A clinical risk prediction model for Barrett esophagus

Area of Science:

• Gastroenterology and Oncology
• Clinical Epidemiology

Background:

• Barrett esophagus is the sole known precursor to esophageal adenocarcinoma.
• Endoscopic diagnosis of Barrett esophagus is invasive and costly.
• Effective risk stratification is needed to guide screening decisions for patients with gastroesophageal reflux symptoms.

Purpose of the Study:

• To develop and validate a clinical risk prediction model for identifying Barrett esophagus.
• To assist in selecting patients with gastroesophageal reflux symptoms for endoscopic screening.

Main Methods:

• Retrospective analysis of patients with incident nondysplastic Barrett esophagus and control patients with esophageal inflammation.
• Development of a risk prediction model using stepwise backward logistic regression, analyzing predictors separately for men and women.
• External validation of the model's discrimination and calibration using an independent dataset.

Main Results:

• The final risk model incorporated age, sex, smoking status, body mass index, education level, and acid suppressant use.
• The model demonstrated moderate discrimination in the development dataset (AUC = 0.70) and external validation dataset (AUC = 0.61).
• The model was well-calibrated, showing a high correlation between predicted probabilities and observed risks.

Conclusions:

• The developed risk prediction model is a potentially effective clinical tool for guiding endoscopic screening decisions in patients with gastroesophageal reflux symptoms.
• The model's performance suggests it can aid in identifying individuals at higher risk for Barrett esophagus.
• Further implementation could optimize resource allocation for Barrett esophagus screening.