Development and Validation of Predictive Models for Gastric Neoplasm Risk Stratification in Screening Esophagogastroduodenoscopy

食管胃十二指肠镜筛查中胃肿瘤风险分层预测模型的开发与验证

阅读:4

Abstract

BACKGROUND/AIMS: Stratifying patients for gastric neoplasm risk before screening esophagogastroduodenoscopy (EGD) is challenging. The aim of this study was to develop a prediction model for assessing gastric neoplasm risk in a screening setting. METHODS: This retrospective cross-sectional study included 21,586 EGD patients from Seoul St. Mary's Hospital, Korea (2009 to 2019). Logistic regression analyses identified risk factors, and score-based prediction models were developed on the basis of these risk factors. These models were evaluated using the area under the curve (AUC) and the Hosmer?Lemeshow goodness of fit test. Internal validation was performed using bootstrapping (1,000 resamples) and a validation cohort. RESULTS: The study included 10,414 patients in the derivation cohort and 11,172 in the validation cohort. Gastric dysplasia and cancer were identified in 49 (0.47%) and 35 (0.34%) patients, respectively. Four models were developed, with Model 4 including age, sex, pepsinogen I/II ratio, anti-Helicobacter pylori immunoglobulin G antibody, smoking, body mass index, alcohol use, and family history of gastric cancer. Model 4 had the highest AUC (0.827) in the derivation cohort, while Model 2 achieved the highest AUC (0.788) after risk scores were assigned. Observed prevalence rates were 0.24%, 1.05%, and 4.08% for low-, medium-, and high-risk groups, respectively (p<0.001). In internal validation, Model 3 demonstrated the highest AUC (0.802), with consistent performance in the validation cohort, and all models passed the Hosmer-Lemeshow test (p>0.8). CONCLUSIONS: The predictive models achieved an AUC of approximately 0.8. Further improvements with additional stratification factors are needed for better diagnostic performance in prescreening.

特别声明

1、本页面内容包含部分的内容是基于公开信息的合理引用;引用内容仅为补充信息,不代表本站立场。

2、若认为本页面引用内容涉及侵权,请及时与本站联系,我们将第一时间处理。

3、其他媒体/个人如需使用本页面原创内容,需注明“来源:[生知库]”并获得授权;使用引用内容的,需自行联系原作者获得许可。

4、投稿及合作请联系:info@biocloudy.com。