This meta-analysis examines the effectiveness of interactive middle school-based drug prevention programs on adolescent cannabis use in North America, as well as program characteristics that could moderate these effects. Interactive programs, compared to more didactic, lecture style programs, involve participants in skill-building activities and focus on interaction among participants. A systematic literature search was conducted for English-language studies from January 1998 to March 2014. Studies included evaluations using random assignment or a quasi-experimental design of interactive school-based substance use prevention programs delivered to adolescents (aged 12-14) in North American middle schools (grades 6-8). Data were extracted using a coding protocol. The outcomes of interest were post-treatment cannabis use, intent to use, and refusal skills compared across intervention and control groups. Effect sizes (Cohen's d) were calculated from continuous measures, and dichotomous measures were converted to the d index. A total of 30 studies yielding 23 independent samples were included. The random effects pooled effect size for cannabis use (k=21) was small ([Formula: see text]= -0.07, p<0.01) and favorable for the prevention programs. The pooled effect sizes for intention to use (k=3) and refusal skills (k=3) were not significant. Moderator analyses indicated significant differences in program effectiveness between instructor types, with teachers found to be most effective ([Formula: see text]= -0.08, p=0.02). The findings provide further support for the use of interactive school-based programs to prevent cannabis use among middle school students in North America.