["Addiction, EarlyView. ", "\nAbstract\n\nBackground and aims\nEarly initiation to cannabis use in adolescence is linked to cognitive and mental health risks. Having friends who use cannabis is a strong risk factor for initiation, but the mechanisms underlying this relationship are unclear. This study examined the role of perceived ease of access to cannabis in the association between close friends' cannabis use and cannabis initiation.\n\n\nDesign\nObservational longitudinal investigation using data from the COMPASS study.\n\n\nSetting\nA convenience sample of 11 high schools in Québec, Canada, from 2017 to 2019.\n\n\nParticipants\n1768 students who had not initiated cannabis use before 2019.\n\n\nMeasurements\nHaving ≥1 close friend who uses cannabis and perceived ease of access to cannabis were measured in 2018, while cannabis initiation in the past 12 months was assessed in 2019. Covariates measured in 2017 included sociodemographic factors, substance use history, lifestyle behaviors and school‐related variables. We compared the risk of cannabis initiation across groups defined jointly by friends' cannabis use and perceived access. Generalised estimating equations with bootstrap resampling, inverse probability weighting for attrition and multiple imputation for missing data were used to estimate these associations. Natural effect models were then applied to estimate total, direct and indirect effects with initiation as the outcome and perceived access as the mediator.\n\n\nFindings\nWithin the COMPASS cohort, 14% of cannabis‐naive adolescents in 2018 initiated use in 2019. Using students with no close friends who use cannabis and who perceive access as difficult as the reference group, the risk difference (RD) for having at least one friend who uses cannabis while still perceiving access as difficult was 3.2 [95% confidence interval (CI) = −3.8 to 10.2]. Perceiving access as easy but having no friends who use cannabis yielded an RD of 8.4 (95% CI = 3.0–13.7). Friends' cannabis use combined with easy access acted synergistically, producing an RD of 21.6 (95% CI = 15.5–27.7). According to the natural effect models, perceived ease of access mediated about 39% of the association between having ≥1 close friend who uses cannabis and subsequent initiation, suggesting a possible explanatory pathway.\n\n\nConclusions\nIn Canada, cannabis initiation risk appears to be highest among adolescents with both exposure to close friends who use cannabis and perceived easy access. Perceived access may help explain part of this association. Preventive strategies should address the influence of friends and accessibility through school policies, public education and stricter enforcement of access restrictions to delay adolescent cannabis use.\n\n"]