Naltrexone and acamprosate reduce relapse in alcohol dependence. They have not yet been compared in a published trial. The aim of this study was to compare the efficacy of these compounds in conditions similar to those in routine clinical practice. Random allocation to a year of treatment with naltrexone (50 mg/day) or acamprosate (1665-1998 mg/day) was made in 157 recently detoxified alcohol-dependent men with moderate dependence (evaluated using the Addictions Severity Index and Severity of Alcohol Dependence Scale). All were patients whom a member of the family would accompany regularly to appointments. Alcohol consumption, craving and adverse events were recorded weekly for the first 3 months, and then bi-weekly, by the treating psychiatrist who was not blinded. At 3-monthly intervals, investigators who were blinded to the treatment documented patients' alcohol consumption based on patients' accounts, information given by the psychiatrists when necessary, and reports from patients' families. Serum gamma-glutamyltransferase (GGT) was also measured. Efforts were made to sustain the blindness of the investigators. The same investigator did not assess the same patient twice. The integrity of the blindness was not checked. There was no difference between treatments in mean time to first drink (naltrexone 44 days, acamprosate 39 days) but the time to first relapse (five or more drinks in a day) was 63 days (naltrexone) versus 42 days (acamprosate) (P = 0.02). At the end of 1 year, 41% receiving naltrexone and 17% receiving acamprosate had not relapsed (P = 0.0009). The cumulative number of days of abstinence was significantly greater, and the number of drinks consumed at one time and severity of craving were significantly less, in the naltrexone group compared to the acamprosate group, as was the percentage of heavy drinking days (P = 0.038). More patients in the acamprosate than the naltrexone group were commenced on disulfiram during the study. Naltrexone patients attended significantly more group therapy sessions, though this could not explain their better outcome. There were non-significant trends for the naltrexone group to comply better with medication, to stay in the study longer, and to show greater improvement over baseline in serum GGT.