A total of 84 patients with hepatocellular carcinoma and cirrhosis were analyzed retrospectively to investigate prognostic factors. All patients received transarterial oily chemoembolization aa the only anticancer therapy. The follow-up range was 1 to 39 mo (median, 9.5 mo). The overall actuarial survival rates at 12, 24 and 30 mo were 62%, 31% and 249, respectively. According to univariate analysis, variables significantly associated with survival were age, Child-Pugh grade, total serum bilirubin, Okuda stage, tumor size, degree of labeling of the tumor with Lipiodol, gelatin foam use, changes with treatment in tumor size and changes with treatment in a-fetoprotein concentration. T w o multivariate analyses were performed. When pretreatment and treatment variables were considered, parameters with independent prognostic value were age, Child-Pugh grade, total serum bilirubin, tumor size and degree of Lipiodol labeling of the tumor. When follow-up variables were also considered, we (a) confirmed the prognostic significance of all these parameters (age, Child-Pugh grade, total serum bilirubin, tumor size) and (b) found the independent prognostic value of the change in tumor size (or change in a-fetoprotein concentration). Both models yielded different risk coefficients for each class of each variable. Two simple prognostic indexes, based on these coefficients, are proposed: an "initial" index (including pretreatment and treatment variables) and a "follow-up" index (also including follow-up variables). According to the two indexes, the patients were classified into three groups with different prognoses: good (93% and 100% actuarial survival at 1 yr for the initial and follow-up indexes, respectively), intermediate (65% and 53%, respectively) and poor (27% for both indexes).