Randomised Controlled Trial Evidence on Medicinal Cannabis for Treatment of Mental Health and Substance Use Disorders: A Scoping Review

Abstract

Background

With shifting perceptions about the therapeutic potential of cannabis and evolving regulatory frameworks, global prescribing of medicinal cannabis is increasing. While some emerging evidence supports its use for conditions like multiple sclerosis and epilepsy, its efficacy and safety profile for the treatment of mental health conditions remains controversial and under-explored. Previous reviews found inconclusive evidence due to heterogeneity in study design and quality. Accordingly, this review was designed as a scoping review, consistent with established methodological frameworks to map and characterise all available randomised controlled trial (RCT) evidence in this emerging and heterogeneous field. It specifically sought to synthesise the highest-quality trial evidence to date, addressing the question: How effective is medicinal cannabis in treating mental health conditions, as classified by the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), and how safe and tolerable is it, as assessed through adverse events and treatment withdrawals?

Methods

A scoping review was conducted comprising RCTs investigating medicinal cannabis for mental health conditions. Eligible studies were required to meet predefined inclusion criteria based on population, intervention, comparator, outcomes, and study design (PICOS framework). PubMed, Web of Science, and PsycINFO databases were searched, supplemented by citation tracking and Google Scholar, for studies published between 1980 and 2024.

Results

The search identified 8061 studies, with 28 RCTs meeting inclusion criteria across 12 DSM-5 mental health conditions. Indications most frequently studied were schizophrenia (n = 5), cannabis use disorder (n = 4), cocaine use disorder (n = 4), post-traumatic stress disorder (n = 3), anxiety disorders (n = 3), and opioid use disorder (n = 2); there were two trials in autism spectrum disorder and single trials in depression, attention-deficit/hyperactivity disorder, obsessive-compulsive disorder, tobacco use disorder, and Tourette syndrome. Sample sizes ranged from 6 to 150 participants (median = 42), and follow-up durations from 1 day to 13 weeks (median = 6 weeks). Interventions included purified cannabidiol (CBD; single doses of 300–800 mg and daily regimens up to 1000 mg/day), nabiximols or other tetrahydrocannabinol (THC)/CBD oromucosal sprays (up to 113 mg THC/105 mg CBD per day), and smoked or vaporised cannabis flower of varying THC/CBD content. Findings showed substantial heterogeneity and variable quality, with some short-term benefits reported (notably in cannabis use disorder, autism spectrum disorder, and schizophrenia), but no trial demonstrated long-term efficacy.

Conclusion

Despite growing interest, substantial heterogeneity limits current evidence for medicinal cannabis in mental health. This review highlights key gaps, underscoring the need for robust, well-powered RCTs with extended follow-up to clarify its role in the management of mental ill health.

Supplementary Information

The online version contains supplementary material available at 10.1007/s40261-025-01501-3.

Key Points

Evidence for the therapeutic potential of medicinal cannabis is highly variable across different mental health conditions. While promising results were observed in some areas, many findings are inconclusive due to study heterogeneity and limitations in trial design.

Despite growing interest, significant gaps remain in research, particularly for conditions such as depression, ADHD, OCD, and panic disorders. The lack of long-term efficacy data for most conditions is a critical concern.

Many studies are limited by small sample sizes, short durations, and variability in dosing and outcome measures. These limitations restrict the generalisability of findings to real-world settings.

Future research should focus on conducting high-quality, adequately powered trials with rigorous methodologies, standardised measures, and longer follow-up periods to provide more conclusive evidence.

Background

With shifting perceptions about the therapeutic potential of cannabis and evolving regulatory frameworks, global prescribing of medicinal cannabis is increasing [1]. Medicinal cannabis, which includes cannabis flower, cannabinoid plant extracts, and cannabis-derived cannabinoid-based products, is prescribed by health practitioners for various therapeutic indications [2]. Whilst clinical evidence is emerging supporting the use of medicinal cannabis for conditions such as spasticity in multiple sclerosis [3] and childhood epilepsy [4–6], its application in treating mental health conditions remains controversial and under-explored [7].

Numerous systematic reviews have found inconclusive evidence regarding the efficacy of medicinal cannabis for treating mental health conditions. For example, evidence remains mixed for anxiety disorders [8–10], post-traumatic stress disorder (PTSD) [11, 12], psychotic disorders [13–15], and substance use disorders [7, 16]. Despite these efforts, the body of evidence continues to be limited by substantial heterogeneity in study design, methodology, and quality [7, 14, 17].

This considered, the present review synthesises evidence exclusively from randomised controlled trials (RCTs), which represent the gold standard of clinical evidence necessary to inform both clinical guidelines and policy [18]. Building on the review by McKee et al. [7], which also focused exclusively on RCT evidence, this study addresses the growing body of recent trials to provide an essential update [7]. In doing so, we aim to identify and evaluate all existing high-quality trial evidence assessing the efficacy and tolerability of medicinal cannabis in the treatment of mental health conditions.

We have framed this as a scoping review rather than a systematic review because our intent was not to pool data to answer a narrowly defined question of effectiveness, but to map, characterise, and critically appraise the breadth of RCT evidence in this emerging and heterogeneous field. Scoping reviews are recommended when evidence is complex, diverse, and methodologically varied, particularly where the aim is to identify knowledge gaps and inform future research directions [19, 20] As Fletcher et al. [21] emphasise, scoping reviews are especially valuable in rapidly evolving clinical areas such as medicinal cannabis, where systematic review approaches may be premature

With shifting perceptions about the therapeutic potential of cannabis and evolving regulatory frameworks, global prescribing of medicinal cannabis is increasing. While some emerging evidence supports its use for conditions such as multiple sclerosis and epilepsy, its efficacy and safety profile for the treatment of mental health conditions remains controversial and under-explored. Previous reviews have found inconclusive evidence due to heterogeneity in study design and quality. Accordingly, this review sought to map and describe the highest-quality trial evidence to date, addressing the question: how effective is medicinal cannabis in treating mental health conditions, as classified by the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5), and how safe and tolerable is it, as assessed through adverse events and treatment withdrawals?

Methods

This scoping review is performed with a systematic approach informed by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) Extension for Scoping Reviews (PRISMA-ScR) statement [22]. This review was conducted as a scoping review of RCTs to systematically map the breadth and characteristics of high-quality evidence on medicinal cannabis for mental health conditions and substance use disorders, identify research gaps, and summarise key findings without undertaking a formal meta-analysis or validated risk of bias grading, in line with PRISMA-ScR methodology. The protocol for this review was registered with the International Prospective Register of Systematic Reviews (PROSPERO), under registration number CRD42024564177 https://www.crd.york.ac.uk/PROSPERO/view/CRD42024564177 accessed 18/08/2025.

Data Sources and Search Strategy

Selection Criteria

We included all published RCTs in humans, investigating the efficacy of medicinal cannabis (including cannabis flower, cannabinoids, and cannabis-derived cannabinoid-based products) for the treatment of mental health symptoms in people with a primary mental health diagnosis recognised in the DSM-5 [23].

Search Strategy

The search strategy informed by PRISMA guidelines, and the PICOS framework (Table 1) was used to identify three key terms – cannabis, RCT and mental health conditions [24]. PubMed, Web of Science and PsycINFO were searched using the search terms outlined in Online Resource 1 for studies published from 1980 to 2024. Citation searching and Google Scholar were used to identify papers that had not yet been indexed and were not identified in the initial search.

Table 1.

Eligibility criteria for a scoping review of medicinal cannabis for mental health

	Inclusion	Exclusion
Population	Humans Primary mental health diagnosis recognised in the Diagnostic and Statistical Manual for Mental Disorders (DSM-5) All ages
Intervention	Medicinal cannabis, including cannabis flower, cannabinoids, and cannabis derived cannabinoid-based products (i.e., Sativex® (Nabiximols) and Epidyolexa) Singularly (single component of cannabis such as cannabidiol [CBD] or tetrahydrocannabinol [THC]) or in combination (CBD and THC, in various ratios) All routes of administration As primary treatment or adjunct	Synthetic cannabinoids Synthetic cannabinoid-based products Non-medicinal cannabis
Comparison	Placebo or active comparator
Outcome(s)	Mental health conditions remission or change in mental health symptoms
Study type	Randomised controlled trials Developed countries (UN development index) English	All non RCT study types Protocols Grey literature Studies where full text was not available

RCT randomised controlled trial

^aEmerge Health Pty Ltd Suite 3, Level 1, 2 Theatre Place, Canterbury VIC 3126

The search strategy was informed by PRISMA guidelines, and the PICOS framework was used to identify three key terms, namely cannabis, RCT and mental health conditions [24]. PubMed, Web of Science and PsycINFO were searched using the search terms outlined in Online Resource 1 for studies published from 1980 to 2024, which covers the period from the onset of medicinal cannabis decriminalisation and shifts in public and scientific perceptions of cannabis [25, 26].

Synthetic cannabinoids were excluded because their pharmacological profiles, potency, and safety risks differ substantially from plant-derived cannabinoids, and they are not classified as medicinal cannabis under most regulatory frameworks. Including them would introduce heterogeneity in intervention types and limit the applicability of findings to current medicinal cannabis prescribing practices.

Covidence software was used for the initial title and abstract screening, full-text review for study selection, and extraction, including the use of the Cochrane RCT classifier automation tools in Covidence [27, 28]. Study selection is reported in the PRISMA flowchart in Fig. 1 and excluded studies are described in Online Resource 2. Screening was conducted independently by three reviewers (SC, CMH, and YAB) to ensure reliability, with disagreements resolved by discussion until consensus was reached.

