Background: Both gemcitabine (GEM) and S-1, an oral fluoropyrimidine, are shown efficacy for advanced and resected pancreatic cancer (PC). However PC has one of the poorest outcomes in various cancers. In this pooled analysis, the prognostic factors in locally advanced and metastatic PC were evaluated using individual patient data from three randomized studies (Phase III: GEST, randomized phase II: JACCRO PC-01 and GEMSAP). Methods: Patients (pts) were treated with GEM alone or GEM and S-1 (GS) chemotherapy for advanced PC. Evaluated prognostic factors were standard laboratory data, tumor marker (CA19-9, CEA), clinical and pathological indices and treatment related outcomes e.g. efficacy and safety. To identify the optimal cutoff value to divide continuous data into binomial data, maximized log likelihood value for the model were evaluated and the validity of the model was evaluated by the martingale residuals. The Cox proportional hazards model was performed for overall survival. Results: 770 pts were included in this analysis (389 pts were treated by GEM and 381 pts were GS) and 738 events for survival were observed (95.8%). In continuous data, baseline WBC, AST, CA19-9 and CRP were detected as independent prognostic factors and the cutoff values were identified. The results of multivariate analysis and cutoff value were shown in the table. Conclusions: In our analysis, treatment, extent of disease, target lesion, WBC, AST, CA19-9 and CRP were detected as independent prognostic factors and cutoff values were estimated in continuous data. The further examination is necessary, but these data may help future studies.