Algorithm Training and Testing for a Nonendoscopic Barrett's Esophagus Detection Test in Prospective Multicenter...

Area of Science:

• Gastroenterology
• Oncology
• Molecular Diagnostics

Background:

• Endoscopic detection of Barrett's esophagus (BE) and esophageal adenocarcinoma (EAC) is invasive and costly.
• Non-endoscopic methods are guideline-recommended alternatives for BE/EAC detection.
• A previously developed 5-methylated DNA marker (MDM) panel utilized cell collection device (CCD) specimens.

Purpose of the Study:

• To train and validate a new algorithm using a 3-MDM panel for BE/EAC detection.
• To assess the performance of the 3-MDM panel algorithm in an independent cohort.
• To establish a less invasive diagnostic tool for BE and EAC.

Main Methods:

• Algorithm training and testing were conducted on samples from two prospective, multicenter cohorts.
• DNA was extracted from cell lysates collected via CCD, bisulfite treated, and MDMs were assayed.
• A logistic regression model was trained and locked using cross-validation, then tested on an independent dataset.

Main Results:

• The final 3-MDM panel (NDRG4, VAV3, ZNF682) demonstrated high sensitivity and specificity in both training and test sets.
• Overall sensitivity for BE detection was 82%-88%, with 100% sensitivity for high-grade dysplasia and EAC.
• Areas under the receiver operating characteristic curves (0.92-0.94) indicate strong diagnostic performance.

Conclusions:

• A validated 3-MDM panel algorithm using non-endoscopic CCD shows excellent sensitivity for detecting high-risk BE.
• This non-invasive approach provides a promising tool for BE/EAC screening and diagnosis.
• The study provides strong evidence for the clinical utility of this novel diagnostic method.