Esophageal varices: correlation of left gastric venography and endoscopy in patients with portal hypertension.

Selective left gastric venography was performed via percutaneous transheptic catheterization in 100 patients with portal hypertension to study hemodynamics and left gastric collaterals. Flow was hepatofugal in 81 patients, while 63 had varices supplied by the left gastric vein; azygos/hemiazygos collaterals were seen in 27 patients, gastrorenal shunts in 18, and other collaterals in 58. In 79 cases, enlarged, tortuous veins arising from thin parallel vessels at the esophagogastric junction on venography corresponded with varices on endoscopy in both size and course. Venography failed to show small varices which were seen endoscopically, whereas endoscopy often interpreted non-tortuous veins as varices. Using endoscopy as a reference, the sensitivity of venography for varices was 76%, specificity 100%, and accuracy 78%. There was a significant correlation between venographic demonstration of varices and previous hematemesis. Left gastric venography is necessary prior to embolization of varices and is also useful in their diagnosis as well as study of hemodynamics and assessment of bleeding risk.