Cannabis and Psychosis: Weeding Out Fact from Fiction

Cannabis use remains common in the United States. The rapidly changing landscape of cannabis regulatory policies may be contributing to these trends, although evidence remains mixed.¹ While studies are conflicted about the impact of recreational (RML) and medical (MML) marijuana laws on cannabis use, per se, there appears to be emerging support for an uptick in emergency department visits, traffic injuries and fatalities, and burgeoning cases of poisoning in toddlers and preschool-aged children, possibly due to increasing availability of edible forms of the drug.^2,3

Beyond these immediate consequences of cannabis exposure and use lies a vast and concerning literature addressing elevated rates of mental health problems in individuals using cannabis. In particular, individuals with schizophrenia, psychotic disorders, and psychosis are considerably more likely to report chronic cannabis use and problems with cannabis compared to those without these mental health conditions.⁴ Similarly, rates of schizophrenia, psychotic disorders, and psychosis-like behaviors are overrepresented in individuals who use cannabis heavily compared to the general population.⁵ It has long been assumed that these associations, which are particularly elevated in individuals who initiate use at an early age, reflect a causal effect of cannabis use and use disorder (CanUD); i.e., a psychiatric diagnosis reflecting tolerance, withdrawal, loss of control, and impairment due to use) on schizophrenia and psychosis. In this selective review, we outline the evidence linking cannabis use and CanUD to schizophrenia and psychosis and detail evidence supporting and refuting causal assertions.

Epidemiological Evidence

The relationship between cannabis and schizophrenia, and whether the former exerts a causal influence on the latter, is an enduring epidemiological hypothesis. In an initial study of >45,000 Swedish conscripts, Andreassen et al.,⁶ found that those who reported heavy cannabis use (more than 50 times) were at six-fold greater relative risk for schizophrenia onset, an association that persisted when accounting for premorbid personality features. These large-scale observations are echoed in US epidemiological surveys as well (e.g.,^4,5). Subsequent contemporary studies have added nuance to this observation, highlighting the exacerbating influence of earlier-onset cannabis use, higher potency cannabis, and heavier use that may meet criteria for CanUD.

For instance, one study reported that daily use of a particular form of high-potency cannabis (> 10% THC) was almost five-fold more common in first-episode psychosis cases in the United Kingdom.⁷ In another study that accessed the extant health records of the Danish population, researchers reported that the population attributable risk fraction for cannabis use disorder in schizophrenia has markedly increased over time, from 2% in the early 1970s to almost 30% in males aged 21–25 by 2020, corresponding to an adjusted incidence hazard ratio of nearly 3.0, which the authors contend is consistent with increases in its availability and potency, and consistent with the rising incidence of schizophrenia.^8,9 This study concluded that young males were particularly vulnerable to schizophrenia in the context of CanUD.

Transient and Induced Psychosis, and Schizophrenia Conversion

Heavy and high potency cannabis use can induce psychosis. Less than half a percent of individuals using cannabis report related psychosis that requires emergency medical treatment, with paranoia being the most prevalent feature; younger age is associated with >2.5 relative risk.¹⁰ In a recent study of ED visits for acute psychosis, urinary positivity for delta-9 tetrahydrocannabinol (THC) was associated with agitation and symptom acuity.¹¹ A mega-analysis of human laboratory studies that involved intravenous THC infusions found clinically meaningful positive symptoms in ~45% of the samples, with conceptual disorganization and hallucinations being common. Notably, negative symptoms (e.g., blunted affect) were also commonly identified.¹² The authors noted a dose-dependent relationship related to both THC concentration and pace of infusion, which roughly corresponded to three-quarters to a full standard joint.¹³ These studies suggest that cannabis-induced behavioral alterations are likely to involve a wide range of symptoms that broadly map onto positive, negative, and cognitive features of schizophrenia. In a recent study conducted by us, where participants self-reported symptoms that they experienced upon using cannabis, decreased social contact, cognitive difficulties, and dysphoria/anhedonia were far more commonly reported than paranoia or hallucinations.¹⁴ In this study, younger age of cannabis initiation, longer duration of daily use, and a CanUD diagnosis were robust correlates of these experiences.

