Psychological and psychosocial interventions for cannabis cessation in adults: a systematic review short report

Overview of cannabis use

Cannabis use may be defined as acute (occasional) or chronic, with chronic use often defined as use on most days over a period of years. 1 Cannabis dependence can develop from chronic use and is characterised by impaired control over use and difficulty in ceasing use. 1 Cannabis dependence is a recognised psychiatric diagnosis, often diagnosed via the Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria and the International Classification of Diseases, 10th Revision (ICD-10). 2,3

Cannabis use has been found to exacerbate the symptoms of psychiatric disorders. 4 In one study, individuals who used cannabis regularly were found to be six times more likely to have a mood or anxiety disorder. 5 The term ‘dual diagnosis’ is used to describe individuals who have a mental health problem and also are dependent on drugs (or alcohol). 6

Epidemiology and prevalence

Cannabis is the most commonly used illicit drug worldwide. 7 In one study reporting cannabis use in European countries, use for 20 or more days per month ranged from 3.5% to 44.1%, with the figure for the UK being 3.9%. 8 In Australia, the prevalence of Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition (DSM-IV)-defined cannabis abuse in the general population over a 12-month period has been estimated at 2.3%, whereas a national survey undertaken in the USA found that 6% of individuals who used cannabis within a 1-year period met the DSM-IV criteria for cannabis dependence. 9,10 This figure, as would be expected, varies by country. In Australia, 31.7% of individuals who used cannabis more than five times in the past year met the criteria for a cannabis use disorder. 11,12

Estimates for the prevalence of patients with a ‘dual diagnosis’ (substance abuse disorder and mental health problems) vary across sources, but it is frequently reported that over 50% of patients with mental health problems also have a substance abuse problem. 6

Impact of health problem and prognosis

The impact of cannabis use on the individual can be classed as acute or chronic. Acute effects include hyperemesis syndrome (recurrent nausea, vomiting and abdominal pain), impaired co-ordination and performance, anxiety, suicidal ideations/tendencies, impaired attention and memory and psychotic symptoms. 2,13 Chronic effects include development of cannabis dependence, cognitive impairment, pulmonary disease and malignancy of the oropharynx. 2,13 There is increasing evidence to suggest the presence of a cannabis withdrawal syndrome, with symptoms (such as dysphoric mood, disturbed sleep and gastrointestinal symptoms) beginning during the first week and continuing for several weeks following the start of abstinence. 4 In a study by Budney et al. ,14 47% of participants withdrawing from cannabis reported four or more severe symptoms, including irritability, craving and nervousness; other symptoms were less severe and included depression, restlessness and headaches.

In a cohort study comparing those seeking treatment with non-treatment seekers, those seeking treatment reported increased cannabis use and more symptoms of dependence but a more positive attitude to treatment. 15 Even when an individual has sought treatment, recovery from substance dependence is hampered by poor adherence to psychological and psychosocial treatments, with factors such as cognitive defects, personality disorder and younger age predicting low treatment adherence. 16

Measurement of disease

Cannabis abuse and dependence is diagnosed using one or more assessment criteria, the most widely used being DSM-IV and ICD-10. There are DSM-IV criteria for both substance dependence and abuse. 3 Dependence is defined as tolerance (a need for increased amounts of the substance to achieve the desired effect), withdrawal (either having withdrawal symptoms or taking another substance to avoid withdrawal symptoms), taking substance in larger amounts than intended, and having persistent desire or unsuccessful efforts to cut down use. Substance abuse is characterised by recurrent use resulting in failure to fulfil obligations, recurrent use in hazardous situations, recurrent substance-related legal problems and continued use despite recurrent social or interpersonal problems. For both sets of criteria, individuals meeting three or more criteria within a 12-month period meet the diagnosis. In 2013, the updated Diagnostic and Statistical Manual of Mental Disorders-Fifth Edition (DSM-5) criteria were released. 17 In the revised criteria, there is no distinction between abuse and dependence, but a spectrum of substance use disorders. 18

Relevant national guidelines

Guidance from the National Institute for Health and Care Excellence (NICE) states that pharmacological interventions for chronic cannabis users are not well developed, so psychosocial interventions are the mainstay of effective treatment. 19 UK Department of Health guidelines for the treatment of chronic users recommend that clinicians should consider motivational interventions in mild cases and structured treatment with key working (when a health professional works with the individual to ensure delivery and ongoing review of care being received) in more heavy users, whereas cognitive–behavioural therapy (CBT) is recommended in cases with comorbid depression and anxiety. 20 European best practice guidance, produced by the European Monitoring Centre for Drugs and Drug Addiction, recommends the use of multidimensional family therapy for regular cannabis users, while individual sessions of CBT are stated to be possibly advantageous. 21

Management of the condition

Providing treatment to chronic users of cannabis to reduce or cease their use is a relatively recent occurrence. Until the 1980s, it was thought that chronic cannabis use did not lead to dependence and treatment was, therefore, not required. 22 Since then, research has looked into evaluating the use of a wide variety of psychological and psychosocial interventions, such as motivational interviewing (MI), CBT and contingency management (voucher incentives). 12 There is limited evidence to suggest which of the many psychological and psychosocial interventions are the most effective at reducing cannabis use.

