INTEROBSERVER AGREEMENT ON ENDOSCOPIC DIAGNOSIS OF DIFFUSE GASTRIC-MUCOSAL FEATURES IN PATIENTS WITH CIRRHOSIS

Objective: Although diffuse gastric mucosal features have frequently been described in upper gastrointestinal endoscopy, they have never been subjected to a reliability study. Therefore, the aim of this study was to measure the interobserver agreement for gastric mucosal features in patients with cirrhosis, due to the high prevalence of these features in this condition, by means of the kappa index. Design: Prospective study on interobserver agreement in gastric endoscopy. Setting: Consecutive patients seen in an endoscopy unit of a university hospital. Patients, participants: Seventy patients with cirrhosis and seven endoscopists (four seniors and three juniors) were enrolled in this study. Intervention: Endoscopic examinations were recorded on videotape with a standardized technique by the same observer. Endoscopists were asked to judge the presence and location of nine elementary signs (erythema, atrophy, large folds, oedema, mosaic pattern, red spots, erosions, dark spots and microbleeding) and to identify the presence, location and type of three syndromes: gastritis, congestive gastropathy and antral vascular ectasia. Moreover, endoscopists were asked to judge the presence of gastric-associated lesions and varices as well as oesophageal varices and red signs. Results: Interobserver agreement was poor for all diffuse gastric mucosal features except for dark spots and microbleeding. Agreement was good for oesophageal features and associated gastric features. Agreement was not related to the experience of the endoscopists or to the quality of the examination. Agreement slightly improved when the presence of the sign was graded into two categories instead of three. According to a questionnaire, endoscopists greatly overestimated the reliability of their judgement. The mean kappa indexes between the first and second portions of patients studied did not vary significantly but the correlation of kappa indexes between these two series was low (r = 0.64) suggesting that agreement studies should involve a large number of patients. Conclusions: The endoscopic diagnosis of diffuse gastric mucosal features is hampered by poor interobserver agreement. These features should be more precisely defined, with a simple and consensual classification. Finally, the training for the identification of gastric mucosal features by endoscopy needs improvement.