A predictive risk-scoring model for multiple synchronous early gastric cancers or gastric dysplasia before initial endoscopic resection.

OBJECTIVE To establish a new and easy-to-use risk-scoring predictive model to help identify high-risk patients with multiple synchronous gastric neoplasms (MSGN), including early gastric cancer (EGC) and gastric dysplasia (GD), before initial endoscopic resection (ER). METHODS We retrospectively enrolled 1361 patients with EGC or GD who had undergone ER from November 2006 to September 2019. The patients were randomly divided into the training (n = 681) and validation cohorts (n = 680). In the training phase a prediction score was constructed to assess the independent predictors of MSGN based on multivariate logistic regression analysis. The performance of the prediction model was evaluated using the receiver operating characteristic (ROC) curve and the Hosmer-Lemeshow test. RESULTS Of the 1361 patients, 122 (9.0%) had MSGN. Three predictors for MSGN were scored and weighted, as follows: elderly male (≥65 y; three points), a family history of gastric cancer (two points) and surface redness (two points). Accordingly, patients were divided into the low (risk score, 0-3 points) or high-risk groups (risk score, 4-7 points). In the validation cohort, the incidence of MSGN in the low-risk and high-risk groups were 6.1% and 32.0%, respectively (P < 0.001). Our predictive risk-scoring model showed good discrimination (the area under the ROC curve [AUROC] 0.719, 95% confidence interval [CI] 0.634-0.794, P < 0.001) and calibration ability (Hosmer-Lemeshow test, χ2  = 6.539, P = 0.587) in the validation group. CONCLUSION This risk-scoring model has a good performance in predicting MSGN before the initial ER.