A review of 265,000 people found cannabis use was linked to about twice the odds of committing violence in the general population and about 2.5 times among psychiatric patients.

A review of 63 studies covering 265,079 people found that people who used cannabis had about twice the odds of having committed violence compared with people who did not. Among people in psychiatric care, the odds were about 2.5 times higher. Cannabis use was also linked to being a victim of violence, more strongly in women than in men. But the size of the effect shifts depending on how the original studies were designed, and an independent violence researcher says that matters for how far the headline numbers can be trusted.

The review, published in Psychological Medicine by researchers at the Institute of Psychiatry, Psychology and Neuroscience at King’s College London, searched three databases for studies up to November 2024. It used the World Health Organization’s broad definition of violence, which covers physical, sexual, and psychological violence toward another person, both inside and outside relationships. It looked separately at violence people reported themselves and violence that led to a criminal conviction, and where a study measured more than one kind, it used the most serious.

What the pooled numbers show

Using random-effects models, the team found that cannabis users were at significantly higher risk of perpetrating violence, with about twice the odds in the general population and about 2.5 times the odds among psychiatric patients. For violence that led to a criminal conviction, cannabis users were three to four times as likely to have been convicted, though the researchers note those findings come from a small number of studies. Cannabis use was also linked to being a victim of violence, with an odds ratio of 1.49 (95% confidence interval 1.38 to 1.60). That link was more pronounced in females and mixed-gender groups than in males.

An odds ratio compares the odds of an outcome between two groups. It describes a pattern across studies, not the chance that any individual cannabis user will be violent.

The effect shrinks when people are followed over time

Prof Simon Moore, who directs the Violence Research Group at Cardiff University and was not involved in the review, pointed to a split in the results. Studies that followed people over time found a much smaller association, an odds ratio of 1.17, compared with the roughly twofold figure that includes cross-sectional studies, which measure cannabis use and violence at a single point. Moore said that suggests some of the larger association “may reflect circumstances that were not measured.” He added that a confounder no stronger than commonly cited factors such as peer group or neighbourhood would be enough to account for the longitudinal estimate entirely.

Results were similar with and without statistical adjustment (odds ratios of 2.01 and 2.19), but Moore noted that only shows the factors adjusted for made little difference, not that context was taken into account.

What the review cannot say

The authors state that the review shows an association and cannot prove that cannabis causes violence, because cannabis use often goes alongside other drug and alcohol use and difficult social circumstances. Few of the studies measured how much or how often people used cannabis.

Moore noted that the review did not look at where, with whom, or in what circumstances cannabis was used. Because cannabis remains illegal in most places, illegal markets can bring users into contact with people and places where violence is more likely. In his words, “the review cannot say whether cannabis itself or the environments in which it is used explain the violence.”

The authors say mental health services should routinely ask about cannabis use when assessing a person’s risk of violence, and they caution against stigmatizing people who use cannabis more broadly.

Source

Trotta G, Rodriguez V, Marino P, et al. “Cannabis use is associated with increased risk of violence: A systematic review and meta-analysis.” Psychological Medicine, 2026;56:e296.
DOI: 10.1017/S0033291726105856