Background: Gemcitabine plus cisplatin combination (GC) therapy is the standard therapy for advanced biliary tract cancer (BTC). In previous trials, gemcitabine plus S-1 combination (GS) therapy and S-1 mono-therapy had shown considerable efficacy in patients with BTC. The aim of this trial is to evaluate the efficacy and safety of the two regimens and to determine which is more promising as a test arm regimen for a subsequent phase III trial. Methods: Chemotherapy-naive patients with recurrent or unresectable BTC (gallbladder [GB], intrahepatic biliary duct [IHBD], extrahepatic biliary duct [EHBD], ampulla of Vater [AV]), an ECOG PS of 0 - 1, and adequate organ function were randomly assigned to receive GS (gemcitabine: 1000 mg/m², iv, days 1 and 8; S-1: 60 mg/m², p.o., days 1 - 14, every 3 weeks) or S-1 (80 mg/m², p.o., days 1 - 28, every 6 weeks). We assumed that %1-year survival of one regimen is 30% and that of the other regimen is more than 40%. To ensure at least 85% probability of correct selection, 98 eligible pts are required. The decision rule was that the regimen with higher %1-year survival will be considered as more promising regimen. Results: From February 2009 to April 2010, 101 pts (GB, n=38; IHBD, n=35; EHBD, n=20; AV, n=8) were randomized (GS, n=51; S-1, n=50). For the GS arm and S-1 arm, %1-year survivals were 52.9% and 40.0%, the median survival time were 12.5 and 9.0 months (hazard ratio 0.86 [95%CI 0.54-1.36]; p=0.52), and the median progression-free survival time were 7.1 and 4.2 months (0.44 [0.29-0.67]; p<0.0001). Grade 3/4 hematological toxicities were more frequent in the GS arm than in the S-1 arm, (percentage in GS/S-1 arms): neutropenia, 60.8/4.0; leukocytopenia, 29.4/2.0; hemoglobin, 11.8/4.0; and thrombocytopenia, 11.8/4.0; respectively. Although two treatment-related deaths occurred in the GS arm (pneumonitis, acute myocardial infarction), other grade 3/4 non-hematological toxicities were infrequent and reversible in both arms. Conclusions: The GS arm was superior in %1-year survival to S-1. Here we consider GS to be more promising as the test arm for a subsequent phase III trial comparing with GC.