Hjorthøj · Psychological medicine 2023 · nationwide register-based cohort study · n=6,907,859

Association between cannabis use disorder and schizophrenia stronger in young males than in females.

Level 3 - non-randomized controlled study

Nationwide observational register-based cohort study.

PubMed 37140715 · doi:10.1017/S0033291723000880 · record verified 2026-08-26

What was done

Nationwide Danish register-based cohort study following 6,907,859 individuals aged 16–49 at some point between 1972 and 2021 (129,521,260 person-years). Diagnoses of cannabis use disorder (CUD) and incident schizophrenia were ascertained from national registers. The authors estimated adjusted hazard ratios (aHR), adjusted incidence risk ratios (aIRR), population attributable risk fractions (PARFs), and joinpoint trend analyses across sex and age groups.

What was found

Across follow-up, 45,327 incident cases of schizophrenia were identified. The overall aHR for CUD on schizophrenia was higher in males (aHR = 2.42, 95% CI 2.33–2.52) than in females (aHR = 2.02, 95% CI 1.89–2.17). Among individuals aged 16–20 years, the risk ratio in males was more than double that in females (males: aIRR = 3.84, 95% CI 3.43–4.29; females: aIRR = 1.81, 95% CI 1.53–2.15). Between 1972 and 2021, the annual average percentage change in PARFs was 4.8% for males (95% CI 4.3–5.3; p < 0.0001) and 3.2% for females (95% CI 2.5–3.8; p < 0.0001). In 2021, the PARF was 15% among males and around 4% among females.

Why it matters

This study shows a substantial and increasing population attributable risk of schizophrenia linked to cannabis use disorder over a 50-year period, with the burden concentrated disproportionately among young men.

Limits

Registry data capture only clinically diagnosed cannabis use disorder, underestimating overall cannabis consumption and missing subclinical use. The observational design cannot rule out residual confounding (such as polygenic risk or other substance misuse) or reverse causation/prodromal self-medication. PARF calculations strictly assume causality, which cannot be proven from observational register records alone. Findings from the Danish population may not fully generalize to settings with different cannabis potencies, demographics, or legal frameworks.

Cited by