A randomised phase III trial of MVAC (methotrexate, vincristine, doxorubicin, cisplatin) vs gemcitabine and cisplatin (GC) (G 1000 mg m(-2) days 1, 8, and 15 plus C 70 mg m(-2) day 2, q 4 wks) indicated GC had similar efficacy and lower toxicity (JCO 2000). Significant haematologic toxicities in the GC arm occurred on day 15, necessitating dose adjustments in 37% of cycles. We conducted a phase I/II dose escalation trial using GC on a 21-day cycle, with G and C split between days 1 and 8. The objective of the study to define maximum-tolerated dose and dose-limiting toxicity (DLT), objective response rate, and overall survival. In all, 32 patients with locally advanced, relapsed, or metastatic disease received: dose level 1, G/C 1000/35; level 2, 1100/35; level 3, 1200/35; level 4, 1200/45 mg m(-2) (G and C given on days 1 and 8 every 3 wks). A total of 19 patients had glomerular filtration rate <60 ml min(-1) and 19 patients had metastatic disease. Dose-limiting toxicity was haematologic (grade 4 thrombocytopenia) at dose level 2. Of 151 cycles, at day 15, platelets were <100 in 61 cycles; neutrophils <0.5, platelets <50 in 26 cycles. Only seven cycles were deferred due to haematological toxicity; four for renal toxicity (chemotherapy instituted posthydration). Overall response rate was 65.5% on an intention-to-treat analysis (75% [21/28] for assessable patients), with four complete responses (12.5%) and 17 partial responses (53%). After the median follow-up of 17.2 months (range 13.1-32.4 months), 12 patients remain alive. The overall median survival was 16 months (range 10.1-26.6 months). G plus C every 3 weeks is active and well tolerated in an outpatient setting, even in patients receiving prior platinum-based regimens and with poor renal reserve.