Abstract
BACKGROUND: Chronic atrophic gastritis (CAG) is a major digestive disorders, and prognosis is determined by many social-demographic and clinicopathologic characteristics. This study aimed to identify risk factors and construct a predictive model for better diagnosis of CAG. METHODS: We utilized a multi-center retrospective analysis, including 539 cases of CAG patients diagnosed and treated in Second Affiliated Hospital of Anhui University of Chinese Medicine from September 2018 to December 2024 as training dataset, and 230 clinical data diagnosed with CAG from Hefei Second People's Hospital from April 2018 to November 2024 as validation dataset to establish the predictive model. Both univariate and multivariate logistic regression analysis were employed to investigate the risk factors of CAG based on R software 4.4.1. After that, our predictive model was evaluated by nomogram, receiver operating characteristic (ROC) curve for discrimination of the predictive model, calibration curves, Hosmer-Lemeshow goodness of fit test for uniformity between the predicted and actual probabilities and decision curve analysis (DCA) curves for clinical validity. RESULTS: Our multivariate logistic regression analysis revealed that depression disorder, drinking consumption, family history of digestive disorders, HP infection, pepsinogen I, pepsinogen II and gastrin 17 were the independent risk factors of our predictive model. A nomogram of CAG was established. The ROC curve revealed that our predictive model showed the best predictive efficacy with an AUC of 0.827 (95%CI = 0.784-0.870), with a specificity of 0.838 and sensitivity of 0.705 in training dataset, and an AUC of 0.970 (95%CI = 0.945-0.995), with a specificity of 0.881 and sensitivity of 0.950 in the validation dataset. Hosmer-Lemeshow goodness of fit test showed that our predictive model had a good fit for the training dataset (X-squared = 3.8293, df = 8, p = 0.8722) and validation dataset (X-squared = 8.9753, df = 8, p = 0.3444). Moreover, calibration and DCA curves demonstrated that our predictive model had a good fit, better net benefit and predictive efficiency in patients with CAG. CONCLUSION: Our predictive model demonstrated that depression disorder, drinking consumption, family history of digestive disorders, HP infection, pepsinogen I, pepsinogen II and gastrin 17 were the independent risk factors of CAG with high accuracy and good calibration.