Interobserver agreement and predictive value of endoscopic findings for H. pylori and gastritis in normal volunteers

Abstract

Background: Endoscopic findings such as erythema are frequently labeled as gastritis. We sought to determine interobserver agreement for specific endoscopic features and assess the diagnostic value of features with good agreement for Helicobacter pylori and histologic gastritis. Methods: Fifty-two healthy subjects without ulcers, erosions, or hemorrhages had a full endoscopy recorded on video tape. Biopsy specimens were examined for H. pylori and gastritis. Two endoscopists independently reviewed the tapes for predefined features (erythema, area gastricae, clefts, and nodularity) in the gastric body and antrum. Diagnostic value of endoscopic features with acceptable agreement (κ > 0.40) was then determined for H. pylori and gastritis. Results: κ was greater than 0.40 only for prominent body area gastricae (0.49), body nodularity (0.65), and antral nodularity (0.68). For antral nodularity, sensitivity was 32%, specificity was 96%, and positive predictive value was 90% for H. pylori . When both antral nodularity and body area gastricae were both present, sensitivity was only 18% but specificity and positive predictive value were 100%. Conclusion: Interobserver agreement is poor for some features such as erythema labeled as gastritis. Antral nodularity is a fairly reproducible finding and is very specific, though not sensitive, for H. pylori gastritis. (Gastrointest Endosc 1995;42:420-3.)