Cannabis use linked with roughly double the odds of perpetrating violence

Design

Systematic review and meta-analysis of 63 observational studies; random-effects models

Population

265,079 people across general-population and psychiatric samples

Primary outcome

Perpetration of, and victimisation by, violence

Effect

Perpetration odds ratio 2.49 (psychiatric) and 2.05 (general); victimisation 1.49 (1.38–1.60)

A meta-analysis of 63 studies and 265,079 people examined cannabis use and violence, both as perpetrator and victim. Cannabis use was associated with about twice the odds of perpetrating violence in the general population and in psychiatric patients, with a larger association for violence resulting in a criminal conviction. It was also linked with a higher risk of being a victim, more so in women and mixed-gender cohorts.

The design is observational, so this is association, not proof that cannabis causes violence — confounding by other substance use, social adversity and pre-existing illness is hard to exclude, and the longitudinal estimate was much weaker than the cross-sectional one.

What it supports is practical: cannabis use is worth asking about routinely in psychiatric assessment, both as a risk marker and as a modifiable target. It does not justify framing patients who use cannabis as dangerous.

• Odds of perpetrating violence roughly doubled — psychiatric patients odds ratio 2.49 (95% CI 1.72–3.61), general population 2.05 (1.74–2.41).
• The association was strongest for violence leading to a criminal conviction (general population odds ratio 3.75, 2.54–5.53).
• Cannabis use was also linked with being a victim of violence (odds ratio 1.49, 1.38–1.60).
• The longitudinal estimate was weak (odds ratio 1.17, 1.07–1.28), so a causal reading is not supported.

Why it matters

It strengthens the case for routine cannabis assessment, while showing why the headline number overstates any causal effect.

Don't overread it

This was observational — it cannot show that cannabis use causes violence, and the weak longitudinal estimate argues against a simple causal link.

The statistics, in plain English

The gap between the cross-sectional odds ratio (2.37) and the much smaller longitudinal one (1.17) is the tell: when you follow people over time rather than snapshot them, the apparent effect shrinks, which is what you expect when confounding and reverse causation are driving part of the association.

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.