A number of systematic reviews have been undertaken to assess the benefits of such interventions for regular cannabis users, many of which included meta-analyses. However, they all had limited scope and, therefore, did not assess all the available evidence, and several further studies have been published since. A review by Denis et al. 12 that excluded studies in populations dependent on drugs other than cannabis analysed six randomised controlled trials (RCTs) via narrative synthesis, involving interventions such as CBT and motivational enhancement therapy (MET), finding that CBT provided improved outcomes over brief interventions, whereas voucher incentives were found to enhance treatment when used in combination with other therapies. Dutra et al. 23 undertook a meta-analysis, identifying five studies assessing the use of psychological treatments (including case management, CBT and relapse prevention), finding a significant difference between outcomes for cannabis use, with a mean effect size of 0.81 [95% confidence interval (CI) 0.25 to 1.36] when comparing intervention treatments with control [which consisted of motivational enhancement, wait list control and treatment as usual (TAU)]. Other reviews have focused on specific interventions to treat regular users of cannabis. Tait et al. 24 assessed the use of internet-delivered interventions, finding that such interventions provided a significant decrease in cannabis use at post treatment [g (Hedges’ bias-corrected effect size) = 0.16, 95% CI 0.09 to 0.22; p-value < 0.001].

Previous reviews have also sought to investigate the effectiveness of such interventions across the spectrum of substance misuse, including alcohol and opioids. The review by Dutra et al. 23 reported that treatments incorporating both CBT and contingency management had the greatest effect sizes on substance use across a range of substances including cocaine, opiates and cannabis (Cohen’s d 1.02), whereas the two treatments alone had smaller effect sizes on the same group of substances (contingency management, Cohen’s d 0.58, 95% CI 0.25 to 0.90; CBT, Cohen’s d 0.28, 95% CI 0.06 to 0.51). In contrast, a review by Hunt et al. 6 included RCTs of patients with a severe mental illness and substance dependence, finding no compelling evidence to suggest a significant decrease in substance use when comparing CBT over TAU [two studies, risk ratio (RR) 1.12, 95% CI 0.44 to 2.86] or of CBT plus MI over TAU (one study, mean difference 0.19, 95% CI –0.22 to 0.60). The use of MI alone compared with usual treatment had positive effects on abstinence from alcohol (one study, RR 0.36, 95% CI 0.17 to 0.75) but no effect on other substances (one study, RR –0.07, 95% CI –0.56 to 0.42). 6 Other reviews have focused on specific interventions for ‘general’ substance misuse. Magill et al. 25 analysed 52 studies assessing the use of CBT (plus pharmacological treatments in a number of studies) on a range of substance dependences (including alcohol, cannabis, opiates and cocaine), reporting a small effect on the reduction of substance use for those studies reporting relevant outcomes (34 studies, g = 0.108, 95% CI 0.051 to 0.165; p-value < 0.005). Wood et al. 26 assessed the use of computer-delivered interventions, finding that drug prevention programmes were effective at reducing use in the mid-term (12 months) but not at post treatment. Mindfulness-based interventions have also been found to be effective for substance abuse. 27

Summary of interventions

This review assesses the clinical effectiveness of psychological and psychosocial interventions aimed at assisting regular cannabis users to reduce or cease their use. Only interventions delivered in an outpatient or community setting are included. A full list is provided in Chapter 3, Methods for reviewing effectiveness.

Population and setting

The relevant population included individuals ≥ 18 years of age who were regular users of cannabis and had participated in a study providing treatment(s) for cannabis use in a community or outpatient setting. Studies focusing specifically on treating cannabis users within prisons or the criminal justice system or in inpatient settings were excluded. Inclusion was not restricted according to level of cannabis use at baseline.

Interventions

Studies involving psychological and psychosocial interventions were included.

Relevant comparators

Comparators included other psychological and psychosocial interventions, waiting list control, TAU or no treatment (comparisons with drug treatments were excluded).

Key outcomes

The key outcomes for this review were frequency and amount of cannabis use; severity of dependence; motivation to change; level of cannabis-related problems (including medical and other); and attendance, retention and dropout rates. The results of the review were also used to formulate recommendations for future research.

Identification of studies

The following electronic databases were searched to February 2014 for published and unpublished research evidence: MEDLINE, EMBASE, Cochrane Controlled Trials Register, Health Technology Assessment (HTA) database, Database of Abstracts of Reviews of Effects, Cochrane Database of Systematic Reviews, NHS Economic Evaluation Database, PsycINFO and Web of Science Conference Proceedings Citation Index. This included reference searching within relevant systematic reviews and included studies, contact with experts and searching clinical trials databases (https://ClinicalTrials.gov and www.controlled-trials.com) and relevant websites, including United Nations Office on Drugs and Crime (www.unodc.org), DrugScope (www.drugscope.org.uk), American Society of Addiction Medicine (www.asam.org), National Institute on Drug Abuse (www.drugabuse.gov), Canadian Centre on Substance Abuse (www.ccsa.ca), and Canadian Society of Addiction Medicine (www.csam-smca.org).

The protocol for this review is available on request from the authors.

Population and setting

The relevant population included participants aged ≥ 18 years, who were regular users of cannabis. Inclusion was not restricted according to level of cannabis use at baseline. The review focused on studies in a community or outpatient setting.

Studies focusing on the following subpopulations were excluded:

• Studies in the setting of the criminal justice system, that is prisons, following release (on parole) or within the court system.

