To evaluate the activity of long-term, single-agent oral etoposide against advanced breast cancer, this study employed etoposide 50 mg/day and 100 mg/day (given over 14 days) in previously treated and chemotherapy-naive patients with histologically confirmed, recurrent or metastatic breast cancer. Of 38 patients, 24 had had chemotherapy, 34 had prior radiotherapy, and 31 had received hormone therapy. Etoposide courses in both treatment groups were repeated every 4 weeks for at least two courses; delays were instituted when patients' total leukocyte nadir fell to or below 3.0 x 10(9)/l. Dose reductions were made in the 100-mg group (to 50 mg/day) if World Health Organization leukopenia grade 3 or higher was present. Plasma pharmacokinetic profiles were measured in selected patients to assess inter- and intrapatient variability in etoposide's oral bioavailability. No complete responses were achieved among the 36 patients evaluable for response, but eight patients had a partial response. Responses were more frequent at the 100-mg dose and in previously untreated patients (seven of eight partial responders had not had previous chemotherapy). Median duration of response was 16 weeks (range, 7 to 46). Myelosuppression (variable and unpredictable) and alopecia (universal) were the notable toxicities. Pharmacokinetic analyses of oral bioavailability revealed significant interpatient variability, but much less intrapatient variability when successive etoposide courses in individual patients were evaluated. Despite the relatively small number of patients in this study, the responses achieved by previously untreated patients suggest etoposide's value against breast cancer. Further trials should use pharmacokinetic studies to assess bioavailability as well as to help define 'target' etoposide doses, based on plasma etoposide levels, during treatment.