Increasingly successful operative management of gastroesophageal variceal hemorrhage has been achieved by newer techniques of portal venous reconstruction. Although it is postulated that the clinical success may be due to more selectivity in portal venous shunting, direct determination of the effect of portasystemic shunt on portal vein blood flow has not been possible. Direct determinations of portal vein blood flow were performed preoperative on unanesthetized, hemodynamically stable cirrhotic patients by observation of radiopaque water-insoluble droplets. Patients were then randomized into elective distal splenorenal (Warren) or mesocaval shunt and determinations were performed postoperatively under similar conditions when clinically possible. Although portal vein blood flow was not significantly different before (929 +/- 147 ml/min) or after 899 +/- 271 ml/min) distal splenorenal shunt, there was a large change in portal vein blood flow after mesocaval shunt, decreasing from 772 +/- 177 ml/min (hepatopetal) to -1021 +/- 310 ml/min (hepatofugal) p < 0.01). After either procedure total hepatic blood flow (as determined by cardiac green clearance) was not significantly changed, nor was renal blood flow; however, cardiac output was significantly increased after mesocaval shunt. Thus the theoretical hemodynamic goals of the selective distal splenorenal shunt, i.e., preservation of the hepatopetal flow within the portal vein, is achieved as determined in the early postoperative period. The correlation between these changes and the eventual clinical outcome remains to be determined.