Model May Predict Barrett Esophagus Diagnosis in GERD Patients

A diagnostic model can help predict diagnosis of Barrett esophagus in patients with symptomatic gastroesophageal reflux disease, according to a study published online in The Lancet Digital Health.

Avi Rosenfeld, Ph.D., from the Jerusalem College of Technology in Israel, and colleagues used data from two case-control studies (BEST2 and BOOST) to develop a risk prediction panel to screen for Barrett esophagus. Questionnaires were analyzed from 1,299 patients from BEST2, including 67.7 percent with Barrett esophagus; the cohort was split (6:4) into training and testing data sets. An external validation cohort was compiled from the BOOST cohort, with 398 patients, including 198 with Barrett esophagus.

The researchers identified 40 diagnostic features in the BEST2 study. After correlation-based feature selection, eight showed independent diagnostic value (age, sex, cigarette smoking, waist circumference, frequency of stomach pain, duration of heartburn and acidic taste, and taking antireflux medication). Except for the frequency of stomach pain, all were correlated with an increased risk for Barrett esophagus. The highest prediction quality was with logistic regression, which had an area under the receiver-operator curve (AUC) of 0.87. In the testing and external validation data sets, the AUCs were 0.86 and 0.81, respectively.

"The results would identify a high-risk group of people at an early stage, who could then go on to have clinical screening to diagnose and treat esophageal cancer at a much earlier stage and significantly improve survival rates," a coauthor said in a statement.

One author designed the cytosponge device with her research team; the technology was licensed to Covidien GI Solutions, which is now part of Medtronic.

Editor's Note:

En-CPhI.CN is a vertical B2B online trade platform serving the pharmaceutical industry,

for any copyright disputes involved in the reproduced articles,

please email: Julia.Zhang@ubmsinoexpo.com to motify or remove the content.