Every year, another cohort of teenagers reaches the age where cannabis becomes available to them: through a dealer, a friend, or increasingly, a legal dispensary. Most try it once or twice and move on. A smaller group use it regularly through adolescence. Experts have argued for decades over what that regular use does to a brain still under construction, usually in terms that generate more heat than light. A study published in February 2026 in JAMA Health Forum does not end that argument, but it does move the conversation about cannabis and the adolescent mind onto firmer ground. It puts a number on the risk, using one of the largest and most carefully built datasets on the subject to date, and it is worth walking through in some detail, because the strength of its conclusions rests on how researchers built it.
A Cohort Built for This Question
Researchers at Kaiser Permanente Northern California, an integrated health system covering more than 4.6 million members, tracked 463,396 adolescents aged 13 to 17 who had completed a confidential screening questionnaire during routine well-child visits between 2016 and 2023. Every teenager in the system answered the same question at each biennial check-up: did you use marijuana in the past year? Researchers recorded that answer as a time-varying exposure, meaning it could change from one visit to the next as a teenager’s habits changed, rather than staying fixed at a single point in time. They then followed adolescents forward through their electronic health records, in some cases as far as age 26, and pulled any subsequent diagnosis of a psychotic, bipolar, depressive, or anxiety disorder from clinical coding rather than a symptom survey. That distinction matters. This was not a study asking teenagers to rate how anxious they felt. It was tracking who actually walked into a doctor’s office and received a formal diagnosis.
The scale of the cohort lets the study do things smaller studies cannot. Of the 463,396 adolescents, just over half were male, the average age at enrolment was 14.5, and the sample was ethnically broad: 29.5 percent Hispanic, 20.2 percent non-Hispanic Asian, 7.6 percent non-Hispanic Black, 33.0 percent non-Hispanic White, and 4.1 percent multiracial or of other backgrounds. At baseline, 5.7 percent of adolescents, 26,345 teenagers, reported past-year cannabis use. That group looked demographically different from their non-using peers. Cannabis users were older on average, more likely to be Hispanic, non-Hispanic Black, or non-Hispanic White, and less likely to be non-Hispanic Asian. They were also more likely to be on Medicaid, to live in more deprived neighbourhoods, and to report drinking alcohol or using other substances, nearly two-thirds of cannabis-using teenagers also reported drinking, compared with just over three percent of non-users. This pattern of overlapping risk is exactly why the researchers built their statistical models to adjust for all of it: sex, race and ethnicity, neighbourhood deprivation, insurance type, and alcohol and other drug use, all entered as covariates rather than left to confound the picture.
Doubled Risk of Psychosis and Bipolar Disorder
Over the follow-up period, the cohort generated 4,105 new diagnoses of psychotic disorder, 4,061 of bipolar disorder, 62,137 of depressive disorder, and 73,096 of anxiety disorder. Psychosis and bipolar disorder were, in absolute terms, rare: around a quarter of a case per thousand person-years each. Depression and anxiety were far more common, at four to six cases per hundred person-years. Set against cannabis exposure, the pattern was consistent across all four. Adolescents who reported past-year cannabis use were 2.19 times more likely than non-users to receive a psychotic disorder diagnosis and 2.01 times more likely to receive a bipolar disorder diagnosis, after the researchers adjusted for the full set of covariates. The links to depression and anxiety were smaller in size but still statistically solid: a 1.34-fold increase in risk for depressive disorder and a 1.24-fold increase for anxiety disorder.
These are adjusted figures: the models already account for the more obvious alternative explanations, drinking, other drug use, socioeconomic disadvantage, and none of them explain the pattern away. The researchers then pushed harder, testing whether teenagers who already showed signs of mental illness before using cannabis were simply more likely to do both: use the drug and later receive a diagnosis, for reasons that had nothing to do with cannabis itself. So they reran the models adjusting for each adolescent’s psychiatric history at baseline, and again after removing anyone with any psychiatric history at all, leaving only a cleaner, lower-risk group. Neither check changed the story much. The risk of psychosis and bipolar disorder stayed roughly doubled in both versions (psychosis: 1.92 and 1.99; bipolar disorder: 1.73 and 2.00), and depression and anxiety held in much the same range as the main result (depression: 1.33 and 1.37; anxiety: 1.19 and 1.22). Whichever way the researchers sliced the data, the pattern survived.
