A randomized trial of hepatic arterial chemoembolization was conducted in 42 patients with unresectable hepatocellular carcinoma. These patients represented 41% of patients with hepatocellular carcinoma seen during the inclusion period. In the remaining 59%, 9% had resectable tumours and 50% had unresectable tumours with contraindication for chemoembolization. Patients received either repeated chemoembolization with gelfoam powder and doxorubicin (group 1) or symptomatic treatment (group 2). There was no difference in age, prevalence of cirrhosis or staging according to Okuda between the two groups of patients. A complete tumour response (assessed by arteriography, ultrasonography and serum alphafetoprotein) was observed in four patients, and a partial response in three other patients from group 1. Actuarial survival rates were 33 and 24% in group 1 and 52 and 31% in group 2 at 6 and 12 months, respectively (differences were not significant--logrank test). With the treatment used in our study, chemoembolization did not prolong the survival time of patients with unresectable hepatocellular carcinoma. There were, however, some complete or partial responses. The high spontaneous 1-year survival rate of untreated patients was probably due to the exclusion of the most severely ill patients. Our results do not support the use of this method of chemoembolization in the treatment of hepatocellular carcinoma.