A phase III multisite randomised controlled trial to compare the efficacy of cannabidiol to placebo in the treatment of cannabis use disorder: the CBD-CUD study protocol | BMC Psychiatry

Research hypothesis and study aims

The research hypothesis is that CBD, compared to placebo, will achieve statistically and clinically significant reductions in cannabis use, as measured by the number of self-reported cannabis-free days and urinary THC-COOH levels, among treatment-seeking patients with moderate-severe CUD.

The primary aim of the CBD-CUD study is to examine the efficacy of CBD, compared to placebo, in reducing cannabis use (as measured by self-report and quantitative measures of cannabis metabolites (THC-COOH) in urine drug screens) during treatment (Weeks 1–12) in participants seeking treatment for moderate-severe CUD, when used in combination with psychological interventions.

Secondary aims include examination of (i) safety, (ii) other cannabis related measures (e.g., cannabis withdrawal and cravings, cannabis-related problems); (iii) tobacco and other substance use; (iv) health and quality of life (QoL) measures, (v) patient experience measures; (vi) treatment retention rates; (vii) cognitive performance; (viii) post-treatment (Week 24) cannabis use, health outcomes and QoL measures.

Study design

The study is a parallel group prospective double-blind Phase 3 randomised controlled trial comparing a 12-week treatment period of oral CBD (400 mg daily) (Experimental) to placebo (Control), with both groups receiving 4 sessions of manualised CBT-based counselling. Research interviews will be conducted at baseline (week 1), 3-weekly during the study intervention (weeks 4, 7, 10 and 13) and 12-weeks after the end of treatment (week 25) (Fig. 1). The study will use a modified intention-to-treat analysis. The expected number of participants is 250, of which we estimate approximately 20% (n = 50) will be of Indigenous background. The study will also include qualitative interviews with Indigenous Australian participants in both control and intervention groups (a total of n = 15–25 Indigenous Australian participants) to examine their experiences in the study.

Fig. 1
figure 1

Ethical statement

The study will be conducted in accordance with the National Statement on Ethical Conduct in Human Research (2007), the CPMP/ICH Note for Guidance on Good Clinical Practice and consistent with the principles that have their origin in the Declaration of Helsinki. The study was approved by the Sydney Local Health District Human Research Ethics Committee (no 2022/ETH02467) and the Aboriginal Health and Medical Research Council’s Human Research Ethics Committee (no 2110/23). The project has an Aboriginal Reference Group that oversees all aspects of the study, including data collection and analysis as they relate to Indigenous Australians. The study has been registered on the Australian and New Zealand Clinical Trial Registry (ACTRN12623000526673).

Setting and study sites

The multicenter trial will be coordinated from the Specialty of Addiction Medicine, Faculty of Medicine and Health, University of Sydney (study sponsor). Treatment will be provided at seven specialist addiction outpatient treatment centers: four in Sydney, one in Newcastle and two in Melbourne, Australia.

Participants and recruitment

Eligibility criteria

The target study population is treatment-seeking adults with moderate to severe CUD under conditions of informed consent. Eligibility will be assessed by an Addiction Medicine or Psychiatry credentialed Study Medical Officer (SMO).

Inclusion criteria:

  1. 1.

    Aged 18 to 65 years.

  2. 2.

    Meeting DSM-5 criteria for moderate or severe CUD (≥ 4/11 criteria) [2], with recent frequent cannabis use (≥ 4 days per week in the preceding 4 weeks).

  3. 3.

    Willing and able to provide informed consent to study procedures.

  4. 4.

    Proficient in English at a conversational level sufficient to participate in a counselling intervention.

Exclusion criteria aim to exclude individuals with conditions that jeopardise safety or confound data interpretation:

  1. 1

    Prescribed medicinal cannabis products (e.g., CBD, THC) for any indication in the previous 4 weeks.

  2. 2

    Another active (past year) moderate-severe substance use disorder other than tobacco; determined on clinical assessment using DSM-5 criteria.

  3. 3

    Active or severe medical (e.g., pain, epilepsy, cardiovascular disease) or psychiatric (e.g., psychosis, severe affective disorder) conditions based on clinical assessment.

  4. 4

    Moderate to severe hepatic disease (transaminase elevations > 3 times, bilirubin > 2 times upper normal limits at screening).

  5. 5

    Pregnant or lactating women (based on urine β-hCG at screening).

  6. 6

    Hypersensitivity to CBD or any excipients of Investigational Product.

  7. 7

    Using medications with known drug-drug hepatic CYP-450 interactions with CBD: 3A4, (e.g., carbamazepine, fluvoxamine, methadone), 2C19 (e.g., rifampin); CYP2B6 (e.g., bupropion), CYP2C9 (e.g., warfarin).

