Background: Cannabis use disorder is the most commonly reported illegal substance use disorder in the general population; although demand for assistance from health services is increasing internationally, only a minority of those with the disorder seek professional assistance. Treatment studies have been published, but pressure to establish public policy requires an updated systematic review of cannabis-specific treatments for adults. Objectives: To evaluate the efficacy of psychosocial interventions for cannabis use disorder (compared with inactive control and/or alternative treatment) delivered to adults in an out-patient or community setting. Search methods: We searched the Cochrane Central Register of Controlled Trials (CENTRAL; 2015, Issue 6), MEDLINE, EMBASE, PsycINFO, the Cumulaive Index to Nursing and Allied Health Literature (CINAHL) and reference lists of articles. Searched literature included all articles published before July 2015. Selection criteria: All randomised controlled studies examining a psychosocial intervention for cannabis use disorder (without pharmacological intervention) in comparison with a minimal or inactive treatment control or alternative combinations of psychosocial interventions. Data collection and analysis: We used standard methodological procedures as expected by The Cochrane Collaboration. Main results: We included 23 randomised controlled trials involving 4045 participants. A total of 15 studies took place in the United States, two in Australia, two in Germany and one each in Switzerland, Canada, Brazil and Ireland. Investigators delivered treatments over approximately seven sessions (range, one to 14) for approximately 12 weeks (range, one to 56). Overall, risk of bias across studies was moderate, that is, no trial was at high risk of selection bias, attrition bias or reporting bias. Further, trials included a large total number of participants, and each trial ensured the fidelity of treatments provided. In contrast, because of the nature of the interventions provided, participant blinding was not possible, and reports of researcher blinding often were unclear or were not provided. Half of the reviewed studies included collateral verification or urinalysis to confirm self report data, leading to concern about performance and detection bias. Finally, concerns of other bias were based on relatively consistent lack of assessment of non-cannabis substance use or use of additional treatments before or during the trial period. A subset of studies provided sufficient detail for comparison of effects of any intervention versus inactive control on primary outcomes of interest at early follow-up (median, four months). Results showed moderate-quality evidence that approximately seven out of 10 intervention participants completed treatment as intended (effect size (ES) 0.71, 95% confidence interval (CI) 0.63 to 0.78, 11 studies, 1424 participants), and that those receiving psychosocial intervention used cannabis on fewer days compared with those given inactive control (mean difference (MD) 5.67, 95% CI 3.08 to 8.26, six studies, 1144 participants). In addition, low-quality evidence revealed that those receiving intervention were more likely to report point-prevalence abstinence (risk ratio (RR) 2.55, 95% CI 1.34 to 4.83, six studies, 1166 participants) and reported fewer symptoms of dependence (standardised mean difference (SMD) 4.15, 95% CI 1.67 to 6.63, four studies, 889 participants) and cannabis-related problems compared with those given inactive control (SMD 3.34, 95% CI 1.26 to 5.42, six studies, 2202 participants). Finally, very low-quality evidence indicated that those receiving intervention reported using fewer joints per day compared with those given inactive control (SMD 3.55, 95% CI 2.51 to 4.59, eight studies, 1600 participants). Notably, subgroup analyses found that interventions of more than four sessions delivered over longer than one month (high intensity) produced consistently improved outcomes (particularly in terms of cannabis use frequency and severity of dependence) in the short term as compared with low-intensity interventions. The most consistent evidence supports the use of cognitive-behavioural therapy (CBT), motivational enhancement therapy (MET) and particularly their combination for assisting with reduction of cannabis use frequency at early follow-up (MET: MD 4.45, 95% CI 1.90 to 7.00, four studies, 612 participants; CBT: MD 10.94, 95% CI 7.44 to 14.44, one study, 134 participants; MET + CBT: MD 7.38, 95% CI 3.18 to 11.57, three studies, 398 participants) and severity of dependence (MET: SMD 4.07, 95% CI 1.97 to 6.17, two studies, 316 participants; MET + CBT: SMD 7.89, 95% CI 0.93 to 14.85, three studies, 573 participants), although no particular intervention was consistently effective at nine-month follow-up or later. In addition, data from five out of six studies supported the utility of adding voucher-based incentives for cannabis-negative urines to enhance treatment effect on cannabis use frequency. A single study found contrasting results throughout a 12-month follow-up period, as post-treatment outcomes related to overall reduction in cannabis use frequency favoured CBT alone without the addition of abstinence-based or treatment adherence-based contingency management. In contrast, evidence of drug counselling, social support, relapse prevention and mindfulness meditation was weak because identified studies were few, information on treatment outcomes insufficient and rates of treatment adherence low. In line with treatments for other substance use, abstinence rates were relatively low overall, with approximately one-quarter of participants abstinent at final follow-up. Finally, three studies found that intervention was comparable with treatment as usual among participants in psychiatric clinics and reported no between-group differences in any of the included outcomes. Authors' conclusions: Included studies were heterogeneous in many aspects, and important questions regarding the most effective duration, intensity and type of intervention were raised and partially resolved. Generalisability of findings was unclear, most notably because of the limited number of localities and homogeneous samples of treatment seekers. The rate of abstinence was low and unstable although comparable with treatments for other substance use. Psychosocial intervention was shown, in comparison with minimal treatment controls, to reduce frequency of use and severity of dependence in a fairly durable manner, at least in the short term. Among the included intervention types, an intensive intervention provided over more than four sessions based on the combination of MET and CBT with abstinence-based incentives was most consistently supported for treatment of cannabis use disorder.