One hundred ninety-six children with acute lymphocytic leukemia were entered into a randomized study of maintenance chemotherapy with either intermittent chemotherapy and immunotherapy with bacillus Calmette-Guerin (BCG) or only intermittent chemotherapy. On admission t o the study, patients were stratified into three prognostic categories on the basis of clinical and laboratory features at presentation. Patients considered t o have a favorable prognosis (PF) were induced with vincristine and prednisone; those with an unfavorable prognosis (PU) received combination chemotherapy consisting of either Ara-C, cyclophosphamide and asparaginase, or vincristine, prednisolone, and daunomycin followed by Ara-C and asparaginase; those with an average prognosis (PA) received vincristine and prednisolone followed by a single course of Ara-C and asparaginase. All patients received central nervous system prophylaxis with 2,400 rad cranial irradiation and four injections of intrathecal methotrexate. One hundred seventy-seven patients (90%) achieved complete remission, and 165 were randomized. Those randomized to Group A (83) received maintenance chemotherapy with six-week courses of 6-mercaptopurine, weekly oral methotrexate, and monthly vincristine. Each six-week course was followed by a two-week interval of n o chemotherapy, and treatment was continued for 36 months. Patients randomized t o Group B (82) received the same maintenance chemotherapy, but during the two-week interval without chemotherapy, they were given BCG inoculation. With a median follow-up of 110 weeks, no significant difference in duration of remission, survival, or CNS relapse was found between Groups A and B in the total patient population or within each prognostic category. In patients classified as PU, a significantly lower proportion (P < 0.02) remained in complete remission. Within this group, patients with a white cell count > 100 X 109/L had a significantly shorter duration of remission (P < 0.05). Patients classified as P F showed a
354Ekert et a1 high incidence of late relapses and fared no better than those classified as PA. It is possible that the 2-drug induction regimen used in PF patients may have contributed to the relatively high late-relapse rate.