General Medicine · Medical Literature
[Correspondence] Efficacy and safety of cannabinoids – Authors’ reply
Abstract
We thank Albert Batalla and colleagues and Jing Yuan and colleagues for their comments on our Article.1 As noted, clinical trial data are limited in the generalisability of findings, because the administration of medicinal cannabis in controlled laboratory settings differs to how it is used and prescribed in the real-world.
Clinical relevance
Do cannabinoids demonstrate efficacy and an acceptable safety profile when used as a primary treatment for mental disorders or substance use disorders in randomized controlled trials, and do RCT findings generalize to real-world clinical prescribing contexts? __SNAPSHOT__ {"studyDesign":"Systematic review and meta-analysis of randomized controlled trials (RCTs). The document submitted for summary is an authors' reply (Correspondence) published August 2026 responding to post-publication commentary on the primary article; no new primary data are introduced in the reply.","setting":"International; multi-database search (Ovid MEDLINE, PsychINFO, Cochrane CENTRAL, Cochrane Database of Systematic Reviews, Embase); coordinating institutions include the University of Sydney (Matilda Centre), University of Queensland (National Centre for Youth Substance Use Research), Monash University (Monash Addiction Research Centre), and University of Bath (Addiction and Mental Health Group).","sampleSize":"54 RCTs; approximately 2,477 total participants across included trials. Median trial size: approximately 31.5 participants (reported in secondary commentary; original abstract value not confirmed from abstract text alone — treat as approximate).","population":"Adults with diagnosed mental disorders or substance use disorders (SUDs) enrolled in RCTs published between January 1, 1980, and May 13, 2025. Conditions evaluated include anxiety disorders, depression, PTSD, psychotic disorders, ADHD, autism spectrum disorder (ASD), insomnia, Tourette's syndrome/tic disorders, cannabis use disorder, opioid use disorder, cocaine use disorder, and other SUDs. Approximately 69% of participants were male; median age approximately 33 years.","intervention":"Plant-based and pharmaceutical cannabinoids (CBD alone, THC alone, or combined THC+CBD formulations) administered as the primary treatment for a mental disorder or SUD. Of 54 included trials: 24 tested CBD in isolation, 18 tested THC alone, and 12 used combined formulations.","comparator":"Placebo or active control, as specified in each included RCT.","primaryOutcome":"Efficacy of cannabinoids in reducing or treating symptoms of mental disorders or SUDs (primary indication), and safety assessed by adverse event rates including serious adverse events and all-cause adverse events.","followUp":"Varied by included trial; a separate scoping review noted durations from 1 day to 13 weeks across included RCTs. Aggregate follow-up duration for the meta-analysis as a whole: not reported as a single figure in available abstract or secondary sources.","fundingSource":"Government (Australian National Health and Medical Research Council [NHMRC]). Some authors received WHO consultation fees or Therapeutic Goods Administration funding for separate evidence reviews; one author received payment for expert testimony on cannabis risks. No pharmaceutical industry sponsorship of the primary study was reported. Approximately 20% of included RCTs raised concerns about conflicts of interest or industry affiliations at the trial level."}
Key findings
The meta-analysis found no significant effect of cannabinoids on anxiety symptoms (SMD = −1.88; 95% CI −4.79 to 1.03; not statistically significant). No RCTs evaluating cannabinoids for major depressive disorder met inclusion criteria, meaning this indication was entirely absent from the evidence base. No significant effect was demonstrated for PTSD, psychotic disorders, OCD, anorexia nervosa, or opioid use disorder. Cannabinoids were associated with a statistically significant increase in all-cause adverse events overall (OR = 1.75; 95% CI 1.25 to 2.46), but serious adverse events and study withdrawals did not differ significantly between cannabinoids and controls across conditions. The NNTH from a 2016 predecessor review for anxiety was reported as 7 (for short-term adverse events); this figure applies to a prior review, not the current meta-analysis — included for historical context only. Limited positive signals were identified for: cannabis use disorder (THC+CBD combinations, reducing cannabis consumption and withdrawal when combined with psychological therapy); insomnia (any cannabinoid); Tourette's syndrome (THC+CBD); and autism spectrum disorder — however, the quality of evidence for all these conditions was rated as low. Cannabinoids were reported to increase cravings in cocaine use disorder, suggesting potential harm in this population. NNT values for conditions showing signal (cannabis use disorder, insomnia, ASD, Tourette's) were not reported in available abstract or secondary sources.
