Perioperative corticosteroids and pancreatic surgery outcomes: a systematic review and meta-analysis combining human expertise and AI support (ChatGPT).

Pancreatic surgery is associated with high postoperative morbidity. The role of corticosteroids (CCS) in pancreatic surgery remains uncertain. Systematic review and meta-analysis conducted in accordance with PRISMA guidelines. Eligible studies included patients undergoing pancreatic resection with perioperative CCS administration. Primary outcomes were mortality, morbidity and surgical site infections (SSI). Data extraction was cross-validated with artificial intelligence (ChatGPT). Nine studies (1913 patients), including five RCTs (554 patients), were included. CCS regimens consisted of hydrocortisone or dexamethasone, with or without postoperative continuation. The analysis showed no significant differences in mortality or major complications. Data from RCTs showed that CCS reduced SSI (OR 0.53, 95%CI 0.32-0.91, p = 0.02). Also, among patients undergoing pancreatoduodenectomy, CCS reduced morbidity (OR 0.61, 95%CI 0.39-0.98), SSI (OR 0.57, 95%CI 0.40-0.82), and shortened hospital stay (MD -1.57 days, p = 0.04). Agreement between manual and AI-assisted data extraction was high (r = 0.97-1.00), but substantial discrepancies were found. Perioperative CCS appear safe in pancreatic surgery and may reduce morbidity and SSI. Large multicentre RCTs are needed to define optimal regimens and identify patients most likely to benefit. AI-assisted review may complement traditional approaches but still require further refinement before they can be considered as reliable as human effort.