Aspects on endoscopic characterization and clinical management of Barrett's esophagus

Abstract: Barrett's esophagus (BE) is considered to result from prolonged gastroesophageal reflux and is the only known precursor of esophageal adenocarcinoma. The clinical management of BE patients aims to control esophageal reflux to reduce mucosal injury and neoplastic progression, and to detect early neoplastic lesions in Barrett’s mucosa, suitable for curative endoscopic treatment. The first part of this thesis evaluates the effect of a stepwise increase in the dose of proton pump inhibitors (PPI), on esophageal acidic reflux, symptoms and histology in long segment BE patients (group 1, n=24). We also compare these outcomes in BE patients under PPI with the results of BE patients after clinically successful fundoplication (group 2, n=30). In all but one patient in group 1, it was possible to normalize acid reflux with PPI, resulting in improvements in symptom scores. However, symptomatic amelioration was only significant in the first step of PPI treatment. Patients with PPI or fundoplication had the same levels of symptom scores. Normalization of the acid reflux in both groups was associated with reductions of papillary length, thickness of the basal cell layers, dilation of intercellular spaces, and acute and chronic inflammation of the squamous epithelium. We did not find a significant change in markers of proliferation and differentiation in Barrett’s mucosa associated with normalization of acid reflux in either group. The second part of this thesis assesses 3 different endoscopic classification systems, Amsterdam, Kansas and Nottingham, developed for the characterization of Barrett’s mucosa. These classifications use magnification endoscopy with narrow band imaging (ME-NBI) for the identification of intestinal metaplasia and dysplasia in Barrett’s mucosa. We used 84 video segments from Barrett’s mucosa, that were randomly selected and blindly evaluated by 9 observers with different expertise in the field. All classifications were feasibly but showed suboptimal accuracy and low inter-observer agreement, with slightly better results for the Amsterdam classification. The last part of this thesis evaluates the role of a structured learning program for the application of the Amsterdam classification system. We used the first 70 videos from the 84 randomly selected videos from the previous study. While, during the learning process, there was a decrease in the time spent for evaluation and an increase in declared certainty of prediction, the accuracy in histological prediction did not improve. This classification system was found to be suboptimal in terms of accuracy and inter- and intra-observer agreements. This thesis shows that, in long segment BE patients, acid reflux and symptom scores correlated through several steps of the PPI treatment process, achieving the same level as after a successful fundoplication. If a single dose of PPI is associated with marked improvement of symptoms, higher doses still may be needed for complete acid suppression. Minor changes were found among morphological markers of reflux disease, both in the glandular and in the squamous epithelium, irrespective of medical or surgical treatment. Our results underscore the questionable utility of ME-NBI classification systems for clinical routine practice in BE.