• Studies for which the majority of participants were young people (< 18 years of age). In studies of mixed age groups, data for subgroups aged ≥ 18 years were extracted if available or, if not, then the study was included if ≥ 80% of participants were aged ≥ 18 years or, if these data were not available, where the mean age of participants was ≥ 18 years, at baseline.

• Studies for which participants were treated in an inpatient setting, that is, the patient received treatment for regular cannabis use while occupying a hospital ward, drug rehabilitation centre or within an emergency department. Studies for which a subset of the participants were residing in inpatient psychological treatment centres were included, provided that the cannabis intervention was delivered as a standalone therapy rather than as an integrated part of psychological treatment.

• Studies in which the intervention, or a component of the intervention, was provided to participants other than the cannabis user (e.g. parents or partners). An example of such an intervention is Multidimensional Family Therapy.

• Studies in very specific subpopulations (such as indigenous communities or human immunodeficiency virus patients).

For studies covering abuse of more than one substance (i.e. poly-substance abuse, involving other drugs or alcohol), the following approach was taken:

• Studies were included only if they reported cannabis-use outcomes (rather than any drug use) for the subpopulation who were cannabis users.

• Studies were excluded if the entire study population was dependent on alcohol, cocaine, opiates, amphetamines or receiving methadone maintenance (as these are quite specific populations and less relevant to cannabis cessation).

Included interventions

Relevant interventions included a range of psychological and psychosocial interventions aiming to reduce or cease cannabis use. Combinations of therapies were included. All possible modes of delivery were included, including individual face-to-face or group sessions, plus interventions provided via the internet or telephone. Relevant interventions included:

• CBT – an approach aiming to manage cannabis use by changing the way the participant thinks or behaves29

• MI – a person-centred approach that aims to improve motivation to change and resolve ambivalence to change30

• MET – a variant of MI that is manual based31

• relapse prevention therapy – based on CBT, enables clients to cope with high-risk situations that may lead to drug taking32

• contingency management – providing patients with tangible rewards (such as monetary vouchers) in return for a reduction or cessation in drug taking20

• case management – a strategy to improve the co-ordination and continuity of the delivery of services to a patient33

• mutual aid therapy – therapy in which people with similar experiences assist each other to overcome or manage their issues (e.g. Self-Management and Recovery Training)

• other psychological and psychosocial interventions as identified within the review process.

Comparators

Comparators included other psychological and psychosocial interventions, waiting list control, TAU or no treatment. Studies comparing a psychosocial intervention with a drug treatment were excluded.

Outcomes

The key outcomes for this review were:

• frequency and intensity of cannabis use, via self-report, with or without confirmation by biological analysis (urinalysis, hair/saliva analysis)

  • number (%) of days used, time periods of use per day, amount per day

  • number (%) reporting abstinence following intervention

• severity of drug-related problems [measured via the Addiction Severity Index (ASI)]34

• severity of dependence [measured via the Severity of Dependence Scale (SDS)]11

• stage of change or motivation/contemplation to change [e.g. as measured by the Readiness to Change Questionnaire (RCQ)]31

• level of cannabis-related problems – medical problems, legal problems, social and family relations, employment and support [assessed via questionnaires such as the Cannabis Problems Questionnaire (CPQ)]35

• attendance, retention and dropout rates, measured as number of sessions attended or number (%) completing whole treatment period

• recommendations for future research.

Included study types

Only RCTs were included in this review.

Excluded study types

The following study types were excluded:

• non-randomised studies

• narrative reviews, editorials, opinion pieces

• reports written in a language other than English or published as meeting abstracts, if insufficient methodological details are reported in the abstract to allow critical appraisal of study quality and extraction of study characteristics and key outcomes.

Data extraction strategy

Titles and abstracts of citations identified by the searches were screened for potentially relevant studies by one reviewer and a 10% sample checked by a second reviewer (and a check for consistency undertaken). Full texts were screened by two reviewers. One reviewer performed data extraction for each included study. All numerical data were checked against the original article by a second reviewer and any disagreements were resolved through discussion. When studies comprised duplicate reports (parallel publications), the most recent and relevant report was used as the main source and additional reports checked for extra information. Excluded studies were tabulated (see Appendix 3).

Methods of data synthesis

Data were analysed via a narrative synthesis. As described by Popay et al. ,36 this method is based around grouping and tabulating the data in meaningful clusters, allowing results to be summarised (in the form of text and tables) to provide an overview of the direction of effect for each relevant subgroup. Within this review, studies were first divided into two main population subgroups (general cannabis users and those with a major psychiatric condition). Second, studies were categorised according to their intervention and comparison groups (e.g. CBT vs. wait list, CBT vs. MI, etc.). Third, results were tabulated for two key time points (post treatment and later follow-up). Within each study, outcomes at each time point were categorised according to whether or not they were significantly different between groups or between baseline and follow-up. Finally, summary tables were populated for each intervention/comparison. Outcomes across studies at each time point were summarised as being mainly significant, mainly not significant or mixed.

There was substantial heterogeneity between studies in terms of populations, interventions, comparators, outcome measures reported and statistics reported. To increase clarity, the main results of this review are presented in the form of an overview of the outcomes reported per study and how many showed a significant difference, as described above. Detailed numerical results per study group are not presented in the main results section, but are provided in Appendix 4 for reference. Meta-analysis was not undertaken, as this would have required restricting each analysis to studies reporting the same outcome in a consistent format with full data and it was felt that the broad results picture might have been lost.

