Cannabinoids for Mental Health: Large Review Finds Mostly Low-Quality Evidence With Few Clear Benefits

A Lancet Psychiatry systematic review and meta-analysis of 54 randomized controlled trials finds limited and low-certainty support for cannabinoids in cannabis withdrawal, insomnia, Tourette’s syndrome, and autism, while demonstrating no benefit for anxiety, PTSD, psychosis, depression, or opioid use disorder, and highlighting a meaningful adverse event burden across indications.

Why This Matters

Mental health and substance use disorders collectively rank among the leading causes of disability worldwide, and the gap between patient need and effective treatment options remains wide across nearly every diagnostic category. Patients living with anxiety, PTSD, insomnia, autism spectrum disorder, and addiction frequently encounter inadequate responses to existing pharmacotherapies, driving intense interest in alternative or adjunctive approaches. Cannabinoids, including cannabidiol (CBD) and tetrahydrocannabinol (THC), have attracted particular attention because the endocannabinoid system modulates many of the neural circuits implicated in mood regulation, fear extinction, reward processing, and sleep architecture, providing a biologically plausible rationale for their therapeutic exploration. Yet despite this mechanistic promise, cannabinoid-based products have been approved and widely prescribed for psychiatric indications in multiple countries without a comprehensive, rigorous synthesis of the evidence base. This review arrives at a critical moment: patient demand is accelerating, regulatory frameworks are expanding, and clinicians are being asked to make prescribing decisions in the absence of the kind of high-certainty evidence they would expect for any other psychotropic medication class.

Clinical Summary

Mental health and substance use disorders affect hundreds of millions of people globally, yet the therapeutic armamentarium for many of these conditions remains limited, poorly tolerated, or insufficiently effective for a substantial proportion of patients. Insomnia, for instance, is commonly treated with medications that carry habituation risk and next-day impairment. Cannabis use disorder lacks any approved pharmacotherapy in most jurisdictions. Tourette’s syndrome and autism spectrum disorder present with symptoms that current medications address incompletely, often with significant side-effect profiles. Anxiety disorders, PTSD, and psychotic disorders have established pharmacological treatments, but response rates are far from universal, and many patients seek alternatives due to adverse effects or partial benefit. It is against this backdrop of unmet clinical need that cannabinoid therapies have entered practice and public discourse, frequently outpacing the evidence that should guide their use.

Published in Lancet Psychiatry in 2026 by Bahji and colleagues, this systematic review and meta-analysis represents the most comprehensive attempt to date to synthesize randomized controlled trial evidence for cannabinoids across the full spectrum of mental health and substance use disorders. The biological rationale for cannabinoid therapeutics in these conditions rests on the endocannabinoid system’s extensive involvement in central nervous system function. Endocannabinoid signaling through CB1 receptors modulates GABAergic and glutamatergic neurotransmission in circuits governing fear learning, reward, and emotional regulation. CB2 receptors, expressed on microglia and immune cells within the brain, participate in neuroinflammatory processes increasingly implicated in depression, psychosis, and substance use disorders. CBD, in particular, has attracted attention for its putative anxiolytic, antipsychotic, and anti-inflammatory properties through mechanisms that extend beyond classical cannabinoid receptor binding, including modulation of serotonin 5-HT1A receptors, TRPV1 channels, and adenosine reuptake. THC, meanwhile, acts as a partial agonist at CB1, with well-characterized effects on sleep onset, appetite, and pain, but also with known anxiogenic and psychotomimetic potential at higher doses. The review was pre-registered on PROSPERO, searched five major databases from 1980 through May 2025, and applied Cochrane Risk of Bias 2.0 and GRADE certainty assessments to all included evidence.

Across 54 included RCTs enrolling 2477 participants, the review identified statistically significant benefits in only four conditions, all rated as low certainty by GRADE criteria. Combined CBD/THC preparations reduced cannabis withdrawal symptoms (standardized mean difference [SMD] of negative 0.29) and weekly cannabis use quantity (SMD negative 1.00) compared with placebo. Any cannabinoid type improved objectively recorded sleep time in insomnia (SMD 0.54) and sleep diary-recorded sleep (SMD 0.55). Cannabinoids reduced tic severity in Tourette’s syndrome (SMD negative 0.68) and autistic traits in autism spectrum disorder (SMD negative 0.36). In contrast, no significant efficacy was identified for anxiety, PTSD, psychotic disorders, anorexia nervosa, or opioid use disorder. No RCT evidence whatsoever was available for depression, one of the most commonly cited reasons patients seek cannabinoid prescriptions. A particularly concerning finding emerged in cocaine use disorder, where cannabinoids significantly increased cocaine craving (SMD 0.69), representing a clear harm signal rather than a therapeutic benefit.

The safety analysis revealed that cannabinoids significantly increased the odds of any adverse event (odds ratio 1.75, with a number needed to treat for harm of 7), though they did not increase serious adverse events or study withdrawal rates. The review’s limitations are substantial and shape the clinical weight that can be assigned to its conclusions. Forty-four percent of included trials had high risk of bias. Trial sample sizes were generally small, with a total of only 2477 participants spread across 14 diagnostic categories. Ethnicity data were unavailable, limiting the generalizability of findings across demographic groups. Heterogeneity in cannabinoid formulations, doses, and treatment durations further complicates interpretation. The authors conclude that the routine use of cannabinoids for mental disorders and substance use disorders is rarely justified at this time and call for large, well-designed RCTs with standardized formulations and longer follow-up before clinical practice should change.

