Background: Immune checkpoint inhibitors (ICI) have limited activity as monotherapy in unselected patients with ABC; the objective response rate (ORR) of PEM was 5.8% in the multicenter phase 2 KEYNOTE-158 trial. GM-CSF modulates immune cells including monocytes and the innate immune response and has demonstrated safety and prolonged survival (OS) in combination with ipilimumab in melanoma. This phase 2 trial was designed to evaluate the efficacy and safety of PEM in combination with GM-CSF in ABC. Methods: Single-center, phase 2 trial with Simon’s 2-stage design and expansion cohort (EC) (NCT02703714). Key eligibility: ABC with progression/intolerance on ≥ 1 standard therapy, no prior ICI, bilirubin ≤1.5xULN. Treatment: PEM 200 mg IV Q21 days plus 2 cycles of GM-CSF 250 µg SC D1-14 Q21 days in cycles 2 and 3 (Stage 1 and EC) or in cycles 1 and 2 (Stage 2). Primary endpoint was objective response rate (ORR) by RECIST 1.1, H₀ 5% vs. H₁ 20%. Key secondary endpoints were safety, progression-free survival (PFS), OS, PD-L1 expression. Exploratory endpoints included relationship between efficacy outcomes and anatomic subsite, risk factors, and tumor genotype. Results: Overall, 42 patients enrolled between 5/2016-12/2019: n = 9 in Stage 1, n = 18 in Stage 2, and n = 15 in EC. Overall: median age 61; female 67%; intrahepatic (ICC) 67%, extrahepatic (ECC) 26%, gallbladder (GBC) 7%; stage IV 90%; hepatitis B/C virus positive (HBV/HCV+) 24%; microsatellite instability (MSI-H)/stable (MSS)/unknown n = 1/35/6. Immune-related (ir)AE occurred in 69%, grade 3/4 treatment-related (tr)AE in 10%, all-cause serious (S)AE in 36%, and trSAE in 7%. ORR was 12% (95% CI: 4, 26), with median PFS of 63 days (95% CI: 55, 125), PFS at 6 months in 27% (95% CI: 14, 43), and median OS 3of 93 days (95% CI: 243, 573). There was no significant difference in ORR, PFS, or OS by anatomic subsite or tumor PD-L1 expression; HBV/HCV+ showed trends toward higher ORR (30% vs 6%, p = 0.08) and longer median PFS (276 vs 63 days, p = 0.06) and OS (1033 vs 323 days, p = 0.052). Conclusions: The combination of PEM plus GM-CSF was safe and well-tolerated with higher ORR than expected for PEM monotherapy in a predominantly MSS ABC population but did not meet target ORR threshold for efficacy. ORR and median PFS and OS were highest in patients with underlying HBV or HCV infection. Immune profiling of on-treatment biopsies and peripheral blood are ongoing to ascertain influence of both PEM and GM-CSF on circulating and tumor immune microenvironment. Clinical trial information: NCT02703714.