Cannabis, schizophrenia genetic risk, and psychotic experiences: a cross-sectional study of 109,308 participants from the UK Biobank

The study investigates whether cannabis use is differentially associated with distinct categories of psychotic experiences (auditory hallucinations, visual hallucinations, persecutory delusions, and delusions of reference) and whether genetic predisposition to schizophrenia modulates these associations. Prior work links cannabis use with psychosis and earlier onset of psychosis, and shows cannabis can acutely induce psychotic-like experiences. Psychotic experiences are more common than diagnosed psychotic disorders but predict adverse outcomes, underscoring clinical relevance. The authors hypothesize that cannabis may have stronger relationships with delusion-related experiences than hallucinations and that schizophrenia polygenic risk increases susceptibility to cannabis-associated psychotic experiences. They further assess whether cannabis users’ psychotic experiences differ in age of onset, distress, and help-seeking compared with non-users.

Epidemiological studies associate cannabis with psychosis and earlier age at onset, though causality remains debated. Cannabis can induce transient psychotic-like experiences in healthy individuals, and the association between cannabis use and psychotic-like experiences appears largely genetically mediated with environmental contributions. Population surveys show self-reported psychotic experiences are more common than psychotic disorders yet predict later psychopathology. Meta-analytic evidence indicates a dose-dependent relationship between cannabis use and psychosis-related outcomes, including self-reported psychotic experiences. Literature suggests potential differential effects of cannabis on hallucinations versus delusions: cannabis was historically labeled a hallucinogen, but genuine cannabis-induced hallucinations are uncommon, whereas delusional ideation (e.g., persecutory ideation and paranoia) is frequently observed, including in clinical trials. Longitudinal adolescent data show stronger associations for paranoia than hallucinations. Genetic studies implicate the endocannabinoid system (e.g., CNR2 locus) in distressing psychotic experiences, and gene-by-environment interactions indicate that schizophrenia polygenic risk may amplify cannabis-associated neuroanatomical risk (e.g., reduced cortical thickness). However, prior to this work, no study directly tested whether schizophrenia genetic predisposition magnifies cannabis-related risk for psychotic experiences.

Design: Cross-sectional analysis using UK Biobank Mental Health Questionnaire data. Participants: From 157,348 MHQ respondents, 109,308 unrelated individuals of genetically defined White British ancestry (61,047 female; 48,261 male), aged 40–69 at recruitment, were included. Inclusion required responses on cannabis use and psychotic experiences and absence of ICD-10 psychotic disorder diagnoses (F20–F29) in linked records. No participants had ICD F19.15 or F19.95 (drug-induced psychosis), which were not used as exclusion criteria. Exposure: Self-reported cannabis use from MHQ—“Ever taken cannabis” (Data-Field 20453) and “Maximum frequency of taking cannabis” (20454). Cannabis use frequency coded as risk units: never=0 (N=87,010), ever=1 (N=14,642), monthly=2 (N=2,671), weekly=3 (N=3,582), daily=4 (N=1,403). Outcomes: Self-reported psychotic experiences: auditory hallucinations (20463), visual hallucinations (20471), persecutory delusions (20468), delusions of reference (20474). Additional attributes: age of onset (20461), distress (20462), help-seeking (20477). Non-missing sample sizes per outcome ranged from 108,174 (visual) to 109,104 (persecutory). Genetics: Schizophrenia polygenic risk score (PRS) computed using summary statistics from an independent GWAS (CLOZUK/PGC2). UK Biobank genotype QC: non-duplicate autosomal SNPs with call rate >95%, HWE p>1e-10, MAF>0.1%, info>0.8; harmonization with GWAS alleles; exclusion of ambiguous SNPs; p-value thresholding at p<0.05 (yielding best discrimination: AUC=0.677 versus stricter thresholds); LD pruning r^2<0.5 in 500 kb windows; PRS scored as the weighted sum of effect alleles per individual (missing genotypes mean-imputed). Samples restricted to White British, no sex chromosome aneuploidy, included in genetic PCs. Statistical analysis: Logistic regression (R glm) estimated adjusted odds ratios (AORs) for each psychotic experience regressed on cannabis ever-use (binary) or cannabis use frequency (ordinal 0–4), adjusting for covariates: birth year, sex, educational qualifications, household income, employment status, Townsend deprivation index, Index of Multiple Deprivation (three fields), smoking status, alcohol intake frequency, assessment center, and top 10 genotype PCs; categorical covariates modeled as indicators. Interaction analyses: Tested additive interactions between cannabis exposure and schizophrenia PRS via linear regression including exposure, PRS, and exposure×PRS term, comparing models with/without interaction via chi-square tests (anova in R). To reduce confounding, included covariate×exposure and covariate×PRS terms in both models. Multiple testing controlled using Benjamini–Hochberg FDR at 10%. Sex-stratified analyses were performed. For comparing logistic coefficients, differences were tested using z-scores computed from standard errors and inverse-normal transformation. Quintile-stratified analyses examined exposure–outcome associations within PRS quintiles.

