Background: The main objective of this prospective study was to evaluate the independent diagnostic accuracy of gastroesophageal endoscopic signs for cirrhosis.
Methods: Endoscopic signs were evaluated in vivo by one observer and on standardized videotape recordings by a consensus opinion of two endoscopists, with the stomach and esophagus examined both separately ("blind video") and together ("unblind video") in 168 consecutive patients with cirrhosis (n = 91), with non-cirrhotic liver disease (n = 29), and without liver disease (n = 48). The results were then tested in 149 different patients.
Results: Discriminant analysis of the "blind video" examination, considered to be the reference examination, showed that esophageal varices and portal hypertensive gastropathy had independent diagnostic accuracy (> or = 87%) for cirrhosis regardless of control group. However, in the "unblind video" and in in vivo examinations with different control groups, these results were altered. Prior knowledge of gastric patterns influenced the assessment of esophageal patterns. There was no independent association between portal hypertensive gastropathy or esophageal varices and the etiology of liver disease or the Child-Pugh score.
Conclusions: When methodologic biases are eliminated, esophageal varices and portal hypertensive gastropathy have independent diagnostic accuracy for cirrhosis: esophageal varices had a diagnostic accuracy of 77% at the first step and increased to 89% with portal hypertensive gastropathy at the second step.