Many Mental Elves have blogged about cannabis and psychosis in the past. For example, we know that cannabis use in adolescence is associated with an increased risk of psychosis diagnoses in adulthood in a dose-response manner (Richardson, 2018). We also know that the more frequent use and higher potency of cannabis are associated with elevated risks of psychosis (Sheridan Rains, 2019).
But what would happen to the risk of psychosis if we greatly reduce the barrier to accessing cannabis at the population level? Here is the great Canadian (quasi-) experiment. Medical cannabis has been available in Canada since 2001 but only for a narrow range of restricted medical conditions (pre-legalisation). In December 2015, the Federal government committed to legalising non-medical cannabis (liberalisation). When the policy came into effect in October 2018, Canada became the first nation in the world to have commercial sales of non-medical cannabis (legalisation). Myran et al. (2025) examined data from Ontario, Canada, to find out what happened to the risk of psychosis after cannabis legalisation.

What would happen to the risk of psychosis if we legalise cannabis? Can we learn from Canada where cannabis was legalised in 2018?
Methods
In this population-based retrospective cohort study, routinely collected health administrative data were used to include all residents in Ontario, Canada, aged between 14 and 65 from 2006 to 2022. The primary exposure was the diagnosis of cannabis use disorder (CUD) upon discharge from either an emergency department visit or inpatient admission. The primary outcome was the first diagnosis of schizophrenia, and the secondary outcome was the first diagnosis of psychosis not otherwise specified (NOS).
The study utilised the interrupted time series design to compare the population-attributable risk fraction (PARF) for CUD associated with schizophrenia and psychosis NOS over three different policy periods:
- pre-legalisation (2006 – 2015),
- liberalisation (2015 – 2018), and
- legalisation (2018 – 2022).
In short, an interrupted time series design collects and compares the data pattern before and after an intervention. Considered one of the strongest quasi-experimental designs, it allows the evaluations of intervention when randomised controlled trials are unethical or impractical to implement (Hudson et al., 2019). The PARF estimates the proportion of cases that would have been eliminated if the exposure was absent. In the study, the PARFs estimate the proportion of new-onset schizophrenia or psychosis NOS if nobody had CUD.
Results
The study included over 13 million people over 17 years. Of which, 118,650 (0.9%) individuals were diagnosed with CUD and 91,106 (0.7%) individuals were diagnosed with schizophrenia. The proportion of people who developed schizophrenia was disproportionately higher among the CUD group compared to the non-CUD group (8.9% v 0.6%).
Over the three policy periods (pre-legalisation, liberalisation, and legalisation):
- The prevalence of CUD in the past three years (per 100,000 person-years) increased from 35.4 to 143.3, then 182.4.
- The incidence of schizophrenia (per 100,000 person-years) remained stable at 53.5, 52.8, and 53.3.
- The incidence of psychosis NOS (per 100,000 person-years), however, increased from 33.9 to 45.8, then 54.3.
When examining the proportion of people with CUD with the new onset diagnosis of schizophrenia or psychosis NOS over the same periods:
- The incidence of schizophrenia among people with CUD increased from 7.0% to 11.8%, then 16.7%.
- The incidence of psychosis NOS among people with CUD also increased from 6.7% to 10.8%, then 14.9%.
The hazard ratio (HR) for developing schizophrenia or psychosis NOS for people with CUD compared to those without remained consistently elevated over the three policy periods:
- For schizophrenia, the HR increased slightly from 2.60 (95% confidence interval [CI]: 2.30 to 2.89) to 2.83 (95% CI: 2.75 to 2.92) and then 3.07 (95% CI: 2.67 to 3.47).
- For psychosis NOS, the HR increased from 4.17 (95% CI: 3.58 to 4.76) to 4.29 (95% CI: 4.09 to 4.49), then 5.13 (4.38 to 5.89).
Finally, the PARFs for incident schizophrenia or psychosis NOS associated with CUD roughly tripled over the same policy periods in both groups:
- The PARF for CUD and schizophrenia increased from 3.68% to 7.26% then 10.27%.
