Reply to Zhou et al’s “Refining Psychosis Research: Insights on Cannabis Use and Data Accuracy”

Dear Editor,

We thank Zhou et al for their letter responding to our recent publication by Brink et al¹ on the role of social determinants in the wider environment on incidence rates of psychotic disorder. Zhou et al highlight 3 important critiques of our work, which require proper contextualization as we seek to understand how the social environments in which we live shape the burden of psychotic disorders within the population. Their points concern (1) cannabis complexity, (2) reliability, and (3) other adversities arising from harmful cannabis use. We respond to each of these points in turn.

Cannabis Complexity

Zhou et al note that the measures we used in the EU-GEI study to estimate the level of cannabis use within the population at risk in each of our 14 settings were relatively simple proxies (for the proportion of the population at risk who were estimated to use cannabis on a daily basis, or who had ever used high-potency cannabis). Our estimates for the population at risk were obtained from population-based controls who took part in the EU-GEI study in each setting, weighted to be representative of the underlying population at risk, by broad age group, sex, and ethnicity.

We agree with Zhou et al that the cannabis exposures that affect individual-level risk of psychosis will be more complex and nuanced than whether someone reports consuming cannabis on a daily basis, or reports having tried high-potency cannabis at least once in their lifetime. Several cannabis-related factors, which may operate synergistically to affect individual-level risk, have been associated with individual risk of psychosis, including age at exposure,² genetic liability,³ potency,^4–7 and frequency of use.^5,6 Zhou et al also speculate that reasons for use, including social vs lone consumption, may play a role in individual risk of psychosis, although earlier work found a strong effect of individual cannabis use on risk of schizophrenia, independent of social integration.⁵ How or whether these factors also influence the population-level incidence rate of psychotic disorders is a potentially distinct issue, which may have a set of overlapping and/or distinct determinants. Our study was an attempt to investigate this issue at the population level, using 2 important measures of cannabis use—frequency and potency. We found no evidence that either daily or high-potency cannabis use in controls—as a proxy for levels of usage in the population at risk—were associated with the population-level incidence of non-affective psychotic disorders, and only limited evidence for an effect of daily usage on incidence of affective psychotic disorders (with no effect of high-potency use in the population at risk). The question then arises, as to why we may see strong individual-level associations between frequent or high-potency cannabis use and individual psychosis risk, but observe null to modest effects on rates of psychotic disorder at the population level.

Well, in part, current levels of frequent or high-potency cannabis use in the population (as estimated in Brink et al¹) may be too low to have detectable changes in incidence at the population level for a set of rare disorders, despite the strong increased individual risk they convey. It is also possible that cannabis consumption itself is determined by other factors that are common causes of both cannabis usage and psychosis outcomes, as we discuss in the paper. Indeed, in Brink et al,¹ we found evidence that owner-occupancy, as a marker of socioeconomic deprivation and/or social fragmentation, was a common cause of both the amount of cannabis use in the population at risk and the incidence of psychotic disorders, and was the larger driver of incidence rates at the population level. Understanding the upstream social determinants of mental health conditions will be vital for developing appropriate points for public mental health intervention.⁸ These concepts tie in with Geoffrey Rose’s prevention paradox⁹—that although heavy or high-potency cannabis use increases individual risk of psychotic disorder, at the population level, we may prevent more cases of psychosis (and reduce cannabis use) by investing in strategies that seek to lift people out of deprived circumstances. A third possibility, is that the true causal association between cannabis use and psychosis risk is more modest than suggested in individual-level observational studies to date, where unmeasured confounding, selection bias, and reverse causation may all feature. Recent genetically informed designs,¹⁰ natural experiments,¹⁰ and other quasi-experimental studies¹¹ provide sobering caution on these issues, leading Gillespie and Kendler¹⁰ to conclude that “claims made about the changes in risk for schizophrenia stemming from changing levels of cannabis use are very likely to be exaggerated and potentially substantially so.” Our findings, are consistent with these possibilities. In our hunt for the causes of psychosis, it is vital we understand the relative causal influence of different risk factors on both individual risk and the population-level expression of psychosis, as this will shape how we design successful universal, selected, and indicated prevention strategies in public mental health. Zhou et al’s letter reminds us of the need to think clearly about how risk affects mental health outcomes at different levels of causation.