Fig. 1.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow diagram. MC mast cells, MHC major histocompatibility complex

The Critical Appraisal Skills Programme (CASP) Randomised Controlled Trial Standard Checklist (Online Resource 3), originally based on JAMA Users’ Guides to the Medical Literature (1994) and updated using the CONSORT 2010 guidelines, was used as a framework to assess the design, reporting, and generalisability of the included RCTs [29]. This process ensured that methodological quality was considered as part of the scoping review, consistent with recent recommendations for enhancing transparency in evidence mapping. The checklist was not used to produce a numeric score but rather to support an overall assessment of study quality, focusing on clarity of aims, robustness of data collection, appropriateness of analytic methods, and validity of interpretations (Online Resource 3). SC critiqued all articles, while CMH and YB each independently reviewed a subset to ensure that at least two reviewers assessed every study. Disagreements were resolved through team discussion until consensus was reached. To enhance transparency, we collated the methodological limitations that were most commonly observed into a descriptive summary table, reporting the proportion of trials affected in each domain (e.g., short duration, small sample size, incomplete outcome data). This addition provides a clear overview of recurrent issues while remaining consistent with scoping review principles by avoiding formal scoring or aggregate ‘risk-of-bias’ ratings.

In line with the scoping review design, we conducted a structured quality assessment of included RCTs, focusing on study design features, sample characteristics, intervention details, and reporting transparency. A validated risk-of-bias tool was not applied, as the primary aim was to map and summarise the breadth of available high-quality evidence rather than to formally grade it. This approach is consistent with scoping review methodology, which prioritises comprehensive evidence mapping over detailed bias quantification, while still providing readers with an appraisal of study quality relevant to interpretation [19, 20].

Results

Overview

Of the 8061 studies identified through database searches, 62 were included for full text review. Of these, 28 peer-reviewed RCTs were selected for inclusion, based on alignment with the inclusion criterion (Table 1).The included studies that investigated the use of cannabinoids across a range of primary mental health conditions (including anxiety disorders, post-traumatic stress disorder, psychotic disorders, major depressive disorder, attention-deficit/hyperactivity disorder, and substance use disorders). Trials used both parallel-group (n = 20) and crossover (n = 8) designs and were conducted across 12 countries, most commonly in the USA, Canada, and Australia. Sample sizes ranged from 6 to 150 participants (median = 42), with study durations varying from acute laboratory-based assessments (1–3 days) to extended treatment periods of up to 13 weeks (median = 6 weeks).

Cannabinoid interventions included purified cannabidiol (CBD), tetrahydrocannabinol (THC), and combination CBD–THC products in varying ratios. Doses ranged from single acute administrations (e.g., CBD 300–600 mg) to repeated daily dosing regimens (e.g., THC 2.5–10 mg twice daily, CBD up to 1000 mg/day). Routes of administration included oral capsules, sublingual sprays, and inhaled vaporised cannabis flower.

Primary outcomes were condition-specific, including validated symptom severity scales (e.g., Positive and Negative Syndrome Scale, Clinician-Administered PTSD Scale), neurocognitive performance measures, quality of life indices, and, where relevant, substance use outcomes such as craving intensity, withdrawal severity and abstinence rates.

The CASP appraisal identified variability across key domains of methodological quality. Several trials demonstrated strengths in randomisation and blinding procedures, particularly in studies using matched placebos and clearly reported allocation concealment. However, incomplete reporting of sequence generation, lack of allocation concealment details, and insufficient information on adherence and protocol deviations were common. Sample sizes were often small and underpowered, limiting the precision of effect estimates. While most studies used validated outcome measures and reported appropriate statistical analyses, the handling of missing data and reporting of adverse events were inconsistent. Few trials provided follow-up data beyond the intervention period, and none included an economic evaluation component.

The diversity in study designs and outcomes reflects the evolving interest in cannabinoids as a potential therapeutic option (Table 2).

Table 2.

