There is no standard treatment for patients with advanced urothelial cancer who are ineligible ("unfit") for cisplatin-based chemotherapy (CHT). To compare the activity and safety of two CHT combinations in this patient group, a randomized phase II/III trial was conducted by the EORTC (European Organisation for Research and Treatment of Cancer). We report here the phase II results of the study. CHT-naïve patients with measurable disease and impaired renal function (30 mL/min < glomerular filtration rate GFR < 60 mL/min) and/or performance status (PS) 2 were randomly assigned to receive either GC (gemcitabine 1,000 mg/m(2) on days 1 and 8 and carboplatin area under the serum concentration-time curve AUC 4.5) for 21 days or M-CAVI (methotrexate 30 mg/m(2) on days 1, 15, and 22; carboplatin AUC 4.5 on day 1; and vinblastine 3 mg/m(2) on days 1, 15, and 22) for 28 days. End points of response and severe acute toxicity (SAT) were evaluated with respect to treatment group, renal function, PS, and Bajorin risk groups. Three of 178 patients who were ineligible or did not start treatment were excluded. SAT was reported in 13.6% of patients on GC and in 23% on M-CAVI. Overall response rates were 42% (37 of 88) for GC and 30% (26 of 87) for M-CAVI. Patients with PS 2 and GFR less than 60 mL/min and patients in Bajorin risk group 2 showed a response rate of only 26% and 20% and an SAT rate of 26% and 25%, respectively. Both combinations are active in this group of unfit patients. However, patients with PS 2 and GFR less than 60 mL/min do not benefit from combination CHT. Alternative treatment modalities should be sought in this subgroup of poor-risk patients.