Background: Pancreatic cancer (PC) is one of the most lethal digestive system tumors. Most new cases are diagnosed based on metastasis or local aggression and are known as "advanced PC." Recently, studies investigating S-1 have indicated that it has a better clinical curative effect on PC. We conducted a meta-analysis to evaluate the efficacy and safety of S-1 monotherapy compared with S-1 combination regimens in patients with gemcitabine (GEM)-refractory PC. Methods: Trials published between 1978 and 2016 were identified by an electronic search of public databases (Medline, Embase, and the Cochrane Library). All prospective studies were independently identified by 2 authors for inclusion. The response rate (RR), progression-free and overall survival (PFS and OS, respectively), and the primary toxicities were extracted for the meta-analysis. Results: Four randomized controlled trials consisting of 623 patients were included in the analysis, among which 315 patients underwent S-1 monotherapy and 308 patients underwent S-1 combination therapy. The pooled data showed a significantly higher response rate and longer PFS in the S-1 combination group than in the S-1 monotherapy group (RR, 1.75; 95% confidence interval [CI], 1.19-2.57; P=.005 and hazard ration [HR], 0.75; 95% CI, 0.62-0.91; P=.005). There were no significant differences in OS or adverse events. Conclusions: Compared with the S-1 monotherapy group, the S-1 combination group had a higher response rate and longer PFS. Both groups had few adverse events, which were balanced between the groups. The subgroup analysis suggested that S-1 combination regimens with leucovorin or irinotecan (CPT-11) provided promising efficacy. These promising combination regimens should be considered for patients with advanced PC who choose S-1 as their second-line therapy.