Study characteristics

Indication	Author(s) and year	Title	Country	RCT design	Intervention	Sample size	Length of study	Primary outcome(s)	Key findings
ADHD	Cooper, Williams, Seegobin, Tye, Kuntsi and Asherson, 2017 [30]	CBDs in attention-deficit/hyperactivity disorder: a randomised-controlled trial	U K	Parallel	Nabiximols (up to 37.8 mg THC/35 mg CBD) or 0 mg (placebo)	30	6 wk	Cognitive performance and activity level using the QbTest	No sig. effect on primary (QbTest; Est = − 0.17, 95% CI − 0.40 to 0.07, p = 0.16). Secondary: nominal ↓ hyperactivity/impulsivity (p = 0.03), ↑ inhibition (p = 0.05). Trends for inattention (p = 0.10), emotional lability (p = 0.11). AEs in both groups
Anxiety	Bergamaschi et al, 2011 [31]	Cannabidiol reduces the anxiety induced by simulated public speaking in treatment-naïve social phobia patients	Brazil	Parallel	CBD 600 mg or 0 mg (placebo)	24	1 day	VAMS, SSPS-N, BSS	CBD 600 mg ↓ public-speaking anxiety in social anxiety disorder: VAMS anxiety (d = 0.95, p < 0.01), cognitive impairment (d = 1.05, p < 0.01), discomfort (d = 1.15, p < 0.01) vs placebo. No effect on physiological measures
	Crippa et al, 2011 [32]	Neural basis of anxiolytic effects of CBD in generalized social anxiety disorder: a preliminary report	Brazil	Crossover	CBD 400 mg or 0 mg (placebo)	10	2 d (1 wk apart)	VAMS, regional cerebral blood flow (rCBF) at rest	CBD 400 mg significantly decreased subjective anxiety compared to placebo (p < 0.001) Reduced rCBF in left parahippocampal gyrus, hippocampus and inferior temporal gyrus (p < 0.001), increased rCBF in right posterior cingulate gyrus (p < 0.001) No correlations found between subjective ratings in VAMS and rCBF in any brain areas
	Kwee et al, 2022 [33]	Cannabidiol enhancement of exposure therapy in treatment refractory patients with social anxiety disorder and panic disorder with agoraphobia: a randomised controlled trial	Netherlands	Parallel	CBD 300 mg or 0 mg (placebo)	80	8-wk (with 3-month and 6-month follow-up)	The Fear Questionnaire (measures level of avoidance because of the anxiety disorder)	CBD (300 mg) did not enhance exposure therapy outcomes in social anxiety disorder or panic disorder with agoraphobia: FQ β = 0.32 (95% CI − 0.60 to 1.25) across groups; β = − 0.11 (95% CI − 1.62 to 1.40) within groups; NS AEs comparable (CBD 10.3%, placebo 15.4%)
ASD	Aran et al, 2021 [34]	CBD treatment for autism: a proof-of-concept randomised trial	Israel	Crossover	Cannabis extract (CBD/THC, 20:1), and purified CBD/THC at the same ratio (up to 420 mg CBD/21 mg THC) or 0 mg (placebo)	150	12-wk efficacy assessment, 4-wk washout, cross-over for another 12 wk to further assess tolerability	Behavioural problems assessed using the Home Situation Questionnaire-ASD (HSQ-ASD) and the CGI-I scale with disruptive behaviour anchor points	No diff. on HSQ-ASD (primary) or APSI (secondary). CGI-I disruptive behaviour improved in 49% on whole-plant extract vs 21% placebo (p = 0.005). SRS total ↓14.9 vs ↓3.6 (p = 0.009). No serious AEs; common AEs: somnolence (28%) and ↓appetite (25%) on extract vs placebo (8%, 15%)
	da Silva et al, 2024 [35]	Evaluation of the efficacy and safety of cannabidiol-rich cannabis extract in children with autism spectrum disorder: randomized, double-blind, and placebo-controlled clinical trial	Brazil	Parallel	Cannabis extract at a concentration of 0.5% (5 mg/mL), in the ratio of 9CBD:1THC, up to 70 drops daily	60	12 wk	Aggressiveness, psychomotor agitation, concentration, meals (no. meals/day), sleep (no. hours of sleep/day), social interaction with peers, verbal language (speech), anxiety; and repetitive and stereotyped movements (stereotypes evaluated using semi-structured interview prepared by the authors containing questions related to ASD symptoms and the ATEC	CBD-rich extract improved social interaction (F1,116 = 14.13, p = 0.0002), ↓anxiety (F1,116 = 5.99, p = 0.016), ↓psychomotor agitation (F1,116 = 9.22, p = 0.003), ↑meals/day (F1,116 = 4.11, p = 0.04), ↑concentration in mild ASD (F1,48=6.75, p = 0.01). AEs in 3/31 (9.7%) children (dizziness, insomnia, colic, weight gain)
Depression	Pinto et al, 2024 [36]	Cannabidiol as an adjunctive treatment for acute bipolar depression: a pilot study	Brazil	Parallel	Up to 300 mg CBD or 0 mg (placebo)	35	12 wk	Changes in the MADRS	CBD (150–300 mg/day) adjunct ↓MADRS scores (placebo − 14.56; CBD − 15.38), no sig. diff. vs placebo. No sig. effects on secondary outcomes. Exploratory: CBD 300 mg/day ↓MADRS from wk 2–8 (placebo − 6.64; CBD − 13.72). No ↑ manic symptoms or AEs. CBD well tolerated
Cannabis use disorder	Allsop et al, 2014 [37]	Nabiximols as an agonist replacement therapy during cannabis withdrawal a randomized clinical trial	Australia	Parallel	Nabiximols (up to 86.4 mg THC/80 mg CBD) or 0 mg (placebo)	51	6-d regimen of Nabiximols with a 28-d follow-up	Severity of cannabis withdrawal and cravings (CWS-C), retention in withdrawal treatment, and AEs using a 4-point severity scale (0, none; 1, mild; 2, moderate; and 3 severe)	Nabiximols (THC 86.4 mg + CBD 80 mg/day, 6 d) ↓withdrawal severity vs placebo (F8377.97 = 2.39, p = 0.01), incl. ↓irritability, depression, cravings; limited benefit on sleep, anxiety, appetite, restlessness. ↑treatment retention (HR = 3.66, 95% CI 1.18–11.37, p = 0.02; NNT = 2.84). No ↑ intoxication (F1,6 = 0.22, p = 0.97). No sig. diff. in AEs (F1,50 = 0.3, p =0.59; severity F1,50 = 2.69, p = 0.10). At follow-up, both groups ↓cannabis use, no advantage vs placebo (F1,48 = 0.29, p = 0.75)
	Trigo et al, 2016 [38]	Effects of fixed or self-titrated dosages of Sativex on cannabis withdrawal and cravings	Canada	Crossover	Sativex (up to 108 mg THC/100 mg CBD) or 0 mg (placebo)	9	8 wk	Withdrawal symptoms and craving assessed using the CWS, MWC and MCQ. Cannabis use assessed via self-reports, vial weight control, toxicology, and metabolite analysis	Sativex (fixed 40 sprays/day; self-titrated up to 108 sprays/day, 6 d): ↓ withdrawal symptoms vs placebo (CWS total p < 0.05). Self-titrated also ↓ anxiety, sleep disturbance. No sig. diff. in retention or cannabis uses at follow-up. AEs mild; no SAEs
	Trigo et al., 2018 [39]	Nabiximols combined with motivational enhancement/cognitive behavioural therapy for the treatment of cannabis dependence: A pilot randomized clinical trial	Canada	Crossover	Nabiximols (up to 113.4 mg THC/105mg CBD) or 0 mg (placebo)	40	12 wk	Tolerability and number of days abstinent	Nabiximols (self-titrated up to 113.4 mg THC + 105 mg CBD/day, 12 w) + MET/CBT: no diff. in abstinence vs placebo. Cannabis use ↓70.5% nabiximols vs ↓42.6% placebo. ↓craving, but no sig. diff. in withdrawal scores. No SAEs; AEs not sig. diff. (F1,39 = 0.205, NS). Participants unable to distinguish active vs placebo (F1,40 = 0.585, NS)
	Freeman et al, 2020 [40]	Cannabidiol for the treatment of cannabis use disorder: a phase 2a, double-blind, placebo-controlled, randomised, adaptive Bayesian trial	UK	Parallel	CBD 200 mg, 400 mg, or 800 mg or 0 mg (placebo) (stage 1), CBD 400 mg, or 800 mg or 0 mg (placebo) (stage 2)	48 (stage 1), 34 additional participants (stage 2)	4 wk	Cannabis use, as measured in urine (THC-COOH: creatinine ratio) and by self-report (days/wk with abstinence from cannabis	CBD (200/400/800 mg/day, 4 w) + MI: 200 mg dropped at interim (inefficacious). At final, 400 mg and 800 mg met efficacy criteria (posterior prob. >0.9). CBD 400 mg ↓urinary THC-COOH/creatinine by − 94.21 ng/mL (95%CrI − 161.83 to − 35.56) and ↑abstinence +0.48 d/wk. (0.15–0.82). CBD 800 mg ↓THC-COOH/creatinine by − 72.02 ng/mL (− 135.47 to − 19.52) and ↑abstinence +0.27 d/wk (− 0.09 to 0.64). Well tolerated; no SAEs; 94% completed treatment.
Cocaine use disorder	Mongeau-Pérusse et al, 2021 [41]	Cannabidiol as a treatment for craving and relapse in individuals with cocaine use disorder: a randomized placebo-controlled trial	Canada	Parallel	CBD (800 mg) or 0 mg (placebo)	78	12 wk	Drug–cue-induced craving during detoxication (VAS) and time-to-cocaine relapse during subsequent outpatient treatment	CBD 800 mg/day (12 w, post-detox): No sig. effect on cue-induced craving (Δ+ 4.69 vs + 3.21; CI − 0.33 to 3.04; p = 0.069; BF=0.498). No diff. in relapse risk (HR = 1.20, 95% CI 0.65–2.20, p = 0.51; BF = 0.152). Nearly all relapsed by Wk12. CBD well tolerated; main AE: diarrhoea
	de Meneses-Gaya et al, 2021 [42]	Cannabidiol for the treatment of crack-cocaine craving: an exploratory double-blind study	Brazil	Parallel	CBD (300 mg) or 0 mg (placebo)	31	10 d	Craving was assessed with the Cocaine Craving Questionnaire-Brief (CCQ-Brief) and the Minnesota Cocaine Craving Scale (MCCS)	CBD 300 mg/day × 4 wk vs placebo. No sig. diff. in cannabis use days (incidence rate ratio [IRR] = 0.93, 95% CI 0.67–1.28, p = 0.65), cannabis withdrawal ( β = − 0.06, 95% CI − 0.26 to 0.14, p = 0.55), craving ( β = − 0.09, 95% CI − 0.29 to 0.11, p = 0.39), or anxiety symptoms ( β =0.04, 95% CI − 0.16 to 0.25, p = 0.65). High attrition (35%; 11/31) CBD well tolerated; AEs mild, no SAEs
	Mongeau-Pérusse et al, 2022 [43]	Cannabidiol effect on anxiety symptoms and stress response in individuals with cocaine use disorder: exploratory results from a randomized controlled trial	Canada	Parallel	CBD (800 mg) or 0 mg (placebo)	78	12 wk	Anxiety symptoms and stress response assessed using the VAS, the Beck Anxiety Inventory (BAI), and cortisol levels	CBD 800 mg/day × 92d vs placebo. No sig. diff. in anxiety (BAI: B = 1.70, p = 0.27; VAS: B = 0.55, p = 0.18). No ↓ anxiety after stress cue (p = 0.14) or cocaine cue (p = 0.89). No sig. diff. in cortisol (B = 3.42, p = 0.76). Subgroup (baseline BAI ≥ 16): no effect (VAS p = 0.08; BAI p = 0.48; cortisol p = 0.89). Both groups ↓ anxiety over time, but equally. Well tolerated; no SAEs; attrition 36%
	Gallassi et al, 2024 [44]	Cannabidiol compared to pharmacological treatment as usual for crack use disorder: a feasibility, preliminary efficacy, parallel, double-blind, randomized clinical trial	Brazil	Parallel	CBD (600 mg) with three control drugs (fluoxetine, valproic acid, and clonazepam)	73	10 wk	Safety and tolerability, frequency of crack use, AEs, physical health symptoms, and craving (urge and recall-induced craving)	CBD (600 mg/d, 10 wk) vs pharmacological treatment-as-usual (fluoxetine, valproic acid, clonazepam). 73 participants randomised; high attrition (only 25 completed). CBD group reported significantly fewer AEs than control (e.g., dizziness [p = 0.001], memory impairment [p = 0.043])]. Control group showed greater improvement in clinical/psychiatric complaints (p = 0.008). Intra-group analysis: CBD group ↓crack use from baseline (p = 0.016, T0–T1) and showed some advantages across parameters, but inter-group differences in efficacy outcomes not significant. Both regimens safe/tolerable; no SAEs
OCD	Kayser, Haney, Raskin, Arout and Simpson, 2020 [45]	Acute effects of cannabinoids on symptoms of obsessive‐compulsive disorder: a human laboratory study	USA	Crossover	Cannabis flower (high THC 7.0% THC/0.18% CBD; or high CBD 0.4% THC/10.4% CBD; or placebo 0%THC/0% CBD)	12	3 d (at least 5 d apart)	Self‐reported OCD symptoms using YBOCCS and the OCD‐VAS, self-reported anxiety symptoms using e Spielberger State‐Trait Anxiety Inventory, state subscale (STAI‐S)	Cannabis flower (7.0% THC/0.18% CBD; 0.4% THC/10.4% CBD; placebo, single-dose, crossover RCT): No sig. effect on OCD symptoms (Y-BOCS total: THC Δ− 0.9, CBD Δ− 0.5, placebo Δ− 0.9; F2,35 = 0.32, p = 0.73). No diff. in craving (F2,35 = 0.13, p = 0.88) or anxiety (F2,35 = 0.19, p = 0.83). THC and CBD ↑ acute intoxication (VAS high: THC mean = 47.6, CBD mean = 17.7, placebo = 4.6; F2,35 = 15.4, p < 0.001). AEs mild; no SAEs; high-THC associated with greater transient anxiety/sedation
Opioid use disorder	Hurd et al, 2019 [46]	Cannabidiol for the reduction of cue-induced craving and anxiety in drug-abstinent individuals with heroin use disorder: a double-blind randomized placebo-controlled trial	USA	Parallel	CBD (400 mg or 800 mg) or 0 mg (placebo)	42	2 wk total (3 consecutive d with 1 wk follow-up)	Cue-induced craving measured using the Heroin Craving Questionnaire, anxiety measured using the visual analogue scale for anxiety (VAS-A) and The Clinical Opiate Withdrawal Scale to identify any signs of opioid withdrawal, Measures of opioid craving (assessed using the VAS for craving [VAS-C])	CBD 400 mg or 800 mg/day ×3 d (post-detox) ↓ cue-induced craving (F2,78 = 5.74, p = 0.0047) and anxiety (F2,78 = 5.15, p = 0.0079). Effects persisted 7 d post-dose. ↓ HR and cortisol during cue session. Well tolerated, no SAEs. Correction (2020): fatigue/tiredness misattributed to placebo
	Suzuki et al, 2023 [47]	Impact of cannabidiol on reward- and stress-related neurocognitive processes among individuals with opioid use disorder: a pilot, double-blind, placebo-controlled, randomized cross-over trial	USA	Crossover	CBD (600 mg) or 0 mg placebo (as adjunct to buprenorphine or methadone)	10	2 d (1 wk apart)	Cue-induced craving, attentional bias (visual probe task), subjective stress-reactivity measured using the negative effect subscale of the PANAS, physiologic stress-reactivity measured with salivary cortisol	CBD 600 mg single dose vs placebo (n = 10, cross-over). ↓ cue-induced craving (0.2 vs 1.3, p = 0.040) and attentional bias (− 80.4 vs 100.3 ms, p = 0.041). No effects on delay discounting, stress-reactivity, or withdrawal. No AEs reported
PTSD	Bonn-Miller et al, 2021 [48]	The short-term impact of 3 smoked cannabis preparations versus placebo on PTSD symptoms: a randomized cross-over clinical trial	USA	Crossover	Cannabis flower (high THC = 12% THC and < 0.05% CBD; high CBD = 11% CBD and 0.50% THC; THC+CBD = 7.9% THC and 8.1% CBD, and placebo = < 0.03% THC and < 0.01% CBD)	74	3 wk (stage 1), 2 wk washout, 3 wk (stage 2)	Change in PTSD symptom severity from baseline to end of treatment in Stage 1 assessed using the Clinician-Administered PTSD Scale (CAPS-5)	No significant between-group differences in CAPS-5 PTSD severity scores after 3 wk of treatment [F (3,73) = 1.85, p = 0.15]. All groups, including placebo, showed significant within-group reductions (placebo mean Δ = − 13.1, SD = 12.1). Active cannabis was well tolerated; most common AEs were cough, throat irritation, and anxiety No active preparation outperformed placebo
	Bolsoni, Crippa, Hallak, Guimaraes and Zuardi, 2022 [49]	Effects of cannabidiol on symptoms induced by the recall of traumatic events in patients with posttraumatic stress disorder	Brazil	Crossover	CBD (300 mg) or 0 mg (placebo) single dose	33	3 d (1 wk apart)	Mood and anxiety were recorded (VAMS and STAI-state), along with physiological correlates of anxiety blood pressure (BP), heart rate (HR), and salivary cortisol (SC)	During trauma-recall, CBD ↓ anxiety (VAMS anxiety factor: − 7.3 vs + 1.2; p = 0.04), ↑ calmness (VAMS tranquility: +10.8 vs − 0.7; p = 0.03), and ↓ STAI-state (− 7.7 vs + 1.6; p 0.02) vs placebo. No effects on HR, BP, or cortisol. Well tolerated; no AEs reported
	Walsh et al, 2023 [50]	A small clinical trial of vaporized cannabis for PTSD: suggestive results and directions for future study	Canada	Crossover	Cannabis flower (mixed 10 ± 2% THC and 10 ± 2% CBD; or high THC 10 ± 2% THC and < 1% CBD; or placebo < 1% THC and < 1% CBD)	6	3 wk (stage 1), 2 wk of washout and 3 wk of ad lib cannabis use, 3 wk (stage 2)	Change in scores on the Clinician Administered PTSD Scale (CAPS-5)	Trend-level within-subject reductions in PTSD symptoms: CAPS‑5 pre 39.0 ± 5.9 vs post 30.7 ± 11.2; t (5) =1.95, p (two-tailed) = 0.11, one-tailed p = 0.06; Cohen’s d = 0.80. PCL‑5 also ↓ (d = 1.02, one-tailed p = 0.03). Small sample; placebo comparators not analysable; suggests moderate within-subject effects
Schizophrenia	D’Souza, 2005 [51]	Delta-9-tetrahydrocannabinol effects in schizophrenia: implications for cognition, psychosis, and addiction	USA	Crossover	THC 2.5 mg or 0 mg (placebo)	13	3 d (at least 1 wk apart)	Cognitive (Hopkins Verbal Learning Test), behavioural PANSS and CADSS, motor (parkinsonism (Simpson Angus), akathisia (Barnes), dyskinesia (AIMS) and neurochemical (Blood sampling: − 9-THC, cortisol, prolactin)	THC 2.5 mg IV (crossover, n = 13, schizophrenia): ↑psychotic symptoms (PANSS + 7 vs placebo, d ≈ 0.8, p < 0.05); ↑dissociative symptoms (CADSS +10 vs placebo, d ≈ 0.9, p < 0.05); ↓verbal learning/memory (HVLT − 3 words, p < 0.05); ↑cortisol and prolactin (p < 0.05); no effect on extrapyramidal/motor symptoms (Simpson Angus, Barnes, AIMS); acute effects resolved within hours; no SAEs
	Leweke et al, 2012 [52]	Cannabidiol and amisulpride improve cognition in acute schizophrenia in an explorative, double-blind, active-controlled, randomized clinical trial	Germany	Parallel	CBD (up to 800 mg) or active control amisulpride (up to 800 mg)	39	4 wk	Positive and Negative Syndrome Scale (PANSS) and Brief Psychiatric Rating Scale (BPRS) for assessment of psychotic symptoms, CGI severity; SAS; EPS	Both CBD and amisulpride improved psychotic symptoms (PANSS, BPRS) with no sig. differences between groups. Cognitive outcomes: CBD and amisulpride both improved visual memory (d ≈ 0.5–0.6) and processing speed (d ≈ 0.4–0.6); sustained attention and visuomotor coordination improved only with CBD, while working memory improved only with amisulpride. No correlation observed with symptom change or anandamide levels. CBD had fewer side effects (less weight gain, prolactin increase, and EPS)
	McGuire et al, 2018 [53]	CBD as an adjunctive therapy in schizophrenia: a multicentre randomized controlled trial	Poland	Parallel	CBD 1000 mg or 0 mg (placebo)	88	8 wk	PANSS, BACS, GAF, and the improvement and severity scales of the Clinical Global Impressions Scale (CGI-I and CGI-S)	Greater reduction in positive psychotic symptoms (PANSS pos Δ= − 3.2 vs − 1.7, p < 0.05; treatment diff − 1.5, 95% CI − 2.5 to − 0.2); higher “improved” ratings (CGI-I improvement +23%; 95% CI +8 to 38%); CGI-S less severe (treatment diff − 0.3, 95% CI − 0.5 to 0.0); trend-level cognitive gains (BACS diff +1.31, 95% CI − 0.10 to 2.72) and functioning (GAF diff +3.0, 95% CI − 0.4 to 6.4); well tolerated, similar AEs to placebo
	Boggs et al, 2018 [54]	The effects of CBD on cognition and symptoms in outpatients with chronic schizophrenia a randomized placebo-controlled trial	USA	Parallel	CBD (600 mg/day) or 0 mg placebo	36	6 wk	Cognition assessed using MATRICS Consensus Cognitive Battery (MCCB), Psychotic symptoms assessed using the PANSS	No significant differences in cognition (MCCB composite: d ≈ 0.02, p = 0.89; no domain-specific effects). No effect on PANSS total (β = − 0.03, p = 0.92) or PANSS positive ( β = − 0.06, p = 0.79), negative ( β = − 0.01, p = 0.97), or general psychopathology subscales. Trend toward reduced motor speed in CBD group (p = 0.07). Safety: CBD well tolerated; AEs mild, no SAEs. Attrition: 14% (5/36)
	Leweke et al, 2021 [55]	Cannabidiol enhances anandamide signalling and alleviates psychotic symptoms of schizophrenia	Germany	Parallel	CBD (up to 800 mg) or amisulpride (up to 800 mg)	39	4 wk	PANSS and Brief Psychiatric Rating Scale (BPRS) for assessment of psychotic symptoms, CGI severity; SAS; EPS	CBD and amisulpride both improved visual memory (Cohen’s d ≈ 0.5–0.6) and processing speed (Cohen’s d ≈ 0.4–0.6). CBD uniquely improved sustained attention and visuomotor coordination, while working memory improved only with amisulpride. No relationship between cognitive gains and changes in symptoms or anandamide. Safety: CBD was well tolerated with fewer extrapyramidal symptoms compared to amisulpride
Tobacco use disorder	Morgan, Das, Joye, Curran and Kamboj, 2013 [56]	Cannabidiol reduces cigarette consumption in tobacco smokers: preliminary findings	UK	Parallel	CBD dissolved in ethanal via MDI (400 μg/inhalation) as often as necessary or desired vs pure ethanol inhaler (placebo)	24	2 d (1 wk apart)	No. cigarettes consumed	Participants in the CBD group smoked significantly fewer cigarettes over the treatment week (− − 40% reduction; mean decrease ≈ − 6 cigarettes/day from baseline) compared to no change in placebo (p < 0.05, d ≈ 0.55). No effect on craving (Questionnaire of Smoking Urges—QSU) or withdrawal (Minnesota Nicotine Withdrawal Scale [MNWS]). CBD was well tolerated; no AEs reported
Tourette syndrome	Müller-Vahl et al, 2023 [57]	CANNA-TICS: Efficacy and safety of oral treatment with Nabiximols in adults with chronic tic disorders—results of a prospective, multicentre, randomized, double-blind, placebo controlled, phase IIIb superiority study	Germany	Parallel	Nabiximols (up to 32⋅4 mg THC/30 mg CBD) or 0 mg (placebo)	97	13 wk	Tic reduction according to the Total Tic Score of the Yale Global Tic Severity Scale	More participants met the responder criterion (≥ 25% tic reduction on YGTSS‑TTS): 21.9% (14/64) with nabiximols vs 9.1% (3/33) with placebo, but the superiority criterion was not statistically met. Secondary analyses suggested trends toward tic improvement, reduced depression, and enhanced quality of life. Exploratory subgroup analyses indicated greater benefit in males, those with more severe tics, and those with comorbid ADHD. Nabiximols was well tolerated; no relevant safety issues

