is provided by Moore et al.[4], who noted that that the findings of associations between cannabis and psychotic symptoms have been replicated by a series of seven cohort studies in which: (a) approximately 60 confounding factors have been controlled; (b) one study controlled non-observed confounding; and (c) in all studies reverse causality had been controlled. All studies produced evidence consistent with the view that there is a modest association between increasing use of cannabis and increased rates of psychotic symptoms. Further, as Hall has pointed out [3], these findings are supported by these additional lines of evidence: laboratory findings in animals; double-blind provocation research with people with schizophrenia; and by evidence of gene × environment interactions.

Although all this evidence could be dismissed on the grounds of inappropriate measurement, failure to control residual confounding and reverse causality, such arguments are neither parsimonious nor compelling. In order to sustain their position that cannabis may not be related causally to psychosis/psychotic symptoms, Macleod & Hickman have to engage in an increasingly elaborate set of arguments which: (a) focus the debate exclusively on schizophrenia rather than psychotic symptoms; (b) dismiss recent studies that have controlled confounding and reverse causality extensively; and (c) discount evidence of underlying biological mechanisms.

As Macleod & Hickman note, their arguments are largely underwritten by analogy with ‘several recent studies much more methodologically robust than those relating to cannabis use to psychosis, which have led to some of the most important mistakes in recent epidemiological history’. The issue to which they refer concerns the discrepancies between observational studies and randomized trial findings in the area of hormone replacement therapy (HRT) [9]. In that area, observational studies suggested positive benefits in reducing coronary heart disease (CHD), whereas randomized control trials led to the opposite conclusion. Here is not the place to review this literature, but two points are worth making. The first is that the discrepancy between randomized controlled trials and observational studies for the HRT and CHD was an exception to the general rule that the conclusions of observational studies have been found to be consistent with randomized controlled trials (RCTs) [10,11]. Secondly, an examination of the reasons for observational studies of HRT producing misleading findings appears to have been due to a failure to control socio-economic factors adequately [9]. These limitations in the assessment of confounders do not apply to the literature on cannabis and psychotic symptoms.

In summary, while it remains possible that associations between cannabis and psychotic symptoms are non-causal and can be explained away as a result of inadequate measurement, residual confounding and reverse causation, the weight of the evidence, including observational research, animal studies, laboratory research and behavioural genetic research all points to a causal process in which increasing use of cannabis is associated with small but detectable increases in the risks of psychosis and psychotic symptoms. While the HRT example posts a warning about the pitfalls of causal reasoning with observational evidence, this example is, in fact, an exception to the general rule that the findings of well-conducted observational research are usually consistent with the findings of well-conducted randomized trials. Despite Macleod & Hickman's reservations, a growing number of well-conducted observational studies suggest the presence of a causal link between increasing use of cannabis and increasing rates of psychosis/psychotic symptoms.


  1. Top of page
  2. Declaration of interests
  3. References