  8. 8

    Not available during treatment or follow-up (e.g., travel, impending residential detoxification or residential rehabilitation admission, impending imprisonment).

  9. 9

    Court-mandated treatment requiring abstinence from drugs.

  10. 10

    Current active (counselling and/or medication-based) treatment for CUD.

  11. 11

    Received an investigational medicinal product within the last 4 weeks (or 5 half-lives if using long-acting investigational drugs).

Participant numbers

Sample size calculations are based on the analysis of the primary outcome, that is, the difference between placebo and CBD groups in total number of Cannabis-free Days over the 12-week intervention period. Ferguson has suggested that the minimum effect size (Cohen’s d) for an effect of practical clinical significance is 0.4167 [55]. To achieve 90% power (two-tailed) and α = 0.05, a total of N = 250 (n = 125 per group) participants are needed to detect a between-group effect size of d = 0.41. Of the target 250 sample, it is estimated approximately 20% (n = 50) of the study sample (N = 250) will be Indigenous Australians.

Participants who discontinue study procedures after commencing study interventions (medication dispensed on Day 1) will not be replaced in the study but will be included in the modified intention-to-treat analyses. Participants enrolled and randomised, but do not commence any treatment (no medication dispensed or other clinical interventions) will not be included in the final analysis.

Recruitment, screening and assessment

Participants will be recruited from people seeking treatment at participating study sites, and/or people interested in the study in response to study advertisements at local health services, social media, and clinical trial recruitment platforms. On initial contact with the service, potential participants will be informed of the study and if interested, referred to a site coordinator to complete telephone screening. Following telephone screening, potentially eligible participants will be scheduled a face-to face assessment with a Study Medical Officer (SMO) to confirm eligibility. Potential participants will sign a medical screen consent form prior to the SMO completing a structured history, clinical examination, and any laboratory investigations with the participants. The SMO will also explain the study requirements to the potential participant and explain the study medication and any potential side effects. Eligible participants are scheduled an appointment (Week 1, Day 1) to attend for enrolment into the study. For those participants who are not eligible or choose not to participate in the study, alternative treatment options will be organised in collaboration with the patient, as clinically appropriate.

Informed consent, randomisation and blinding

Written informed consent is obtained on Week 1, Day 1 of the study prior to the commencement of all subsequent study procedures. Consent is obtained with the site coordinator independent of treating clinicians, to minimise ‘pressure’ to participate in the study.

The randomisation schedule has been developed by an independent statistician, with eligible participants randomised in a 1:1 ratio between groups using variable block randomisation to help maintain blinding, with subjects stratified by (a) site (to achieve approximately equal numbers of active and placebo at each site) and (b) Indigenous Australian status (to allow direct between-group statistical comparisons within Indigenous Australian participants).

Participants, clinicians, and researchers involved in service delivery, data collection and analysis will remain blinded to study conditions using matched placebos manufactured by the same manufacturer. CBD and placebo will be packaged in identically labelled containers with the participant’s ID number and site. Aside from site trial pharmacists (who have no direct contact with participants), all other members of the clinical or research teams will be blinded to group allocation.

Unblinding will occur after all data are collected, entered, cleaned and the trial database has been locked. In circumstances where allocation needs to be unblinded (e.g. severe adverse event), the principal investigator will authorise the local site investigator to break the blind (via the site trial pharmacist).



The experimental condition will receive 12 weeks of CBD oral 400 mg daily, administered as 200 mg liquid administered twice a day (BD). The CBD used in the trial is a plant-extracted pharmaceutical product (registered in Australia as Epidyolex® for the treatment of paediatric epilepsy), and is an oral liquid (clear, colourless to yellow solution) containing 100 mg per ml, dispensed in 105 ml bottles. The placebo is identical in composition and appearance (with the exception of the CBD). Both CBD and placebo are manufactured and supplied by Jazz Pharmaceuticals.

The dose is selected based on the findings of the Phase IIa RCT [29] that identified a daily dose of 400 mg CBD being more efficacious than placebo at reducing cannabis use during 4-week treatment and follow up.

Nicotine dependent participants will be offered smoking cessation counselling during the trial, with prescriptions and supply of nicotine replacement therapy (NRT) either in the form of 16-h topical patches (7, 14 or 21 mg) and/or nicotine chewing gum or lozenges provided.


All participants will receive four structured 40–50-min counselling sessions over the 12-week medication phase, based on cognitive behavioural therapy (CBT) and motivation enhancement for relapse prevention, consistent with identified ‘best practice’ for cannabis cessation interventions [56]. Available evidence suggests 4-sessions of CBT deliver comparable outcomes to 6 or more sessions for treating CUD [57]. Counselling will be delivered by psychologists experienced in CUD treatment and trained to deliver manualised counselling interventions. Study Counsellors will keep a log of attendance at counselling sessions.

