Cannabinoids Show Modest, Low-Certainty Benefits for a Few Mental Health Conditions, but Widespread Clinical Use Is Not Justified

The largest systematic review of randomized controlled trials to date finds limited evidence that cannabinoids help with cannabis use disorder, insomnia, tic disorders, and autism spectrum disorder, while documenting meaningful adverse event risks and exposing vast gaps in the evidence base for most psychiatric conditions where cannabinoids are already being prescribed.

Why This Matters

Mental health conditions represent some of the most common reasons patients seek cannabinoid-based therapies, and clinicians encounter these requests with increasing frequency. Depression, anxiety, PTSD, insomnia, and substance use disorders collectively affect hundreds of millions of people worldwide, and the perception that cannabis-based medicines offer a gentler or more natural alternative to conventional psychotropics has become deeply embedded in public discourse. The endocannabinoid system is a scientifically legitimate target in this space: CB1 receptors are densely expressed in the amygdala, prefrontal cortex, and reward circuitry, and endocannabinoid signaling modulates stress reactivity, sleep architecture, and emotional processing in ways that make cannabinoid pharmacology a reasonable line of inquiry. However, a widening gap has opened between the rate at which cannabinoids are being authorized and prescribed for mental health indications and the rigor of the clinical evidence supporting those authorizations. This review, published in The Lancet Psychiatry in 2026, represents the most comprehensive attempt yet to reconcile what has actually been tested in randomized trials with what is being claimed, prescribed, and marketed, making it a critical inflection point for the field’s relationship with its own evidence base.

Clinical Summary

Psychiatric and substance use disorders remain among the leading causes of disability globally, and the limitations of existing pharmacotherapies are well documented. Many patients with anxiety, depression, PTSD, or insomnia experience incomplete responses to first-line treatments, and medications for substance use disorders are underutilized, poorly tolerated, or simply unavailable for certain substances. In this context, the appeal of cannabinoid-based medicines is understandable. Patients and advocacy communities have pushed for expanded access, and regulatory frameworks in numerous jurisdictions now permit cannabinoid prescribing for a range of psychiatric indications. The clinical reality, however, is that the evidence supporting these prescribing patterns has never been comprehensively assembled and critically evaluated at the level of individual conditions, until now.

This systematic review and meta-analysis, led by McCartney and colleagues and published in The Lancet Psychiatry in 2026, synthesized data from 54 randomized controlled trials encompassing 2477 participants across more than a dozen mental health conditions and substance use disorders. The mechanistic rationale for investigating cannabinoids in this space is substantial. The endocannabinoid system, operating through CB1 and CB2 receptors and their endogenous ligands anandamide and 2-arachidonoylglycerol, plays a well-characterized role in modulating GABAergic and glutamatergic neurotransmission in circuits central to mood regulation, threat appraisal, reward processing, and sleep-wake cycling. Phytocannabinoids such as THC and CBD interact with this system through distinct and sometimes opposing mechanisms: THC as a partial CB1 agonist and CBD through a more diffuse pharmacological profile involving serotonin 5-HT1A receptor modulation, TRPV1 activation, and fatty acid amide hydrolase inhibition. The diversity of these targets provides biological plausibility for therapeutic effects across multiple psychiatric domains, but biological plausibility is not clinical evidence.

What the review actually found was far more circumscribed than the breadth of clinical prescribing would suggest. Statistically significant benefits were identified for only four conditions. In cannabis use disorder, the combination of THC and CBD reduced withdrawal symptoms and weekly cannabis use. In insomnia, cannabinoids modestly increased objective and subjective sleep time. In tic and Tourette’s syndrome, cannabinoids reduced tic severity with a moderate effect size. In autism spectrum disorder, cannabinoids reduced autistic trait severity. However, for anxiety disorders, PTSD, psychosis, anorexia nervosa, and opioid use disorder, no significant benefit was detected. For cocaine use disorder, cannabinoids actually increased cocaine craving, representing a potential harm signal. Perhaps most strikingly, no randomized controlled trial evidence exists at all for depression, despite its status as one of the most common indications for which cannabinoids are prescribed in real-world clinical settings.

The limitations of the evidence base are as important as the findings themselves. The 2477 participants were distributed across 54 trials addressing numerous conditions, meaning most condition-specific pooled estimates rested on very few trials with small sample sizes. Forty-four percent of included trials carried a high risk of bias, and GRADE certainty of evidence was rated low for most primary outcomes. The sex distribution was skewed toward male participants at 69 percent, and ethnicity data were largely absent, limiting the generalizability of findings to broader clinical populations. Adverse events were notably elevated: cannabinoids were associated with 1.75-fold higher odds of any adverse event compared to placebo, yielding a number needed to treat for harm of just seven, though serious adverse events and study withdrawal rates did not differ significantly. The authors’ conclusion was unambiguous: given the scarcity of evidence, the routine use of cannabinoids for the treatment of mental disorders and substance use disorders is currently rarely justified, and large, well-powered, condition-specific RCTs are urgently needed before clinical recommendations can be made with confidence.

Dr. Caplan’s Take

This review does something genuinely valuable: it forces the field to confront the distance between what we think we know about cannabinoids in mental health and what has actually been demonstrated in rigorous trials. The mechanistic architecture connecting the endocannabinoid system to psychiatric symptom domains is real and increasingly well-characterized. The scientific rationale for studying cannabinoids in anxiety, sleep, PTSD, and substance use disorders is not speculative. What this review makes painfully clear is that the clinical evidence has not kept pace with the biological plausibility, and certainly not with the pace of regulatory authorization and prescribing.

