The relationship between treatment exposure, drug use, psychosocial variables and non-fatal heroin overdose was examined among a cohort of 495 heroin users, re-interviewed at 12 months. The 12-month overdose rate declined from 24% to 12%, and the proportion administered naloxone declined from 15% to 7%. There were significant reductions in overdose among those who entered maintenance therapies (22% to 4%) and residential rehabilitation (33% vs. 19%) at baseline, but not among those who entered detoxification or were not entering treatment. The total number of treatment days received over the follow-up period was associated independently with a reduced risk of overdose. Each extra treatment day was associated with a 1% reduction in risk of overdose over the follow-up period. By contrast, more treatment episodes were associated with an increased risk of overdose (OR 1.62). Other independent predictors of overdose over follow-up were more extensive polydrug use (OR 1.40), and having overdosed in the year preceding the study (OR 7.87).