The mortality within a cohort of 115 street heroin addicts was studied for 5-8 years using the Kaplan-Meier survival estimate technique. This differed markedly from the relatively low mortality of 166 comparable heroin addicts given methadone maintenance treatment (MT). The street addicts' mortality rate was 63 times that expected, compared with official statistics for a group of this age and sex distribution. When 53 patients in MT were involuntarily expelled from treatment, due to violation of programme rules, they returned to the high mortality of street addicts (55 times that expected). A group of 34 rehabilitated patients who left MT with medical consent retained the low mortality of MT patients (their mortality rate was 4 times that expected). Despite this great improvement in survival, even patients in MT showed a moderately elevated mortality (8 times that expected), mainly due to diseases acquired before entering the treatment programme. It is concluded that MT exerts a major improvement in the survival of heroin addicts.