A clinician-led analysis of causation, confounding, and mental health risk interpretation
| Audience | Primary care physicians, psychiatrists, cannabis clinicians, patients considering medical cannabis |
| Primary Topic | Cannabis effects on mental health conditions including PTSD, anxiety, depression, bipolar disorder, and psychosis |
| Source | Read the full article | Read the PDF |
Cannabis and Mental Health: When Observational Associations Get Mistaken for Causation
A JAMA Internal Medicine review concludes cannabis poses substantial mental health risks while offering no therapeutic benefits, but the evidence quality reveals a more nuanced picture. Understanding the difference between correlation and causation is essential for evidence-based patient counseling.
| Study Type | Narrative review of systematic reviews and meta-analyses |
| Population | Adults and adolescents with or at risk for mental health conditions (PTSD, anxiety, depression, bipolar disorder, psychosis, ADHD, cannabis use disorder) |
| Exposure or Intervention | THC-predominant cannabis, CBD, and cannabinoid medications (widely varying potency, frequency, duration, and age of onset across studies) |
| Comparator | Placebo (in RCTs) or non-users (in observational studies) |
| Primary Outcomes | Symptom severity, psychotic episodes, mania, depression, anxiety, cognitive function, cannabis use disorder incidence |
| Sample Size or Scope | Review of multiple systematic reviews; individual RCTs ranged from N=10 to N=80; observational cohorts included thousands to tens of thousands |
| Journal | JAMA Internal Medicine |
| Year | 2026 |
| DOI | 10.1001/jamainternmed.2025.8215 |
| Funding or Conflicts | Supported by US Department of Veterans Affairs and NIH; authors report no conflicts related to cannabis industry |
The authors reviewed evidence from living systematic reviews examining cannabis and cannabinoids for PTSD (2 small RCTs), anxiety (3 CBD trials, sparse THC data), depression (secondary outcomes only), bipolar disorder (7 observational studies), psychosis (multiple observational cohorts and meta-analyses), ADHD (1 small trial), cannabis use disorder prevalence (cross-sectional surveys), and cognitive effects (experimental and observational studies). They also summarized drug-drug interactions and provided clinical counseling recommendations.
For PTSD, the largest trial (N=80) found no symptom improvement with smoked cannabis versus placebo. For anxiety, low-certainty evidence suggests high-dose CBD (150-300 mg daily) may reduce symptoms, while THC effects are uncertain. No RCT evidence exists for depression. Seven observational studies found cannabis use associated with worse bipolar outcomes (more mania, lower recovery rates). Multiple observational studies link heavy adolescent cannabis use with 2- to 11-fold increased psychosis risk. About 30% of past-year cannabis users meet criteria for cannabis use disorder. Acute THC impairs cognition in dose-dependent fashion, with uncertain long-term effects in adults.
The RCT evidence for therapeutic benefits is extremely limited (tiny trials, short duration, negative or inconclusive results) and appropriately rated as low or very low certainty. The harm evidence is predominantly observational and rated as low certainty, yet treated more confidently in the clinical recommendations. The psychosis associations come from cohort studies that cannot prove causation because they do not adequately control for genetic confounding, trauma exposure, prodromal symptoms, or other substance use. The bipolar data suffers from severe selection bias (sicker patients may use more cannabis). Only the acute cognitive impairment evidence reaches moderate certainty through experimental studies.
First, asymmetric evidence standards: the authors demand RCT-level proof for benefits but accept observational associations as proof of harms. Second, causality overreach: associations are presented as if cannabis causes psychosis, mania, and self-harm without ruling out reverse causation (people with emerging mental illness self-medicating) or shared genetic vulnerability. Third, absolute risk is never reported. An 11-fold increased risk sounds catastrophic, but if baseline psychosis risk is 0.5%, even an 11-fold increase yields 5.5% absolute risk—meaningful but not inevitable. Fourth, exposure is treated as monolithic when “cannabis use” includes everything from monthly low-potency use to daily high-potency concentrate use. Fifth, the review conflates CBD isolate trials (which show possible benefit for anxiety) with THC-dominant recreational cannabis. Sixth, outcome measurement varies wildly across studies (what counts as “cannabis-induced psychosis” ranges from transient paranoia to hospitalization). Finally, critical confounders are unmeasured: trauma, poverty, family psychiatric history, concurrent substance use, and prodromal symptoms.
