Transcatheter arterial chemoembolization with lipiodol (TACE) is widely performed in patients with hepatocellular carcinoma (HCC) unsuitable for curative treatment. It has recently been recommended for patients with 2 or 3 tumors >3 cm or ≥4 tumors in a treatment algorithm proposed by Japanese guidelines. However, the best indication and appropriateness of the algorithm for TACE are still unclear. In 4966 HCC patients who underwent TACE, survival was evaluated based on tumor number, size and liver function; and the adequacy of the algorithm for TACE was validated. Exclusion criteria were: vascular invasion, extrahepatic metastasis, and prior treatment. The mean follow up period was 1.6 years. The overall median and 5-year survivals were 3.3 years and 34%, respectively. Multivariate analysis revealed that Child-Pugh class, tumor number, size, alpha-fetoprotein, and des-gamma carboxy-prothrombin were independent predictors. The survival rate decreased as the tumor number (p=0.0001) and size increased (p=0.04 to p=0.0001) in all but one subgroup in both Child-Pugh-A and -B. The stratification of these patients to four treatments in the algorithm showed potential ability to discriminate survivals of the resection and ablation (non-TACE) groups from those of the TACE group in Child-Pugh-B and partially in A. TACE showed higher survival rates in patients with fewer tumor numbers, smaller tumor size, and better liver function. The treatment algorithm proposed by the Japanese guidelines might be appropriate to discriminate the survival of patients with non-TACE from TACE therapy.