Dysplasia as a predictive marker for invasive carcinoma in Barrett esophagus: A follow-up study based on 138 cases from a diagnostic variability study

The objective of endoscopic surveillance in Barrett esophagus (BE) is to as sess the risk of subsequent development of invasive carcinoma. Criteria for morphologic evaluation of dysplasia, the presumed precursor lesion, have b een established, although there are surprisingly few data in the literature correlating biopsy diagnosis of dysplasia with outcome. We collected follo w-up information on 138 patients with BE whose initial endoscopic biopsy sp ecimens had been selected for submission in an interobserver variability st udy performed by 12 pathologists with special interest in gastrointestinal pathology and reviewed blindly twice each by all the participants. Cases we re scored as BE with no dysplasia, atypia indefinite for dysplasia (IND), l ow-grade dysplasia (LGD) high-grade dysplasia (HGD), intramucosal carcinoma , and frankly invasive carcinoma, thus generating 24 scores on each biopsy specimen. Clinical follow-up was obtained and correlated with both the subm itting diagnoses and majority diagnoses. Kaplan-Meier statistics were used to compare both the submitting and majority diagnoses with outcome using de tection or documentation of invasive carcinoma as the endpoint. Using the s ubmitting diagnoses, no invasive carcinomas were detected in 44 Eases diagn osed as BE (median follow-up, 38.5 months). Carcinomas were detected in 4 o f 22 (18%) cases submitted as IND (median progression-free survival of 62 m onths), in 4 of 25 (15%) cases of LGD (median progression-free survival of 60 months), in 20 of 33 cases of HGD (median progression-free survival, 8 m onths), and all 13 (100%) cases submitted as adenocarcinoma. Grade on initi al biopsy correlated significantly with progression to invasive carcinoma ( log-rank P =.0001). Majority diagnosis was achieved in 99 of the cases. Usi ng the majority diagnoses, no invasive carcinomas were found in 50 cases of BE (median follow-up, 48 months), and carcinomas were detected in 1 of 7 ( 14%) IND cases (80% progression-free survival at 2 months), 3 of 15 (20%) L GD (median progression-free survival, 60 months), 9 of 15 (60%) HGD (median progression-free survival, 7 months), and all 12 (100%) carcinoma. Initial grading again correlated significantly with progression to invasive carcin oma (log-rank P =.0001). However, there were 39 cases without a majority di agnosis. Among these, no carcinomas developed in 8 cases with an average sc ore between BE and IND. Carcinomas were detected in 9 of 21 (43%) cases wit h an average score between IND and LGD, and 7 of 10 (70%) cases with an ave rage score between LGD and HGD. There were ulcers in 8 of 39 cases (20%) of the "no-majority" group and in 13 of 99 (13%) of the majority cases. Of 21 total ulcerated cases, cancer was demonstrated in 15 (71%) of these on fol low-up. These data support combining the IND and LGD categories for surveil lance purposes. Cases without dysplasia may be followed up conservatively. The data obtained from submitted diagnoses as opposed to those from blind r eview suggest that knowledge of the clinical findings aids in diagnosis. Th e data also support the assertion that HGD is strongly associated with inva sive carcinoma. Rebiopsy of ulcerated areas should be considered because th ey may harbor malignancy. Histologic grading of dysplasia using established criteria is a powerful prognosticator in BE.