A total of 248 analyzable patients with Stages III-IV ovarian epithelial cancer (114 with and 134 without prior chemotherapy) were randomized to one of four cisplatin (DDP)-hexamethylmelamine (HMM) regimens. In each, HMM, 200 mg/m2 was given orally daily on days 8-21 of each 21-day cycle. DDP was given i.v. on Day 1 at a dose of 37.5 mg/m2 (regimens A and B) or 75 mg/m2 (regimens C and D). In addition, since pyridoxine administration has been reported to reduce the neurotoxicity of HMM, that agent was given at a dose of 300 mg/m2 orally on Days 1-21 in regimens B and D. Randomization was stratified for performance status (0-1, 2-3) and largest tumor diameter at entry (greater than 2- less than or equal to 10 cm, greater than 10 cm) for previously untreated patients, and for performance status and time from initial diagnosis to entry on study (less than or equal to 1 year, greater than 1 year) for previously treated patients. The overall response rate (PR + CR) was 54%, with 25% of patients achieving a complete response. The 61% response rate with the higher dose DDP regimens was significantly greater than the 47% response rate with the lower dose regimens (p = 0.031). Multivariate analysis identified higher DDP dose, age less than 60 years, no prior chemotherapy, small tumor bulk and favorable tumor grade as significant prognosticators for response. The overall median response duration was 8.3 months (range 1-70 months). Prior chemotherapy, pyridoxine administration, recent diagnosis, and large tumor size were identified by multivariate analysis as factors adversely affecting response duration. Patients treated with the higher dose DDP regimens had more severe nausea, vomiting, and neurotoxicity. This study demonstrates that the combination of DDP + HMM is an effective regimen for advanced ovarian carcinoma that yields response rates comparable to other more complex regimens, and that there is a dose-response relationship for DDP in ovarian cancer. Although pyridoxine administration significantly reduced neurotoxicity, its adverse effect on response duration suggests that the agent should not be administered with DDP or HMM. The mechanism by which pyridoxine may unfavorably affect response duration deserves further investigation.