Development and validation of an intraoperative perforation prediction scoring system for endoscopic resection of gastric gastrointestinal stromal tumors.

Background: Endoscopic resection (ER) is a minimally invasive treatment for gastric gastrointestinal stromal tumors (gGISTs), but intraoperative perforation risk remains a significant concern. Currently, no validated tools exist to preoperatively identify high-risk patients. This study aimed to develop and validate a practical scoring system for predicting intraoperative perforation during ER of gGISTs.

Methods: A multi-center retrospective study analyzed 1,124 patients undergoing ER for gGISTs, divided into training (n = 496), internal validation (n = 212), and external validation cohorts (n = 416). Independent risk factors for intraoperative perforation were identified through univariate and multivariate logistic regression and weighted by regression coefficients to construct the scoring system. Model performance was assessed using the area under the curve (AUC), sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV).

Results: Intraoperative perforation occurred in 463 patients (41.2%). Three independent predictors were identified: tumor location (gastric fundus, 2 points), size (≥ 2.0 cm, 2 points), and growth pattern (intraluminal-dominant, 1 point; extraluminal-dominant, 3 points). The scoring system demonstrated strong discrimination, with AUCs of 0.805 (internal validation) and 0.811 (external validation). Accuracy was 0.726 and 0.805, respectively. Risk stratification revealed intraoperative perforation rates of 8.3% (low-risk: 0-1 points), 38.5% (intermediate-risk: 2-3 points), and 80.0% (high-risk: 4-7 points) in the internal validation cohort, and 4.8%, 36.6%, and 85.5% in the external validation cohort.

Conclusion: This novel scoring system, incorporating tumor location, size, and growth pattern, reliably stratifies intraoperative perforation risk prior to ER for gGISTs. It offers clinicians a practical tool to optimize patient selection and procedural planning, potentially improving safety outcomes. Further prospective studies are warranted to validate its clinical utility.