The effectiveness of endoscopic nasobiliary drainage (ENBD) for postoperative bile leakage after hepatic resection was investigated retrospectively. Between 1997 and 2002 a series of 486 hepatectomies without biliary reconstruction were performed. Bile leakage was divided into two categories. Type A was defined as bile leakage communicating with the main bile tree fistulographically or endoscopic cholangiographically, and type B was bile leakage without such a patency of bile flow. Bile leakage developed in 31 patients (6.4%) (types A/B = 16/15). Type A frequently occurred at the major Glisson's sheath. In contrast, most type B cases occurred at the peripheral bile duct at the cut surface of the liver. Among the type A patients, 10 of 11 were effectively treated with ENBD. For the type B patients, 12 of 15 patients were successfully treated with intraabdominal drainage via surgical drains inserted during the operation or percutaneous tubes newly inserted for biliary fluid collection. ENBD was effective in two of three type B patients. The duration of bile leakage significantly shortened after initiation of ENBD in type A patients (15.3 +/- 6.9 vs. 25.8 +/- 13.2 days, p < 0.05). The classification based on communication with the main bile tree is useful for determining therapeutic strategy. Type A leakage has a good indication for ENBD, whereas type B can be treated with intraabdominal drainage in most cases, although ENBD may be effective in some intractable type B cases. It is preferable to initiate ENBD as early as possible to shorten the duration of bile leakage and the subsequent hospital stay.