Background: Endoscopic screening is an effective approach for detecting early gastric cancer. The interval of endoscopic surveillance should be defined based on the risk stratification of gastric cancer, so as to improve the screening efficiency. Aims: To investigate whether gastric cancer risk could be stratified according to endoscopic atrophic gastritis grading. Methods: Subjects who underwent gastroscopy at least two times between Jan. 2015 and Dec. 2019 at the 900th Hospital of Joint Logistics Support Force, PLA were enrolled in a retrospective study. The demographic data and information on endoscopy were recorded. The extent and degree of gastric atrophy under endoscopy was graded using Kimura-Takemoto classification system. Then the subjects were allocated into mild group (none atrophy and C-1), moderate group (C-2 and C-3) and severe group (O-1, O-2 and O-3) based on the grading. The correlation of endoscopic grading of atrophy with the risk of gastric cancer was analyzed. Results: A total of 8 736 subjects were enrolled, 4 154 were in mild group (47.6%), 2 409 in moderate group (27.6%), and 2 173 in severe group (24.9%). The mean endoscopic follow-up time was (1 052±643) d, and the mean endoscopic surveillance interval was (518±271) d. The overall coincidence rate of endoscopic diagnosis and pathological diagnosis for atrophy was 88.9%. During the follow-up period, gastric cancer was detected in 41 cases; the detection rates of mild group, moderate group and severe group were 0.07%, 0.54% and 1.15%, respectively (P<0.001). Conclusions: The risk of gastric cancer can be stratified according to the endoscopic atrophic gastritis grading, which is helpful for the decision of individualized endoscopic surveillance interval.