I see the consequences of this evidence gap regularly. Patients come to me having already tried cannabinoids for depression or anxiety, often self-directed, sometimes with a medical authorization obtained from a provider who may not have communicated the limitations of the evidence. They want to know whether what they are doing is supported by science. In many cases, the honest answer is that it has not been adequately tested, which is a fundamentally different statement than saying it does not work. Patients deserve to hear that distinction articulated clearly, and too often they do not.

The specific findings for cannabis use disorder and insomnia are encouraging but remain preliminary. A single pooled estimate based on a handful of small trials with low GRADE certainty cannot anchor a treatment recommendation. The cocaine craving finding is particularly important because it reminds us that cannabinoids are pharmacologically active agents with real risks, not benign supplements. When I see clinicians treat cannabinoids as though they sit outside the normal standards of evidence-based prescribing, this is the kind of data I point to: a measurable harm signal in a vulnerable population, detected only because someone ran a placebo-controlled trial.

In practice, what I do with this evidence is use it to have more honest conversations. For patients with cannabis use disorder who are interested in managed reduction, the combination THC-CBD approach has at least preliminary trial support, and I discuss it in that framing. For insomnia, I discuss the modest effect sizes and the fact that the evidence does not yet compare favorably with established behavioral and pharmacological treatments. For depression and anxiety, I am transparent that no RCT evidence supports cannabinoid use, and I ensure patients understand that absence of evidence is not license to assume benefit. I document these conversations carefully, and I watch for adverse effects with the same vigilance I would apply to any pharmacological intervention with a number needed to harm of seven.

Clinical Perspective

This review sits at a pivotal point in the maturation of cannabinoid medicine as a clinical discipline. For years, the field has been characterized by a paradox: widespread clinical use running far ahead of the evidence base that should govern it. McCartney and colleagues have now provided the most comprehensive accounting of what randomized trial evidence actually exists for cannabinoids across the full spectrum of mental health conditions, and the picture is sobering. The conditions for which statistically significant benefits were found represent a narrow subset of the indications for which cannabinoids are currently prescribed, and even those findings carry low certainty ratings that make them provisional rather than definitive.

Clinicians communicating with informed patients about this evidence need to be precise. It is accurate to say that preliminary trial evidence suggests cannabinoids may help with cannabis withdrawal symptoms, may modestly increase sleep duration, and may reduce tic severity and some autism-related symptoms. It is not accurate to say that cannabinoids have been shown to treat anxiety, depression, PTSD, or psychosis, because the trial evidence either shows no benefit or does not exist. The distinction between “not yet adequately studied” and “studied and found ineffective” matters enormously in clinical communication. For anxiety, for instance, the absence of significant benefit in pooled estimates does not prove cannabinoids are unhelpful, but it does mean that any recommendation to use them for anxiety currently rests on something other than RCT evidence, and patients should be told that explicitly.

The safety data demand particular attention. The number needed to treat for harm of seven for any adverse event is a clinically meaningful figure, especially in a population where many patients are already managing complex medication regimens and psychiatric comorbidities. The pharmacological diversity of the cannabinoid products studied, ranging from pure CBD to THC-dominant preparations to combination formulations, means that adverse event profiles are not uniform across products. Clinicians should be aware that THC-containing preparations carry greater psychoactive risk and may interact with psychiatric medications through cytochrome P450 inhibition, particularly CYP3A4 and CYP2C19 pathways. CBD at high doses can elevate hepatic transaminases and may alter the metabolism of common psychotropics including clobazam, valproate, and certain SSRIs. The cocaine craving finding, while based on limited data, should serve as a specific caution against assuming cannabinoids are broadly helpful across substance use disorders.

The most actionable recommendation for clinicians right now is to document the evidence basis, or lack thereof, when discussing cannabinoid therapies for mental health conditions with patients. If a patient is using or requesting cannabinoids for depression, the clinical record should reflect that no RCT evidence currently supports this use and that the decision is being made on a shared basis with full informed consent. Clinicians should monitor for adverse events systematically, particularly in the first weeks of use, and should be prepared to discontinue cannabinoid therapy if no measurable benefit is observed within a defined timeframe. Where possible, referring patients to ongoing clinical trials contributes to the evidence base that this review demonstrates is urgently needed.

Study at a Glance

Study Type

Systematic review and meta-analysis of randomized controlled trials

Databases Searched

Ovid MEDLINE, PsychINFO, Cochrane CENTRAL, Cochrane DSR, Embase

Search Period

January 1, 1980 to May 13, 2025

Trials Included

54 RCTs

Total Participants

2477 (69% male, 31% female; median age 33.3 years)

Conditions Covered

Cannabis use disorder, insomnia, tic/Tourette’s syndrome, autism spectrum disorder, anxiety, PTSD, psychosis, anorexia nervosa, opioid use disorder, cocaine use disorder; insufficient data for ADHD, bipolar disorder, OCD, tobacco use disorder; no RCT data for depression

Risk of Bias

24 of 54 trials (44%) rated high risk of bias

GRADE Certainty

Low for most primary outcomes

Statistical Approach

Random-effects meta-analysis; standardized mean differences for continuous outcomes, odds ratios for dichotomous outcomes; NNTH for safety

Registration