Thus, it is known that cannabis use can result in transient or acute psychosis—but what proportion of the individuals who experience psychotic symptoms have a prior history of psychotic disorders or eventually convert to schizophrenia? In our study, even though active psychotic disorders were exclusionary, genetic liability to schizophrenia (measured using published estimates of genome-wide genetic associations with schizophrenia) was associated with cannabis induced experiences, although the association with cognitive difficulties was much stronger than that with paranoia. Another retrospective study found that medical authorizations for cannabis were associated with 1.63 hazards of a psychosis-related ED visit in patients with no record of prior psychotic disorders.¹⁵ Similarly, use of high potency cannabis at age 16–18 was associated with a doubling of the odds of incident psychosis by ages 19–24.¹⁶ A study of the Danish Civil Registry indicates that 47% of those with cannabis-induced psychosis convert to schizophrenia within three to four years.¹⁷

Meta-Analytic Evidence

A recent umbrella review of 32 eligible studies weighed support for and against several research questions related to cannabis and psychosis (e.g., schizophrenia, schizotypy, psychosis-like experiences).¹⁸ They concluded that while there was notable support for an association between cannabis use and psychosis along the continuum, a majority of the studies were cross-sectional and thus, limited in their inferential capacity. An earlier meta-analysis of 18 studies estimated a meta-analytic odds ratio of ~4.0, alongside a dose-response effect.¹⁹ Examining facets of schizophrenia, one meta-analysis of 65 eligible studies reported that frequent cannabis use was associated with elevated likelihood of positive symptoms, and excitement, but lower severity of negative symptoms and not associated with disorganization.²⁰ Here as well, the authors concluded that the evidence was not sufficient to prove causation.

Causal or Confounded?

There are numerous alternative explanations that must be considered. From a clinical perspective, researchers have argued that early onset of cannabis use, especially heavy, high frequency use or chronic use of high potency cannabis, may be associated with behavioral traits associated with schizophrenia (e.g., early social disengagement, negative affect), premorbid symptoms, or correlated psychopathology (e.g.,²¹). In other words, individuals may start using cannabis at a young age to cope with behaviors that are harbingers of schizophrenia, and this use may further the course of the illness.²² Others argue that psychosis consequent to heavy cannabis use may be a form of acute psychosis, which, while of prognostic value in schizophrenia,²³ may also be distinct from it.²⁴ Younger age and male gender appeared to be risk factors for conversion from onset of psychosis to schizophrenia diagnosis within a three-year period, and this was most pronounced for cannabis relative to other drugs.^17,25 Yet, according to projection modeling, heavy cannabis use would have to be prevented in 2,800 males aged 20–24 to prevent a single case of schizophrenia, which suggests that cannabis-related increases in schizophrenia will be subtle.^26,27 However, given the increasing availability of high potency cannabis, the impact on cannabis-induced acute psychosis, if not schizophrenia diagnoses, may well be expected to shift.²⁸ Thus, while plausible alternatives to the causal hypothesis exist, no epidemiological data have been able to completely refute an independent effect of (certain forms of heavy or high potency) cannabis use, nor prove it.

Genetic Studies Provide Clues

Cannabis use and CanUD are moderately heritable, with 50% of population differences due to the additive effect of segregating genes, while schizophrenia is highly heritable with genetics implicated in up to 80% of population variability.^29,30 Given this heritability, some of the same genes that predispose someone to schizophrenia might also increase risk of heavy cannabis use, particularly if it is used to cope with early behavioral features. To estimate the degree to which the same genetic factors may influence both schizophrenia and cannabis use, researchers typically turn to studies of twins. Identical twins share 100% of their segregating genetic material, while fraternal twins and siblings share, on average, 50% of their DNA. Comparing the degree of correlation between these types of relatives offers preliminary insight into the role of shared genetics, with studies reporting genetic correlations between cannabis use and psychosis-like experiences of 0.40 – 0.56.^31,32 However, as schizophrenia is rare and thus poorly captured in population-based twin studies, more persuasive evidence for shared genetic mechanisms arises from recent genome-wide association studies (GWAS), which are “hypothesis-free” queries of the association between a phenotype and common genetic variation across the genome as indexed by single nucleotide polymorphisms (SNPs, i.e., changes in individual nucleotide base pairs). The SNP-explained correlation between the current largest GWAS of schizophrenia³³ and CanUD³⁴ is 0.37, and is statistically greater than the corresponding genomic correlation between schizophrenia and cannabis use [rG = 0.24].^34–36 These findings highlight the importance of considering shared genetic liability, which can also contribute to dose-response patterns of association.