Reliability

A second concern raised by Zhou et al was that our estimates of cannabis use were based on self-report. This is, of course, a limitation of any studies which use questionnaire-based instruments to elicit answers on cannabis use, and something we include in our list of limitations.¹ To collect cannabis use data we used the modified Cannabis Experiences Questionnaire (CEQ), which has established content, construct, and face validity, and has been reliably deployed in several settings.⁶ As specified in Brink et al,¹ we also attempted to minimize potential selection bias (of controls who may have been interested in participating in the study in part because of a focus on cannabis use) by not mentioning cannabis in documents used to recruit controls. All EU-GEI data were pseudonymized and analyzed at group level to reduce possible social desirability bias in answering the questionnaire. We agree with Zhou et al that any methodological advances that result in unbiased estimates of cannabis use history are welcome. We believe we have taken adequate measures to ensure reliability of our cannabis measures.

Other Mental Health/Social Outcome Impacts of Cannabis Use

We agree with Zhou et al that there may be wider societal and public health consequences associated with population-level cannabis use that require identification, prioritization, and prevention, including “the identification and management of cannabis-related psychiatric symptoms.” However, public mental health must also balance the harmful effects of any putative target for intervention alongside any potential beneficial impact on a range of outcomes. Cannabis use, eg, appears to have documented harmful effects beyond psychosis, particularly for young people, and extending to effects on other psychiatric symptoms,^12,13 suicidal outcomes,¹³ and cognition.¹³ But it may also offer beneficial effects in terms of symptom or pain relief, and improvements in quality of life for other health conditions, including inflammatory bowel disease, multiple sclerosis, and epilepsy (though side effects also exist).¹³ Designing appropriate and effective intervention strategies that prevent harms, promote positive outcomes, and protect benefits requires a nuanced understanding of impact across a very broad set of mental, physical, and social outcomes; designing such strategies based solely on their impact on a single outcome (ie, psychosis) will miss bigger opportunities to improve lives.

Fundamentally, however, we still need more robust data on how levels of cannabis use in the population at risk affect the incidence of psychotic disorders to inform evidence-based public mental health decision-making. Our work provides one such attempt, but we should also leverage strong, quasi-experimental approaches in epidemiology. Recent evidence from Ontario, Canada, by Anderson et al¹¹ provides one such example. They found there was no initial increase in the incidence of psychotic disorders or psychosis-related health service visits up to 18 months after cannabis legalization in the province.¹¹ Nonetheless, longer follow-up data from this group suggests that emergency department visits for cannabis-induced psychosis have increased among young adults (19–24 years old) over the pre-legalization period by up to 63%, though such effects were absent for younger adolescents (below the legal age to purchase cannabis in the province).¹⁴ We agree with Zhou et al that careful surveillance of psychiatric morbidity that may be potentially associated with cannabis use is now required to develop commensurate clinical and public health interventions positioned against the wider public health impact of cannabis usage.

Funding

This work was supported by the European Community’s Seventh Framework Program (HEALTH-F2-2010-241909 [Project EU-GEI]) and the São Paulo Research Foundation (2012/0417-0). V.B. was supported by an internal research grant from the University Medical Centre Groningen (MD-PhD 18-41). J.B.K. was supported by the National Institute for Health and Care Research (NIHR), University College London Hospitals NHS Foundation Trust (UCLH), Biomedical Research Centre (BRC). D.Q. was supported by a Medical Research Council (MRC)/UK Research and Innovation(UKRI) Clinical Academic Research Partnership (CARP) (MRC CARP grant MR/W030608/1).