Three combination chemotherapy regimens each with or without the methanol-extracted residue of bacillus Calmette-Guérin (BCG) (MER) were compared for efficacy. After stratification for disease-free interval and dominant sites of disease, patients were randomized to either CMF (cyclophosphamide [CYC], 100 mg/m2 orally, days 1 through 14; methotrexate [MTX], 40 mg intravenously [IV], days 1 and 8; 5-fluorouracil [5-FU], 500 mg/m2 IV, days 1 and 8), or CAF (CYC, 100 mg/m2 orally, days 1 through 14; doxorubicin [DOX], 25 mg/m2 IV, days 1 and 8; 5-FU, 500 mg/m2 IV, days 1 and 8), or CAFVP (CAF as above plus vincristine [VCR], 1.0 mg/m2 IV, days 1 and 8; and prednisone [PRED], 40 mg/m2 orally, days 1 through 14). Nonspecific immunotherapy with MER was administered in five sites at 100 micrograms or at the lowest tenfold dilution that produced a 1-cm indurated lesion. A total of 432 patients were entered, but 37 were disqualified, leaving 395 evaluable for treatment results and toxicities. One hundred thirty-five evaluable patients were randomized to chemoimmunotherapy until October 28, 1978. One hundred twenty-six evaluable patients were randomized to chemotherapy alone in the same time period. For the entire study, a total of 260 evaluable patients were randomized to chemotherapy. Chemoimmunotherapy patients were compared with the initial 126 chemotherapy patients. Chemotherapy regimens were compared among all 260 patients. Patient characteristics were similar between regimens and between chemotherapy and chemoimmunotherapy treatment groups. For patients on chemotherapy plus MER, there was no significant differences between the regimens for response frequencies: 43%, 41%, and 32%, respectively for CMF, CAF, and CAFVP. The comparable chemotherapy alone group had 36%, 58%, and 63% response, respectively. The response rates, adjusted for chemotherapy regimen, were 52% and 38% (P = .02) for chemotherapy and chemoimmunotherapy, respectively. MER was associated with painful ulcers and fevers. Thus, MER produced toxicity without response or survival benefit and further randomization after October 28, 1978 was to chemotherapy alone. For 260 evaluable patients on chemotherapy alone, the complete (CR) and partial responses (PR) were 37%, 55%, and 58%, respectively for CMF, CAF, and CAFVP. These response rates for CAF and CAFVP were significantly better than CMF (P = .01 and P less than .01, respectively). These comparisons were consistent within subgroupings such as dominant sites of disease.(ABSTRACT TRUNCATED AT 400 WORDS)