Background: The aim of this systematic review and meta-analysis was to evaluate perioperative outcomes and survival in patients undergoing an artery-first approach to pancreatoduodenectomy in comparison with those having standard pancreatoduodenectomy. Methods: A systematic search of PubMed, MEDLINE, Embase and the Cochrane Database of Systematic Reviews was performed in accordance with PRISMA guidelines. Comparative studies including patients who underwent artery-first pancreatoduodenectomy and standard pancreatoduodenectomy were analysed. Results: Seventeen studies were included in the final analysis. There were 16 retrospective cohort or case–control studies and one RCT. A total of 1472 patients were included in the meta-analysis, of whom 771 underwent artery-first pancreatoduodenectomy and 701 had standard pancreatoduodenectomy. Intraoperative blood loss (mean difference –389 ml; P < 0·001) and the proportion of patients requiring intraoperative transfusion (10·6 per cent (54 of 508) versus 40·1 per cent (186 of 464); P < 0·001) were significantly lower in the artery-first group. Although rates of perioperative mortality were comparable between the two groups, perioperative morbidity (35·5 per cent (263 of 741) versus 44·3 per cent (277 of 625); P = 0·002), and the incidence of grade B/C pancreatic fistula (7·4 per cent (26 of 353) versus 12·8 per cent (42 of 327); P = 0·031) were significantly lower in the artery-first group. The R0 resection rate (75·8 per cent (269 of 355) versus 67·0 per cent (280 of 418); P < 0·001) and overall survival (hazard ratio 0·72, 95 per cent c.i. 0·60 to 0·87; P < 0·001) were significantly higher in the artery-first group. Conclusion: The artery-first approach to pancreatoduodenectomy may be associated with improved perioperative outcomes and survival.