Introduction: Neoadjuvant therapy may improve survival compared with upfront surgery in patients with resectable and borderline resectable pancreatic cancer, but high-quality evidence is lacking. Methods: We systematically searched for randomised trials comparing neoadjuvant therapy with upfront surgery for resectable and borderline resectable pancreatic cancer published since database inception until December 2020. The primary outcome was overall survival (OS) by intention-to-treat with subgroup analyses for resectability status. Meta-analyses using a random-effects model were performed. Certainty of evidence was assessed using the GRADE approach. Results: Seven trials with 938 patients were included. All trials included a neoadjuvant gemcitabine-based chemo(radio)therapy arm. None of the studies used adjuvant FOLFIRINOX. Neoadjuvant therapy improved OS (hazard ratio [HR] 0.66, 95% confidence interval [CI] 0.52–0.85; P = 0.001; I2 = 46%) compared with upfront surgery. This represents an increase in median OS from 19 to 29 months. In the subgroup of resectable pancreatic cancer (i.e., venous contact ≤180°, no arterial contact), no statistically significant difference in OS was observed (HR 0.77, 95% CI 0.53–1.12; P = 0.18; I2 = 20%). In the subgroup of borderline resectable pancreatic cancer (i.e. venous contact >180°, any arterial contact), neoadjuvant therapy improved OS (HR 0.61, 95% CI 0.44–0.85; P = 0.004; I2 = 59%). The GRADE certainty of evidence was high for the outcome of OS. Conclusions: Neoadjuvant therapy improves OS compared with upfront surgery in patients with borderline resectable pancreatic cancer. More evidence is required on whether neoadjuvant therapy improves survival for patients with resectable pancreatic cancer.