Clinical reviews

Participants will have 3-weekly medical reviews with the SMO over the 12-week intervention (Weeks 1, 4, 7, 10 and Week 13). At each appointment, the SMO will review cannabis and other substance use since the last appointment, other health and social issues, and client goals, complete Concomitant Medications and Adverse Events assessments, collect UDS, and supply medications dispensed by the trial pharmacist.

Outcome measures

The primary outcomes are illicit cannabis use during the 12-week intervention period, operationalised using two endpoints:

  1. 1)

    Cannabis-free Days over the 12-week intervention period, producing a continuous measure between days 1 and 85. Details regarding number of days of cannabis use will be collected at each research interview (baseline week 1, weeks 4, 7, 10, week 13 and 25) using the Time Line Follow Back (TLFB) approach, a reliable and validated measure of cannabis use, particularly when combined with biological assays (e.g. UDS) and confidentially reported to independent researchers74.

  2. 2)

    Urinary quantitative analysis of THC-COOH (creatinine adjusted). Urine samples will be collected at weeks 1, 4, 7, 10, 13 and 25, and analysed using liquid chromatography-tandem mass spectrometry (LCMS). As THC-COOH can remain ‘positive’ using qualitative thresholds (e.g. 20 ng/ml in LCMS assays) for more than 30 days after abstinence in chronic heavy cannabis users [58], we will analyse quantitative levels of THC-COOH (creatinine adjusted) to detect differences in cannabis use between the two study groups, replicating the approach used in the pilot RCT [29] (see below).

Secondary outcomes (Table 1) include a range of measures that relate to cannabis use (including rates of abstinence or reduced frequency of cannabis use, cannabis withdrawal and cravings, cannabis related problems, severity of CUD), safety (adverse events), health outcomes (including mental health, physical health, QoL), consumer experience of the medication, cognitive performance, other substance use and post-treatment outcomes (12 weeks after the intervention). The relationship between experiences of racial discrimination (using the modified Everyday Discrimination Scale) and outcomes for Indigenous Australians will also be explored.

Table 1 Table of all primary and secondary outcome measures included in study

Research interviews

The schedule of trial procedures and assessments for participants, including the timing of research and clinical interventions is shown in Table 2. Participants are scheduled to have interviews with researchers at 3-weekly intervals during the 12-week intervention (Weeks 1, 4, 7, 10 and 13), and again 12 weeks after the intervention (Week 25). These interviews will be face-to-face with a researcher, although they can be undertaken by telehealth if required. All data collected at researcher interviews will be entered directly into an electronic database, REDCap, and kept confidential from treating clinicians. Participants will be reimbursed with shopping vouchers for time, inconvenience, and expenses of attending research interviews [59].

Table 2 Schedule of trial procedures and assessments

Qualitative interviews with indigenous participants

To gain insights into the experience of Indigenous Australian participants, semi-structured in-depth interviews will be conducted by Aboriginal researchers (part of the study team) at around week seven with Indigenous participants in both control and interventions groups until data saturation occurs—estimated at 15 to 25 participants. These interviews will examine topics such as (a) how participants perceived their cannabis use and identified their treatment goals, and how participants are supported by their family and community; (b) how participants engage with the study treatment procedures (medication and counselling) providing insights into future implementation. The interviews will take approximately 40 to 60 min and be digitally recorded and transcribed. A yarning methodology will be used for the data collection and analysis [60].

Data management and monitoring

Confidentiality of participant data will be secured by removing all identifiable data and replacing it with a unique identifier. The principal investigator and coordinating researcher will have access to key files that link the unique identifier to identifiable data if unblinding is necessary.

Trial data will be electronically entered and stored on REDCap on the research drive of the University of Sydney, with regular data back-up. After the trial, the data will be stored for a minimum of 15 years in a secured study-specific folder on the research drive of the University of Sydney, and access to de-identified data will be considered upon request by the principal investigator.

An independent Data Safety and Monitoring Committee (IDSMC), comprising of an addiction medicine specialist, a statistician and clinical pharmacologist will oversee the safety monitoring of the trial, involving ongoing reviews of any adverse events arising from the administration of CBD (unblinded data). The committee will also monitor aspects of study integrity and design should any protocol changes need to be made.

Data analysis

All data analysis will be performed using Bayesian models instead of frequentist. Bayesian methods can quantify evidence for both effects and the absence of effects, are less prone to non-convergence (due to regularisation), and sample from a joint posterior distribution, hence no family- or experiment-wise correction of regression coefficients for multiple comparisons is necessary [61]. We will use a modified intention-to-treat approach for data analysis, with group membership fixed as the medication type (placebo vs CBD) participants receive on their first study day. Missing data will be imputed via hierarchical multiple imputation [62].

