Background: Stereotactic body radiotherapy (SBRT) has been incorporated into multi-modality treatment of locally advanced and borderline resectable pancreatic ductal adenocarcinoma (PDAC). For non-metastatic inoperable PDAC patients (pts) who receive magnetic resonance-guided SBRT (MRgSBRT), baseline features that predict for treatment response remain unsettled. In localized PDAC, multiple studies have investigated the KRAS oncogene as a prognostic factor with mixed findings. In this single institution retrospective study of PDAC pts treated with MRgSBRT, we hypothesized that KRAS mutation status would be predictive of meaningful clinical outcomes. Methods: From an IRB approved dataset of PDAC pts treated with MRgSBRT between 2016 and 2022, 39 pts with non-metastatic inoperable pancreatic cancer, known KRAS mutation status, ≥ 3 months of neoadjuvant systemic therapy (NST), and at least 3 months post-RT follow up were extracted for analysis. Baseline demographics, tumor and treatment characteristics, and clinical endpoints including conversion to resectability, best imaging response per RECIST v1.1, pathologic response (PR) per TRG-CAP, and overall survival (OS) after MRgSBRT were collected. Objective response (ORR) on both imaging and pathology was defined as complete response (CR) + partial response (PR); disease control (DC) was defined as CR + PR + stable disease (SD). Logistic regression was used to assess correlation between baseline variables and ORR. Cox proportional hazard models were utilized to determine association with OS. Results: Out of 39 pts, 21 (53%) were KRAS-mutated (KRAS-mt) and 18 (47%) were KRAS wild-type (KRAS-wt). Median age was 62, 54% were male, only 1 pt had ECOG > 1, and CA 19-9 at diagnosis was 197. Median duration of NST was 9 cycles, and common agents included Gemcitabine and Abraxane (21%); FOLFIRINOX (28%); a combination of the two regimens (41%); or other NSTs (10%). Median MRgSBRT dose fractionation was 50 Gy (range 35 – 50) in 5 fractions; all fractions underwent adaptive optimization. Thirty pts (77%) experienced DC on imaging, 16 (43%) had ORR, and 12 (31%) were converted to resectable following MRgSBRT. Cohort median OS was 12.3 months, and pts who had surgery had a non-significant median OS advantage (18 vs 12 months; p=0.19). KRAS-mt was associated with worse ORR (p=0.04; AUC=0.73) and decreased OS (HR: 2.15; p=0.03). In a clinical model incorporating baseline characteristics (age, ECOG, radiographic staging, CA 19-9 level, and KRAS), only KRAS predicted OS (HR: 3.02 for KRAS-mt; p=0.01). KRAS-mt status was also predictive of OS among pts who underwent surgery (HR: 5.24 for KRAS-mt; p=0.04). Conclusions: For localized PDAC pts inoperable at diagnosis who received NST followed by MRgSBRT, KRAS was the only significant predictor of ORR and OS, highlighting the need for further treatment intensification in KRAS-mt pts.