Predictive model for pancreatic fistula in minimally invasive surgery for gastric cancer

Background Postoperative pancreatic fistula (POPF) is one of the potentially serious complications after gastrectomy for gastric cancer (GC). Drain amylase level is a predictor of POPF in open and laparoscopic gastrectomy, but no study has focused on minimally invasive surgery (MIS), including robotic gastrectomy (RG). This study assesses the effect of drain amylase levels for POPF in MIS and develop a prediction model in the MIS era. Methods This single-institutional retrospective study, conducted from January 2011 to December 2021, included 1,353 who underwent standard MIS for GC. We placed a drain in all patients undergoing MIS gastrectomy and measured the drain amylase level on the first postoperative day (D1Amy). The predictive accuracy of D1Amy for POPF was assessed. Additionally, the entire cohort was randomly categorized into the training (1,048 patients) and validation sets (305 patients) to establish the nomogram. Results Of the 1353 patients, 530 underwent a robotic approach. POPF and intraabdominal infectious complications of Clavien-Dindo classification grade ≥ II were observed in 80 (5.9%) and 145 (10.7%) patients, respectively. Median D1Amy was 812 U/L. The receiver operating characteristic analysis of D1Amy for POPF revealed an area under the curve (AUC) of 0.888. Multivariate analysis revealed age, tumor location, splenectomy, and D1Amy as significant risk factors for POPF. The AUC of the nomogram was 0.8960, validated with AUC of 0.9259. Conclusions We revealed the utility of D1Amy in predicting POPF in MIS gastrectomy. Furthermore, the nomogram, incorporating D1Amy and other clinical factors, was additionally used as a predictive model for POPF.