PURPOSE: We analyzed the long-term results of a Childrens Cancer Group (CCG) randomized study comparing cyclophosphamide, vincristine, methotrexate, and prednisone (COMP) versus LSA2L2 as treatment for childhood non-Hodgkin's lymphoma. The initial results were previously reported (N Engl J Med 308:559, 1983). PATIENTS AND METHODS: A total of 429 patients are reported here, 68 with localized disease and 361 with disseminated disease. The distribution of disseminated-disease patients by histologic type was 164 lymphoblastic, 60 large-cell, and 137 undifferentiated lymphomas. Median follow-up duration of surviving patients is 8 years. RESULTS: Event-free survival (EFS) of patients with localized disease was 84% at 5 years. No differences were seen between the two treatment regimens. Results for patients with disseminated disease was dependent on histologic subtype: patients with lymphoblastic lymphoma did better when treated with LSA2L2 (5-year EFS of 64% v 35% for COMP); COMP produced better results for patients with undifferentiated lymphoma (5-year EFS of 50% v 29% for LSA2L2). Results for large-cell lymphoma patients were similar (5-year EFS of 52% for COMP v 43% for LSA2L2). Five percent of patients died of treatment-related complications while on therapy (primarily infections). Only four deaths without progression have been observed off-therapy (two from restrictive lung disease, one from an acute asthma attack, one from colon cancer). Patient survival rates after recurrence were poor. CONCLUSION: Treatment success can be expected in 84% of pediatric patients with localized non-Hodgkin's lymphoma. For patients with disseminated disease, treatment success can be expected in 64% of those with lymphoblastic and 50% of those with undifferentiated or large-cell disease. To date, late adverse events are rare.