There is uncertainty regarding the value of endoscopic biopsy surveillance in Barrett's esophagus because, in retrospective studies, some patients with high-grade dysplasia in endoscopic biopsy specimens have had unexpected advanced adenocarcinoma discovered at the time of esophageal resection. We compared the accuracy of preoperative endoscopic biopsy diagnoses with the final pathologic diagnoses in esophagectomy specimens in 4 patients who had both high-grade dysplasia and intramucosal carcinoma and 4 other patients who had only high-grade dysplasia preoperatively. The histologic lesions in all 8 patients were documented in intact mucosa with no gross evidence of neoplasia by endoscopy. The preoperative diagnoses were defined with an endoscopic biopsy protocol in which specimens were taken with large-channel biopsy forceps at least every 2 cm throughout the length of Barrett's epithelium. Final pathologic diagnoses derived from detailed analysis of the resected specimens confirmed high-grade dysplasia without carcinoma in 4 patients and intramucosal carcinoma in 2 patients. The remaining 2 patients with a preoperative diagnosis of intramucosal carcinoma had focal submucosal invasion by carcinoma in the resected specimens, but no involvement of the muscularis propria or adventitial lymph nodes. Because the natural history of high-grade dysplasia is not known, the decision to operate on patients with this lesion must be carefully weighed and individualized for each patient. Two of our patients who underwent esophageal resection for high-grade dysplasia without cancer died, one immediately postoperatively and the other 9 mo later after a postoperative stroke. Once intramucosal carcinoma is documented, surgery should be considered if the patient is an acceptable operative risk. We conclude that systematic preoperative endoscopic biopsy of intact mucosa in Barrett's esophagus can correctly detect high-grade dysplasia, either alone or in combination with early, treatable adenocarcinoma.