Eighty-five heroin addicts who were unwilling to receive methadone maintenance or enter therapeutic communities were assessed, single-blind, for the lowest sublingual dose of buprenorphine that blocked heroin craving (8.0 mg max). All doses were administered daily under observation. After maintenance for 4 to 12 weeks, abstinent subjects (confirmed by urine drug screens) entered a double-blind discontinuation trial and were randomly assigned to receive dose reductions (10% twice weekly for 5 weeks to zero dose, then placebo for 2 weeks) or a stable dose for 7 weeks. Subjects were terminated from discontinuation if heroin was used or they had increased craving/symptoms. Subjects completed the trial if they did not use heroin and had no increase in craving/symptoms. A wide dose range (1.5-8.0 mg/day) was effective in reducing heroin craving and use. Of 73 subjects who received buprenorphine for 4 to 52 weeks, 40 had no prior treatment, despite high levels (mean $/day heroin = 70.5 +/- 94.7) and many years (mean years = 10.7 +/- 8.6) of dependence. Subjects who received dose reductions developed abstinence symptoms, low energy most commonly, associated with drug-seeking behavior. Discontinuation trial outcome (n = 51) shows a highly significant difference between 29 subjects who received dose reductions (28 terminated, 1 completed) and 22 subjects who received no dose reductions (3 terminated; 19 completed) (chi-square = 36.08; p less than .00001). The findings suggest that buprenorphine could be an important medication for reducing demand for heroin by many heroin addicts who remain outside the present health-care system.(ABSTRACT TRUNCATED AT 250 WORDS)