Gastroesophageal endoscopic features in cirrhosis. Observer variability, interassociations, and relationship to hepatic dysfunction.

Nowadays, gastroesophageal endoscopic features of portal hypertension are the recognized predictive factors for bleeding and consequently allow the selection of patients for prophylactic therapies. The aim of this prospective study was to investigate the interobserver agreement, the interassociations between these features, and the relationship between these signs and the degree of hepatic dysfunction. In 100 consecutive cirrhotic patients (84% with alcoholism) without history of digestive bleeding, gastroesophageal endoscopic examination was performed and recorded using a videoendoscope. Four independent observers evaluated the following endoscopic features: the size, extent, color, and red signs of esophageal varices, the mosaic pattern, congestive gastropathy, fundic varices, and associated lesions of the stomach. Agreement was assessed using kappa statistics (kappa) and a quantitative score. The size of esophageal varices was significantly associated with their extent and the presence of red signs, whereas no relation was found either between gastropathy or mosaic pattern and fundic varices, or between esophageal and gastric features. Agreement between observers was good for the size of esophageal varices (kappa = 0.59), the presence of red signs (kappa = 0.60), and of gastric-associated lesions (kappa = 0.68) and gastropathy (kappa = 0.50), while it was poor for the extent (kappa = 0.37) and the color (kappa = 0.28) of esophageal varices as well as for the mosaic pattern (kappa = 0.38). The Child-Pugh score significantly increased along with the presence or the size of esophageal varices as well as with the presence of red signs; no relationship could be shown between this score and the presence of gastric features. We conclude that (1) interobserver agreement was good for the main endoscopic features, especially for the size and the red signs of esophageal varices; (2) esophageal patterns were significantly associated between themselves and related to hepatic dysfunction; and (3) gastric patterns were related neither to esophageal features nor to hepatic dysfunction and were not associated between themselves.