Background: Ipilimumab (IPI) at 3mg/kg has been tested in pancreatic ductal adenocarcinoma (PDA) with a low but detectable response rate. This supports a role for immunotherapy in PDA. Melanoma (MEL) studies demonstrate a dose-response relationship with IPI. Preclinical studies show synergy between anti-CTLA-4 antibodies and GM-CSF cell-based vaccines. This study evaluated IPI 10mg/kg alone (Arm 1) versus IPI 10mg/kg + vaccine (Arm 2) in PDA. Methods: 30 subjects with previously treated, locally advanced or metastatic PDA were randomized (1:1) to Arm 1 or 2. Induction treatments were administered every 3 weeks (wk) X 4 doses. Subjects with stable disease (SD) or better between wk 14 and 22 were eligible for maintenance doses every 12 wks. Activity was assessed by response and overall survival (OS). IFN-g Elispots and ELISAs were used to detect induction of mesothelin-specific T cells and galectin-3 antibodies (Abs), respectively. Results: Baseline characteristics were similar (87% metastatic, 100% gemcitabine-treated) with the exception that Arm 1 had a lower % of subjects with > 1 prior therapies (60 vs 100%). Arm 1 had 2 SD (7 and 22 wks). Arm 2 had 3 SD (1 delayed regression starting at wk 14 and maintained until wk 31, 1 delayed stabilization starting at wk 22 and ongoing at wk 57, 1 SD ongoing at wk 64). Transient CA19-9 declines were seen only in Arm 2 (7/15). Median OS was 3.3 vs 5.5 months in Arm 1 vs Arm 2 (p=0.12) with a 12 month OS of 7 vs 27%. 73% (Arm 1) and 80% (Arm 2) of subjects experienced immune related adverse events (IRAE). 20% of subjects on both arms had Grade 3-4 IRAEs including rash (1), colitis (2), Guillain-Barre (1), nephritis (1), and pneumonitis (1). Rash was less frequent in Arm 1 (53 vs 73%). IRAEs (p=0.037), mesothelin T cell responses (p=0.012), and enhancement of the T cell repertoire (p=0.026) were associated with OS in both arms. Galectin-3 Ab responses were associated with OS in Arm 2 (p=0.002). Conclusions: Immunotherapy has potential even in advanced PDA. IRAEs in PDA are similar to IRAEs in MEL. Delayed responses suggest that standard RECIST criteria will underestimate activity and targeting a population with a longer life expectancy may improve efficacy.