- Dose-dependent associations: Compared with never-users (4.1% any psychotic experience), prevalence rose with cannabis use: ever 7.0% (AOR=1.54 [1.43, 1.65]); monthly 8.4% (AOR=1.69 [1.54, 1.87]); weekly 8.8% (AOR=1.69 [1.51, 1.89]); daily 9.6% (AOR=1.79 [1.52, 2.20]). Each one-step increase in use frequency (“risk unit”) associated with 20% higher odds of any psychotic experience (AOR per unit=1.20 [1.16, 1.24]). - By type of psychotic experience (ever vs never; AORs, 95% CIs): • Auditory hallucinations: ever AOR=1.57 [1.40, 1.77]; monthly 1.84 [1.57, 2.16]; weekly 1.85 [1.55, 2.21]; daily 1.79 [1.38, 2.33]; per unit 1.21 [1.15, 1.27]. • Visual hallucinations: ever 1.58 [1.45, 1.73]; monthly 1.69 [1.50, 1.91]; weekly 1.66 [1.45, 1.91]; daily 1.76 [1.44, 2.15]; per unit 1.21 [1.16, 1.26]. • Persecutory delusions: ever 1.59 [1.34, 1.89]; monthly 1.95 [1.56, 2.44]; weekly 2.14 [1.68, 2.74]; daily 2.44 [1.75, 3.40]; per unit 1.24 [1.15, 1.33]—the strongest association among the four. • Delusions of reference: ever 1.39 [1.15, 1.68]; monthly 1.57 [1.21, 2.03]; weekly 1.65 [1.24, 2.19]; daily 1.67 [1.10, 2.53]; per unit 1.18 [1.08, 1.28]. - Sex differences: Associations were stronger in females than males for any psychotic experience (AOR 1.59 vs 1.44; FDR=6%), auditory hallucinations (1.69 vs 1.40; FDR=5%), and delusions of reference (1.67 vs 1.20; FDR=5%); differences for visual hallucinations and persecutory delusions were not significant after FDR correction. - Age of onset, distress, and help-seeking: Cannabis ever-use associated with both adult-onset (AOR=1.52 [1.38, 1.66]) and early-onset <18 years (AOR=1.90 [1.64, 2.20]) psychotic experiences, with a significantly stronger association for early-onset (p≈1×10−5; FDR=0.5%). Ever-use associated more with distressing experiences (AOR=1.62 [1.45, 1.81]) than non-distressing (AOR=1.50 [1.37, 1.64]; FDR=9% for difference). Associations did not differ for help-seeking vs not (AOR=1.45 [1.25, 1.70] vs 1.55 [1.43, 1.68]; FDR=81%). - PRS interactions: Significant interactions between schizophrenia PRS and cannabis exposure. For ever-use: interactions for auditory hallucinations (p=0.02, FDR=9%), delusions of reference (p=0.04, FDR=9%), and any psychotic experience (p=0.05, FDR=9%). Using frequency improved power: significant interactions for auditory hallucinations (p=0.01, FDR=2%), delusions of reference (p=0.0007, FDR=0.4%), any psychotic experience (p=0.01, FDR=2%), and suggestive for visual hallucinations (p=0.06, FDR=7%). - PRS quintile stratification: Ever-use associated with higher odds among top PRS quintile versus bottom: any psychotic experience AOR 1.58 [1.36, 1.58] vs 1.39 [1.16, 1.65]; auditory 1.73 [1.36, 2.22] vs 1.36 [1.02, 1.82]; visual 1.66 [1.38, 1.99] vs 1.46 [1.18, 1.81]. For delusions of reference, association with cannabis use was significant only in the top two PRS quintiles; illustrative contrast from abstract: ever-use linked to 67% greater adjusted odds in top fifth vs 7% in bottom fifth.

Findings demonstrate a robust, dose-dependent association between cannabis use and self-reported psychotic experiences across categories, with the strongest effect for persecutory delusions. Cannabis users’ psychotic experiences tended to occur earlier and were more distressing, underscoring clinical significance beyond mere prevalence. Gene–environment interplay is supported by significant interactions between schizophrenia polygenic risk and cannabis exposure: individuals with higher PRS show stronger cannabis–psychosis associations, particularly for auditory hallucinations and delusions of reference, and to a lesser extent visual hallucinations. These results address the hypotheses that cannabis relates differentially to types of psychotic experiences and that genetic liability to schizophrenia heightens susceptibility to cannabis-associated psychotic experiences. The study illustrates how population-scale biobanks enable detection of nuanced interaction effects, with potential translational implications for targeted harm reduction in genetically at-risk groups.

This cross-sectional UK Biobank study shows that cannabis use is a predictive risk factor for multiple types of self-reported psychotic experiences, especially persecutory delusions, with dose-response relationships. Cannabis users report earlier-onset and more distressing experiences, and schizophrenia polygenic risk amplifies cannabis–psychosis associations, supporting a gene–environment interaction. These insights highlight the utility of polygenic risk scores for identifying individuals who may benefit from personalized prevention or harm-reduction strategies. Future research should employ longitudinal designs to assess causality and temporality, explore biological mechanisms linking endocannabinoid signaling, genetic risk, and psychotic experiences, and evaluate whether PRS-informed interventions can mitigate risk.

- Cross-sectional design limits causal inference and temporality between cannabis use and psychotic experiences. - Reliance on self-reported cannabis exposure and psychotic experiences may introduce recall or reporting biases. - Sample restricted to unrelated White British UK Biobank participants aged 40–69 at recruitment, potentially limiting generalizability to other ancestries, age groups, and populations. - Residual confounding is possible despite extensive covariate adjustment and inclusion of covariate×exposure/PRS terms. - PRS captures only a portion of genetic liability and may not reflect rare variants or gene–environment correlations. - Exclusion of individuals with diagnosed psychotic disorders may attenuate associations at the severe end of the spectrum.