Abstract:

Objective To explore the risk factors of delayed bleeding in patients with gastrointestinal mucosal tumors after endoscopic mucosal resection (EMR), and to construct the nomogram prediction model. Methods A retrospective analysis was performed on clinical data of patients with gastrointestinal mucosal tumors who underwent EMR in Harbin Second Hospital between May 2020 and September 2023. Among all patients, 160 cases with delayed bleeding within 30 d after surgery were enrolled as the bleeding group, while 320 cases from patients without bleeding during the same period were enrolled as the non-bleeding group according to 1∶2 simple and random sampling method. The clinical data were compared between the two groups. The risk factors of delayed bleeding after EMR were analyzed by multivariate Logistic regression analysis. The nomogram model for postoperative delayed bleeding was constructed. Its predictive efficiency was analyzed by receiver operating characteristic (ROC) curves, calibration degree of the model was evaluated by calibration curves, and the net benefit of the model was evaluated by clinical decision curve analysis (DCA). Results Compared with the non-bleeding group, patient age, proportion of hypertension, tumor diameter, resection area, intraoperative blood loss and postoperative usage rate of anticoagulant agents were higher, and preoperative prothrombin time (PT), withdrawal time of anticoagulant drugs and postoperative duration of anticoagulant drugs were longer, and frequency of preoperative biopsy was higher in the bleeding group (P<0.05). Results of Logistic regression analysis showed that increased age and frequency of preoperative biopsy and intraoperative bleeding were risk factors of delayed bleeding within 30 d after surgery (P<0.05). The area under ROC curve (AUC) of the nomogram model constructed based on the above factors was 0.906 (95%CI: 0.804-0.951), showing high predictive and discriminative efficiency. The calibration curve showed that the consistency between the model and actual condition for predicting bleeding rate was good. Hosmer-Lemeshow test with P=0.352 confirmed the stability of model calibration. DCA revealed that the model achieved clinical net benefit across a wide range of threshold probabilities. Conclusion The nomogram prediction model constructed based on independent risk factors has high predictive efficiency and discriminative ability, which can provide reference for preoperative risk stratification and perioperative individualized management in clinical practice.

Key words: gastrointestinal tumor, mucosa, endoscopic mucosal resection, nomogram, delayed bleeding