They also checked whether their definition of psychosis was doing too much work. Clinicians coded roughly two-thirds of the psychotic disorder diagnoses in the cohort as unspecified psychosis rather than a named condition like schizophrenia, so the researchers tried both a broader definition, one that also captured substance-induced psychosis, and a narrower one limited to unspecified psychosis with no drug involvement coded. Both versions landed almost exactly where the main finding did, around a doubling of risk (2.20 with the broader definition, 2.21 with the narrower one). The result wasn’t an artefact of how the researchers happened to define psychosis.
Correlation, Causation, and Timing
None of this proves cannabis causes these disorders the way smoking causes lung cancer. What a cohort study can show is sequence: cannabis use came first, and reliably so. Among adolescents who went on to receive a diagnosis, the average gap between first reporting cannabis use and receiving a psychiatric diagnosis ranged from 1.7 years for depressive disorder to 2.3 years for bipolar disorder. That is not proof of causation, but it rules out the simplest alternative explanation, that the diagnosis came first and the cannabis use was somehow a consequence of it, at least for the majority of cases captured in the data.
What the study cannot fully rule out is that something else entirely, a genetic vulnerability, an adverse childhood experience, an undiagnosed mood problem already forming beneath the surface, is driving both the cannabis use and the later diagnosis simultaneously. Statisticians have a tool for stress-testing exactly this worry, called an E-value. It asks a simple question: how powerful would a hidden factor have to be, in its own effect on both cannabis use and the outcome, to erase the association entirely? A low number means a fairly ordinary confounder could do the job, and readers should treat the finding cautiously. A high number means the hidden factor would have to be doing something dramatic, which makes coincidence a less comfortable explanation. Here, the numbers came out high for the two more severe conditions and lower for the others: 3.79 for psychosis and 3.44 for bipolar disorder, meaning a hidden factor would need to nearly quadruple the odds of both cannabis use and the disorder to explain the link away, against 2.02 for depression and a more modest 1.79 for anxiety. That pattern is part of why the psychosis and bipolar findings carry more weight in the paper than the depression and anxiety ones, and part of why this study has become a reference point for anyone examining cannabis and the adolescent mind rather than just another entry in the literature.
Younger Users Carry the Greater Risk
One of the more striking patterns in the data only appeared once the researchers checked whether the strength of the association held steady across age. For depression and anxiety, it did not. The risk tied to cannabis use was highest in early adolescence and steadily faded as the cohort got older, almost like a countdown. For depressive disorder, it started at a 78 percent increase in risk for 13 to 15 year olds and eased with each older bracket until it disappeared entirely by the early twenties, no longer statistically distinguishable from zero (1.78, then 1.35, then 1.21, then a non-significant 0.97). Anxiety traced almost the same arc, starting at a 47 percent increase and fading out by ages 21 to 25 (1.47, then 1.24, then 1.24, then a non-significant 0.95). Psychosis and bipolar disorder did not follow this pattern. Whatever is driving the elevated risk for those two conditions appears to work differently to whatever links cannabis to depression and anxiety, and it does not fade out simply because the person using cannabis has grown older.
The Endocannabinoid System and a Brain Still Under Construction
The clinical numbers line up with a body of laboratory and neuroimaging research the Dalgarno Institute, an Australian drug policy and prevention body, has been compiling for years, work that has quietly built the biological case for cannabis and the adolescent mind long before the Kaiser Permanente cohort gave it a population-level number to point to. The mechanism centres on the endocannabinoid system, a receptor network deeply involved in how the adolescent brain wires itself, particularly the circuits governing motivation, emotional regulation, and affective processing. THC, cannabis’s main psychoactive compound, acts directly on cannabinoid receptors that are especially dense in the adolescent brain. The Institute’s archive holds translational work, combining human neuroimaging with animal studies, that argues high-THC products can disrupt this developmental process, with effects on emotional regulation that outlast the period of use itself. That literature frequently cites one figure putting the increased risk of adolescent psychotic disorder at over elevenfold for cannabis users, a considerably larger effect than the JAMA Health Forum figure. The two studies used different populations, definitions, and methods, so readers shouldn’t read the numbers side by side as apples to apples. But they point the same way: adolescence as a particularly exposed period, with the size of the effect scaling with the potency of what teenagers are using. That detail is worth sitting with given where the JAMA cohort came from. Northern California’s legal cannabis market has an average THC content in flower products exceeding twenty percent, several times higher than the products that much of the older research in this field studied.
