Rates and correlates of cannabis-associated psychotic symptoms in over 230,000 people who use cannabis

Cannabis is among the most widely used psychoactive substances globally, with around one in four adults having tried it. While many PWUC report pleasurable acute effects, epidemiological evidence links cannabis use to a long-term increased risk of psychosis, and rising cannabis potency has raised concerns about acute adverse psychiatric effects. Acute reactions can include anxiety, panic, and psychosis-like experiences (paranoia and hallucinations), largely attributable to THC’s action on CB1 receptors. Transient cannabis-associated psychosis-like experiences are relatively common during intoxication, but a subset of PWUC experience severe episodes requiring medical attention. There is growing interest in identifying which individuals are most sensitive to these psychotomimetic effects, with suggestions that young users and those with pre-existing mental health vulnerabilities may be at higher risk. Cannabis use characteristics (e.g., dose, potency, tolerance) may interact with individual-level factors to influence risk. However, evidence on the rates and correlates of CAPS severe enough to require emergency medical treatment in cannabis users specifically remains limited. This study aimed to address three questions: (1) What are the rates of CAPS requiring emergency medical treatment in a large international sample of PWUC? (2) Do rates vary by demographics, cannabis/substance use patterns, and individual vulnerabilities? (3) How are CAPS events characterized in terms of substance use just before treatment and subsequent outcomes?

Prior work has documented psychotomimetic effects of THC and cannabis, with reports of paranoia (15–53% of PWUC) and hallucinations (3–27%) during intoxication. Although severe cannabis-induced psychosis requiring treatment has been reported, most studies assessed broader substance-induced outcomes and did not restrict samples to cannabis-exposed individuals. Registry-based studies have estimated rates of cannabis-induced psychosis, but often lack detailed individual-level predictors (e.g., frequency, potency, route). Evidence suggests higher risk of psychosis-like experiences in younger users and in individuals with mental health vulnerabilities or psychosis liability. High-potency cannabis has been implicated in increased psychosis risk, and tolerance may modulate acute psychotomimetic effects. There is also literature indicating rising THC potency over time, with particularly marked increases in resin. Together, these findings motivated a focused assessment of CAPS requiring medical attention and their correlates within PWUC.

Design: Cross-sectional analysis of pooled data from five Global Drug Survey (GDS) waves (2014, 2015, 2016, 2017, 2019). Inclusion criteria: history of cannabis use; residence in countries with ≥1,000 PWUC respondents; complete CAPS data. Final sample n=233,475 (from total 529,574 participants). Ethics: Approvals from King’s College London (141/02), University of Queensland (2017001452), and University of New South Wales (HREC HC17769); informed consent obtained. Measures: Primary outcome was CAPS requiring emergency medical treatment, assessed for last year and lifetime via self-report items about seeking emergency treatment following cannabis use. Symptom profiles were captured from a checklist; CAPS defined as emergency presentations with hallucinations and/or paranoia. Correlates included demographics (age, gender, country), cannabis use characteristics (preferred type e.g., herbal, high-potency, resin; route of administration; frequency in last 12 months; mixing with tobacco), and other substance use (alcohol, MDMA, cocaine, amphetamines, ketamine, LSD). Mental health history was assessed via self-reported lifetime diagnosis (psychosis, bipolar, anxiety, depression). Items also characterized CAPS events: type/amount of cannabis used before treatment, co-ingested substances, time to recovery, hospital admission, and changes in cannabis use after CAPS. Statistical analysis: Analyses in R 4.0. Estimated lifetime and last-year CAPS rates with 95% CIs via Clopper-Pearson exact method (PropCIs). Correlates assessed in last-year subsample (n=148,109), estimating subgroup rates and risk ratios (RRs) with 95% CIs and p-values (fmsb). Multiple testing controlled using Benjamini-Hochberg FDR. Subgroup comparisons interpreted only when ≥15 CAPS cases per subgroup. Sensitivity analysis excluded respondents reporting prior participation in earlier GDS years to assess potential bias from repeated participation.

