BACKGROUND: Despite the high prevalence of co-morbid substance use among mental health service users (dual diagnosis), very few mental health workers in the UK have had training and/or clinical experience to equip them to deliver targeted interventions to this client group. METHOD: In a randomised controlled trial of training for dual diagnosis interventions, 79 case managers from 12 community mental health teams in South London were randomly allocated to either receive training and follow-up supervision (experimental group) or no training and supervision (control group). Baseline measures of attitude, self-efficacy and knowledge were collected prior to randomisation, and were repeated at 18 months post-training. An intention to treat analysis of follow-up data (adjusted for baseline score for that outcome and team) was performed. RESULTS: At 18 months post-training, the AAPPQ (The Alcohol and Alcohol Problems Perception Questionnaire) total score was did not differ significantly between the two groups (adjusted difference 7.43 [95% CI -0.86 to 15.71], p = 0.08). There were significant differences in favour of the experimental group on 2 of the 6 subscales of the AAPPQ: 'adequacy of knowledge and skills in working with alcohol" (adjusted difference 3.598 [95% CI 1.03 to 6.16], p = 0.007) and "self-esteem in working with alcohol" (adjusted difference 3.00 [95% CI 0.46 to 5.54], p = 0.021). In addition there were significant improvements for the experimental group on "Knowledge About Dual Diagnosis" (adjusted difference 2.00 [95% CI 0.80 to 3.22], p = 0.002) and "Self-Efficacy Scale" (adjusted difference 13.55 [95% CI 8.00 to 26.86], p = 0.001). The effect of membership of teams was added to the analysis of covariance and this changed the results for only one variable: "self-esteem working with drinkers" was no longer significant. CONCLUSION: A brief training course in dual diagnosis interventions had a significant effect on secondary measures of knowledge and self-efficacy that was detectable at 18 months post-training. Improvements in attitudes towards working with drinkers and drug users in mental health settings failed to reach statistical significance. Future research should explore the effects of dose of dual diagnosis training, and the successful integration of skills gained into routine care. TRIAL REGISTRATION: ISRCTN98891022 14th March 2007.