Prediction model for curative endoscopic submucosal dissection of undifferentiated-type early gastric cancer

Background: Endoscopic submucosal dissection (ESD) is an effective treatment for early gastric cancer (EGC); however, its curative resection rate is low for undifferentiated-type EGC. We developed and externally validated a prediction model for curative ESD of undifferentiated-type EGC. Methods: In this cross-sectional study, we included 448 patients who underwent ESD for undifferentiated-type EGC at 18 hospitals in Korea between 2005 and 2015 in the development cohort and 1342 patients who underwent surgery at two hospitals in the validation cohort. A prediction model was developed using the logistic regression model. Results: Endoscopic tumor size 1–2 cm (odds ratio [OR], 2.40; 95% confidence interval [CI] 1.54–3.73), tumor size > 2 cm (OR, 14.00; 95% CI 6.81–28.77), and proximal tumor location from the lower to upper third of the stomach (OR, 1.45; 95% CI 1.03–2.04) were independent predictors of non-curative ESD. A six-score prediction model was developed by assigning points to endoscopic tumor size > 2 cm (five points), tumor size 1–2 cm (two points), upper third location (two points), and middle third location (one point). The rate of curative ESD ranged from 70.6% (score 0) to 11.6% (score 5) with an area under the receiver operating characteristic curve (AUC) of 0.720 (95% CI 0.673–0.766). The model also showed good performance in the validation cohort (AUC, 0.775; 95% CI 0.748–0.803). Conclusions: This six-score prediction model may help in predicting curative ESD and making informed decisions about the treatment selection between ESD and surgery for undifferentiated-type EGC.