• Rates: Lifetime CAPS requiring emergency medical treatment: 0.47% (95% CI 0.42–0.52); last-year CAPS: 0.19% (95% CI 0.17–0.21). Symptom profiles last year: paranoia only 0.10%; both hallucinations and paranoia 0.07%; hallucinations only 0.02%. Sensitivity analysis excluding repeat GDS participants yielded comparable rates.
• Demographic correlates (last-year CAPS): Higher risk in PWUC <21 vs older (RR 2.66). Higher risk in Denmark (RR 3.01), lower in the United States (RR 0.40).
• Cannabis use characteristics: Elevated risk in those predominantly using high-potency resin vs herbal cannabis (RR 2.11). Mixing cannabis with tobacco associated with higher risk vs not mixing (RR 2.15). No significant elevation with frequent cannabis use, predominant use of high-potency herbal (e.g., skunk), route of administration, or last-year use of other psychoactive substances (alcohol, MDMA, cocaine, amphetamines, ketamine, LSD).
• Mental health correlates: Strongly elevated risk with self-reported diagnoses: psychosis RR 14.01 (absolute risk 1.69%), bipolar RR 4.30, anxiety RR 2.92, depression RR 2.68.
• Characterization of CAPS events (last year): Before CAPS, most used high-potency herbal (44%) or resin (24%); 84% used ≤1 g; 43% reported no other drugs; alcohol 21%, MDMA 10%; excessive alcohol (>15 drinks) 3%. Recovery: ≤1 day 56%, >4 weeks 21%. Hospital admission: 36% of CAPS cases; highest among those with lifetime psychosis diagnosis (76%) and with prolonged CAPS (>1 week, 54%). Post-event behavior: 57% reduced cannabis use; 43% did not.
• Post-hoc: Elevated rates in Denmark may reflect popularity of high-potency resin (THC ~23%+).

Findings demonstrate that severe acute CAPS requiring emergency medical treatment occur in a small but meaningful subset of PWUC (~1 in 200 lifetime). Younger age and pre-existing mental health diagnoses, particularly psychosis, markedly increase risk, supporting developmental and vulnerability models of cannabis-related psychotomimetic effects. Country differences, notably higher rates in Denmark, likely relate to patterns of use of high-potency resin. The lack of association with frequency, route, or many other substances suggests that acute risk may be driven more by potency (especially resin) and individual vulnerability than by overall use frequency. Characterization of CAPS events indicates that many occur without co-use of other drugs, often after modest quantities, and that while most resolve within a day, a substantial minority experience prolonged symptoms and require hospitalization. These results inform clinical risk stratification and public health messaging amid changing cannabis policies.

This large international study is the first to systematically estimate rates and correlates of cannabis-associated psychotic symptoms requiring emergency medical treatment among PWUC. It shows that acute CAPS do occur and identifies higher-risk groups, including young users, those consuming potent forms (notably resin), and individuals with mental health problems, especially psychosis. As cannabis legalization expands, efforts to educate young people and the public about risks are needed, alongside further research on predictors and long-term consequences of CAPS to guide prevention and intervention.

• Outcome captures only events warranting emergency medical treatment; milder acute psychotic symptoms are likely more common and not reflected in these rates.
• Reliance on retrospective self-report introduces potential recall and subjective reporting biases; however, the severity of events may mitigate recall concerns. Predictors were restricted to past-year behaviors to reduce bias.
• Non-representative, volunteer web-survey sampling limits population generalizability, though such designs can enhance reporting accuracy on sensitive topics and PWUC samples resemble those from probability sampling in key respects.
• Association between psychosis diagnosis and CAPS may be underestimated due to exclusion of cases with prolonged psychosis post-CAPS and potential underrepresentation of individuals with psychosis in volunteer surveys.