Long-term oncological outcomes following algorithm-based care versus usual care for the early recognition and management of complications after pancreatic resection: a post-hoc analysis of a nationwide, stepped-wedge cluster-randomised trial

Summary Background Studies have shown that severe postoperative complications after resection of pancreatic and periampullary cancer might have a negative effect on long-term oncological outcomes. The PORSCH trial showed that an algorithm for early recognition and minimally invasive management of complications after pancreatic resection significantly improved short-term clinical outcomes. The algorithm's effect on long-term survival remains to be determined. Methods Long-term oncological outcomes were evaluated post hoc in all patients included in the nationwide, stepped-wedge cluster-randomised PORSCH trial (Jan 8, 2018, to Nov 9, 2019, Netherlands Trial Register, NL6671) who underwent pancreatic resection for pancreatic ductal adenocarcinoma, distal cholangiocarcinoma, ampullary carcinoma, or duodenal carcinoma. Cox proportional hazard regression was used to compare overall survival between algorithm-based care and usual care. The crude analyses accounted for the stepped-wedge design by adjusting for calendar time, hospital volume, and hospital as a frailty term. Adjusted analyses were adjusted for baseline factors associated with long-term survival (age at diagnosis, sex, American Society of Anesthesiologists score, preoperative carbohydrate antigen 19-9, neoadjuvant therapy, vascular resection, postoperative tumour size, number of positive lymph nodes, perineural invasion, tumour differentiation, and resection margin status). Adjusted differences in restricted mean survival time (RMST) were estimated between algorithm-based care and usual care, using the same adjustment set as the Cox models. A predefined subgroup analysis was performed for patients with pancreatic ductal adenocarcinoma. Findings In total, 1090 patients were included: 644 (59%) patients with pancreatic ductal adenocarcinoma, 175 (16%) patients with distal cholangiocarcinoma, 188 (17%) patients with ampullary carcinoma, and 83 (8%) patients with duodenal carcinoma. As of Jan 1, 2024, median follow-up was 56 months (IQR 48–63) for the 549 patients who received usual care and 48 months (35–53) for the 541 patients who received algorithm-based care. Unadjusted median overall survival was 24 months (95% CI 22–28) in the usual care group and 26 months (24–30) in the algorithm-based care group. The adjusted difference in RMST for overall survival over 36 months was 2·1 months (95% CI 0·6–3·7, p=0·0080), favouring algorithm-based care. Algorithm-based care was associated with improved overall survival (crude hazard ratio [HR] 0·85 [95% CI 0·71–1·02], p=0·076; adjusted HR 0·76 [0·62–0·93], p=0·0089). Overall survival differences between algorithm-based care and usual care were most pronounced in patients with pancreatic ductal adenocarcinoma (crude HR 0·78 [95% CI 0·62–0·97], p=0·028; adjusted HR 0·71 [0·56–0·90], p=0·0052). For these patients, the adjusted RMST differences up to 36 months favoured algorithm-based care by 2·5 months (95% CI 0·8–4·2, p=0·0046). Interpretation The PORSCH algorithm for early recognition and minimally invasive management of postoperative complications after pancreatic surgery is associated with increased long-term overall survival after resection of pancreatic or periampullary cancer. Funding Dutch Cancer Society and St Antonius Research Fund.