ADHD attention-deficit hyperactivity disorder, ATEC Autism Treatment Evaluation Checklist, BACS Brief Assessment of Cognition in Schizophrenia, BOCS Brief Obsessive Compulsive Scale, BSS Bodily Symptoms Scale, CADSS Clinician-Administered Dissociative Symptoms Scale, CBD cannabidiol, CBT Cognitive Behavioural Therapy, CGI-I Clinical Global Impression-Improvement, CWS Cannabis Withdrawal Scale, EPS Extrapyramidal Symptoms GAF Global Assessment of Functioning scale, MADRS Montgomery-Åsberg Depression Rating Scale, MCQ Marijuana Craving Questionnaire, MDI metered dose inhaler, MET Motivational Enhancement Therapy, MWC Marijuana Withdrawal Checklist, NSPS-N Negative Self-Statement scale, OCD obsessive-compulsive disorder, OCD‐VAS Obsessive‐Compulsive Visual Analogue Scale, PANSS Positive and Negative Syndrome Scale, PTSD post-traumatic stress disorder, rCBF regional cerebral blood flow, RCT randomised controlled trial, SAEs serious adverse events, SAS Simpson-Angus Scale, THC tetrahydrocannabinol, VAS visual analogue score, Y-BOCS Yale-Brown Obsessive-Compulsive Scale