- The PARF for CUD and psychosis NOS increased from 4.47% to 8.00% then 11.64%.
Although the PARFs showed a general pattern of increase over the three policy periods for both sexes and across all the age groups, there were notable age and sex disparities. For example:
- The PARF for CUD and schizophrenia during the legalisation period was the highest among males aged between 19 and 24 at 18.88%, compared to the lowest group (females aged between 45 and 65) at 1.81%.
- In both sexes, the PARFs for CUD and schizophrenia were much higher for the younger groups (25 and under), and for all the age groups, the PARFs were higher for males than females.

In this large population study, rates of schizophrenia were much higher in people with cannabis use disorder (8.9%) compared to those without (0.6%).
Conclusions
The authors concluded:
the proportion of incident cases of schizophrenia associated with CUD almost tripled during a period encompassing ongoing liberalisation of medical and non-medical cannabis. Although the proportion of cases of schizophrenia associated with CUD increased fairly linearly over time, incident cases of psychosis NOS and the proportion associated with CUD accelerated after cannabis liberalisation.
The proportion of incident cases of schizophrenia associated with cannabis use disorder almost tripled during a period encompassing ongoing liberalisation.
Strengths and limitations
The study has many strengths. The authors formulated and addressed a clearly focused question using valid and reliable exposure and outcome variables, capturing almost all the eligible people in Ontario. They were able to adjust for many potential confounders, including income, rural residence, immigration status and past hospital utilisation for mental health or substance-related issues. The study also analysed the data stratified by age and sex, given the differences in age of onset between the two sexes. The findings showed a consistent pattern of increasing rates among the proportions, HRs and PARFs for schizophrenia and psychosis NOS associated with CUD over the three policy periods.
As acknowledged by the authors, the main limitation was related to the PARF estimations. To state with confidence the proportion of cases that would have been eliminated if the exposure was absent, all the relevant confounders need to be adjusted for. There were some obvious confounders, such as family history, education status and potency of available cannabis, that the study could not adjust for. Further, not all people who experience psychosis end up in the hospital, so those who were diagnosed with either schizophrenia or even psychosis NOS were likely to have experienced severe and/or enduring psychosis. Similarly, few people who consume cannabis would end up in hospital, and even fewer would be diagnosed with CUD (even if they meet the diagnostic criteria). There is also likely to be a large degree of uncertainty and inconsistency around these diagnoses in practice. The diagnosis of CUD, in particular, relies heavily on the amount of information available to the clinician to the degree the patient is willing to share. Thus, for both exposure and outcomes measures, the study probably only captured those on the pointy ends of the spectrum. This could have either underestimated or overestimated the psychotogenic impact of cannabis. Nevertheless, the data in the current study provide a good enough estimation of such diagnoses in a volume that would be much larger than possible with a controlled study.
Finally, for most of the legalisation period (between October 2018 and December 2022), Canada (and the rest of the world) was preoccupied by a global pandemic. Given that the impact of any policy change has a significant time lag, it is unlikely that the findings from the study have captured the full impact of cannabis legalisation.
Implications for practice
So, what would happen to the risk of psychosis if we greatly reduce the barrier to accessing cannabis? It looks like the risk of psychosis goes up.
In the invited commentary about the current study, Gilman (2025) argues that the challenge associated with finding causation in a quasi-experimentation is that cannabis legalisation is not a simple intervention. It usually takes years after the legalisation for fully functional commercial markets to emerge, and the social acceptance and consumption of cannabis use often precipitates the legalisation. As her concluding paragraph states; “Policy does matter; allowing unfettered commercial markets to exist exposes more individuals to a greater variety of readily available, high-potency cannabis products”.