Subgroup analyses were undertaken with regard to number of treatment sessions, group/individual treatment, high/low cannabis use at baseline, recruitment method (referral vs. voluntary), participant age and use of other substances (tobacco and alcohol) at baseline.

Quality assessment of included studies

Methodological quality of included RCTs was assessed using an adapted version of the Cochrane risk of bias assessment criteria. This tool addresses specific domains, namely sequence generation, allocation concealment, blinding of participants and personnel, blinding of outcome assessment, incomplete outcome data and selective outcome reporting. 37 Outcome assessment was considered to be blinded if the person assessing or interviewing the participants was blinded to group allocation (although participants were not blinded and many of the data were self-reported). We made two adaptations to these criteria in order to aid quality assessment. First, we utilised the ‘5-and-20 rule’ for incomplete outcome data, as proposed by Schulz and Grimes. 38 Schulz and Grimes38 state that a lower than 5% loss in participants probably leads to little bias, while a greater than 20% loss potentially poses serious threats to validity. We therefore defined < 5% attrition as ‘low risk’, between 5% and 20% attrition as ‘intermediate risk’ and > 20% attrition as ‘high risk’. Attrition was defined as the percentage of patients not followed up at the final time point reported. The second adaptation we made to the Cochrane criteria was to add an ‘overall risk’ criterion, aiming to summarise the overall risk of studies. We categorised studies as low risk, high risk or unclear risk, determined using the following criteria. Low risk was allocated to studies where randomisation, allocation concealment, blinded outcome assessment and incomplete data were all determined to be low risk. High risk was allocated to studies deemed to have undertaken inadequate randomisation (self-selection, sequential patients, odd and even), and/or when allocation was not concealed, and/or when incomplete data were deemed to be high risk. Unclear risk was allocated to all other studies.

Patient and public involvement

In order to seek patient and public input into the review, we recruited a service user through liaison with the project’s clinical advisors, who was currently acting as a ‘service ambassador’ within their treatment service (an individual who has completed a treatment regime, ceased their primary substance use and is now involved in supporting patients at the treatment centre).

A short ‘briefing document’ using non-academic language was developed (see Appendix 5) in order to introduce the individual to the research. The briefing document included sections describing the basic principles of a systematic review, the general area in which the research is being undertaken (i.e. psychological/psychosocial treatments for regular users of cannabis) and the input required from the service user. The service user was compensated for the time spent at meetings and for travel expenses.

The review team met with the service user twice. The first meeting was scheduled once the protocol had been written. The service user provided valuable input into the following areas of the protocol:

• an additional intervention not already identified in the protocol (mutual aid therapy)

• two additional outcome measures that were felt to be important (daily time periods of cannabis use and contemplation to change)

• general approval of the focus of the review.

The review team then met with the service user after the draft report had been written, when the service user had the following inputs:

• suggested amendments to the plain English summary

• reviewed the suggested research priorities

• reviewed the section describing factors relevant to the NHS.

Quantity of research available

The searches identified 1087 citations (1079 via database searches and eight via other sources). Of these, 919 citations were excluded at the title/abstract stage and 168 full-text articles were screened. Of these, 126 were excluded: 65 did not include relevant outcomes, 41 evaluated irrelevant populations, nine were not RCTs, five did not involve a relevant intervention, three detailed a non-relevant secondary analysis or study characteristic and three were not in English (excluded studies are listed in Appendix 3). In total, 42 articles relating to 33 RCTs were included in this review. The PRISMA flow chart is shown in Figure 1.

FIGURE 1.

Study selection process: PRISMA flow diagram.

All titles and abstracts were screened for inclusion by one reviewer and a check for consistency was undertaken. A second reviewer screened approximately 10% of the references (n = 100) during the initial screening stage. No discrepancies were found.

Characteristics of included studies

The 33 studies included in this review were undertaken in a range of countries: the USA (13 studies39–51), Australia (seven studies52–58), Germany (three studies59–61), Brazil (two studies62,63), Canada (two studies64,65), Switzerland (two studies66,67), Denmark (one study68), Ireland (one study69) and worldwide (two studies, one utilising internet-based interventions70 and the other undertaken in a number of locations worldwide71) (Tables 1 and 2).

Population

Risk of bias in included studies

Table 3 summarises the risk of bias for each of the included studies. Most studies used an appropriately generated randomisation sequence, with 21 studies being deemed ‘low risk’, 10 ‘unclear risk’ and two ‘high risk’. Allocation concealment followed a similar pattern. No studies blinded study participants to group allocation and we deemed this form of blinding to be impossible for the interventions under review. As many of the outcome measures were self-reported, outcomes were deemed to have been blinded if the outcome assessors were blinded to group allocation. This form of blinding was poorly reported; in 18 studies, blinding of outcome assessment was unclear or unreported. 40,44–46,49,50,52,55,56,58–63,66,75,81 Participant attrition was well reported but high, ranging from 6% to 79% (mean 30.2%, median 25.5%); 22 studies were rated as high risk for this attribute (with attrition of > 20% at the final follow-up time point). Regarding overall risk, 24 studies40,41,43,48,49,52–55,57–63,65,67–69,70,71,75,81 were deemed to be ‘high risk’, in nine studies39,44–47,50,51,56,66 the risk was unclear and no studies were deemed to be ‘low risk’. In the general population subgroup, 18 studies40,41,43,48,49,53,55,59–62,63,65,67,70,71,75,81 were deemed to be at high risk of bias, whereas in eight studies39,44–47,50,51,56 the risk was unclear. In the psychiatric population studies, six52,54,57,58,68,69 were deemed to be at high risk and in one study66 the risk was unclear. Twenty-one of the studies40,41,43,48,49,53–55,57–63,65,67–69,75,81 were deemed to be at high risk owing to incomplete outcome data (high level of attrition) and three studies52,70,71 were deemed to be at high risk owing to poor random sequence generation or allocation concealment.

