Does Cannabis Cause an Earlier Onset of Psychosis? Comments on a Recent Meta-Analysis

Written by Michael Lynskey, Drug Science Chief Research Officer

Psychosis is a severe and debilitating condition with relatively few effective treatments. For this reason, there is ongoing interest in identifying factors that may make someone more likely to develop psychosis, with the aim of developing intervention and prevention strategies.

One particular area of interest is cannabis use, especially use that begins during adolescence. A landmark study of Swedish conscripts showed that individuals who reported cannabis use by age 18 were more likely to later be admitted to hospital for psychosis. Since then, considerable research has explored whether cannabis use, and particularly early cannabis use, may increase the risk of, or even ‘cause’, psychosis. One reason for this interest is that several of the established risk factors for psychosis, including genetic influences and urbanicity, are difficult or impossible to change.

Given the relatively low lifetime prevalence of psychosis (estimated at 1-2% of the population), traditional prospective designs have been difficult to conduct because they require exceptionally large sample sizes. These studies have been highly informative in understanding trajectories of more common mental health conditions, such as major depression, and are often considered the ‘gold standard’ for studying risk over time. Instead, much of the research on psychosis has relied on case-control studies. These look  retrospectively at potential risk factors and compare individuals with an established diagnosis of psychosis with individuals without one.  

A recent meta-analysis by Carly Stevens and colleagues examined the associations between the age of first cannabis use and the age of onset of psychosis. This was a substantial piece of work: a systematic review identified over 5,000 publications, of which 149 had sufficient information to be included in the meta-analysis. Together, these studies included data from 71,073 people with psychosis, 18,272 of whom reported lifetime cannabis use. Overall, cannabis use was associated with the onset of psychosis occurring an average of 2.5 years earlier.

As the authors note, this finding replicates previous results, and they argue that this replication indicates the association is ‘robust’.  However, several aspects of the findings deserve further consideration.

Firstly, when subdividing participants according to their age at onset of psychosis, they found that the association between early onset cannabis use and risk of psychosis was limited to people whose psychosis began after the age of 20. In other words, cannabis use during adolescence was not associated with earlier onset psychosis among people who developed psychosis as a young age.

Instead, the association became stronger the later psychosis developed. Among cannabis users whose psychosis began between the ages of 20 to 25, onset occurred an average of  1.60 years earlier.  For those whose psychosis began between 25 and 30, the difference was 4.43 years, rising to 6.30 years among those whose psychosis began after age 30.

Stevens et al note that this finding was unexpected and suggest the possibility that it was due to a ‘floor effect’, meaning that for individuals who develop psychosis very early, there may be little scope for cannabis use to bring its onset forward any earlier.

They also suggest that the association between early cannabis use and later (but not early) psychosis may be attributed to the residual effects of adolescent exposure to cannabis. However, this argument ignores the fact that many people who use cannabis do so infrequently (perhaps only a handful of times in the lives). In her landmark study on the natural history of drug use during adolescence and young adulthood, Denise Kandel (Chen & Kandel, 1995) found that 75% of those who had ever used cannabis had not used it in the previous year when they were re-interviewed at age 34-35. Further, the analyses reported by Stevens et al controlled for ‘lifetime extent of cannabis exposure.’ If the association between early cannabis use and later psychosis was driven by continued use of cannabis, controlling for overall exposure would be expected to weaken that association.

It is also worth considering cannabis alongside other drugs. Studies have found that tobacco and alcohol are also associated with an increased subsequent risk of psychosis, and that the magnitude of these associations is similar across all three drugs. Coupled with findings that even infrequent cannabis use may be associated with subsequent risk of psychosis, this suggests that the associations between cannabis, alcohol, and tobacco and psychosis may not reflect a direct impact of these drugs themselves. Instead, it could partly reflect other factors shared by people who use substances but which have not been adequately accounted for in studies. It could also reflect aspects of a person’s environment or consequences associated with drug use (e.g., school exclusion), which may themselves be related to later psychosis risk, independently of the pharmacological properties of cannabis, tobacco, or alcohol.

Finally, the studies included in the meta-analysis, like much of the research in this field, rely on people retrospectively reporting when they first began using cannabis, which could have been years or even decades later. Importantly, errors in this reporting may be related to later adverse outcomes, and retrospective reports may therefore inflate the apparent associations between early onset and later outcomes.  

Stevens et al are appropriately cautious about potential limitations of their study while also noting potential advantages of delaying the onset of cannabis use if this association is indeed causal. However, a continued emphasis on the potential influence of cannabis on risks of psychosis may neglect other potential causes and the development of appropriate prevention, intervention and treatment options.