PURPOSE: To prospectively evaluate the accuracy of esophagography with barium in diagnosis of esophageal varices (EV) in patients with compensated cirrhosis, with endoscopic gastroduodenoscopy as the reference standard. MATERIALS AND METHODS: In this study, which was approved by the local Helsinki Committee and in which all patients consented to participate, 61 patients with cirrhosis (34 men, 27 women; mean age, 61 years; range, 36-76 years) received a diagnosis clinically or with liver biopsy. In 87% (n = 53) of patients, Child-Pugh classification was A; in 13% (n = 8), Child-Pugh classification was B. They were evaluated with endoscopic gastroduodenoscopy, according to Japanese general criteria. Esophagography was performed within 3 weeks of endoscopic gastroduodenoscopy, and EV were assigned grades as follows: 0, no EV were seen; 1, EV manifested as very mild irregularities of the folds; and 2, the irregularity of the folds (EV) was clearly present. They were also assigned grades for shape and size: grade F0, no EV detected; grade F1, small straight EV; grade F2, slightly enlarged tortuous EV occupying less than one-third of esophageal lumen; and grade F3, large coil-shaped EV that occupied more than one-third of esophageal lumen. The sensitivity and specificity and positive and negative predictive values of esophagography for identification of each grade of EV were calculated separately, as was the 95% confidence interval. RESULTS: All large EV (grades F2 and F3) were diagnosed at esophagography. Sensitivity declined with small EV (grade F1) to 71. The overall sensitivity of esophagography was 89% (95% confidence interval: 75.9%, 96.5%), the overall specificity was 83% (95% confidence interval: 64.5%, 94.7%), the overall positive predictive value was 89%, and the overall negative predictive value was 83% (95% confidence interval: 64.5%, 94.7%). Overall accuracy was 87%. CONCLUSION: Esophagography is highly accurate in diagnosis of EV and can be considered a viable noninvasive alternative for determination of patients who should be selected for prophylactic treatment.