Overview of effectiveness section

Results are presented for each intervention/comparator category (e.g. CBT vs. wait list, CBT vs. brief MI, etc.). An overall summary of results is provided in Tables 4 and 5. This is followed by more detailed results for each intervention/comparator category (see Tables 6–23). Owing to the large number of studies and the variability in outcomes and data format, detailed numerical results are not presented here. Instead, this section provides an overview of the outcomes reported per study and how many showed a significant difference, both between intervention groups and in terms of changes from baseline, at different follow-up time points. Full extracted data per study are provided in Appendix 4.

Cognitive–behavioural therapy compared with wait list control

Cognitive–behavioural therapy or psychotherapy compared with brief motivational interviewing

Cognitive–behavioural therapy compared with other interventions (or different cognitive–behavioural therapy format or duration)

Telephone- or internet-based cognitive–behavioural therapy or counselling compared with wait list or other interventions

Brief motivational interviewing compared with wait list or assessment only

Brief motivational interviewing compared with other interventions

Contingency management (vouchers for abstinence) versus other interventions

Cognitive–behavioural therapy plus treatment as usual compared with treatment as usual

Cognitive–behavioural therapy compared with other interventions

Number of sessions, and comparison of longer cognitive–behavioural therapy, compared with shorter motivational interviewing

Two sets of data imply that longer courses of CBT may be somewhat more effective than shorter courses of MI, but findings were mixed. First, four studies directly compared CBT (6–14 sessions) against brief MI (1–4 sessions). 39,40,50,53 Of these, some showed better results for CBT than MI (one39 post treatment, two at later follow-ups39,53), whereas others showed no significant differences (two40,50 post treatment, one at later follow-ups50). Second, studies comparing brief MI with wait list showed some significant effects, but these were not as positive overall as results for CBT compared with wait list which were significant on nearly all outcomes. The somewhat better results for CBT than MI could be due to the nature of CBT treatment, the fact that more sessions were provided, or a combination of the two.

Within six studies of CBT compared with wait list,39,50,53,59,60,63 all durations of CBT (4–14 sessions) appeared effective. There were slightly fewer significant effects in the study of four-session CBT,63 but this may have been related to the small number of participants. One study63 compared four sessions of CBT over either a 4- or 12-week period; post-treatment results significantly favoured 12-week treatment on some but not all outcomes. Within studies of brief MI compared with wait list, there was no clear difference between studies of one-session or two-session MI, with both durations showing mixed results.

Group or individual treatment

Of the 33 included studies, 2739–48,51–55,57–63,66,68,70,71,75 provided individualised treatments, whereas three49,67,69 provided group treatment and two50,65 compared group with individual treatment (not reported for one). 56 There were insufficient group treatment studies within most intervention/comparator categories to meaningfully compare group with individual treatment. Within studies of CBT compared with wait list, the one study of group CBT50 had similar positive outcomes to the five individual CBT studies. Within studies of CBT compared with brief MI, one study50 showed little benefit of group CBT compared with individual MI, and studies of individual CBT compared with MI showed mixed results. Within studies of brief MI compared with wait list, studies of individual MI showed mixed results, whereas the one study67 of group MI showed no significant effect; however only 1 session of MI was provided and results were not measured until 6 months and then only on one outcome (cannabis days of use). One study65 directly compared individual with group CBT (4 sessions each), but only 17 cannabis users were analysed and only one relevant outcome reported (days abstinent); results non-significantly favoured individual CBT post treatment, but this effect was not maintained at 12 months. Overall, group treatment may possibly be less effective than individual treatment, but this is based on very limited data.

High compared with low baseline cannabis use/dependence

The impact of high or low baseline cannabis use/dependence could be assessed to some extent within certain intervention/comparator categories: CBT compared with wait list, computer-delivered/telephone-delivered CBT compared with other, and brief MI compared with wait list. Within each of these categories, studies with low baseline use43,45,46,48,55,60,63,67,70,71 appeared slightly less likely to show significant differences on all outcomes than studies of high use. 39,50,51,53,59,61 However, this difference was not substantial or conclusive. This potential difference could be owing to the interventions having a greater effect on participants with higher baseline use/dependence, either because a greater effect could be demonstrated or because these participants may have been more motivated to reduce use; however, this conclusion should be treated with caution.

In addition, two studies with low baseline use reported subgroup effects. 45,48 One study45 reported no significant difference between internet-based personalised feedback and the control group across all participants, but a significant difference for participants with a higher contemplation to change use or a family history of drug problems. Another study,48 in which participants were not seeking treatment for their cannabis use, reported no significant difference between brief MI and AO across all participants, but significant differences for those with a desire to cease use.