Dr. Caplan’s Take

This review does exactly what the field has needed for some time: it imposes methodological discipline on a landscape that has been driven far more by patient enthusiasm, commercial interest, and regulatory permissiveness than by rigorous evidence. The finding that only four conditions show any signal of benefit, and that all four carry low GRADE certainty, is not a failure of the research question. It is a clear-eyed acknowledgment that the endocannabinoid system’s biological relevance to psychiatric conditions has not yet translated into reliable therapeutic utility in controlled settings. The architecture of this review, with its Cochrane risk-of-bias scoring and GRADE framework, gives clinicians a transparent tool for evaluating exactly where the evidence stands.

In practice, I encounter patients every week who have already decided that cannabinoids should help their anxiety, their PTSD, their insomnia, or their depression. They arrive with conviction shaped by personal experience, social media, or well-meaning but scientifically unsupported recommendations. What they are actually asking for is validation of an approach they have already started, or permission to try something that feels more natural or tolerable than their current medications. An honest clinical response requires acknowledging that the biological rationale is real while being direct that the controlled trial evidence does not yet confirm the clinical benefit they are hoping for, and in some cases actively contradicts it.

The reason this gap persists is not that cannabinoids have been proven ineffective. It is that they have been inadequately studied. Most trials are small, short, and use formulations that differ meaningfully from what patients access in dispensaries or over the counter. The 44% high-risk-of-bias rate means that nearly half of the studies contributing to these pooled estimates have methodological weaknesses that could inflate or distort effect sizes in either direction. The absence of any RCT evidence for depression is particularly striking given how frequently cannabinoids are prescribed or self-administered for depressive symptoms. Until we have adequately powered trials with standardized products, consistent dosing, and clinically meaningful follow-up durations, the honest position is that we do not know enough to recommend cannabinoids for most psychiatric indications.

What I do in the meantime is maintain transparency with patients about the evidence hierarchy and work within it. For patients with cannabis use disorder who are struggling with withdrawal, the data on combined CBD/THC preparations are at least directionally encouraging, and I discuss that signal with appropriate caveats. For insomnia, I acknowledge the modest sleep-duration findings while being clear that the evidence base is thin compared with established behavioral and pharmacological treatments. For anxiety and PTSD, I am direct that the RCT evidence does not support cannabinoid use, even as I validate the patient’s experience and explore what may be driving their perceived benefit, whether that is a true pharmacological effect, a placebo response, or the anxiolytic properties of a self-directed ritual. And for every patient, I document the conversation, the evidence reviewed, and the shared decision reached, because the standard of care in cannabinoid medicine right now is informed consent, not confident recommendation.

Clinical Perspective

This review arrives at a point where clinical practice in cannabinoid medicine has outstripped its evidence base in ways that are difficult to defend by conventional pharmacological standards. Cannabinoids have been approved for psychiatric indications in several countries, and clinicians in jurisdictions with medical cannabis programs routinely authorize their use for anxiety, PTSD, insomnia, and chronic pain with comorbid mood disorders. The Bahji review now provides the most comprehensive and methodologically rigorous synthesis to date, and its central finding is that the evidence justifying these practices is thin. It does not close the door on cannabinoid therapeutics in psychiatry, but it makes unmistakably clear that the door was opened prematurely and without the evidentiary foundation clinicians should demand.

For patient-facing conversations, the most important clinical skill this review demands is the ability to distinguish between “biologically plausible” and “clinically demonstrated.” Patients who ask about cannabinoids for anxiety or depression deserve to know that the endocannabinoid system is genuinely involved in mood and stress regulation, and that this is an active area of scientific inquiry. They also deserve to know that no randomized controlled trial has demonstrated cannabinoid superiority over placebo for generalized anxiety, social anxiety, PTSD, or depression, and that for cocaine use disorder, the evidence suggests cannabinoids may actually worsen craving. Clinicians should resist the temptation to frame absence of evidence as evidence of absence, but they must be equally firm in resisting the opposite error: treating mechanistic plausibility as a substitute for demonstrated clinical benefit. Communicating this distinction honestly, without condescension, is what separates responsible cannabinoid medicine from both dismissive prohibition and uncritical advocacy.

The pharmacological and safety considerations raised by this review warrant specific attention. The adverse event profile, with a number needed to treat for harm of 7 for any adverse event, is clinically significant and should be disclosed to every patient. While serious adverse events were not elevated, the breadth of non-serious adverse effects (sedation, dizziness, gastrointestinal symptoms, cognitive blunting) can meaningfully impair function, particularly in patients already managing psychiatric symptom burdens. The cocaine craving finding demands particular caution in polysubstance use populations: clinicians should not assume that cannabinoids are pharmacologically inert or benign in the context of active substance use disorders. Drug-interaction considerations also remain underexplored in these trials. CBD is a potent inhibitor of CYP3A4 and CYP2D6, and patients on SSRIs, benzodiazepines, antipsychotics, or anticonvulsants require careful monitoring for altered drug levels when CBD is introduced, a complexity that most trial designs in this review did not address.

The most actionable recommendation for clinicians at this moment is to treat cannabinoid prescribing for psychiatric indications with the same evidentiary rigor applied to any other psychotropic agent. This means documenting the clinical rationale, the evidence discussed with the patient, the formulation and dose selected, the monitoring plan, and the criteria for discontinuation if benefit is not observed. Clinicians should strongly consider referring interested patients to ongoing clinical trials rather than initiating off-label cannabinoid therapy in the absence of supporting evidence. For the four conditions where modest signals exist (cannabis use disorder, insomnia, Tourette’s syndrome, and autism spectrum disorder), shared decision-making with full disclosure of the low certainty of evidence and the adverse event profile is the appropriate framework. Routine prescribing for anxiety, PTSD, depression, or psychotic disorders cannot be supported by the current evidence and should not be pursued outside of a research context.