HYPOTHESIS High-grade dysplasia (HGD) of the gastric epithelium is associated with high prevalence of invasive carcinoma, and distinction by endoscopic biopsy is difficult. DESIGN Cohort study, 1996 to 2003. SETTING Tertiary care center. PATIENTS Consecutive sample of 22 patients with initial diagnosis of gastric HGD by endoscopic biopsy. Biopsy specimens were separately reviewed by 3 experienced pathologists. Clinical management was individually decided. MAIN OUTCOME MEASURES Strength of interpathologist agreement (kappa) and final pathological diagnosis. RESULTS The diagnosis was revised to intramucosal carcinoma in 14% to 32% of patients or suspicious for invasive carcinoma in 23% to 41%. The strength of agreement between any 2 pathologists for distinguishing between dysplasia and invasive carcinoma was fair (kappa = 0.35-0.36). A diagnosis of intramucosal carcinoma or suspicious for invasive carcinoma by 2 pathologists correlated strongly with subsequent detection of invasive carcinoma. Three patients underwent gastrectomy for HGD, and invasive carcinoma was detected in all (2 patients, T1 N0; 1 patient, T2 N0). Six patients had invasive carcinoma on endoscopic surveillance at a median of 15 months (range, 3-34 months) after diagnosis of HGD and underwent endoscopic mucosal resection (2 patients, T1 NX), gastrectomy (2 patients, T1 N0), or no resection (2 patients). Another patient had metastatic gastric adenocarcinoma despite having a diagnosis of only HGD by endoscopy. Seven patients (32%) died of unrelated causes, without invasive carcinoma, at a median of 19 months (range, 1-38 months). Three patients were alive with persistent HGD at 26 to 61 months. Two patients had no dysplasia on follow-up. CONCLUSIONS Experienced pathologists often disagreed in distinguishing invasive carcinoma from HGD in gastric biopsy specimens. One third of patients with gastric HGD died of causes unrelated to cancer. Invasive carcinoma was detected in 67% of the remainder.