This also helps explain why the psychiatric associations were strongest for psychosis and bipolar disorder rather than depression and anxiety. Researchers think psychotic and bipolar disorders have stronger neurodevelopmental roots, conditions that a disrupted brain circuit can tip into during a sensitive window. Depression and anxiety have a wider range of triggers, many environmental and temporary, which may be part of why their link to cannabis use fades as the brain finishes developing and other life factors take over as the dominant influence. A mechanism that specifically interferes with neurodevelopment would likely leave a bigger and more durable mark on the former than the latter, and that is roughly what both datasets show.
TikTok, Micro Influencers, and the New Pathway to Use
The Kaiser Permanente data collection ended before the current wave of social-media-driven exposure had fully taken hold, so it cannot speak to a more recent question: how cannabis first reaches today’s teenagers. A 2025 analysis that the Dalgarno Institute cites, drawing on JAMA Network Open research tracking more than 7,600 California high school students, found that teenagers who saw cannabis content on platforms like TikTok were substantially more likely to start using cannabis within a year, and far more likely to become dual users of cannabis and e-cigarettes. Exposure through micro influencers, social media personalities with modest followings who read to teenagers as peers rather than advertisers, carried the strongest associations of all, with exposed teenagers showing a markedly higher risk of recent cannabis use. That matters for the mental health picture too: the same research flagged heightened depression risk among dual users specifically, on top of whatever burden multiple substances place on a developing brain. It suggests the exposure landscape driving the next cohort of teenagers into cannabis use looks meaningfully different from the one that shaped the adolescents in the Kaiser Permanente study, more engineered, more peer-coded, and harder for parents to see happening.
What the Data Can’t Tell Us
The Kaiser Permanente cohort measured past-year use, not frequency, potency, or method of consumption, all of which likely matter a great deal and this dataset could not tease apart. Diagnosis also depended on a teenager actually presenting to a doctor and clinicians coding the case correctly, so milder or unrecognised cases of depression or anxiety may not appear in the data at all; the two thirds of psychotic disorder diagnoses that clinicians left as unspecified point to genuine diagnostic caution in adolescent psychiatry rather than a fully resolved clinical picture. The Dalgarno Institute material leans heavily on animal models and smaller clinical samples: useful for establishing mechanism, but no substitute for the statistical power of a half-million-person cohort. Each source covers a gap the other leaves open. Neither would be persuasive alone.
A Case for Taking Adolescence Seriously
Taken together, the evidence suggests adolescence isn’t a neutral backdrop against which cannabis use just happens to occur. It looks more like a period when the brain is unusually responsive, for better or worse, to what’s introduced into it. That is, at its core, what the research on cannabis and the adolescent mind is now converging on. The JAMA Health Forum associations were sizeable, survived repeated attempts to explain them away through confounding, and held up under three separate sensitivity checks and two alternative outcome definitions. Set against a growing mechanistic literature on how THC interacts with a developing endocannabinoid system, and newer evidence on how easily social platforms now engineer that exposure, adolescent cannabis use looks less like a rite of passage and more like a genuine public health question, one worth treating as such.
That’s not an argument about legalisation for adults, which involves a different set of trade-offs. It’s a narrower claim: cannabis and a still-developing brain don’t mix cleanly, and the data now backs that up from more than one direction, in populations as different as a Californian health system’s electronic records and a laboratory studying receptor density in animal models. It points toward earlier screening in routine paediatric care, clearer messaging aimed at parents and teenagers rather than adult consumers, and marketing and packaging rules that take seriously how differently a fifteen-year-old responds to this drug compared with a thirty-five-year-old. (Source: WRD News – Dalgarno Institute)