Cannabis use is associated with both committing and suffering violence

Design

Systematic review and meta-analysis of 63 observational studies

Population

265,079 people, psychiatric and general populations

Primary outcome

Violence perpetration and victimisation

Effect

Perpetration OR 2.49 (psychiatric), 2.05 (general); longitudinal OR 1.17 (1.07 to 1.28); victimisation OR 1.49

This meta-analysis pooled 63 studies (265,079 people) on cannabis use and violence. Cannabis users had higher odds of perpetrating violence among psychiatric patients (OR 2.49, 95% CI 1.72 to 3.61) and in the general population (OR 2.05, 1.74 to 2.41). The link was strongest for violence leading to a criminal conviction (OR 4.21 in psychiatric and 3.75 in general populations). Cannabis users were also more likely to be victims of violence (OR 1.49, 1.38 to 1.60), especially women.

The design matters. In longitudinal studies — which can at least establish that cannabis came first — the association shrank to OR 1.17 (1.07 to 1.28). Cross-sectional studies gave OR 2.37. That gap suggests a large part of the headline figure reflects shared risk factors such as other substance use, personality and social adversity, rather than cannabis itself.

For practice, the finding supports routine, specific questioning about cannabis in every psychiatric assessment and risk formulation, including asking whether the person has been harmed as well as whether they have harmed others.

• Ask about cannabis frequency and potency at every psychiatric assessment, not only in suspected psychosis
• Include cannabis use in violence risk formulations, alongside other substances
• Ask cannabis users, especially women, whether they have been assaulted or threatened
• Offer or refer for cannabis reduction treatment as part of risk management, not only for addiction

Why it matters

Cannabis is usually assessed for psychosis risk; this puts it in the violence and victimisation assessment too.

Don't overread it

These are observational studies; the much smaller longitudinal estimate (OR 1.17) suggests much of the association is confounding, not cause.

The statistics, in plain English

An odds ratio of 2 means about double the odds, but the longitudinal studies — the better design here — found only 17% higher odds. When stronger designs give smaller effects, confounding is usually part of the explanation.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.