We evaluated factors affecting long-term survival after transcatheter arterial chemoembolization (TACE) for hepatocellular carcinoma (HCC) complicating cirrhosis. One hundred eighty-two patients with Child's class A or B cirrhosis and an HCC, not amenable to surgery or percutaneous ethanol injection, underwent 346 TACEs (mean 1.9) with epirubicin, iodized oil, and gelatin sponge. Many prognostic factors were subjected to univariate analysis and thereafter, when significant, to the Cox's hazard proportional model. Finally, the significant indices in the Cox's model were used to estimate the accuracy of the probability of death with computation of the area under the receiving operative characteristic (ROC) curve. The cumulative survival rates at 1, 2, 3, and 5 years were 0.83, 0.52, 0.40, and 0.16, respectively. According to Cox's model, the factors associated with significantly worse survival were the presence of ascites (p = 0.0027), elevated bilirubin levels (p = 0.0163), elevated alpha-fetoprotein (alphaFP) values (p = 0.0067), a tumor greater than 5 cm in diameter (p = 0.0001), and the absence of a tumor capsule-like rim (p = 0.0278). According to these parameters, the accuracy of the probability of death estimated with ROC analysis was 0.63. Minor and major complications occurred in 82 patients (45%) and caused death in 2 patients. Long-term prognosis after TACE for HCCs in patients with Child's class A or B cirrhosis depends on the presence of ascites, the bilirubin level, the alphaFP value, the diameter of the tumor, and the presence of a tumor capsule-like rim. However, when considered altogether, these variables are poor predictors to evaluate survival, and other factors should be investigated to identify subjects more responsive to TACE. Complications occur in a high percentage of patients, but they do not affect long-term prognosis.