Receiver operator characteristic analysis of endoscopy as a test for gastritis

Abstract

Endoscopic evaluation of the presence or absence of gastritis is often performed in lieu of biopsy and histologic diagnosis. The purpose of our study was to assess the value of endoscopic examination as a diagnostic test for gastritis. Two endoscopists prospectively assessed the antrum of 73 patients undergoing upper gastrointestinal endoscopy and graded, on a scale of 0-4 (0 = completely absent, 4 = definitely present), the likelihood of gastritis. The following features were also assessed at the time of endoscopy: erythema, nodularity, erosion, edema, and friability. Two concomitant antral biopsies (3 cm from the pylorus on the greater curvature of the stomach) were performed regardless of the endoscopic impression. The histologic findings were graded independently on a scale of 0-3 by two pathologists who were not aware of the endoscopic findings. The following histologic features were graded: acute inflammation, chronic inflammation, lymphoid aggregates, intestinal metaplasia, and quantity of Helicobacter pylori organisms. Receiver operator characteristic analysis, a method derived from signal detection theory, assesses the trade-off of sensitivity and specificity over all cutoff points of a test and is considered the best method by which to compare tests and determine the diagnostic utility of a given test. Receiver operator characteristic analysis gave an area of 0.65 +/- 0.01 SE for endoscopy as a test for gastritis (0.5 = chance, 1 = perfect) as defined by the histologic presence of inflammation. Additionally, endoscopy as a test for the presence of histologically proven Helicobacter pylori gave an area of 0.55 +/- 0.01 SE. All endoscopically graded features treated as separate tests for gastritis and/or H. pylori gave areas of approximately 0.44-0.61, indicative of a poor test. While H. pylori was always associated with at least some degree of inflammation, linear regression analysis revealed no correlation among any of the histologic features or of any histologic feature with any endoscopic feature. We conclude that a tissue diagnosis is essential for the proper diagnosis of gastritis.