The adverse health outcomes are not limited to psychosis. For instance, Wilson recently blogged about another Danish delight that showed that CUD was associated with an increased risk of both unipolar and bipolar disorders (Wilson, 2024). Another blog reported the increased CUD diagnosis among veterans with chronic pain following cannabis legalisation (Williamson and Leightley, 2024). In another blog, Turner reported that social environment, especially peer deviance, was more influential in cannabis use among university students than genetic vulnerabilities (Turner, 2022). Cannabis commercialisation is perhaps the strongest peer pressure that a society can put on its young people.
Yet, it is crucial to be mindful that health is only a small part of politics. Cannabis is a luxurious business that lubricates communities financially. Sure, they may over-sell the benefits and under-report the risks of their products, but what companies don’t? Only a small proportion of people who smoke cannabis will develop CUD, and even a smaller proportion of people will subsequently develop psychosis. Certainly, many patients vote with their feet. Nowadays, I rarely get through my work week without at least one patient telling me how amazing medicinal cannabis is for their mental health condition.
I can’t help but wonder if we may have missed an opportunity to conduct a proper scientific experiment with cannabis. We could have scientifically investigated the right dose and the right potency for different mental disorders to optimise individual responses to cannabis. Instead, what we now have looks more like pseudo-science based on marketing trends and consumer preferences through quasi-experiments. In his 2022 blog, Hamilton proposed that “the acid test is whether you would be happy for your son or daughter to use cannabis in their adolescence” (Hamilton 2022). To rephrase his question; as a nation, would we be happy for the next generation to think that it is acceptable to use cannabis in our society? Is the question still relevant, or has the train left the station already?

Would we be happy for the next generation to think that it is acceptable to use cannabis in our society? Or has that train already left the station?
Statement of interests
None.
Links
Primary paper
Myran DT, Pugliese M, Harrison LD, et al (2025) Changes in Incident Schizophrenia Diagnoses Associated With Cannabis Use Disorder After Cannabis Legalization. JAMA Netw Open. 2025 8(2):e2457868. https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2829840
Other references
Gilman J, Association of Cannabis Legalization With Prevalence of Schizophrenia—Challenges of Attributing Biological Causality to Policy Change. JAMA Netw Open. 2025 8(2):e2457876. https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2829844
Hamilton I. Young adults do not quit cannabis because of psychosis symptoms, according to new Europe-wide research. The Mental Elf, 12 Jan 2022. https://www.nationalelfservice.net/mental-health/substance-misuse/cannabis-discontinuation-psychosis/
Hudson J, Fielding S, and Ramsay CR (2019) Methodology and reporting characteristics of studies using interrupted time series design in healthcare. BMC Med Res Methodol 19, 137. https://bmcmedresmethodol.biomedcentral.com/articles/10.1186/s12874-019-0777-x
Richardson T. Adolescent cannabis use increases risk of an adult psychotic diagnosis. The Mental Elf, 29 May 2018. https://www.nationalelfservice.net/mental-health/psychosis/adolescent-cannabis-use-increases-risk-of-an-adult-psychotic-diagnosis/
Sheridan Rains L. Daily skunk cannabis use associated with a 5-fold increase in psychosis risk. The Mental Elf, 30 April 2019. https://www.nationalelfservice.net/mental-health/substance-misuse/daily-skunk-cannabis-use-associated-with-a-5-fold-increase-in-psychosis-risk/
Turner S. Cannabis use in college: genetic predispositions less influential than social environment. The Mental Elf, 21 Oct 2022. https://www.nationalelfservice.net/mental-health/substance-misuse/cannabis-use-in-college/
Williamson G and Leightley D. Cannabis use and its legalisation: analysing chronic pain in US veterans using electronic health records. The Mental Elf, 9 Feb 2024. https://www.nationalelfservice.net/mental-health/substance-misuse/cannabis-chronic-pain-veterans/
Wilson J. Cannabis use disorder associated with increased risk of both psychotic and nonpsychotic unipolar depression and bipolar disorder. The Mental Elf, 12 June 2024. https://www.nationalelfservice.net/mental-health/substance-misuse/cannabis-use-disorder-depression-bipolar-disorder/