Recruitment method (voluntary compared with referrals)

Among the 26 general population studies,39–51,53,55,56,59–63,65,67,70,71,75 1640,42–44,47–50,51,53,55,61,62,67,70,75 used voluntary recruitment via advertisement, while four45,46,60,71 used referrals and four39,41,59,65 used a combination of voluntary and referrals (not reported for two studies56,63). Results for the general population studies showed no clear difference in results according to recruitment method, although this comparison is based on limited data. No clear difference was observed within studies of CBT compared with wait list (two voluntary,50,53 one referral,60 two combination;39,59 all positive results) or within studies of brief MI compared with wait list (six voluntary,43,48,50,51,53,67 three referral,45,46,71 one combination;39 all mixed results). Nevertheless, it should be noted that the majority of the general population studies recruited volunteers via advertisement and, therefore, may reflect a more motivated group when compared with the ‘average’ cannabis user.

In contrast to the general population studies, all seven studies in psychiatric populations recruited patients via referral (four studies52,54,66,68) or a combination of referral and voluntary methods (three studies57,58,69). Across four studies of CBT plus TAU compared with TAU in psychiatric populations, three studies52,66,68 used referrals and one69 used a combination of voluntary and referrals; all four showed little difference between CBT plus TAU and TAU alone. Comparisons within the other three studies54,57,58 were too heterogeneous to assess the effects of recruitment method.

Participant age

Within most intervention/comparator categories, mean ages were similar across studies, so the effect of age could not be meaningfully assessed. This was true for comparisons of CBT compared with wait list (mean age 24–36 years), CBT compared with brief MI (mean age 32–36 years), telephone-/internet-delivered CBT (mean age 25–36 years) and CBT plus TAU vs. TAU (mean age 26–29 years). For brief MI compared with wait list (10 studies39,43,45,46,48,50,51,53,67,71), mean age ranged from 18 years to 36 years. Five studies43,45,46,48,67 assessed relatively young populations (mean age 18–23 years, upper age range in teens or twenties) and these studies were all classed as low baseline use/dependence. There were no clear differences in effects according to age, with the five studies of younger populations showing mixed results in a similar manner to other studies.

Baseline use of other substances

There were insufficient data to assess the effect of baseline alcohol and tobacco use; these were only reported in ten studies (alcohol)39–41,44,51,54,55,59,60,63 and two studies (tobacco)55,60 and were reported via very different measures (see Tables 1 and 2).

General population studies

Of 26 studies39–51,53,55,56,59–63,65,67,70,71,75 assessing the general population of cannabis users (7643 randomised participants), 1640,42–44,47–50,51,53,55,61,62,67,70,75 recruited via advertisement and eight39,41,45,46,59,60,65,71 via referrals or both. Baseline use/dependence was high in 13 studies39–42,44,47,49–51,53,56,59,61 and low in 10. 43,45,46,48,55,60,63,65,70,75 Across six studies39,50,53,59,60,63 of CBT (4–14 sessions) compared with wait list, CBT was significantly better on most outcomes (cannabis use, severity of dependence, cannabis problems) post treatment (in all five studies39,50,59,60,63 with data) and at 9 months (in the one study53 with later follow-up). Four studies39,40,50,53 comparing CBT (6–14 sessions) with briefer MI/MET (1–4 sessions) gave mixed results, with two studies40,50 showing better results for CBT post treatment and at 9–16 months, while two further studies39,53 showed few between-group differences; both CBT and MI gave significant improvements from baseline. In one small study,56 supportive–expressive dynamic psychotherapy (16 sessions) gave significant improvements over one-session MI. One study49 of CBT compared with social support group (10 sessions each) and another44 of CBT compared with case management (nine sessions each) showed no significant differences between groups but all groups significantly improved from baseline with changes maintained at 14–15 months. One study each of telephone-delivered CBT, internet-delivered CBT and internet-delivered counselling all showed significant improvements over wait list or education control post treatment and at 3 months. 55,61,70

Ten studies39,43,45,46,48,50,51,53,67,71 assessing brief MI/MET (one or two sessions) compared with wait list or AO gave mixed results, with brief MI appearing significantly better on some outcomes but not others, post treatment and at 3–9 months. Results were similar for three studies51,62,75 comparing brief MI against education controls. Five studies40–42,44,47 assessed contingency management (monetary vouchers for abstinence). Vouchers alone and CBT plus vouchers gave better results than CBT or MET alone post treatment (three studies40,42,44), while at 14–15 months positive results were maintained for CBT plus vouchers but less so for vouchers alone (two studies41,44).

Psychiatric population studies

Seven studies52,54,57,58,66,68,69 (525 randomised participants) assessed psychiatric populations (schizophrenia, psychosis, bipolar disorder or major depression); all recruited via referrals or referrals plus advertisements. Baseline use/dependence was high in three studies66,68,69 and low in four. 52,54,57,58 Across four studies52,66,68,69 assessing CBT (6–24 sessions) plus TAU compared with TAU alone, there were few significant between-group differences post treatment and none at 10–12 months (small studies; limited data), with little change from baseline in either group. Two studies54,58 reported no significant difference between different types of 10-session therapy (one compared CBT, computer-delivered CBT and PCT; the other compared CBT and psychoeducation), although the latter reported significant improvements from baseline in both groups (limited data). 54 A further study reported improvements for 10-session CBT or computer-delivered CBT over single-session MI at 12 months’ follow-up on one outcome (daily cannabis use). 57

Strengths

This report systematically reviews the evidence for a range of psychological and psychosocial treatments for regular users of cannabis. Thirty-three studies were included and the scope of the review is inclusive, covering a wide range of populations, interventions and outcomes. The majority of previous reviews in this topic area have restricted their scope to a subtype of intervention or population, or are not specific to cannabis users. 6,33 The present review has included all psychosocial or psychological interventions undertaken in the adult, community-dwelling population of cannabis users and has only included RCTs, ensuring that only the highest quality available evidence has been included. Robust methods were used, including a search methodology with wide scope, grey literature searching and contact with clinical experts in the area. Data were double-checked for accuracy.

