OBJECTIVE Inflammatory bowel disease surveillance strategies are designed to identify patients at greater than average risk for the development of invasive colonic carcinoma. Colonoscopic detection of mucosal dysplasia is considered the best available surveillance tool. However, the usefulness of dysplasia as a marker for cancer is uncertain. Furthermore, when dysplasia is found some suggest immediate colectomy, whereas others opt for continued surveillance. The aim of this study is to determine whether an association between dysplasia grade and cancer exists in patients with chronic ulcerative colitis, to ascertain the sensitivity, specificity, and positive predictive value of dysplasia as a cancer marker, and to clarify what action to take once dysplasia is discovered. METHODS The pathology reports of 590 patients who underwent total proctocolectomy or restorative proctocolectomy for chronic ulcerative colitis were reviewed for dysplasia, grade of dysplasia, presence of carcinoma, and tumor stage. One hundred sixty of these patients had undergone colonoscopic examination within the year before surgery. Findings from these studies were also reviewed. RESULTS Seventy-seven specimens (13.1 percent) contained at least one focus of dysplasia. Invasive cancers were found in 38 specimens (6.4 percent). Cancers were significantly more common among specimens with dysplastic changes (33/77 vs. 5/513; P < 0.001). Specimens with dysplasia of any grade were 36 times more likely to harbor invasive carcinoma. Stage III disease was found in association with indefinite or low-grade dysplasia in 5 of 26 (19.2 percent) of cases. Tumor stage did not correlate with dysplasia grade. Preoperative colonoscopy identified neoplastic changes in 57 (69.5 percent) cases. Dysplasia, cancer or both were missed in 25 cases. Lesions were correctly identified in only 31 (39.7 percent) of cases. Colonoscopically diagnosed dysplasia as a marker for synchronous cancer had a sensitivity of 81 percent and a specificity of 79 percent. The positive predictive value of a finding of preoperative dysplasia of any grade was 50 percent. The positive predictive value of a finding of low-grade dysplasia was 70 percent. CONCLUSIONS Dysplasia is an unreliable marker for the detection of synchronous carcinoma. However, when dysplasia of any grade is discovered at colonoscopy, the probability of a coexistent carcinoma is relatively high. Colonoscopic evidence of low-grade dysplasia has a higher positive predictive value than either dysplasia associated mass or lesion or high-grade dysplasia. Dysplasia grade does not predict tumor stage. Because advanced cancer can be found in association with dysplastic changes of any grade, confirmed dysplasia of any grade is an indication for colectomy.