Patients with epithelial ovarian cancer (EOC) referred to our institution are stratified into risk groups based on their stage, grade and presence of residual cancer, with a specific treatment policy for each group. One-hundred and thirty-one patients with no visible residual tumor following primary surgery and either stage I, grade 3; stage II, grade 3; or stage III, any grade EOC were treated between November 1983 and the end of December 1991. Regimen A (cisplatin 75 mgm-2 and cyclophosphamide 600 mgm-2 intravenously every 4 weeks for 6 cycles with abdominopelvic irradiation between cycles 3 and 4) was used until April 1989 and was then replaced with Regimen B (cisplatin 75 mgm-2 intravenously every 3 weeks for 6 cycles). The 5-year actuarial overall and failure-free survivals were 78% and 64% respectively. Multivariate analysis identified increasing stage and treatment with Regimen B as independent adverse prognostic factors for failure-free survival. The importance of treatment regimen reached statistical significance for the stage I patients (P = 0.04) but not stage II (P = 0.11) or stage III (P = 0.79). It is possible to undertreat EOC as shown by the inferior results achieved with Regimen B (single agent cisplatin) compared to Regimen A (cisplatin-cyclophosphamide, irradiation). This effect of treatment regimen was particularly important for the lower-stage patients. Our postulate is that treatment resistant clones are less regularly present in lower-stage patients, and that a certain minimum amount of treatment is required to eliminate all the sensitive cancer.