To provide readers with a clear overview of recurrent methodological issues, we collated the most common limitations into a descriptive summary (Table 3). The quality of the studies, as assessed using the CASP checklist, was highly heterogeneous, with variability both across domains (e.g., randomisation, blinding, outcome reporting) and between conditions studied. This overview is followed by the detailed study-by-study CASP checklist results presented in Table 4.

Table 3.

Common methodological limitations across included RCTs (descriptive CASP synthesis)

Domain	Number of RCTs (n = 28)	Proportion (%)	Examples/issues
Short duration (≤ 6 weeks)	17	60	Limited follow-up; uncertain durability of effects
Small sample size (≤ 40 participants)	16	57	Underpowered to detect clinically meaningful effects
Incomplete/missing outcome data	11	39	Attrition bias; incomplete reporting of outcomes
Placebo, absent or not described	7	25	Weak internal validity; unclear blinding
Limited generalisability	8	29	Inpatient/subclinical populations; exclusion of comorbidities
No economic evaluation	28	100	No cost-effectiveness data reported

CASP Critical Appraisal Skills Programme, RCTs randomised controlled trials

Table 4.

Critical appraisal of included studies by indication

Indication	Study	Strengths	Limitations
ADHD	Cooper, Williams, Seegobin, Tye, Kuntsi and Asherson, 2017 [30]	Power calculation reported Placebo described Intention to treat analysis Precision of the estimate of the intervention effect reported (CI) Sound method with a focused research question, randomisation, and blinding	Short study duration (6 weeks) Small sample size (N = 30) limits generalisability
Anxiety	Bergamaschi et al, 2011 [31]	Combination of subjective (i.e., VAMS) and objective outcome (heart rate and blood pressure) measures for anxiety Placebo described Precision of the estimate of the intervention effect reported (CI)	Short 1 day study which only examines the acute effects of a single dose of CBD. Population was treatment naïve individuals with social anxiety disorder, so limited generalisability Power calculation not reported
	Crippa et al, 2011 [32]	Placebo described Within-subject between-condition comparisons performed Combination of subjective (i.e., VAMS) and objective outcome (regional cerebral blood flow) measures for anxiety Precision of the estimate of the intervention effect reported (SEM)	Power calculation not reported Randomisation treatment order not specified Short trial length (2 days, 1 week apart), small sample size (n = 10) Incomplete outcome data
	Kwee et al, 2022 [33]	Power calculation reported Multi-centre four-year trial Validated outcome measures Intention to treat analysis Precision of the estimate of the intervention effect reported (CI)	Only assessed a single dose of CBD Subclinical population limits generalisability
ASD	Aran et al, 2021 [34]	Power calculation reported (sample size [n = 150]) Placebo described Strong double blinded crossover design (12-wk efficacy assessment, 4-wk washout, cross-over for another 12 wk to further assess tolerability) Comprehensive reporting of results	Treatment order effect Missing outcome data (some parts of standardised questionnaires were inapplicable for the lowest functioning participants) Unclear if withdrawals after randomisation were included in analysis Precision of the estimate of the intervention effect not reported
	da Silva et al, 2024 [35]	Power calculation reported Placebo described Randomisation with true random number service Intention to treat analysis Precision of the estimate of the intervention effect reported (SD)	Variability in dose as parents could choose how many drops of cannabis extract up to 70 drops Trial disrupted by covid, (known that routine is crucial for children with ASD) country wide quarantine in Brazil so laboratory tests could not be completed at end of the RCT, resulting in missing data
Depression	Pinto et al, 2024[36]	Power calculation reported Effects reported using validated scales (MADRS) Placebo described Intention to treat analysis Precision of the estimate of the intervention effect reported (CI)	Small sample size (n = 35) Focus on acute improvements in depression but no measurement of other relevant outcomes such as quality of life
Cannabis use disorder	Allsop et al, 2014 [37]	Power calculation reported Placebo described Robust double-blind, randomised design Detailed reasons for dropouts Intention to treat analysis Where randomisation did not distribute participants equally (i.e., Sheehan Disability Scale scores) factors were included as covariates in analysis to minimise bias Precision of the estimate of the intervention effect reported (CI)	Short trial length (6-day regimen of Nabiximols with a 28-d follow-up) Narrow study population (treatment-seeking cannabis dependent individuals with no current other drug dependence) Inpatient setting so generalisability to an outpatient setting is limited
	Trigo et al, 2016 [38]	Within subject design (ABACADA) adapted from a prior published study Placebo described Comprehensive reporting of primary and secondary outcomes Detailed account of dropouts and adherence Precision of the estimate of the intervention effect reported (SEM)	Small sample size (n = 9) of mostly Caucasian males limits generalisability Participants self-titrated their dose of the different treatments, therefore comparison between fixed and self-titrated conditions is difficult Self-reported measures
	Trigo et al, 2018[39]	Power calculation reported Investigated a comprehensive potential treatment approach (Combination of Nabiximols and MET/CBT) Intention to treat analysis Precision of the estimate of the intervention effect reported (SD and SEM)	High attrition rate (only 27 of 40 participants completed the study), which introduces risk of selective outcome reporting Homogenous sample of mostly Caucasian males
	Freeman et al, 2020 [40]	Power calculation reported Placebo described Intention to treat analysis Precision of the estimate of the intervention effect reported (CI) Adaptive Bayesian design allowed identification of efficacious dose (efficient in terms of resources and participant burden) Rigorous blinding of randomisation (R command blockrand) Unmasking of investigators only occurred after database had been locked by statistician	Short trial length (4 wk) Trial not necessarily designed to estimate efficacy
Cocaine use disorder	Mongeau-Pérusse et al, 2021 [41]	Intention to treat analysis Precision of the estimate of the intervention effect reported (CI)	High attrition rate (36%), introducing possible attrition bias Subjective outcome measures and participants knew when their urine test would be so could potentially plan their drug use around these
	de Meneses-Gaya et al, 2021 [42]	Robust outcome measures Placebo described Intention to treat analysis Reporting adherence to CONSORT guidelines for randomised trials Precision of the estimate of the intervention effect reported (SD)	Power calculation not reported Small sample size (n = 31) and short trial length (10 d) Inpatient setting so generalisability to an outpatient setting is limited
	Mongeau-Pérusse et al, 2022 [43]	Intention to treat analysis Precision of the estimate of the intervention effect reported (CI)	High attrition rate (36%), introducing possible attrition bias Subjective outcome measures and participants knew when their urine test would be so could potentially plan their drug use around these
	Gallassi et al, 2024 [44]	Randomised, double blind design Placebo described Comprehensive safety and tolerability measures	Precision of the estimate of the intervention effect not reported Power calculation not reported Participants responsible for compliance and dosing High attrition rate, only 34 out of 73 participants completed at least half the study (likely due to COVID pandemic)
OCD	Kayser, Haney, Raskin, Arout and Simpson, 2020 [45]	Within subject crossover design and treatment order randomisation Extensive reporting of baseline characteristics Precision of the estimate of the intervention effect reported (SE and SD)	Small sample size (n = 12) and short trial length (3 d, at least 5 d apart) Incomplete outcome data Smoking route of administration so generalisability is limited
Opioid use disorder	Hurd et al, 2019 [46]	Effective blinding and randomisation Placebo described Inclusion of a diverse demographic Precision of the estimate of the intervention effect reported (SD)	Power calculation not reported No assessment of abstinence rates Short trial length (3 consecutive d and 1 wk follow-up) Selective outcome reporting (only 86% of the randomised sample included in analysis, indicating a high risk of bias)
	Suzuki et al, 2023 [47]	Double blind, placebo controlled, crossover design Precision of the estimate of the intervention effect reported (SD)	Placebo not described Power calculation not reported Small sample size (n = 10) and short trial length (2 d, 1 wk apart) Investigated single dose of CBD so cannot provide information on long-term effects, nor dose response Whilst participants were asked to abstain from cannabis use throughout the RCT, five tested positive, potentially introducing confounders
PTSD	Bonn-Miller et al, 2021[48]	Power calculation reported Placebo described Examines multiple concentrations of THC and CBD Extensive safety monitoring Precision of the estimate of the intervention effect reported (CI)	Smoking route of administration so generalisability is limited Short trial length (3 wk [stage 1]), 2 wk washout, 3 wk [stage 2]) Prescence of several confounding factors including 43% of the sample not adhering to the two-week abstinence from cannabis Stage two did not include a placebo arm so crossover analysis possible was limited Unclear if withdrawals after randomisation were included in analysis
	Bolsoni, Crippa, Hallak, Guimaraes and Zuardi, 2022 [49]	Power calculation reported Placebo described Comprehensive psychological and psychological outcome measures Precision of the estimate of the intervention effect reported (SD)	Short trial length (3 d, 1 wk apart) Treatment groups were not matched for comorbidities resulting in CBD group with significantly more participants with comorbidities
	Walsh et al, 2023 [50]	Withdrawals accounted for in CONSORT flowchart Revised analytic approach to within-subject analysis	Very low power due to under recruitment (n = 6) and short trial length (two 3-wk stages) Issues with effective blinding and absence of placebo control Precision of the estimate of the intervention effect not reported
Schizophrenia	D’Souza, 2005 [51]	Comprehensive cognitive and psychological assessments	No power calculation Small sample size (n = 13) Placebo not described Short trial length (3 d at least 1 d apart) Randomisation, allocation concealment and blinding processes not described IV route of administration, which limits variability within and between participants plasma THC levels
	Leweke et al, 2012 [52]	Comprehensive outcome measurement including validated scales for psychosis and measurement of anandamide levels Modified intention-to-treat analysis	Short trial length (4 wk) Whilst the study used an active control (amisulpride) there was no placebo arm, so unclear in improvements were related to drug treatment or hospitalisation
	McGuire et al, 2018 [53]	Power calculation reported Multi-centre trial Placebo described Robust randomisation and allocation concealment process Validated outcome measures Intention to treat analysis Precision of the estimate of the intervention effect reported (CI) Appropriate measurement and complete reporting of outcome	Extensive inclusion and exclusion criteria, which may have limited sample diversity and generalisability of study results
	Boggs et al, 2018 [54]	Power calculation reported Use of validated and widely used scales for measurement of cognition (MCCB) and symptom severity (PANSS) Precision of the estimate of the intervention effect reported (SD and SE)	Short trial length (6 wk) Placebo not described Withdrawals not accounted for in analysis Missing outcome data Lower dose than investigated in previous literature for this indication
	Leweke, Rohleder, Gerth, Hellmich, Pukrop and Koethe, 2021 [55]	Comprehensive outcome measurement including validated scales for psychosis and measurement of anandamide levels Modified intention-to-treat analysis	Short trial length (4 wk) Whilst the study used an active control (amisulpride) there was no placebo arm, so unclear in improvements were related to drug treatment or hospitalisation
Tobacco use disorder	Morgan, Das, Joye, Curran and Kamboj, 2013 [56]	Directly measures the impact of CBD on cigarette consumption, with clear and relevant outcomes for smoking cessation	Small sample size (n = 24) Short trial length (2 d, 1 wk apart) Randomisation process not described Self-reported cigarette consumption without biochemical verification may have resulted in reporting bias Precision of the estimate of the intervention effect not reported
Tourette syndrome	Müller-Vahl et al, 2023 [57]	Power calculation reported, sample size (n = 97) Multicentre trial Robust randomisation and allocation concealment process Placebo described Validated outcome measures Intention to treat analysis Precision of the estimate of the intervention effect reported (CI) Appropriate measurement and complete reporting of outcome	Specific inclusion criteria may limit generalisability to all patients with tic disorders At least 13% of participants unblinded themselves (either intentionally or accidentally), which may have influenced results High attrition rate (19.4%)

