Supervised consumption ensures patients take methadone as prescribed and prevents diversion, however, the influence of supervised consumption on retention is unclear. We examined association between supervised consumption and retention across multiple treatment episodes. Cohort study of persons experiencing ≥1 MMT episodes in primary care (2004-2010), excluding ongoing episodes at the start of follow-up. Length of treatment episodes based on methadone prescriptions, retention classified as no interruption in prescribed methadone lasting >7 days. When a patient did not receive a new prescription within seven days after the end of coverage of a prescription, they were considered to have ceased treatment. We evaluated the relationship between supervised consumption and time to discontinuation of treatment using proportional hazards gamma frailty models to account for recurrent MMT episodes. Age, gender, median daily methadone dose, and comorbidities included as potential confounders. 6393 patients experienced 19,715 treatment episodes over the six-year follow-up period. A J-shaped relationship was observed; having between 20 and 60% of methadone scripts supervised (compared to <20%) associated with reduced time to discontinuation (20-39% HR=0.88, 95% CI 0.81-0.95; 40-59%: HR=0.87, 95% CI 0.81-0.94). Beyond a threshold of 60%, retention reduced (60-79% of scripts: HR=1.28, 95% CI 1.20-1.36;>80% of scripts: HR=3.59, 95% CI 3.38-3.81). Median daily dose between 60 and 120mg/per day, and multiple treatment episodes also associated with longer time to discontinuation of treatment. A J-shaped relationship was observed between supervised consumption and retention in treatment. Additionally, patients experiencing multiple treatment episodes tend to stay in treatment for progressively longer periods of time.