Pancreatic ductal adenocarcinoma (PDAC) has dismal outcomes, and preoperative tools that refine resectability and survival estimates are urgently needed. Radiomics converts routine CT into quantitative features that capture tumor heterogeneity, while serum CA19-9 is the most widely used biomarker in PDAC. We systematically synthesized original studies evaluating CT-based radiomics, alone or combined with CA19-9, for predicting resectability, early recurrence, or overall survival (OS) in PDAC. Following PRISMA principles, nine original radiomics studies were included. Across cohorts (n=68–326), radiomics improved prognostication beyond clinicopathologic models: for survival, radiomics increased C-index versus clinical factors (from 0.68 to 0.74), and radiomics nomograms frequently outperformed TNM staging. Vessel-centric radiomics yielded high resectability discrimination (AUC =0.92; external sensitivity/specificity up to 100%/88%), and models predicted resectability after neoadjuvant therapy in locally advanced disease (AUC =0.85–0.94). Integrating CA19-9 with radiomics added value: preoperative nomograms (radiomics + CA19-9) exceeded CA19-9 alone, and delta-radiomics combined with CA19-9 achieved strong performance (c-index =0.87) for treatment response and survival. Limitations include retrospective designs, heterogeneous CT protocols, variable segmentation/reproducibility reporting, and scarce multi-center validation. Overall, CT radiomics, especially when combined with CA19-9, shows consistent promise for preoperative risk-stratification in PDAC and warrants prospective, standardized, multi-centric validation.