Prediction model for non-curative resection of endoscopic submucosal dissection in patients with early gastric cancer

Background and Aims Endoscopic submucosal dissection (ESD) is a useful method for complete resection of early gastric cancer (EGC). However, there are still some patients who undergo additional gastrectomy after ESD because of non-curative resection. There is no model that can accurately predict non-curative resection of ESD. We aimed to create a model for predicting non-curative resection of ESD in patients with EGC. Patients and methods We reviewed the medical records, including all gross findings of EGC, of patients who underwent ESD for EGCs. We divided the patients into a non-curative resection group and a curative resection group. The clinicopathologic characteristics were compared between the groups to identify the risk factors for non-curative resection of ESD. We created a scoring system based on logistic regression modeling and bootstrap validation. Results Of 1639 patients who had undergone ESD for EGCs, 272 were identified as being treated non-curatively with ESD. A large tumor size (≥20 mm), tumor location in the upper body of the stomach, the presence of ulcer, fusion of gastric folds, the absence of mucosal nodularity, spontaneous bleeding, and undifferentiated tumor histology were associated with non-curative resection of ESD. Points of risk scores were assigned for these variables based on the β coefficient as follows: tumor size (≥20 mm), 2 points; tumor location in the upper body of the stomach, 1 point; ulcer, 2 points; fusion of gastric folds, 2 points; absence of mucosal nodularity, 1 point; spontaneous bleeding, 1 point; and undifferentiated histology, 2 points. Our risk scoring model showed good discriminatory performance on internal validation (bootstrap-corrected area under the receiver operating characteristic curve, 0.7004; 95% confidence interval, 0.6655-0.7353). Conclusions We developed a validated prediction model that can be used to identify patients who will undergo non-curative resection of ESD. Our prediction model can provide useful information for making decisions about the treatment of EGC before performing ESD. Endoscopic submucosal dissection (ESD) is a useful method for complete resection of early gastric cancer (EGC). However, there are still some patients who undergo additional gastrectomy after ESD because of non-curative resection. There is no model that can accurately predict non-curative resection of ESD. We aimed to create a model for predicting non-curative resection of ESD in patients with EGC. We reviewed the medical records, including all gross findings of EGC, of patients who underwent ESD for EGCs. We divided the patients into a non-curative resection group and a curative resection group. The clinicopathologic characteristics were compared between the groups to identify the risk factors for non-curative resection of ESD. We created a scoring system based on logistic regression modeling and bootstrap validation. Of 1639 patients who had undergone ESD for EGCs, 272 were identified as being treated non-curatively with ESD. A large tumor size (≥20 mm), tumor location in the upper body of the stomach, the presence of ulcer, fusion of gastric folds, the absence of mucosal nodularity, spontaneous bleeding, and undifferentiated tumor histology were associated with non-curative resection of ESD. Points of risk scores were assigned for these variables based on the β coefficient as follows: tumor size (≥20 mm), 2 points; tumor location in the upper body of the stomach, 1 point; ulcer, 2 points; fusion of gastric folds, 2 points; absence of mucosal nodularity, 1 point; spontaneous bleeding, 1 point; and undifferentiated histology, 2 points. Our risk scoring model showed good discriminatory performance on internal validation (bootstrap-corrected area under the receiver operating characteristic curve, 0.7004; 95% confidence interval, 0.6655-0.7353). We developed a validated prediction model that can be used to identify patients who will undergo non-curative resection of ESD. Our prediction model can provide useful information for making decisions about the treatment of EGC before performing ESD.