The included studies reported a heterogeneous set of data; in many cases, similar outcome measures (e.g. cannabis use) were reported in different ways. Narrative synthesis was used to analyse and explore the data. Results are presented for each intervention/comparator category (e.g. CBT vs. wait list or CBT vs. MI) and by population (general vs. psychiatric). This allows studies with similar comparisons and populations to be analysed together, to provide an overview of the direction of effects for each category at different time points. This approach also minimises loss of data, as any meta-analysis would have been restricted to studies reporting the same outcome in a consistent format and reporting full data [including standard deviations (SDs), etc.].

Limitations

There was substantial heterogeneity between studies in terms of their populations, interventions, comparators, outcome measures and data format, and limited time was available to conduct this systematic review short report. Therefore, results are presented as an overview of the outcomes reported per study and how many showed a significant difference, both between intervention groups and in terms of changes from baseline, at different follow-up time points. Detailed numerical results per group are not presented in the main results section (these are provided in Appendix 4) and meta-analysis was not undertaken. This approach has the limitations that (1) it was not possible to present effect sizes for outcomes and (2) data were not pooled across studies. However, the narrative synthesis approach was thought to provide benefits in terms of interpretability as described above. In addition, owing to time constraints, we were unable to include studies written in languages other than English.

Only RCTs were included in this review. Although this ensures that only the highest quality of evidence is included in the synthesis, it does potentially result in informative studies being rejected. Nine potentially relevant articles were excluded owing to this;83–91 however, most of these studies would have been rejected for other exclusion criteria (four of the studies included a population comprising people aged < 18 years). 84,88,90,91

We excluded studies that were undertaken within the criminal justice setting (e.g. studies undertaken within the court system, prison, or while study participants are on parole). We also excluded studies that treated individuals other than the cannabis user (e.g. a family member). Although risking the exclusion of potentially valid trials, excluding these populations reduced the many sources of heterogeneity. This review focused on community-delivered interventions. Interventions carried out within the criminal justice setting are unlikely to be replicable when delivered to cannabis users outside that setting, owing to differences in recruitment, intervention delivery and outcome assessment.

The studies included in this review utilised a variety of recruitment methods, involving voluntary recruitment, referral by a health-care professional or a combination of both. Studies in the general population of cannabis users mostly used voluntary recruitment methods (most often via advertisements); therefore, these may have reflected more motivated populations and may not be generalisable to all cannabis users. Conversely, this may reflect practice in that psychological interventions are likely to be provided to those willing to receive them. In addition, the included studies recruited cannabis users with varying frequencies of cannabis use at baseline.

Results obtained from the psychiatric population studies may have been affected by provision of TAU to both groups, which in two studies included psychosocial interventions. 68,69 Although these presumably focused on the psychiatric condition rather than cannabis use, they may indirectly have affected cannabis outcomes in both groups. Another study provided a self-help book on substance abuse as part of TAU. 52

The following topics are outside the scope of this systematic review, but could form aspects of future work in this area. Although this short report focused on treatment of adult cannabis users, there is a large amount of literature on treatment of cannabis use in adolescents, including the effects of preventative strategies as well as interventions involving families or schools. In addition, assessment of the effects of factors such as therapist type and treatment fidelity, which are important factors when considering psychosocial interventions, may form a part of future reviews.

Intervention delivery

• Availability of CBT and other treatments within the NHS, and which type of health professional would provide these. Department of Health guidance suggests that psychosocial interventions for substance misuse may be delivered by a key worker (when a health professional works with the individual to ensure delivery and ongoing review of care being received) with the required competencies or by a drug worker or psychologist. 20

• UK guidelines also emphasise the importance of person-centred care, consideration of family and carer involvement, links between services to avoid loss of contact and importance of key working. 19,20 They also advise that treatment and information should be accessible to people with disabilities and to those who do not speak or read English. 19

Patient identification

• How cannabis users would be identified/diagnosed and referred for treatment (e.g. via a general practitioner, social worker or a variety of routes).

• The level of cannabis use or dependence at which treatment is required.

