© 2017 Elsevier Ltd Background Maintenance therapy improves outcomes in various tumour types, but cumulative toxic effects limit the choice of drugs. We investigated whether maintenance therapy with vinflunine would delay disease progression in patients with advanced urothelial carcinoma who had achieved disease control with first-line chemotherapy. Methods We did a randomised, controlled, open-label, phase 2 trial in 21 Spanish hospitals. Eligible patients had locally advanced, surgically unresectable, or metastatic transitional-cell carcinoma of the urothelial tract, adequate organ function, and disease control after four to six cycles of cisplatin and gemcitabine (carboplatin allowed after cycle four). Patients were randomly assigned (1:1) to receive vinflunine or best supportive care until disease progression. We initially used block randomisation with a block size of six. Four lists were created for the two stratification factors of starting dose of vinflunine and presence of liver metastases. After a protocol amendment, number of cisplatin and gemcitabine cycles was added as a stratification factor, and eight lists were created, still with a block size of six. Finally, we changed to a minimisation procedure to reduce the risk of imbalance between groups. Vinflunine was given every 21 days as a 20 min intravenous infusion at 320 mg/m 2 or at 280 mg/m 2 in patients with an Eastern Cooperative Oncology Group performance status score of 1, age 75 years or older, previous pelvic radiotherapy, or creatinine clearance lower than 60 mL/min. The primary endpoint was median progression-free survival longer than 5·3 months in the vinflunine group, assessed by modified intention to treat. Comparison of progression-free survival between treatment groups was a secondary endpoint. This trial is registered with ClinicalTrials.gov, number NCT01529411. Findings Between April 12, 2012, and Jan 29, 2015, we enrolled 88 patients, of whom 45 were assigned to receive vinflunine and 43 to receive best supportive care. One patient from the vinflunine group was lost to follow-up immediately after randomisation and was excluded from the analyses. One patient in the best supportive care group became ineligible for the study and did not receive treatment due to a delay in enrolment, but was included in the intention-to-treat efficacy analysis. After a median follow-up of 15·6 months (IQR 8·5–26·0), 29 (66%) of 44 patients in the vinflunine group had disease progression and 24 (55%) had died, compared with 36 (84%) of 43 patients with disease progression and 32 (74%) deaths in the best supportive care group. Median progression-free survival was 6·5 months (95% CI 2·0–11·1) in the vinflunine group and 4·2 months (2·1–6·3) in the best supportive care group (hazard ratio 0·59, 95% CI 0·37–0·96, p=0·031). The most common grade 3 or 4 adverse events were neutropenia (eight [18%] of 44 in the vinflunine group vs none of 42 in the best supportive care group), asthenia or fatigue (seven [16%] vs one [2%]), and constipation (six [14%] vs none). 18 serious adverse events were reported in the vinflunine group and 14 in the best supportive care group. One patient in the vinflunine group died from pneumonia that was deemed to be treatment related. Interpretation In patients with disease control after first-line chemotherapy, progression-free survival exceeded the acceptable threshold with vinflunine maintenance therapy. Moreover, progression-free survival was longer with vinflunine maintenance therapy than with best supportive care. Vinflunine maintenance had an acceptable safety profile. Further studies of the role of vinflunine are warranted. Funding Pierre-Fabre Médicament.