A novel model might predict the risk of chronic atrophic gastritis: a multicenter prospective study in China

Objective. To find a new way to predict the risk of chronic atrophic gastritis (CAG). Material and methods. All the participants received endoscopy and histological examination as well as a standard questionnaire. Multivariate analysis was performed by logistic regression to build the CAG risk model. The accuracy was evaluated by 1418 subjects recruited from six medical centers. 63 subjects received another endoscopy after 1-year follow-up and divided into three groups according to the comparison of the histological results (improved, no change and worse). Results. The model showed relatively good discrimination, with an AUROC of 0.888 (95% CI 0.852-0.925). A final probability cut-off score of 0.73 was used to predict the presence (> 0.73) or absence of CAG (<= 0.73). Sensitivity, specificity, PPV and NPV were 82.8%, 74.7%, 91.8% and 56%, respectively. The predicted results of 1418 subjects compared with the histological results were quite similar. There was a significant difference of the scores between three groups who were followed-up for 1 year (F = 3.248, p = 0.046). In multiple comparisons, a significant difference existed between Group A (the histological results had improved after 1-year follow-up) and Group C (the results were worse) (p = 0.019). Conclusions. This is the first demonstration of the use of a mathematical model for CAG risk screening. Endoscopy should be recommended to those who are positive according to the model, to detect CAG early and conserve medical resources. In those who have a high-risk score, closer follow-up is needed.