Both CanUD and schizophrenia are also polygenic; that is, many genetic variants of small effects, in combination, shape the genetic etiology of these disorders. We can estimate a score that aggregates across these small effects (e.g., by taking the effect sizes from a large GWAS and weighting a person’s own genotype by these effects).³⁷ One study found that such a polygenic score (or PGS) for cannabis use and CanUD improved prediction of psychotic disorders beyond the corresponding PGS for psychotic disorders themselves, indicating genetic overlap.³⁸ Another study found that individuals with high PGS for schizophrenia showed stronger associations between cannabis use and risk for psychotic-like experiences, such as auditory hallucinations, suggesting a gene-environment interaction effect between genetic risk for schizophrenia and cannabis use.³⁹

Another way to understand what underlies the genetic correlation is by pinpointing the specific genetic variants that influence both cannabis use and schizophrenia. We identified 327 independent genetic signals that were associated with both CanUD and schizophrenia, as well as tobacco smoking, although the genetic correlation between CanUD and schizophrenia was statistically larger than that between tobacco smoking and schizophrenia (rG = 0.37 vs. 0.17).⁴⁰ Interestingly, the genetic signals at the intersection of CanUD and schizophrenia appeared to be particularly associated with executive functioning, such that genetic variants associated with greater risk of CanUD and schizophrenia are also associated with liability for lower executive functioning. The strongest pleiotropic (i.e., same genetic signal influencing multiple behaviors) locus included CHRNA2 (neuronal nicotinic acetylcholine receptor, subunit alpha 2), which persisted as a signal even after accounting for tobacco genetics.

According to a study in the American Journal of Psychiatry, individuals who have even one psychotic episode following cannabis use have a 47 percent higher chance of developing schizophrenia or bipolar disorder. And adolescents and young adults are at the highest risk.

Causal Inference from Genetic Studies

True pleiotropy can exist alongside causation, and in some cases, even be indicative of it. Genetically informed studies, especially those that can estimate genetic correlations, are also well suited for the study of causal inferences, using both twin and GWA studies. In one such study, we compared the level of psychosis-like experiences reported by pairs of twins and similarly-aged siblings of whom one reported high levels of cannabis use while their co-twin only reported low levels or no cannabis use.³¹ We found that even after accounting for shared genetics, there was a statistically significant residual elevation in psychosis like experiences reported by those who used cannabis frequently or met criteria for CanUD. Thus, evidence for both genetic correlation and non-genetic contributors, which may potentially be causal, was noted.

Similar approaches for causal inference have also been developed to leverage measured genomic data. Borrowing instrumental variable analysis techniques and theory from epidemiology and econometrics, Mendelian Randomization (MR) and related approaches use measured genomic data to estimate causal relationships between traits.⁴¹ MR can be thought of as analogous to a randomized controlled trial, in that an individual is randomly “assigned” to their genotype at birth, so confounders should be randomly distributed across genotypes (in theory). Generally, recent well-powered studies of heavy use or CanUD have found evidence for bidirectional causal effects of cannabis on schizophrenia and vice versa.^{34,36,38,40–42}

The Environment, Prenatal Cannabis Exposure and Psychosis

In the genetic studies reviewed above, the residual association, after accounting for shared genetic liability, is considered to be indicative of “third variable” effects, which may include causation. One such environmental/epigenetic mechanism relates to in utero cannabis exposure. Research by our team has found that self-reported cannabis use subsequent to knowledge of pregnancy in the birthing person is associated with increased likelihood of psychosis-like behaviors during middle childhood, although the continued influence of prenatal cannabis exposure on psychosis during adolescence appears less clear.^43,44 Some have theorized that endogenous cannabinoid receptors, notably the ubiquitous CB1 receptors that are expressed during seven to eight weeks of gestation,⁴⁵ may serve as a mechanism potentiating the impact of cannabis on other biological systems that subsequently contribute to psychosis. As more birthing individuals turn to cannabis for pregnancy-related nausea, sleep, and mood management,^46,47 we might expect a shift in psychosis within the next two decades, if indeed cannabis contributes in utero to psychosis-related fetal reprogramming. However, controlling for maternal psychopathology, which might increase the likelihood of cannabis use during pregnancy, will have to be accounted for when ascribing causal explanations to such findings. Similarly, other postnatal early life adversities, which are likely to shape risk for cannabis use, CanUD and schizophrenia, are worth studying as potential alternate explanations for causality.⁴¹

