Hepatic arterial interruption inevitably leads to fatal liver hypoxia when all the collateral arteries to the liver have been eradicated. To prevent such hypoxia, we aimed to determine the appropriate flow of arterial blood in the arterio-portal shunt (APS). After division of all the arteries to the liver, we created three types of APSs between the common hepatic artery (CHA) and the portal vein in dogs, using catheters which were adjusted to pass blood flows of approximately half (group I), equal to (group II) and twice (group III) the CHA blood flow before shunting, except in the control (no-shunt) group. Postoperatively, at 1 and 48 h, we examined the hemodynamics of the liver biochemically and pathologically. After shunting, portal blood flow and oxygen saturation markedly increased, whereas portal venous pressure did not rise significantly. The serum alanine aminotransferase level was significantly higher in the no-shunt group and group I than in group II. Only in group II was the preoperative energy charge maintained, and light- and electron-microscopic examinations revealed no degeneration of the hepatocytes. APS blood flow similar to the original CHA (as in group II) is most appropriate for preventing liver hypoxia.