AIM: To study characteristics of collateral circulation of gastric varices (GVs) with 64-row multidetector computer tomography portal venography (MDCTPV). METHODS: 64-row MDCTPV with a slice thickness of 0.625 mm and a scanning field from 2 cm above the tracheal bifurcation to the lower edge of the kidney was performed in 86 patients with GVS diagnosed by endoscopy. The computed tomography protocol included unenhanced, arterial and portal vein phases. The MDCTPV was performed on an AW4.3 workstation. GVs were classified into three types according to Sarin's Classification. The afferent and efferent veins of each type of GV were observed. RESULTS: The afferent venous drainage originated mostly from the left gastric vein alone (LGV) (28/86, 32.59%), or the LGV more than the posterior gastric vein/short gastric vein [LGV > posterior gastric vein/short gastric vein (PGV/SGV)] (22/86, 25.58%), as seen by MDCTPV. The most common efferent venous drainage was via the azygos vein to the superior vena cava (53/86, 61.63%), or via the gastric/splenorenal shunt (37/86, 43.02%) or inferior phrenic vein (8/86, 9.30%) to the inferior vena cava. In patients with gastroesophageal varices type 1, the afferent venous drainage of GV mainly originated from the LGV or LGV > PGV/SGV (43/48, 89.58%), and the efferent venous drainage was mainly via the azygos vein to the super vena cava (43/48, 89.58%), as well as via the gastric/splenorenal shunt (8/48, 16.67%) or inferior phrenic vein (3/48, 6.25%) to the inferior vena cava. In patients with gastroesophageal varices type 2, the afferent venous drainage of the GV mostly came from the PGV/SGV more than the LGV (PGV/SGV > LGV) (8/16, 50%), and the efferent venous drainage was via the azygos vein (10/16, 62.50%) and gastric/splenorenal shunt (9/16, 56.25%). In patients with isolated gastric varices, the main afferent venous drainage was via the PGV/SGV alone (16/22, 72.73%), and the efferent venous drainage was mainly via the gastric/splenorenal shunt (20/22, 90.91%), as well as the inferior phrenic vein (3/23) to the inferior vena cava. CONCLUSION: MDCTPV can clearly display the afferent and efferent veins of all types of GV, and it could provide useful reference information for the clinical management of GV bleeding. (C) 2010 Baishideng. All rights reserved.