The therapeutic effect of transcatheter arterial chemoembolization (TAE) performed on 31 patients with primary liver cancer was evaluated using the following procedures: (1) the alpha-fetoprotein (AFP) reduction rates and prognoses were analyzed according to the tumor reduction rates (TR), and (2) the AFP reduction rates and prognoses were also analyzed according to the tumor necrosis rates (TN) estimated by regarding every region with Lipiodol retention as being necrotic. The following results were obtained. The AFP level was 400 ng/ml or higher in 15 patients (48%). Their AFP reduction rates were as favorably high as 65.4%-99.8% (mean, 88.1%), and the AFP level was normalized in 3 patients. The cumulative survival rates after the initial treatment were relatively high, i.e., 78.4% in the 1st year, 58.1% in the 2nd year, and 38.7% in the 3rd year. These results suggested the effectiveness of the TAE treatment undertaken in this study. Regarding the TR, the tumor was reduced in size by 50% or more in only 5 patients (16%), and most patients had a TR of less than 25%. On the other hand, the majority, 25 patients (81%), had a TN ranging between 50% and less than 100%, including 7 who had a TN ranging between 50% and less than 90% and 18 who had a TN ranging between 90% and less than 100%. There was no significant correlation between the AFP reduction rate and the TN or TR. Regarding evaluation of the cumulative survival rates by TR and TN, the 1-year survival rate was lower in patients having a TR of less than 25% than in those having a TR of 25% or more. Patients having a TN of less than 50% showed a poor outcome as compared with those having a TN of 50% or more. Although the TR was found to be less than 50% in a majority of the patients when the therapeutic effect of TAE on the liver cancer was evaluated according to the TR, many of these patients showed a good outcome. Thus, the conventional efficacy evaluation, in which a tumor reduction of 50% or more is considered to be effective, should be reconsidered. On the other hand, the TN was found to be 50% or more in most of the patients, suggesting the necessity of a more detailed classification of TN. In relation to the survival rate, patients having a TN of less than 50% showed a poor outcome.