We investigated the impact of S-1 on the prognosis of patients with gemcitabine-refractory pancreatic cancer. A total of 108 patients with gemcitabine-refractory pancreatic cancer were divided by the time of S-1 introduction in our institution: 47 patients who experienced progressive disease before February 2005 (pre-S-1 group) and 61 patients showed progressive disease after February 2005 (post-S-1 group). Introduction rates of second-line chemotherapy and survival were compared. Prognostic factors for residual survival were analyzed using the Cox proportional hazards model. Introduction rates of second-line chemotherapy were 12.8% in the pre-S-1 group and 45.9% in the post-S-1 group. Second-line chemotherapy was administered to 34 patients: 29 using S-1, 4 using 5-fluorouracil-based chemoradiation and 1 using 5-fluorouracil. The objective response rate, progression-free survival and overall survival for second-line chemotherapy with S-1 were 17.2%, 2.5 and 7.7 months, respectively. By the introduction of S-1 in our institution, residual survival was prolonged from 3.1 months in the pre-S-1 group to 6.7 months in the post-S-1 group (P < 0.001). Overall survival from the initiation of gemcitabine was 8.8 months in the pre-S-1 group and 11.3 months in the post-S-1 group (P = 0.013). Multivariate analysis identified the post-S-1 group (hazard ratio, 0.43; P = 0.001), gender, performance status, liver metastasis, and lactate dehydrogenase and C-reactive protein levels at progressive disease for gemcitabine to be prognostic factors for residual survival. The introduction of S-1 might improve the prognosis of patients with gemcitabine-refractory pancreatic cancer.