Risk Prediction of Barrett's Esophagus in a Taiwanese Health Examination Center Based on Regression Models

Area of Science:

• Gastroenterology
• Oncology
• Medical Informatics

Background:

• Barrett's esophagus (BE) is a precancerous condition for esophageal adenocarcinoma.
• Identifying target populations for BE screening in Asia presents a significant challenge.

Purpose of the Study:

• To develop and compare risk prediction models for BE using logistic regression (LR) and artificial neural network (ANN) methods.
• To assess the models' predictive performance for identifying individuals at high risk of BE.

Main Methods:

• Retrospective analysis of 9,646 adults (≥20 years) undergoing upper gastrointestinal endoscopy in Taiwan.
• Development of LR and ANN risk prediction models for BE.
• Evaluation using 10-fold cross-validation to determine discriminative power (Area Under Curve - AUC).

Main Results:

• Both LR and ANN models demonstrated good discriminative power.
• Comparable AUC values were achieved for both models, with AUC = 0.702.
• Models were developed using data from individuals with and without clinical indications for endoscopy.

Conclusions:

• The developed risk prediction models can effectively identify high-risk individuals for BE screening.
• These models offer a practical tool for endoscopic screening of the general population in Asia.
• Further application of these models can aid in early detection and prevention of esophageal adenocarcinoma.