Primary outcomes

We will model the effects of CBD on number of cannabis free days (out of 84 days) via single-level Gaussian regression with the outcome regressed on the main covariate experimental group (placebo vs CBD). Number of cannabis-free days in the 28 days prior to baseline will be included as a covariate to control for variation in participants’ prior frequency of use entering the study. Two treatment factors that could plausibly influence the primary outcomes and which vary across participants will also be included as covariates: number of counselling sessions attended during the study period (count variable range 0–4), and whether or not NRT was taken (binary variable measured at baseline: did not undertake NRT vs undertook NRT). If residuals are distributed normally, we will report the results from this analysis. If residuals are not distributed normally, we will treat cannabis-free days/84 as a bounded count instead of a numeric variable and use aggregated binomial regression with a logit link.

We will model urinary THC-COOH levels (a continuous outcome, in ng/MoL) via random-slopes mixed-effects models with the group, time (6-level categorical ordered predictor; Weeks 1 (baseline), 4, 7, 10, 13), the group × time interaction, number of counselling sessions attended, and whether or not NRT was undertaken as the fixed factors, and participant ID as the random factor.

Secondary outcomes

All repeated measures of secondary outcomes will be modelled using the same approach as for urinary THC-COOH. That is, random slopes mixed-effects models for repeated measures regressions with group, time, the group × time interaction, number of counselling sessions attended, and whether or not NRT was taken as the fixed factors. These models will all be based on the generalised linear model, with link functions differing depending on the form of the outcome, as follows:

  1. (a)

    Numeric (e.g., PROMIS-29 scores, marijuana craving questionnaire scores): Gaussian regression with identity link function

  2. (b)

    Ordinal (e.g., motivation to change cannabis use): ordinal logistic regression with logit link function

  3. (c)

    Bounded count (e.g., severity of CUD) or binary (e.g., participant rating of group allocation): binomial logistic regression with logit link function

  4. (d)

    Unbounded count (e.g., adverse event count): negative binomial regression with log link function

See Table 1 for the form of each outcome measure.

Several secondary outcomes are single observations per individual. Group, number of counselling sessions attended, and whether or not NRT was taken will be the sole predictors in these models. Total abstinence from cannabis during weeks 10–13 (non-abstinent vs abstinent) and 50% increase in cannabis-free days during weeks 10–13 relative to cannabis-free days prior to baseline (< 50% reduction vs ≥ 50% reduction) will be modelled with binary logistic regression with logit link function, relative risk of adverse events during the trial period with negative binomial regression with log link function, hazard of treatment dropout via discrete-time hazard model with complementary log–log link function.

Statistical methods for Indigenous Australian focussed outcomes

Indigenous Australian and non-Indigenous participants will be compared on baseline participant characteristics (e.g., age, gender, frequency of substance use, scores on quality-of-life scales) via simple regression: Gaussian for continuous measures, Logistic for binary and count variables, and multinomial logistic for multilevel categorical data. For the main study analyses, comparing the frequency of illicit cannabis use between placebo and CBD groups, all participants will be pooled and included in main analyses, irrespective of Indigenous status. However, an additional regression will be performed where Indigenous status and the interaction between Indigenous status and the study drug (Placebo vs CBD), along with the primary predictor study drug, will be included in the regression.

The study will stratify randomisation according to Indigenous status, to achieve an approximately equal number of Indigenous Australian participants on active and placebo conditions, thus requiring no additional statistical procedures beyond those outlined in the previous paragraph.

The effect of the experience of discrimination on outcomes related to cannabis use disorder will be estimated via regressing various outcomes related to cannabis use on scores on the modified Everyday Discrimination Scale (m-EDS). Two regressions will be performed for each outcome, with a different version of the m-EDS as the primary predictor in each: (i) a continuous version of the scale (i.e., total score) and (ii) a three-level categorical version of the scale (no vs low vs moderate-to-high). The outcomes that the m-EDS will be regressed on will be: (i) (baseline characteristics (e.g., years of regular cannabis use, scores on quality-of-life scales), (ii), treatment engagement (e.g., treatment retention, number of counselling sessions) and (iii) outcomes during the trial (e.g., frequency of cannabis use, health measures). As described above, the type of regression will depend on the type of outcome: Gaussian for continuous, logistic for binary or bounded count, ordinal logistic for ordered categorical, and negative binomial for unbounded count. The effect of m-EDS on treatment retention will be estimated via Kaplan Meier plots discrete-time hazard model with complementary log–log link function.

Read more here: Source link

Leave a Reply

Your email address will not be published. Required fields are marked *