A randomized phase II study of intraperitoneal (ip) mitoxantrone or floxuridine (FUDR) was performed for the treatment of minimal residual epithelial ovarian cancer found at second-look laparotomy after initial platinum-based chemotherapy. Entry was to take place within 30 days of reassessment laparotomies, with documentation of peritoneal metastases either microscopic or gross with cytoreduction to less than or equal to 1 cm in largest diameter. Patients were stratified by the site of the largest disease present (microscopic to 0.5 cm maximum diameter versus greater than 0.5 to 1 cm maximum diameter), by time of registration (< 14 days versus up to 30), and by serum CA-125 (< or = 35 versus >35 units/ml) prior to randomization to either ip mitoxantrone 10 mg/m2 every 2 weeks X 9 or ip floxuridine (FUDR) 3 g (total dose)/ day X 3 days every 3 weeks X 6 cycles. Implantable ip systems and 1.5-2 liters of normal saline were used to deliver the drugs of 83 patients registered between December 1988 and January 1994; there were 6 pathology exclusions and 9 surgical exclusions, and 1 nonevaluable patient for a total of 39 evaluable on mitoxantrone and 28 on FUDR being evaluable. FUDR is the choice for further study because of a progression-free survival exceeding 15% at 1 year over mitoxantrone and a median overall survival of 38 months. It should be emphasized again that the goal of a randomized phase II selection design is to select a winner for phase III testing should there be a substantial difference between the treatments with respect to the primary endpoint. Comparative conclusions between the treatment arms should not be attempted due to the inherently much smaller sample sizes. This should reemphasize the limitations in a comparison of efficacy; however, the toxicologic differences still emerge quite clearly.