BACKGROUND AND STUDY AIMS:: There is a growing trend toward the use of minimally invasive endoscopic methods to treat early esophageal cancers. Although there is continuing controversy regarding the management of Barrett's esophagus and the value of surveillance programs continues to be debated, the ultimate goal is to eradicate all of the foci of intestinal metaplasia and hence the risk of developing an adenocarcinoma. A number of ablative techniques have so far been applied, but none has yet been shown to be superior and entirely satisfactory. The present study evaluates the feasibility, efficacy, and safety of a promising new method of endoscopic mucosal resection (EMR) in a sheep model, based on the use of a modified rigid esophagoscope. MATERIALS AND METHODS: The resectoscope consists of a rigid esophagoscope with a distal transparent window through which the mucosa and part of the submucosa are sucked in and then resected with a wire loop. The sheep model was chosen because of its similarities to human anatomy with regard to the thickness and histological structure of the esophagus. Fifty-five separate hemicircumferential resections and 11 circumferential resections were carried out in 21 and 11 animals, respectively. Mitomycin C, an agent inhibiting fibroblast proliferation, was administered at different time intervals after eight circumferential resections to prevent the development of esophageal strictures. Results : All of the specimens of hemicircumferential resections were obtained as single distinct pieces and were easily examined histologically. The surface of the specimen correlated with the size of the window and ranged from 6 to 12 cm (2). In circumferential resections, the specimens were obtained in two pieces. An accurate resection depth through the submucosa was achieved in 58 of 65 resected specimens. No complications occurred after hemicircumferential resections. Complications after circumferential resections (stenosis or perforation, or both) were minimized after appropriate timing of mitomycin C administration. CONCLUSIONS: This EMR method offers a promising approach in comparison with other options currently available. It appears to be superior in terms of the size of the resected specimen, the precision and regularity of the resection depth, and the accuracy of histological diagnosis with safety margins. Hemicircumferential EMRs have been shown to be safe in the sheep model. This new technique warrants further animal studies before being used for circumferential EMR in humans