Cannabis is not supported for mental health treatment by current RCT evidence. Adolescents, people with psychotic or bipolar disorders, pregnant individuals, and those with substance use vulnerabilities face higher risks and should avoid regular use. For other adults, risks from occasional or moderate use are uncertain but likely lower than this review suggests. The key clinical skill is risk stratification: who is this patient, what is their psychiatric and family history, what potency and frequency are they using, and what are they trying to achieve? One-size-fits-all warnings based on observational data from high-risk populations do not serve patients well. Harm reduction (lower potency, less frequent use, avoiding smoking, monitoring for dependence) is more clinically useful than abstinence-only messaging for patients who will use regardless. Finally, recognize that absence of evidence is not evidence of absence—the lack of RCTs reflects research barriers, not proof that cannabis cannot help some patients with some conditions.
? Join the Conversation
Have a question about how this applies to your situation? Ask Dr. Caplan →
Want to discuss this topic with other patients and caregivers? Join the forum discussion →
Have thoughts on this? Share it:
Frequently Asked Questions
Frequently Asked Questions
Does cannabis cause psychosis?
Current evidence shows a consistent association between heavy cannabis use, especially during adolescence, and higher rates of psychotic disorders. However, most studies are observational and cannot fully separate causation from confounding variables such as genetics, trauma exposure, prodromal symptoms, or other substance use. The strongest concern applies to people with existing psychotic vulnerability or family history of schizophrenia-spectrum disorders.
Why is observational research limited when studying cannabis and mental health?
Observational studies can identify associations but cannot definitively prove cause and effect. Many cannabis studies rely on self-reporting, inconsistent exposure definitions, and incomplete adjustment for socioeconomic, psychiatric, or genetic factors. This means correlations may reflect shared vulnerabilities rather than direct biological harm from cannabis itself.
What populations appear most vulnerable to cannabis-related psychiatric risks?
Adolescents, individuals with psychotic disorders, people with bipolar disorder, and those with strong family psychiatric histories appear to face the greatest risk from frequent THC exposure. Evidence consistently suggests worse outcomes in these populations, particularly with high-potency and daily use patterns.
Does CBD carry the same mental health risks as THC?
CBD and THC behave very differently pharmacologically. Preliminary evidence suggests CBD may reduce anxiety in some populations, while THC can worsen anxiety or psychotic symptoms at higher doses in vulnerable individuals. Many reviews incorrectly group all cannabinoids together despite important biological differences between compounds.
What does “11-fold increased psychosis risk” actually mean?
Relative risk can sound dramatic without context. If baseline psychosis risk is approximately 0.5%, an 11-fold increase corresponds to roughly 5.5% absolute risk. That represents a meaningful increase, but still means most heavily exposed individuals do not develop psychosis.
Is occasional adult cannabis use well studied?
Not adequately. Most psychiatric risk studies focus on heavy, near-daily, or adolescent use. Evidence for occasional adult use without psychiatric vulnerability remains sparse, making it difficult to draw confident conclusions about low-frequency exposure patterns.
Why do cannabis studies often produce conflicting conclusions?
Cannabis research suffers from inconsistent exposure measurement, changing product potency, legal barriers to randomized trials, and substantial population differences across studies. One study may examine daily adolescent concentrate use while another evaluates regulated CBD products in adults, yet both may be discussed under the broad label of “cannabis.”
What is the difference between relative risk and absolute risk?
Relative risk compares groups proportionally, while absolute risk reflects the real-world likelihood of an outcome occurring. Relative increases can appear alarming even when the underlying baseline rate is small. Responsible clinical interpretation requires understanding both numbers together.
Does this review prove cannabis has no medical value for mental health?
No. The review mainly demonstrates a lack of large, high-quality randomized controlled trials for psychiatric indications. Absence of strong evidence is not identical to proof of ineffectiveness, particularly in a field historically constrained by legal and regulatory barriers to research.
What is the most clinically responsible takeaway from this JAMA review?
Clinicians should avoid simplistic conclusions. Certain populations likely face elevated risks from THC exposure, especially adolescents and individuals with psychotic vulnerability. At the same time, observational associations should not be overstated as definitive proof of causation for all adults or all patterns of cannabis use.
