This study examined the impact of a school-based preventive intervention on cannabis use in adolescence, using a cluster-randomized trial of a multilevel intervention aimed at improving social relationships within schools by promoting change in school environment. Four waves of data were collected at baseline (1997, Year 8: mean age 13 years) and six, 18, and 30 months later (1999, Year 10: mean age 16 years). Self-reported substance use, school engagement, and sociodemographic data were collected using computer-administered questionnaires. Some 2,678 (74%) Year 8 students participated (wave 1) with minimal attrition (10% by wave 4). Adjusting for baseline use, weak evidence existed for an intervention effect on the prevalence of any use at Year 10 (OR 0.75, 95% CI 0.54, 1.05) and incident weekly use (OR 0.72, 95% CI 0.39, 1.33). These effects were reduced after adjusting for confounders. Moderate evidence suggested an interaction effect between intervention group and tobacco use (p = 0.04), suggesting the intervention was more effective for non-smokers at baseline (Adj. OR 0.50, 95% CI 0.26, 0.98). This study indicates that a multilevel school-based program may provide an innovative direction for sustainable school interventions with the potential to reduce substance use.