Gene-Environment Interplay

Schizophrenia itself (and psychosis less so) is so highly heritable that the assumption of a purely environmental form—that is, solely attributable to cannabis—seems implausible. Therefore, genetic liability for schizophrenia/psychosis may be a prerequisite and cannabis a provocateur for the development of schizophrenia. Several studies of “candidate genes”, i.e., the study of investigator-selected genes or genetic variants, which suffer from notoriously high false discovery rates, have yielded no clarity on this issue of gene-environment interplay.⁴⁸ More recent studies that can estimate the genome-wide risk for schizophrenia, while finding much support for common genetic mechanisms, have also resulted in mixed support for gene-environment interactions with cannabis use. While several studies found independent but not interactive contributions of cannabis use and schizophrenia genetic liability (e.g.⁴⁹) on psychosis, one study reported that individuals at the highest genetic susceptibility to schizophrenia were more likely to report psychotic experiences if they were cannabis users.³⁹

We have previously contended that, if cannabis use is to be viewed as an “environmental” risk factor, then the appropriate explanation may lie in tests of gene-environment correlation rather than gene-environment interaction. Put another way, the use of cannabis—at an earlier age, of a higher potency, and with greater chronicity—in individuals with psychosis is not a coincidence or a chance observation. Rather, individuals who are at high genetic liability for schizophrenia are more likely to use higher potency cannabis (and at an earlier age, and with greater chronicity) because shared genetic factors elicit this heritable “environmental” exposure. Causal studies refute this possibility by documenting the absence of psychotic illness or related personality traits in those with cannabis-related incident psychosis. However, they do not account for all other markers of psychopathology that genetic studies tell us may be related to schizophrenia.

Can We Ever Prove Causation, and Do We Need To?

The dilemma with the causal hypothesis is that—in humans—it is challenging to prove without equivocation. There are guidelines, for instance, the Hill criteria.^50,51 However, such criteria can be unsatisfactory for the etiologically complex relationship between cannabis and schizophrenia.⁵² We must then ask if it is vital that causation be proven. Put another way, what do we expect if cannabis use exerts a causal influence? We may find then that as higher potency forms become more accessible, our physicians may anticipate increased ED presentations for cannabis-induced psychosis. There is no clear evidence of this as yet (e.g.,^53,54 but see55). Perhaps the period of vulnerability matters and only early exposure (e.g., prenatal or adolescent) is a risk factor? If this is the case, then we should be concerned as rates of prenatal use are on the rise,^56,57 and policies that vigorously deter use in vulnerable populations is fundamentally important. Relying on industry self-regulation (e.g., “we card” programs) is unlikely to be efficacious and significant investment in sensible policies is necessary.^58,59

Are individuals with certain genetic risk factors more vulnerable? Genetic prognostication comes with its own challenges, but one approach might be to focus additional prevention efforts towards youth with family histories of mental illness, potentially in routine medical settings. Reflecting on these expectations, two observations are evident. First, there is no instance in which a purely causal explanation exists. For instance, even if policy change were to be the sole reason for increasing cannabis use (and this is unlikely), then the impact on schizophrenia would still be titrated by genetic load for the disorder; that is, we do not expect a non-genetic form of schizophrenia, that is solely due to chronic cannabis use. Second, rational regulatory policy that deters cannabis use in youth and birthing persons, and medical vigilance of use in general is simply sensible practice that does not require a causal bedrock for implementation.

Conclusion

Without a doubt, more research into mechanisms by which cannabis exposure might amplify genetic risk for schizophrenia is needed. It is worth noting that cannabis is federally a schedule I drug, which severely restricts scientific access, especially for federally-funded research. Rescheduling cannabis could greatly facilitate such research, allowing for real-world experiments with greater generalizability. However, lack of mechanistic understanding is not a reason for complacency. The state of Missouri is now part of the national cannabis experiment. It is likely a good time for greater vigilance surrounding cannabis use, which should be routinely queried in medical settings, similar to alcohol and tobacco. Realistic, scientifically-backed information should serve as the foundation to youth prevention, which should be widespread (e.g., pediatrician’s office, educational settings). However, many individuals with established patterns of cannabis use are reluctant to quit, because they ascribe self-medication attributes to the drug. In such instances, and for a drug now commonplace, harm reduction approaches may prove to be of considerable utility.⁶⁰