ADHD attention-deficit hyperactivity disorder, ASD Autism Spectrum Disorder, CBD cannabidiol, CI confidence interval, IV intravenous, MADRS Montgomery-Åsberg Depression Rating Scale, MCCB™ MATRICS™ Consensus Cognitive Battery (MCCB™), MET Motivational Enhancement, PANSS Positive and Negative Syndrome Scale, RCT randomised controlled trial, SE standard error, SEM standard error of the mean, THC tetrahydrocannabinol, VAMS Visual Analogue Mood Scale

Figure 2 illustrates the growth in research interest in the use of medicinal cannabis for mental health conditions over time. From 2005 to 2016, only a small number of studies were conducted, which reflects the limited amount of research interest in this area prior to changes in legislation relating to legalisation of medicinal cannabis across the globe. However, from 2017 onwards, there was an increase in the number of studies that peaked in 2020. From 2020, there was consistent output from 2020 to 2024, which indicates a growing recognition of its potential therapeutic potential and a shift toward the need for more rigorous clinical investigations

Fig. 2.

Included studies of use of medicinal cannabis for mental health conditions. MC mast cells, MHC major histocompatibility complex

The RCTs included in this review that investigate the therapeutic potential of medicinal cannabis across mental health conditions provide insights into treatment outcomes, tolerability, and the nuances of efficacy across different conditions. The following subsections present condition-specific findings, summarising the evidence for medicinal cannabis and cannabidiol in treating psychiatric and substance use disorders.

Conditions

Attention Deficit Hyperactivity Disorder

One double-blinded parallel RCT that examined the efficacy of nabiximols on cognitive performance and activity level in 30 adults with ADHD [30]. The 6-week trial found no significant differences between nabiximol treatment and placebo [30].

Anxiety Disorders

One parallel RCT (n = 24, CBD 600 mg) [31], and one crossover RCT (n = 10, CBD 400 mg) [31], investigated CBD for the treatment of social anxiety disorder. Cannabidiol was found to significantly reduce self-reported anxiety on VAMS compared with placebo, in both studies. A more recent 8-week parallel RCT investigated whether CBD; 300 mg would enhance the efficacy of exposure therapy in treatment-resistant patients (n = 80) with anxiety disorders [33]. This study found no difference in exposure therapy treatment outcome between CBD and placebo on Fear Questionnaire scores [33].

Autism Spectrum Disorder (ASD)

The efficacy of CBD-rich cannabis extracts for the treatment of ASD in children was tested in one crossover RCT (n = 15) and one parallel RCT (n = 60) [34, 35]. In the crossover trial, change in behavioural problems did not differ between placebo and treatment groups [34], whereas the parallel trial found CBD was significantly more effective at improving social interaction (p = 0.0002) relative to placebo [35].

Depression

A single 12-week parallel RCT investigated CBD as an adjunctive treatment for depression and found no significant difference in depression rating (Montgomery-Åsberg Depression Rating Scale) between participants treated with CBD and those treated with placebo [36].

Cannabis Use Disorder

One placebo controlled parallel RCT (n = 51) [37] and two crossover RCTs (n = 9, n = 40) [38, 39] investigated the efficacy of nabiximols for ameliorating cannabis withdrawal in participants with cannabis use disorder. A significant reduction in withdrawal symptoms was observed in participants receiving nabiximols (up to 86.4 mg THC, 80 mg CBD) in the six-day parallel trial, although there was no significant difference in abstinence rates between groups at the 28-day follow-up [37]. The two crossover RCTs—by the same investigators—used higher doses of nabiximols (up to 108 mg THC/100 mg CBD and up to 113.4 mg THC/105 mg CBD) [38, 39]. In comparison with placebo, nabiximols significantly reduced the severity of cannabis withdrawal symptoms in the first study, but not craving [38]; and significantly reduced cannabis craving in the second study but not abstinence rates at the end of the 12-week trial [39].

Additionally, in a Phase 2a trial using an adaptive Bayesian design, CBD (400 mg and 800 mg) was found to be more efficacious than placebo at reducing cannabis use in cannabis-dependent participants (n = 51) [40]. Cannabidiol 400 mg increased cannabis abstinence by 0.48 days/week (95% CI 0.15 to 0.81) and decreased THC-COOH:creatine ratio by − 94.21 ng/mL (95% CI − 161 to − 35.56) [40]. Cannabidiol 800 mg increased cannabis abstinence by 0.27 days/week (95% CI − 0.09 to 0.64) and decreased THC-COOH:creatine ratio by − 72.02 ng/mL (95% CI − 135.47 to − 19.52)[40].

