A phase I study was performed to determine the maximum tolerated doses of intravenous etoposide and paclitaxel for women with previously treated persistent or recurrent ovarian cancer. Starting doses were paclitaxel 135 mg/m2 during 24 hours and etoposide 50 mg/m2/day for 3 consecutive days. The study was designed to escalate first the dose of etoposide, and then the dose of paclitaxel, in successive cohorts of patients. In an attempt to determine whether toxicity was affected by sequence of the drugs, the order of administration of the two drugs was reversed on alternate cycles. The starting doses of paclitaxel (135 mg/m2/24 hours) and etoposide (50 mg/m2/day x 3) caused severe neutropenia even with the addition of granulocyte colony-stimulating factor, and the trial was amended to administer the paclitaxel during 3 hours. However, this also proved too myelosuppressive without growth factor support. Twenty-one women were treated. A complete response was observed in one of nine patients with measurable disease, and a major decrease in CA-125 was noted in two patients who did not have measurable disease. Because of the severe myelosuppression observed in most patients, dose reduction was often required after the first cycle. The power to detect sequence-dependent variation in toxicity was minimal; however, no large differences were observed. A combination of the usual doses of these drugs will be difficult to administer in patients who have received previous chemotherapy for ovarian cancer.