After disease recurrence or dissemination, patients who had been treated previously with radiation with or without surgery for cancer of the head and neck were given either cisplatin (16 patients), cisplatin/etoposide (15 patients), or cisplatin/etoposide/5-fluorouracil (5-FU) (19 patients) in an ambulatory care clinic. Intravenous (i.v.) cisplatin 25 mg/m2 was given weekly, while etoposide was given i.v. (80 mg/m2) on day 1 and orally (160 mg/m2) on day 2 of every odd-numbered week. In the three-drug regimen, 5-FU 500 mg/m2 i.v. was given every even-numbered week. Patients in all three groups received daily oral prednisone to decrease myelosuppression and oral co-trimoxazole and ketoconazole to prevent infection. The supportive drugs were given to all groups to keep these variables constant. As expected, myelosuppression did not occur in the cisplatin group, while the rates of severe neutropenia (less than 1.0 x 10(9)/L) in the two- and three-drug groups were 26% and 74%, respectively. The incidence of infection requiring hospitalization was low (2.5%). The response rate (complete plus partial) was lowest in the cisplatin group (6%) and higher in the cisplatin/etoposide (47%) and cisplatin/etoposide/5-FU (53%) groups. Because of the low response rate and the short time to progression (5 weeks) in the cisplatin group, 9 of these 16 patients were treated subsequently with cisplatin/etoposide. Time to progression and response duration were similar in the cisplatin/etoposide and cisplatin/etoposide/5-FU groups--12 and 14 weeks, and 12 and 9 weeks, respectively. Median survival times of the cisplatin and cisplatin/etoposide/5-FU groups were 36 and 34 weeks, respectively.(ABSTRACT TRUNCATED AT 250 WORDS)