A risk prediction model for gastric cancer based on endoscopic atrophy classification.

Abstract

Backgrounds: Gastric cancer (GC) is a prevalent malignancy affecting the digestive system. We aimed to develop a risk prediction model based on endoscopic atrophy classification for GC.

Methods: We retrospectively collected the data from January 2020 to October 2021 in our hospital and randomly divided the patients into training and validation sets in an 8:2 ratio. We used multiple machine learning algorithms such as logistic regression (LR), Decision tree, Support Vector Machine, Random forest, and so on to establish the models. We employed the Least absolute shrinkage and selection operator (LASSO) to screen variables for the LR model. However, we chose all the variables to construct the models for other machine learning algorithms. All models were evaluated using the receiver operating characteristic curve (ROC), predictive histograms, and decision curve analysis (DCA).

Results: A total of 1156 patients were selected for the analysis. Five variables, including age, sex, family history of GC, HP infection status, and Kimura-Takemoto Classification (KTC), were screened using LASSO analysis. The area under the curve (AUC) of all the machine learning models ranged from 0.762 to 0.974 in the training set and from 0.608 to 0.812 in the validation set. Among them, the LR model exhibited the highest AUC value (0.812, 95%CI: 0.737-0.887) in the validation set with good calibration and clinical applicability. Finally, we constructed a nomogram to demonstrate the LR model.

Conclusions: We established a nomogram based on endoscopic atrophy classification for GC, which might be valuable in predicting GC risk and assisting clinical decision-making.