A model of predicting infiltration depth of early gastric cancers based on conventional white light endoscopy and magnifying endoscopy with narrow band imaging findings

Objective To analyze the characteristics of early gastric cancer (EGC) with submucosal infiltration under conventional white light endoscopy (C-WLE) and magnifying endoscopy with narrow band imaging (ME-NBI), and to improve the diagnostic accuracy of EGC infiltration by combining C-WLE and ME-NBI findings. Methods Data of patients who received endoscopic submucosal dissection or surgical treatment for EGC at Beijing Friendship Hospital from January 2015 to December 2017 were retrospectively analyzed. The basic information, lesion characteristics, and postoperative pathology of patients were collected. The characteristics of EGC with submucosal infiltration were analyzed, and a model for predicting the depth of EGC invasion was constructed by combining independent risk factors of submucosal infiltration. Results A total of 207 lesions in 195 patients were included in the study, divided into the modeling group (138 lesions) and the testing group (69 lesions) in the ratio 2∶1. In the modeling group, the lesions located in the upper third of the stomach (OR=12.949, 95%CI: 2.148-78.070, P=0.005), middle third of the stomach (OR=7.534, 95%CI: 1.044-54.360, P=0.045), >2 cm in size (OR=6.828, 95%CI: 1.657-28.136, P=0.008) and presence of dilated blood vessel (OR=6.856, 95%CI: 1.577-29.805, P=0.010) were independent risk factors for submucosal infiltration. Based on the above independent risk factors, the infiltration depth predicting scoring system (DPSS) was constructed (5 points for the lesion located in upper third of the stomach, 4 points for the lesions located in the middle third of stomach, 4 points for lesions >2 cm in size, and 4 points for the presence of dilated vessels). The areas under the receiver operating characteristic curve for predicting the infiltration depth of DPSS were 0.884 (95%CI: 0.809-0.960) in the modeling group and 0.799 (95%CI: 0.684-0.914) in the testing group. The sensitivities of the modeling group and the testing group were 83.3% and 71.4% respectively; and the specificities were 76.2% and 74.5%, respectively in the two groups at 8 as the cut-off score. Conclusion The DPSS based on C-WLE and ME-NBI findings can predict the invasion depth of EGC. Key words: Stomach neoplasms; Early diagnosis; Tumor infiltrating; Forecasting models; Magnification endoscopy with narrow band imaging