One hundred sixty-three patients with advanced non-Hodgkin's lymphoma including adult T cell leukemia/lymphoma (ATL) were treated from 1981 to 1983 with VEPA (vincristine, cyclophosphamide, prednisolone, and doxorubicin) or VEPA-M (VEPA plus methotrexate) in randomized fashion after stratification by surface marker. The complete response (CR) rate and the 4-year survival rate of patients treated with VEPA-M was 62.2% and 36.9%, respectively, while for those treated with VEPA the rates were 51.9% and 26.6, respectively. The difference was not statistically significant, but pretreatment characteristics predictive for response and survival were interesting. Three factors, leukemic change, poor performance status (PS), and T cell marker, were negatively associated with both CR and survival rates, and high-grade pathology was adversely associated with survival rate in a multivariate analysis. These prognostic factors are somewhat different from those in Western lymphomas. This may be reflection of major differences in patients' characteristics between Japanese and Western lymphomas: in this study, there was a high incidence of T cell lymphoma/leukemia (50%) including ATL (33%), leukemic manifestation (34%), poor PS (34%), and a low incidence of follicular lymphoma (9%). The statistically significant three factors for both CR and survival rates were used to construct a model containing eight categories of patients at increasing risk for poor response and shortened survival. These categories were divided into four groups, with respective CR and 4-year survival rates of 91% and 73%, 67% and 35%, 27% and 7%, and 10% and 5%. Ninety-three patients in whom CR was induced by VEPA or VEPA-M therapy were evaluated for prognostic factors predictive for disease-free survival. A shorter period (less than 28 days) required to achieve CR, a clinical diagnosis of ATL, and a lower hemoglobin level were found to affect disease-free survival adversely. These results have important implications for both the design of prospective randomized therapeutic trials and the determination of optimal therapy for individual patients.