A randomized clinical trial of combination chemotherapy for patients who relapsed following primary radiation therapy for Hodgkin's disease was conducted from 1975 to 1981 by the Cancer and Leukemia Group B (CALGB). One hundred thirteen patients were prospectively randomized to receive 12 cycles of either CVPP (CCNU, vinblastine, procarbazine, and prednisone), ABOS (bleomycin, vincristine [Oncovin; Lilly, Indianapolis], doxorubicin [Adriamycin, Adria Laboratories, Columbus, Ohio], and streptozotocin), or alternating cycles of CVPP and ABOS. The median length of observation for patients in this report is 4 years. Toxicities of the three treatment programs were primarily hematologic. Frequencies of complete response were 72% for CVPP, 70% for ABOS, and 82% for CVPP/ABOS (P = .37). Females and patients who had nodular sclerosing disease at initial diagnosis had significantly higher complete response rates. The 5-year disease-free survival for the complete responders was 55%; the 5-year overall survival was 60%. There were no significant differences among the treatments on disease-free survival (P = .78) or overall survival (P = .18). Age under 40 years was the only significant positive prognostic factor for disease-free survival (P = .095) and overall survival (P = .003). This study demonstrates no statistically significant advantage for alternating cycles of combination chemotherapy in affecting complete response frequency, disease-free survival, or overall survival as compared with therapy with CVPP or ABOS alone. However, the power to detect differences in these outcome parameters is somewhat limited by the sample sizes.(ABSTRACT TRUNCATED AT 250 WORDS)