Practice implications
This is the largest RCT-only meta-analysis on cannabinoids for mental disorders and SUDs to date. It establishes that current high-quality controlled trial evidence does not support routine prescribing of cannabinoids for anxiety, depression, PTSD, psychotic disorders, OCD, anorexia nervosa, or opioid use disorder. The complete absence of RCT data for depression is itself a clinically important finding, indicating that widespread prescribing for this indication has occurred without any controlled trial foundation. The study did not establish that cannabinoids are ineffective — absence of evidence is not evidence of absence — but rather that the evidence base is too small, too heterogeneous, and of too low certainty to justify routine use for most psychiatric indications. The authors' August 2026 reply (the document submitted for this summary) explicitly acknowledges that RCT findings have limited generalizability because controlled laboratory administration differs substantially from real-world prescribing and use patterns. Populations not adequately represented include patients with complex multi-morbid presentations, those using high-THC unregistered products, those self-medicating, and real-world titration scenarios with variable cannabinoid ratios. The review's findings are most robustly applicable to pharmaceutical-grade cannabinoid formulations in carefully selected, relatively homogeneous trial populations. Clinicians should exercise caution in extrapolating findings in either direction: the RCT evidence does not permit confident conclusions about what happens when patients use high-potency, variable-composition products over extended periods in routine care. Only one abstract was available for the correspondence itself; the clinical interpretation of the correspondence relies primarily on the original article and publicly available secondary sources. Practice should not change toward increased cannabinoid prescribing for the conditions examined; the signal for cannabis use disorder warrants cautious optimism pending confirmatory trials. __LIMITATIONS__ 1. Only RCTs were included, deliberately excluding observational and registry data that may better reflect real-world prescribing — a methodological choice that strengthens internal validity but limits external generalizability. 2. Aggregate sample size across 54 RCTs was only approximately 2,477 participants, with a median trial size of approximately 31 participants — this is an extremely thin evidence base for a widely prescribed class of medicines. 3. Complete absence of RCT data for depression, the most common mental health indication for which cannabinoids are prescribed. 4. Substantial clinical and methodological heterogeneity across included trials in terms of cannabinoid type, dose, route of administration, formulation, patient population, and outcome measurement. 5. Most included trials used pharmaceutical-grade registered cannabinoids (e.g., Sativex), which differ substantially from high-THC unregistered products dominant in current real-world markets. 6. Blinding integrity is inherently compromised in THC trials due to psychoactive effects, increasing risk of performance and detection bias. 7. Outcome measurement for cannabis use relied predominantly on self-report rather than objective verification in many included trials. 8. Approximately 20% of included RCTs raised concerns about conflicts of interest or industry affiliations, though the primary study itself was government-funded. 9. Short follow-up durations (up to 13 weeks) preclude conclusions about long-term efficacy or safety. 10. The correspondence (authors' reply) is available only as a brief abstract excerpt, meaning the full scope of the authors' responses to specific methodological criticisms raised by Batalla et al. and Yuan et al. cannot be assessed in this summary. __BOTTOM_LINE_CATEGORY__ potentially_practice_informing __BOTTOM_LINE__ This landmark meta-analysis of 54 RCTs finds no controlled trial evidence supporting cannabinoid use for anxiety, PTSD, depression, psychotic disorders, OCD, or opioid use disorder — and no RCTs for depression exist at all — while demonstrating a statistically significant increase in all-cause adverse events. Limited low-certainty signals favor THC+CBD combinations for cannabis use disorder, and cannabinoids for insomnia, Tourette's syndrome, and ASD, but these findings require confirmatory trials before routine prescribing is justified. The authors' reply appropriately acknowledges that RCT findings have constrained generalizability to real-world cannabis prescribing, underscoring the urgent need for pragmatic, adequately powered trials that reflect how medicinal cannabis is actually used in clinical practice.