Cocaine Use Disorder

Three recent parallel RCTs have evaluated the efficacy of CBD for managing craving during detoxification in participants with cocaine use disorder [41, 42, 44]. Two of these studies (300 mg and 600 mg CBD) found significant reductions in craving over 10 days and 10 weeks, respectively, although no significant differences were observed between CBD and placebo groups [42, 44]. The third of these trials reported that CBD 800 mg did not reduce cocaine craving or relapse to cocaine over 12 weeks [38]. Additional analysis of data from the same trial explored the effect of CBD on anxiety symptoms and stress response and found that CBD was no more effective than placebo in reducing anxiety symptoms or cortisol levels in individuals with cocaine use disorder [43].

Obsessive Compulsive Disorder

A double-blind, randomised, placebo-controlled, crossover trial investigated 14 patients with obsessive-compulsive disorder (OCD) who smoked cannabis. Participants received cannabis high in THC (7.0% THC, 0.18% CBD), cannabis high in CBD (10.4% CBD, 0.4% THC), or placebo cannabis (< 0.01% THC/CBD). Neither the THC dominant nor CBD dominant cannabis significantly improved OCD symptoms (measured with the Yale-Brown Obsessive-Compulsive Scale [Y-BOCS]) compared with placebo. However, anxiety (State Anxiety subscale of the State-Trait Anxiety Inventory [STAI-S]) scores were significantly lower after placebo than after both THC (p = 0.0002) and CBD (p = 0.039) [45].

Opioid Use Disorder

Two short trials have reported on the effects of CBD on cue-induced craving in drug-abstinent adults with opioid use disorder [46, 47]. The first was a three-day parallel trial (n = 42) that found both 400 mg and 800 mg CBD to be efficacious in reducing both craving (VAS-C) and anxiety (VAS-A) as compared with placebo; these effects were still evident at the follow-up session 1-week after drug administration [46]. The second RCT had a crossover design and assessed stress in addition to craving [47]. This two-day pilot study (n = 10) also found CBD (600 mg as an adjunct to buprenorphine or methadone) to be more efficacious (p = 0.040) in decreasing craving than placebo, although no impact on stress was observed [47].

Post Traumatic Stress Disorder

Two crossover RCTs have tested the effect of cannabis flower on the severity of PTSD symptoms, assessed with the Clinician-Administered PTSD Scale (CAPS-5) [48, 50]. In the first trial (n = 74), no active cannabis preparations (12% THC and < 0.05% CBD; 11% CBD and 0.50% THC; 7.9% THC and 8.1% CBD) statistically outperformed the smoked placebo on CAPS-5 scores (< 0.03% THC and < 0.01% CBD) on CAPS-5 scores after 3 weeks of treatment, indicating no sustained therapeutic effect. Acute within-session changes were not reported, and no effect sizes were provided [48]. In the second trial, which investigated two concentrations of vaporised cannabis, some symptomatic improvement was observed during treatment (acute effects), but only five participants completed the study, leaving insufficient statistical power to evaluate sustained outcomes or placebo effects [50]. Additionally, a three-day parallel RCT investigated the effects of CBD (300 mg) on mood and anxiety in 33 individuals with PTSD [49]. In comparison to placebo, CBD was found to significantly improve mood (VAMS scale) but not anxiety (STAI-state) [49].

Schizophrenia

For schizophrenia, four unique RCTs were identified: three parallel-group trials and one crossover trial. One of the parallel-group trials was reported in more than one publication, which may explain discrepancies in previous counts. Four RCTs with participants diagnosed with schizophrenia met the inclusion criteria for this review. This included one small crossover RCT assessing the efficacy of THC and four parallel RCTs assessing the efficacy of CBD. The crossover RCT (n = 13) demonstrated an acute worsening of schizophrenia symptoms during the three-day trial in participants treated with THC, suggesting a short-term detrimental effect without sustained follow-up data.

Of the CBD trials, a 6-week (n = 36) trial found administration of 600 mg CBD did not significantly improve cognition assessed using The MATRICS^TM Consensus Cognitive Battery (MCCB) or psychotic symptoms assessed using Positive and Negative Syndrome Scale (PANSS) [54]. No acute effects were reported.

A higher dose of CBD (1000 mg) as an adjunct to participants’ usual antipsychotic medication was found to significantly improve psychotic symptoms (PANSS positive treatment difference = − 1.4, 95% CI − 2.5, to − 0.2) as compared with placebo in an eight-week RCT (n = 88). This reflects a modest sustained benefit; however, no clinically significant difference (≥ 20% improvement in PANSS total scores) was reported [53].

The remaining trial (n = 39), tested the effects of CBD and the antipsychotic drug amisulpride (both drugs up to 800 mg) on psychotic symptoms (PANSS) [52]. Both groups showed symptomatic improvement over the 4-week period (sustained effects), although the absence of a placebo arm limits attribution of change to treatment [52]. Further analysis of this trial investigated neurocognitive outcomes and found no difference in post-treatment outcomes between groups [55].

Tobacco Use Disorder

A single parallel RCT has investigated inhaled CBD for reducing cigarette consumption in 24 participants with tobacco use disorder. The number of cigarettes smoked was significantly reduced in CBD-treated participants as compared with placebo-treated participants (p = 0.002) in the two days of treatment (1 week apart). There was some maintenance of this effect, albeit no significance difference between groups at the two-week follow-up [56].

Tourette Syndrome

A single placebo controlled parallel RCT (n = 97) investigated the efficacy of nabiximols for reducing tics in adults with chronic tic disorders [57]. The superiority of nabiximols over placebo was not demonstrated in the 13-week trial [57].

Tolerability

Medicinal cannabis was well tolerated in 24 of the 28 studies reviewed, with the remaining four studies not reporting on tolerability. Five studies reported treatment-related severe or serious AEs, although most studies reported no difference in rates of AEs between medicinal cannabis and control groups. The most frequently reported side effects included nausea, diarrhoea, dry mouth and throat and tiredness.

Summary of Evidence

Across conditions, the evidence base for medicinal cannabis in primary mental health disorders remains limited by small sample sizes, short intervention durations, and substantial heterogeneity in study populations, interventions, and outcomes. While CBD was the most frequently investigated compound, studies varied markedly in dose, formulation, and route of administration, complicating cross-trial comparisons. Evidence for THC was sparse and more often associated with symptom exacerbation in psychotic disorders. Trials in anxiety-related conditions, post-traumatic stress disorders, and substance use disorders generally reported modest or no benefit, with occasional signals of efficacy in specific symptom domains. Adverse event profiles were inconsistently reported but suggested tolerability issues at higher THC doses. The overall pattern of findings indicates a lack of consistent, clinically meaningful benefit across conditions, underscoring the need for larger, well-controlled trials with standardised outcome measures and longer follow-up periods.

Discussion

This scoping review identified 28 RCTs investigating the efficacy and tolerability of medicinal cannabis across a range of mental health conditions. Evidence was heterogeneous, with variable study designs, interventions, and outcome measures, and mixed findings across indications. While some trials reported modest benefits, most notably for specific symptom domains in psychotic disorders, results were frequently inconsistent, and effect sizes were often small or clinically uncertain.

Of the 28 studies reviewed, five were both high in quality and reported significant results for at least one primary outcome. These studies covered three main indications: cannabis use disorder, autism spectrum disorder, and schizophrenia. For cannabis use disorder, acute reduction of cannabis use was demonstrated in one study investigating CBD [40] and another study investigating nabiximols [37]. Regarding autism spectrum disorder, one of the two studies reported significant improvement in social interaction in children treated with CBD [35] (da Silva, Medeiros et al. 2024). For schizophrenia, high-dose CBD was found to improve psychotic symptoms in two RCTs; however, the lack of a placebo group in one of these studies limits definitive conclusions [52, 53].

The selection of the 28 papers included in this review was guided by the recognition that RCTs represent the highest level of evidence in clinical research. This standard of evidence, required to inform both clinical decisions and policy, is particularly important as the medicinal cannabis research field is still in its infancy. The methodology inherent to RCTs aims to minimise bias and provide robust evidence, ensuring that findings from this review contribute to a more informed understanding of the efficacy of medicinal cannabis in treating mental health conditions.

Of the high-quality papers, outcome measures were comprehensive and included validated scales, and there were strong randomisation and blinding procedures, which minimised allocation bias. However, with a considerable number of studies being either low quality or unclear, there is a considerable risk of bias. Common issues included short trial length and small sample sizes; the duration of 17 RCTs (60%) was 6  weeks or shorter and 16 RCTs (57%) had 40 or fewer participants. Small sample sizes limit power and the ability to detect statistically significant differences and increases the likelihood of type two errors. Additionally, in seven studies, placebo was not included or not described, 11 studies had missing outcome data, and eight studies had subclinical populations or were conducted in an inpatient setting. All these factors limit the validity of the results and the generalisability to the broader population who may be prescribed medicinal cannabis for their mental health conditions. Most of the studies excluded participants on other psychiatric medication or with multimorbidity, which again limits the external validity of the results. Studies investigating medicinal cannabis for PTSD were particularly low in quality for reasons including short trial length, issues with blinding and treatment adherence, and the lack of a placebo control in two of the studies [48, 50]. Overall, the review revealed significant heterogeneity in study quality and high levels of potential bias, which necessitates caution in the interpretation of the overall findings.

