Sixty-three studies, 265,079 people. Giulia Trotta and nine colleagues at King’s College London and the South London and Maudsley NHS Foundation Trust brought them together to answer an old question: are cannabis users more violent? The meta-analysis, which combines the results of separate studies, was published open access in Psychological Medicine on 24 September 2026.
The authors kept two groups apart, the general population and psychiatric patients, and two roles, those who commit violence and those who suffer it. Violence meant physical, sexual or psychological violence against another person, as each study defined it. Where a study measured several forms, they took the most serious.
The result is expressed as an odds ratio, the ratio between the odds of an event in two groups. A value of 1 means no difference. In the general population, cannabis users have an odds ratio of 2.05 for having committed violence, with a 95 per cent confidence interval from 1.74 to 2.41. Among psychiatric patients it rises to 2.49, from 1.72 to 3.61. For violence that led to a criminal conviction it reaches 3.75 in the general population and 4.21 among psychiatric patients, but each of those two estimates rests on just three studies. Cannabis users also have an odds ratio of 1.49 for suffering violence, more marked in women than in men.

The press releases turn 2.05 into «twice as likely». That is not the same thing. An odds ratio approximates the ratio of risks only when the event is rare; when the event is common it exaggerates it.
The figure that matters most is another one. In cross-sectional studies, which capture use and violence at the same moment, the odds ratio for committing violence is 2.37. In longitudinal studies, which measure use first and violence afterwards, it falls to 1.17, from 1.07 to 1.28. For suffering violence the drop is steeper: 1.49 across all studies, 1.05 in the longitudinal ones, from 1.01 to 1.10, barely above one. The longitudinal design is the one that can establish what comes first. It is also the one in which the effect becomes small.
The studies differ widely. The I² index, which measures how far results diverge beyond chance, is 98.6 per cent for violence committed in the general population. The funnel plots are asymmetric and Egger’s test is significant, a possible sign of publication bias, meaning null results left in the drawer. Corrected with the trim-and-fill method, the general-population estimate stays similar at 2.12; the psychiatric one falls to 1.85.

The authors say it plainly: the association does not show that cannabis causes violence. Cannabis use often travels with alcohol, other drugs, impulsivity, antisocial traits, poverty and violent surroundings. The most obvious confounder is alcohol, and most of the included studies did not account for the use of other substances. Few studies measured how much cannabis people used or how often, and almost all predate the spread of high-THC products, THC being the psychoactive compound. The search was also limited to papers in English.
One of the included studies is Italian. In 2013 Carabellese and colleagues followed 1,582 psychiatric patients in southern Italy; in the meta-analysis their study gives an odds ratio of 3.70, with a wide interval, from 1.48 to 9.26.
The authors draw two practical conclusions. Mental health services should always ask about cannabis use when assessing a patient’s risk of violence. Public information on cannabis should cover violence, committed and suffered, alongside the harm to the user.

The study was funded by the UK National Institute for Health and Care Research and the Medical Research Council; the authors declare no competing interests. They also declare that they used ChatGPT 4.0 and Gemini to improve the language of the text, which they then reviewed and take responsibility for.
What holds is this. Where there is cannabis there is more violence, committed and suffered. How much of that violence comes from cannabis, and how much from what comes with it, the pooled studies do not say.

References
Trotta G., Rodriguez V., Marino P., Gurmesa M., Spinazzola E., Li Z., Alameda L., Di Forti M., Murray R., Vassos E., «Cannabis use is associated with increased risk of violence: A systematic review and meta-analysis», Psychological Medicine, 56, e296, published online 24 September 2026, open access under CC BY 4.0. doi: 10.1017/S0033291726105856
Carabellese F., Candelli C., Martinelli D., La Tegola D., Catanesi R., «Cannabis use and violent behaviour: A psychiatric patients cohort study in Southern Italy», Rivista di psichiatria, 48, 1, 43–50, 2013. doi: 10.1708/1228.13614 (cited as included in the meta-analysis; not read directly).
King’s College London, press release, republished by ScienceDaily on 30 September 2026 as «Cannabis users were twice as likely to commit violence, review finds».