ABSTRACT
BACKGROUND: The published prevalence of Barrett's esophagus (BE) varies from 0.9% to 25%, in part because of differences in the endoscopic interpretation of the disease. OBJECTIVE: We studied the accuracy of diagnosis in 130 patients previously labeled as having BE. Our aim was to determine the interobserver consistency of endoscopic findings and assess the percentage of patients with confirmed BE versus those with a revised diagnosis. DESIGN/SETTING/PATIENTS: Patients previously diagnosed with BE of any length and due for surveillance endoscopy were eligible for study. INTERVENTIONS: After intensive consensus anatomic and endoscopic review, study patients underwent endoscopy and biopsy by 1 of 3 endoscopists. BE was defined as any length of columnar-lined esophagus with goblet cells. MAIN OUTCOME MEASUREMENTS: Patients were photographed/videotaped for review by the other 2 endoscopists, and BE was either confirmed or revised. RESULTS: Eighty-eight patients (67.7%) had confirmed BE, and 42 (32.3%) had their diagnosis revised to no BE (95% confidence interval, 24.4%-41.1%) because there was no visible columnar-lined esophagus proximal to the gastric folds or no goblet cells were found on biopsy. BE length, site of previous endoscopy, age, sex, and hiatal hernia size were predictors of revision. All 3 endoscopists agreed on all confirmed BE cases and 38 of 42 of those revised. LIMITATIONS: Retrospective analysis, possible sampling error. CONCLUSIONS: BE is overdiagnosed in clinical practice with important implications for patient care including increased costs, reduced insurability, and psychological stress. The true BE cancer risk may also be underestimated. This study suggests the need for a better definition of the gastroesophageal junction, stricter accountability for BE diagnosis, and improved endoscopic education.