Objectives: To investigate the effectiveness of screening for liver cancer in reducing mortality from the disease in a high-risk population in China. Setting: A randomised controlled trial was carried out among men aged 30-69 who were chronic carriers of hepatitis-B virus (HBsAg positive) during the period 1989-1995 in Qidong county, Jiangsu Province, China. Methods: 5581 HBsAg carriers were identified by population screening and randomly assigned to a screening group (group A, 3712 men), and controls (group B, 1869 men). Screening was planned to be six monthly alpha-fetoprotein (AFP) assays, with follow-up of subjects having an abnormal (≥20 μg/l) test. All subjects were followed up for liver cancer and/or death until 31 December 1995. Results: The overall sensitivity and specificity of the programme was 55.3% and 86.5%, respectively; in subjects who complied with all scheduled screening tests, the values were 80.0% and 80.9%. Three hundred and seventy-four primary liver cancer (PLC) cases were diagnosed. The percentage of cases in stage I was significantly higher in group A (29.6%) than in group B (6.0%). The one-, three-, and five-year relative survival rates were 23.7%, 7.0%, and 4.0% in group A, and 9.7%, 4.0%, and 4.1% in group B respectively, with no difference in five-year survival between the groups. The mortality rate in the screened group (1138 per 100,000 person-years) was not significantly different from that in the controls (1114 per 100,000). A Poisson regression model showed that the probability of death (rate ratio) in the screening group was 0.83 (95% CI 0.68-1.03) relative to the control group. Conclusions: Screening with AFP resulted in earlier diagnosis of liver cancer, but the gain in lead time did not result in any overall reduction in mortality, because therapy for the patients found by screening was ineffective. Further studies using improved methods of screening, diagnosis and treatment are indicated.