The feasibility of sequential carboplatin followed by docetaxel-based therapy for untreated ovarian cancer was determined. Patients received four q3w cycles of carboplatin AUC 7, then four q3w cycles of either docetaxel 100 mg m(-2) (day 1) (arm A); docetaxel 75 mg m(-2) (day 8) and gemcitabine 1250 mg m(-2) (days 1,8) (arm B) or docetaxel 25 mg m(-2) and gemcitabine 800 mg m(-2) (both given weekly (days 1,8,15)) (arm C). A total of 44 patients were randomised to each treatment arm. None of the arms demonstrated an eight cycle completion rate (70.5/72.7/45.5% in arms A/B/C, respectively), which was statistically greater than 60% (P=0.102, P=0.056, P=0.982) which was our formal feasibility criteria, although only the completion rate in arm C was clearly worse than this level. The overall response rate (ORR) after carboplatin was 65.7% in 70 evaluable patients. In evaluable patients, ORRs after docetaxel-based cycles were: arm A 84.0% (21 out of 25); arm B 77.3% (17 out of 22); arm C 69.6% (16 out of 23). At follow-up (median 30 months), median progression-free survival times were: arm A 15.5 months (95% CI: 10.5-20.6); arm B 18.1 months (95% CI: 15.9-20.3); arm C, 13.7 months (95% CI: 12.8-14.6). Neutropenia was the predominant grade 3-4 haematological toxicity: 77.8/85.7/54.4% in arms A/B/C, respectively. Dyspnoea was markedly increased in both gemcitabine-containing arms (P=0.001) but was worse in arm C. Although just failing to rule out eight cycle completion rates less than 60%, within the statistical limitations of these small cohorts, the overall results for arms A and B are encouraging. Larger phase III studies are required to test these combinations.