The actual utility of a new classification system of acute myeloid leukemia (AML) recently introduced by the World Health Organization (WHO) has not been thoroughly investigated yet. In this study, we evaluated long-term outcomes of unselected AML patients categorized according to the new WHO classification. Between 1990 and 2002, 109 adult AML cases were referred to our hospital. For the entire population, the median survival duration was 1.2 yr with a 5-yr survival rate of 31%. AML with recurrent genetic abnormalities accounted for 26%, AML with multilineage dysplasia for 29%, therapy-related AML for 13%, and AML not otherwise categorized for 32% of classifiable cases. Among the four groups, a significant difference was observed in terms of overall survival (P < 0.0001). Univariate analysis showed that six variables affected survival: cytogenetic risk, age, multilineage dysplasia, prior chemo/radiotherapy, type of treatment (intensive or palliative), and transplantation. However, in multivariate analysis no adverse prognostic impact of multilineage dysplasia and prior chemo/radiotherapy was detected (P = 0.4979 and 0.8702), whereas cytogenetic risk and patient age maintained their prognostic value (P = 0.0005 and 0.0100). These results indicate that outcomes for AML patients appear to be distinguished on the basis of the WHO classification, but the prognostic significance of multilineage dysplasia and prior therapy is lost after adjusting for cytogenetic risk and age. Our findings suggest that the WHO classification may be strengthened by greater emphasis on genetic/cytogenetic information.