Between 1974 and 1977, 652 patients with non-Hodgkin's lymphoma without prior chemotherapy were randomized to 1 of 3 combination chemotherapy programs designed to induce complete remission (CR): COP-bleomycin (180 patients), CHOP-bleomycin (232 patients) or CHOP plus immunotherapy with Bacillus Calmette Guerin (BCG) (240 patients). With mature follow-up, the major effect of BCG immunotherapy was observed in patients with large cell lymphomas (diffuse or nodular "histiocytic") and not in other common lymphoma subtypes. CR rate for 65 patients with large cell lymphoma treated with CHOP-BCG was 68% compared to 48% in 61 patients treated with CHOP-bleomycin (P = 0.02) (two-tailed test) or 44% for 45 patients treated with COP-bleomycin (P = 0.02). CR duration for both CHOP-based regimens was similar and superior to that produced by COP-bleomycin (P = 0.03). Survival of patients with large cell lymphoma treated with CHOP-BCG was better than that observed with CHOP-bleomycin (P = 0.02) or COP-Bleomycin (P = 0.002). Although the explanation for the favorable effect of BCG remains unclear, further clinical trials to evaluate the combination of chemotherapy and other "biologic response modifiers" is warranted for patients with lymphoma.