Three hundred thirty-two eligible patients with advanced (Ann Arbor stage III or IV) non-Hodgkin's lymphoma of aggressive histologic subtype (Rappaport classification diffuse histiocytic [DH], diffuse poorly differentiated lymphocytic [DPDL], diffuse mixed [DM], or diffuse undifferentiated [DU]) were randomly assigned to receive induction chemotherapy with one of three intensive regimens in a clinical trial conducted by the Eastern Cooperative Oncology Group (ECOG) between 1978 and 1983. Chemotherapy regimens consisted of cyclophosphamide, vincristine, prednisone, and doxorubicin (Adriamycin; Adria Laboratories, Columbus, OH) (COPA) administered in 3-week cycles; cyclophosphamide plus doxorubicin plus prednisone beginning day 1, with vincristine plus bleomycin day 15 of each 3-week cycle (COPA + Bleo); or cyclophosphamide plus doxorubicin plus procarbazine beginning day 1, and bleomycin plus vincristine plus prednisone beginning day 15 of each 4-week cycle (CAP-BOP). The median patient follow-up from study entry for patients still alive is 5 years. The three regimens were not significantly different with respect to complete response (CR) rates (43% to 46%), time to progression of malignant disease (median, 1.0 to 1.7 years), or survival (5-year survival, 34% to 45%), although duration of complete remission appeared to be shorter in patients receiving COPA (P = .03). COPA + Bleo and CAP-BOP were significantly more toxic than the COPA regimen. This study did not demonstrate any substantial therapeutic advantage associated with the addition of a fifth or sixth chemotherapy drug, or with treatment administered on a more frequent administration schedule, compared with the COPA regimen in this population of patients with advanced diffuse non-Hodgkin's lymphoma. The relatively small proportion of long-term disease-free survivors treated with COPA underscores the need for prospective clinical trials of new and more effective treatments for patients with these potentially curable tumors.