Results from small randomised trials on tamoxifen in the treatment of hepatocellular carcinoma (HCC) are conflicting. We studied whether the addition of tamoxifen to best supportive care prolongs survival of patients with HCC. Patients with any stage of HCC were eligible, irrespective of locoregional treatment. Randomisation was centralised, with a minimisation procedure accounting for centre, evidence of disease, and time from diagnosis. Patients were randomly allocated best supportive care alone or in addition to tamoxifen. Tamoxifen was given orally, 40 mg per day, from randomisation until death. 496 patients from 30 institutions were randomly allocated treatment from January, 1995, to January, 1997. Information was available for 477 patients. By Sept 15, 1997, 119 (50%) of 240 and 130 (55%) of 237 patients had died in the control and tamoxifen arms, respectively. Median survival was 16 months and 15 months (p=0.54), respectively. No differences were found within subgroups defined by prognostic variables. Relative hazard of death for patients receiving tamoxifen was 1.07 (95% CI 0.83-1.39). Our findings show that tamoxifen is not effective in prolonging survival of patients with HCC.