Aflatoxins have long been suspected to be human hepatic carcinogens but no direct study was feasible until assays to measure individual aflatoxin exposure became available. We have used assays for urinary aflatoxin B1, its metabolites AFP1 and AFM1, and DNA-adducts (AFB1-N7-Gua) to assess the relation between aflatoxin exposure and liver cancer, as part of an ongoing prospective study of 18,244 middle-aged men in Shanghai, People's Republic of China. After 35,299 person-years of follow-up, 22 cases of liver cancer had been identified. For each case, 5 or 10 controls were randomly selected from cohort members without liver cancer on the date the disorder was diagnosed in the case and matched to within 1 year for age, within 1 month for sample collection, and for neighbourhood of residence. Subjects with liver cancer were more likely than were controls to have detectable concentrations of any of the aflatoxin metabolites (relative risk 2.4, 95% confidence interval 1.0-5.9). The highest relative risk was for aflatoxin P1 (6.2, 1.8-21.5). In an analysis adjusting for the effects of hepatitis B surface antigen seropositivity, level of education, cigarette smoking, and alcohol consumption, the relative risk for the presence of aflatoxin metabolites was 3.8 (1.2-12.2). There was a strong interaction between serological markers of chronic hepatitis B infection and aflatoxin exposure in liver-cancer risk. Reduction of aflatoxin exposure may be a useful intermediate goal in prevention of liver cancer, since the benefits of wide-scale hepatitis B vaccination will not be apparent for many years.