The value of endoscopic surveillance of Barrett's esophagus and the appropriate management of high-grade dysplasia remain unclear. Seventeen patients who were referred from endoscopic surveillance programs for management of high-grade dysplasia or adenocarcinoma developing in Barrett's esophagus were compared with 35 patients who had a newly recognized Barrett's adenocarcinoma, who had not been in a surveillance program. The referral diagnosis in the surveyed group was adenocarcinoma in six and high-grade dysplasia in 11. After repeat endoscopy with aggressive biopsy, two additional patients with adenocarcinoma were identified. Of the nine patients who underwent esophagectomy for high-grade dysplasia, five had invasive adenocarcinoma in the esophagectomy specimen, which had been missed before the operation, despite the fact that the median number of biopsy specimens obtained per 2 cm of Barrett's mucosa was 7.8 (range 1.5 to 15.0). Overall, 13 patients in the surveyed group had adenocarcinoma, 12 staged early and one staged intermediate by the WNM classification. Surveyed patients were operated on at an earlier stage than the nonsurveyed patients (10 early, 14 intermediate, and 11 late stage tumors; chi 2 = 15.6, p < 0.01). Despite the presence of adenocarcinoma in 13 of the 17 surveyed patients, their survival was significantly better than that of the nonsurveyed group (chi 2 = 5.8, p < 0.05). Patients referred from surveillance programs for Barrett's esophagus have a better outcome and earlier stage tumors than nonsurveyed patients. Inasmuch as multiple biopsy procedures do not exclude the presence of adenocarcinoma, continued surveillance of high-grade dysplasia is dangerous and potentially destructive to surveillance efforts.