Risk stratification of Barrett's esophagus: Updated prospective multivariate analysis

OBJECTIVES: Prospective evaluation of Barrett's esophagus (BE) in order to determine what demographic, endoscopic, and histologic features are predictive of the prevalence and incidence of Barrett's high-grade dysplasia (HGD) and adenocarcinoma (Ca). METHODS: Newly diagnosed BE patients were entered into and followed in a standardized surveillance protocol. The following features were examined using either forward, stepwise multiple regression analysis, or Cox proportional hazards to determine their ability to predict the presence of HGD or Ca at index BE diagnosis as well as their ability to predict progression of BE during follow-up: age, race, gender, length of BE in cm, size of a hiatal hernia, severity of dysplasia at index diagnosis as well as during surveillance, gastric Helicobacter pylori infection status, and type of medical acid-reflux treatment. RESULTS: A total of 550 patients were diagnosed with BE over the study period. Stepwise multiple regression analysis showed three factors significantly associated with index diagnosis of HGD or Ca: hiatal hernia (larger size), Barrett's length (longer length), and absence of H. pylori infection. Three hundred and twenty-four BE entered the surveillance protocol. Cox proportional hazards models revealed a significant and independent association for five factors predictive of the time to progression of BE: presence of dysplasia at index diagnosis (p < 0.001), severity of dysplasia during surveillance (p < 0.001), length of Barrett's epithelium (p = 0.012), size of hiatal hernia (p 0.006), and gastric H. pylori infection status (p = 0.023). CONCLUSIONS: Endoscopic and histologic features of BE at initial diagnosis are predictive of index HGD and cancer as well as with risk of BE progression.