Risk factor analysis and prediction model construction for the progression risk of type Ⅱ+Ⅲ mixed intestinal metaplasia

Objective To establish a prediction model for assessing the histological progression risk in patients with biopsy-confirmed type Ⅱ+Ⅲ mixed gastric intestinal metaplasia (GIM) using a prediction scoring tool based on clinical-endoscopic-pathological parameters.

Methods A total of 374 patients with type Ⅱ+Ⅲ mixed GIM who visited the Digestive Endoscopy Center of Zhongda Hospital Affiliated to Southeast University, from November 2019 to November 2024 were enrolled according to the principle of convenience sampling. Patients were randomly divided into training set (n=262) and validation set (n=112) at a ratio of 7∶ 3. The primary outcome was histological progression during a follow-up of ≥6 months, including an expansion of the scope of GIM or an increase in the grade at the original site (subtype transformation/neoplasia/canceration). Baseline data, serological indicators, endoscopic feature indicators, and pathological indicators were included in this study. Univariate Logistic regression analysis was used for initial screening of these indicators, followed by dimensionality reduction using the least absolute shrinkage and selection operator (LASSO) algorithm (λ_min=0.035). Finally, non-zero variables were included in multivariate logistic regression analysis to construct the prediction model and draw a nomogram. Model performance was evaluated using the receiver operating characteristic curve, calibration curve, decision curve analysis, and Cox-Snell R² and Nagelkerke R².

Results In the training set, histological progression occurred in 92 patients. LASSO regression retained 10 non-zero coefficient variables. These variables were included in multivariate logistic regression analysis, which showed that age, smoking history, gastric antrum GIM, cardiac GIM, and Helicobacter pylori (Hp) infection were independent influencing factors for histological progression intype Ⅱ+Ⅲ mixed GIM (P < 0.05). In the training set, the area under the curve of the model was 0.748 (95%CI, 0.686 to 0.810). Decision curve analysis showed that within the risk threshold range of 0 to 0.8, the clinical net benefit of the prediction model was higher than that of the full-intervention or non-intervention strategies.

Conclusion The nomogram based on age, smoking history, involvement of the gastric antrum and cardia, and Hp infection status can individually predict the short-term histological progression risk in patients with type Ⅱ+Ⅲ mixed GIM. The model has good discrimination and calibration and can providea basis for risk-stratified follow-up and personalized management.