• Whether all patients with a particular level of use or dependence would be referred or, for example, only those who wished to receive a psychosocial intervention and/or those expressing a desire to reduce or cease use (note that the majority of data for ‘general population’ studies involved participants voluntarily responding to advertisements). Existing UK guidelines advise that service users should be allowed to make informed decisions about their treatment in partnership with their health professionals. 19

• Other relevant interventions exist that do not explicitly target cannabis users, but are aimed at individuals with addictive behaviours. Although no relevant RCTs of such therapies were identified in this review, they may be worthy of further consideration or research. For example, mutual aid therapies (such as Self-Management and Recovery Training) involve people with similar experiences assisting each other to overcome or manage their issues. Department of Health guidance advises that self-help and mutual aid groups should be recommended for all drug misusers seeking to achieve and maintain abstinence. 20 Other interventions not within the scope of this review may increase the effectiveness of psychosocial or psychological interventions for cannabis cessation. For example, nicotine replacement therapy may increase the ability of regular cannabis users to reduce their use of the drug, although a recently conducted pilot study found that this was not the case for a group of 12 cannabis-dependent individuals. 92

Comparison of results from this review with relevant national guidelines

• Existing UK guidelines from NICE advise that CBT should not be offered routinely for treatment of cannabis abuse. Department of Health guidance advises that brief motivational interventions may be considered in mild cases of cannabis use, whereas more heavily dependent users may require structured treatment with key working. 19,20 Our review found that CBT appeared effective for routine treatment of cannabis abuse, but it was unclear how much more effective it was than briefer interventions.

• In terms of people with co-existing psychiatric conditions, NICE and Department of Health guidelines suggest that CBT should be considered for treatment of cannabis users with comorbid depression and anxiety disorders. 19,20 Our review found that studies with people suffering from psychological conditions appeared to show less promising effects of CBT,52,54,57,58,66,68,69 but we focused on studies of CBT aimed at treating the cannabis abuse rather than the psychological condition and these studies may have been confounded by both groups receiving TAU for the psychological condition (including psychosocial treatments in some cases). 52,66,68,69

Interventions

Future studies may wish to test other interventions in addition to CBT. The use of contingency management (vouchers for abstinence) improved long-term outcomes when added to CBT. Future studies assessing CBT (and/or brief interventions) may also wish to include a group receiving CBT and contingency management. In addition, mutual aid therapies and self-help groups (for which no RCTs were identified in this review) may be worthy of future study. Other treatments not within the scope of this review, such as nicotine replacement therapy, could be assessed in conjunction with psychological and psychosocial interventions in order to increase effectiveness. All studies should aim to report the included interventions in sufficient detail to allow replication.

The current review has identified that CBT may be effective when delivered to cannabis-dependent individuals, but effectiveness has not been demonstrated when such treatment is provided to patients with psychological comorbidities. This lack of effectiveness for the dually diagnosed population confirms findings from a previous review. 6 These findings suggest that patients with a dual diagnosis may require separate treatments for their substance abuse and psychological problem. Alternatively, if this lack of effectiveness in the dually diagnosed population was due to the participant themselves being unable to respond to the CBT treatment owing to their psychological condition, further research may be necessary to identify interventions that are effective in such populations.

Populations

Future studies may also wish to consider the potential effects of recruitment method. Most existing general population studies recruited via advertisement. Studies using other methods of recruitment may be more generalisable to a clinical setting. Conversely, a somewhat selected study population may be reasonable (e.g. those wishing to cease use and/or willing to receive the intervention) as current guidance suggests that these populations would be the most relevant to receive psychosocial interventions (as opposed to, for example, users expressing no desire to cease use).

The included studies did not report the effect of the psychopharmacology of cannabis. Recent findings indicate that certain strains of cannabis containing high levels of cannabidiol are associated with less cognitive impairment and positive therapeutic potential in psychosis and other disorders (such as Parkinson’s disease). 93–96 This pharmacological factor could be an important modulating factor in treatment outcomes and, therefore, should be taken into account in future studies.

Outcomes

Outcomes reported in most studies could be classified into four main groups: (1) level of attendance, (2) cannabis use, (3) severity of dependence and (4) cannabis-related problems. Cannabis use covered a range of specific outcomes including point abstinence rates, abstinence over a specified period, number of days using cannabis or number of days abstinent (over a specified period), amount of cannabis use per day, and number of periods of use per day (e.g. of four daily periods). All the above outcomes showed significant effects in at least some studies. Future studies may wish to consider which outcomes have been most commonly used in existing studies making similar comparisons, to improve comparability between studies. In terms of the above outcomes, level of cannabis use can be difficult to measure owing to the different levels of active ingredients in different cannabis products. Abstinence is a frequently used outcome but may not be desirable or attainable for all users. Severity of dependence addresses the impact on a person’s life rather than focusing on quantities of use. Cannabis-related problems may also be a useful measure but in populations with other issues (such as psychiatric conditions) it may be difficult to distinguish between the causes of problems. Patient preference for different types of psychological intervention may also be useful to assess.

Methodology

Future studies should carefully consider trial methodology. The studies included in this review utilised a range of comparison groups, including active treatments (e.g. a variation of CBT or MI), less specific controls for time and attention (e.g. cannabis education) and inactive controls (wait list or AO). We recommend that wait list controls are included as a group in future studies, even those comparing different types or durations of active treatments, to indicate whether or not active treatments are effective when compared with no treatment (as well as with each other).

In addition, it would be useful to consider carefully future study designs for populations with psychiatric conditions. If TAU (for the psychiatric condition) is provided to all groups, this should be reflective of current clinical practice and consideration should be given to whether or not it may confound the study intervention (e.g. whether or not it includes psychosocial interventions). It may also be worth considering whether or not a single psychosocial intervention can be tailored to address jointly both the psychiatric condition and cannabis use.

The studies included in this review followed up participants over various time periods, ranging, overall, between 12 weeks and 16 months. Future studies should aim to follow-up patients over the long term; wait list controls with long-term follow-up are also valuable to assess fully long-term effects of treatment; however, this needs to be balanced against ethics considerations and acceptability to trial participants.