In addition to efficacy outcomes, tolerability and safety profiles were variably reported. Across most RCTs, medicinal cannabis preparations were generally well tolerated, with AEs typically mild to moderate in severity [35, 37, 40, 53, 54]. Commonly reported side effects included fatigue, dizziness, somnolence, dry mouth, and gastrointestinal discomfort. Trials involving THC-containing products more frequently reported psychoactive effects such as euphoria, anxiety, and transient worsening of psychiatric symptoms, particularly in schizophrenia [53, 58]. The CBD-dominant interventions were associated with fewer psychoactive effects but were not entirely free from AEs, with diarrhoea and somnolence among the more common complaints [52, 54]. Withdrawal due to AEs was uncommon but did occur in a small proportion of participants, often in higher-dose or THC-containing arms [37, 53]. Importantly, few studies conducted systematic safety monitoring or long-term follow-up, limiting conclusions about sustained tolerability or the risk of rare but serious AEs. This underscores the need for future RCTs to incorporate rigorous, standardised AE reporting and extended monitoring periods.

Recent meta-analyses provide important context for interpreting these findings. Black et al (2019) conducted the most comprehensive synthesis to date, analysing 83 studies, 40 of which were RCTs of cannabinoids for psychiatric disorders [14]. They reported that while some evidence suggests potential benefits for conditions such as anxiety and psychosis, overall effect sizes were small and the certainty of evidence was consistently low, largely due to high risk of bias, short trial duration, and heterogeneity in interventions and outcomes. Similarly, Bahji et al. [8] reviewed trials of cannabinoids for anxiety disorders and found that although cannabinoids, particularly CBD, were associated with reductions in anxiety symptoms in some studies, the pooled results did not demonstrate consistent or clinically meaningful effects. The authors emphasised that the evidence was insufficient to support routine clinical use and highlighted the urgent need for larger, high-quality trials with long-term follow-up. Kock et al. [59] positioned cannabinoid trials within the broader evidence landscape of psychiatric interventions, concluding that the current evidence base for cannabinoids is far less robust than for established pharmacological and psychological therapies. They argued that without adequately powered and methodologically rigorous RCTs, cannabinoids cannot yet be considered a reliable treatment option in psychiatry. Taken together, these reviews reinforce the findings of our scoping review: despite rapidly growing prescribing rates, the clinical evidence for medicinal cannabis in mental health remains weak, inconsistent, and insufficient to inform evidence-based practice.

Despite the increasing number of RCTs exploring medicinal cannabis for mental health conditions, this review found no RCTs investigating medicinal cannabis for panic disorders and only a single RCT was identified for each of the following indications: depression, ADHD, OCD, tobacco use disorder, and Tourette syndrome. Of the 28 studies reviewed, five were both high in quality and reported significant results for at least one primary outcome. These studies covered three main indications: cannabis use disorder, autism spectrum disorder, and schizophrenia. For cannabis use disorder, acute reduction of cannabis use was demonstrated in one study investigating CBD [40] and another study investigating nabiximols [37]. Regarding autism spectrum disorder, one of the two studies reported significant improvement in social interaction in children treated with CBD [35]. For schizophrenia, high-dose CBD was found to improve psychotic symptoms in two RCTs; however, the lack of a placebo group in one of these studies limits definitive conclusions [52, 53]. Importantly, no RCTs were identified for two of the most prevalent and burdensome mental health conditions—depression and panic disorder. This gap is concerning given the frequency with which these disorders are treated with medicinal cannabis in clinical practice. The absence of trial evidence for these core indications highlights the disconnect between current prescribing patterns and the available evidence base and underscores the need for well-designed RCTs to address these populations.

Taken together, no trials included in this review provide evidence to support the long-term efficacy of medicinal cannabis for any of the mental health conditions investigated. Given this, the limited and heterogeneous evidence base highlights the need for caution when considering use in clinical practice. While medicinal cannabis prescribing has been reported to be increasing in psychiatric populations, there remains minimal supporting RCT evidence—particularly for core indications such as depression and anxiety. This mismatch between available evidence and areas of prescribing emphasis underscores the importance of prioritising high-quality trials in these conditions before broad uptake in routine care.

From a clinical perspective, the current evidence base does not support routine prescribing of medicinal cannabis for any psychiatric indication. Given the variability in study quality, heterogeneity of interventions, and limited long-term safety data, prescribers should approach use in mental health conditions with caution, ensuring decisions are made on a case-by-case basis and in the context of shared decision-making with patients. Policymakers should avoid prematurely expanding access pathways without parallel investment in high-quality research, robust pharmacovigilance systems, and prescriber education. Care is warranted for conditions such as depression and anxiety, where prescribing is increasing despite minimal supporting RCT evidence. Aligning clinical practice and policy with the strength of evidence will be essential to safeguard patient safety and optimise therapeutic outcomes. For clinical practice, prescribers should prioritise established first-line therapies with stronger evidence bases and consider medicinal cannabis only when these options have been exhausted or are poorly tolerated. Where medicinal cannabis is considered, it should be prescribed cautiously, with attention to product type (CBD-dominant vs THC-psychoactive-dominant), close monitoring of psychiatric symptoms and AEs, and clear communication with patients about the current uncertainty of evidence. Shared decision-making is particularly important, ensuring that patients are informed of potential benefits and risks. From a research perspective, priority should be given to adequately powered trials in common conditions such as depression and anxiety, longer-term studies that incorporate systematic safety monitoring, and investigations that include economic evaluation. These steps are needed to generate evidence that can meaningfully inform prescribing frameworks and clinical decision-making.

Strengths and Limitations

The strength of this review is its protocol-driven design and adherence to PRISMA-ScR guidance, which ensured methodological rigour and transparency throughout. By focusing exclusively on RCTs, the review captured the highest-quality evidence available in this emerging field. A further strength is the comprehensiveness of the search strategy, which spanned multiple databases and included citation searching to minimise the risk of missing eligible studies. In addition, the review applied a structured appraisal of study quality using the CASP checklist, which provided a transparent framework for identifying recurrent methodological limitations across trials. This goes beyond the minimum requirements for scoping reviews and enhances the interpretability of the evidence base. Finally, the use of both a summary table of common methodological issues (Table 3) and a detailed study-level appraisal (Table 4) improved the clarity and accessibility of quality assessments for readers. This review has some limitations that should be considered when interpreting its findings. First, the decision to include only RCTs excluded other study designs that may offer valuable preliminary or qualitative insights into the use of medicinal cannabis for mental health conditions. Second, although inclusion criteria required the use of DSM-IV (or equivalent) diagnostic classifications, several trials did not clearly describe how these criteria were applied, and in some cases used less rigorous or outdated diagnostic approaches for specific subpopulations. This variability may have reduced comparability across studies. Third, there was considerable heterogeneity in cannabinoid formulations, dosing regimens, outcome measures, and study populations, limiting the capacity for direct comparison and precluding meta-analysis. Fourth, small sample sizes and short follow-up periods were common, restricting generalisability and limiting conclusions about long-term efficacy and safety. Finally, the review was restricted to English-language publications, which may have introduced language bias. These limitations underscore the need for standardised, adequately powered, and methodologically rigorous research to strengthen the evidence base in this field. A formal statistical assessment of publication bias was not undertaken because the number of RCTs per indication was below the minimum threshold (generally 10) recommended for reliable analysis. As advised by the Cochrane Handbook, funnel plots and associated tests lack power and can be misleading when applied to small study sets. While the possibility of unpublished negative trials cannot be excluded, our broad database and citation searches aimed to minimise this risk.

Conclusion

Overall, whilst there may be some potential therapeutic benefits for medicinal cannabis in certain psychiatric disorders, the literature reviewed here showed highly variable outcomes, with studies showing substantial heterogeneity in quality and high levels of potential bias. In agreement with previous systematic reviews, the findings of the present review reiterate the need for more high-quality, adequately powered RCTs investigating the use of medicinal cannabis for mental health conditions. The studies reviewed do not recommend medicinal cannabis for any psychiatric indication, and it is recommended that clinicians and policymakers integrate new high-quality research as it becomes available. Future studies should focus on proper definition of disorders. Although this review required that included studies use DSM-IV (or equivalent) diagnostic criteria to ensure a consistent baseline for comparison, several included trials did not provide detailed reporting on how these criteria were applied or used less rigorous or outdated diagnostic approaches for specific subpopulations. This variability in the precision and application of diagnostic definitions may have influenced participant selection and study outcomes. Therefore, the recommendation for future research to prioritise proper and standardised definitions of disorders reflects the need to ensure greater consistency and transparency in diagnostic methodology, beyond simply stating alignment with a given classification system. In addition, future trials should use established rating scales, ensure blinding, and include longer-term follow up to provide more conclusive evidence in this field.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

No funding or support was received for the project. This unfunded project was undertaken as part of Ms Sophie Cooling’s Honours year in the Faculty of Medicine, Dentistry and Health Sciences at the Department of General Practice and Primary Care, University of Melbourne.

Declarations

Conflict of interest

All authors declare that they have no conflicts of interest.

Ethics approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Availability of data and materials

Data are available on request.

Code availability

Not applicable.

Author contributions

SC, YB, and CMH designed the review and drafted the initial manuscript. SC, YB and CMH contributed to the abstract, screening, and data extraction for the included studies. SC, YB and CMH provided critical review of the included studies and interpretation of the result. All authors SC, YB, DC and CMH contributed to manuscript development. SC and CMH prepared the manuscript for submission. CMH undertook the major revision of the manuscript, with input from all authors. All authors approved manuscript.