Background: The PDAC tumor microenvironment is notoriously immune suppressive, and PD-L1 expression is one mechanism by which tumor cells evade immune surveillance. In PDAC, clinical trials evaluating immunotherapy have not been successful in microsatellite stable tumors, and PD-L1 has not been shown to be predictive of immunotherapy response. Studies suggest that high PD-L1 expression may be associated with worse clinical outcomes and more advanced disease. We aimed to evaluate PD-L1 as a prognostic marker in routine clinical practice using a real-world population of patients with PDAC. Methods: We conducted a retrospective study of patients with PDAC at Mount Sinai Health System from 5/2017 to 12/2021 for whom PD-L1 status by immunohistochemistry (22c3) and combined positive score (CPS) were reported. We assessed the association between PD-L1 expression and overall survival (OS) using Kaplan-Meier estimates and multivariable Cox proportional hazards regression models. Chi-squared tests were used to evaluate the association between PD-L1 expression and mutations detected by routine next generation sequencing (NGS) using a standard panel. Results: 107 patients were evaluable. At diagnosis, median age was 68 years, 43% of patients were male, 83% had ECOG performance status 0-1, 38% had resectable disease, 14% had borderline resectable disease, and 48% had unresectable disease. Ultimately, 45 patients (42%) underwent surgery. 44% patients had gemcitabine-based initial therapy, perioperatively or as first-line, and 56% had 5FU-based initial therapy. 35% of patients had radiation therapy (RT) at any time. In the entire cohort, 44 (41%) patients had PD-L1 CPS < 1, and 63 patients had CPS ≥ 1. 93 patients had NGS, but no associations were found between PD-L1 status and commonly mutated genes, such as KRAS and SMAD4. For patients who underwent surgery, the median OS (mOS) of patients with PD-L1 CPS ≥ 1 was 46.1 months, and mOS of PD-L1 CPS < 1 was 29.8 months (P = 0.13). For patients who did not receive surgery, mOS was 12.3 months for CPS ≥ 1 and 15.8 months for CPS < 1 (P = 0.07). When adjusted for age, gender, ECOG score, stage, type of chemotherapy, RT, and baseline CA 19-9 level, patients who received surgery and had a CPS ≥ 1 had better OS than those who had CPS < 1 (HR 0.22, 95% CI 0.06-0.77, P = 0.02). For patients without resection, CPS ≥ 1 was not associated with OS in the adjusted model (HR 1.52, 95% CI 0.73-3.15, P = 0.26). However, patients with PD-L1 CPS ≥ 10 without surgery had worse OS compared to patients with CPS < 1 in the adjusted model (HR 11.02, 95% CI 2.89-41.96, P< 0.001). Chemotherapy type and receipt of RT did not modify OS based on PD-L1 expression. Conclusions: In resected PDAC, PD-L1 CPS ≥ 1 was independently associated with improved OS, whereas in unresectable PDAC, PD-L1 CPS ≥ 10 was independently prognostic of worse OS. PD-L1 expression is a potential prognostic factor in PDAC and may be a useful and context-dependent biomarker.