The research team, based at UNSW Sydney, searched MEDLINE, Embase, and PsycInfo from inception through April 2025 for peer-reviewed cohort and case-control studies reporting the mean and standard deviation of age at onset of psychosis in cannabis users compared with nonusers, or other usable effect size data.

The final analysis pooled 149 studies representing 181 distinct samples, 18,272 cannabis users, and 52,801 nonusers, using random-effects meta-analysis to calculate a pooled standardized mean difference, with subgroup analysis by each sample’s mean age of psychosis onset and formal meta-regression to test whether age explained the differences in effect size between studies.

Across the full pooled sample, cannabis use was associated with a mean age at onset of psychosis 2.50 years earlier than nonusers, a figure that lines up closely with the same senior author’s smaller 2011 meta-analysis of 83 studies, which found a 2.70-year difference.

What is new here is the age-stratified breakdown. When the researchers grouped samples by their mean onset age, the association was not evenly spread. It was statistically absent under age 20, present but modest at 1.60 years earlier for the 20-to-under-25 group, larger at 4.43 years earlier for the 25-to-under-30 group, and largest at 6.30 years earlier for the 30-and-older group, a progressive pattern the authors had not been able to test in earlier, smaller analyses.

The researchers used multiple meta-regression to test whether sample mean onset age explained the size of the earlier-onset effect independent of other differences between studies, and found it did, a strong and statistically significant result (z = -7.59, p < .0001).

That statistical independence matters because it means the age pattern is not simply a byproduct of which countries, decades, or study designs happened to include older samples. The senior author has described this as evidence for a cumulative-exposure effect, where ongoing cannabis use over years progressively lowers the threshold for psychosis onset, though the meta-analysis itself, being observational, cannot directly test that mechanism.

The absence of a significant difference in the under-20 group could be misread as reassuring for adolescents, but the authors offer a different explanation, what they and outside commentary have called a possible biological floor. Psychosis is rare before the mid-teens, which leaves a narrow window in which any cannabis-related shift in onset age could be detected, and strong neurodevelopmental and genetic risk factors may dominate at that age regardless of cannabis exposure.

This finding does not undercut the well-established, separately documented risks of adolescent cannabis exposure to neurodevelopment; it simply reflects a limitation in how an age-of-onset shift can be measured in a population where the outcome itself is still uncommon.

This is a meta-analysis of observational age-at-onset data, not a randomized trial, and the authors are explicit that it cannot prove cannabis causes psychosis or that reducing use would delay onset in any individual, since the experimental research needed to establish that directly would be unethical to conduct.

The analysis also did not capture cannabis potency, product type, frequency, or route of administration across the included studies, and most of the underlying literature focused on people who went on to develop schizophrenia specifically, which limits how confidently the findings generalize to the full spectrum of psychotic disorders or to today’s higher-potency products.

Senior author Matthew Large, a practicing psychiatrist and Conjoint Professor at UNSW’s School of Clinical Medicine, said the field has focused heavily on adolescent risk while the data suggest risk does not disappear in the twenties, describing the pattern as consistent with a cumulative effect similar to alcohol or nicotine, and said he believes the threshold for stronger public health warnings about cannabis and psychosis has been met.

Lead author Carly Stevens said the findings challenge the assumption that waiting until the mid-20s to use cannabis removes the psychiatric risk. Jack Wilson, a research fellow at the University of Sydney’s Matilda Centre who was not involved in the study, called it valuable new evidence on age-specific psychosis risk, while noting that the absence of potency and product-type data is a real limitation for translating the findings into specific clinical guidance.

The association between cannabis use and earlier psychosis onset has been documented since at least the same senior author’s 2011 meta-analysis, and a 2015 analysis separately found psychosis typically emerged about 6.3 years after cannabis use began, though that earlier work lacked the statistical power to test whether the association varied by age group.

This new analysis, roughly double the size of the 2011 study, closes that gap by showing the association is not uniform across the lifespan but grows progressively larger with age, shifting the clinical conversation away from adolescence as the sole period of concern and toward a cumulative-exposure model that implicates sustained adult cannabis use as well. That shift matters directly for cannabis medicine, where the patient population using cannabis regularly for years, often well into their thirties, forties, and beyond, has not historically been the focus of psychosis-risk counseling.

This is a well-powered, carefully analyzed meta-analysis, and the age-stratified finding is genuinely new information, not a restatement of what we already knew about adolescent risk. I take it seriously, and it changes a specific part of how I counsel patients: I can no longer treat a patient’s age alone as reassurance that their psychiatric risk from ongoing cannabis use is low. The size of the difference in the 30-and-older group, more than six years, is large enough that it belongs in an informed consent conversation with any adult patient who has a personal or family history of psychosis, mania, or schizophrenia, regardless of how long they have already been using cannabis without incident.

At the same time, I want to be precise about what this study is and is not. It is a meta-analysis of age at onset drawn from cohort and case-control literature, most of it focused on schizophrenia, with no data on potency, dose, or product type, and the authors are explicit that it cannot prove causation. It does not tell me that a specific patient in front of me, using a specific product at a specific dose, faces this exact risk. What it does tell me is that the old reassurance, that if you make it past your twenties without a problem you are largely in the clear, is not supported by the best current evidence, and that psychiatric screening and honest risk conversations need to extend across the full adult lifespan of cannabis use, not taper off after adolescence.

A Four-Step Reading Frame