Since Whipple's successful resection of the head of the pancreas and duodenum in 1935, pancreaticoduodenectomy has become a standard operation for periampullary malignancies. Although the operative mortality has decreased dramatically in the recent years, it continues to be associated with high morbidity; with anastomotic leakage remaining a major problem.One hundred and seventy-six pancreaticoduodenectomies performed for periampullary lesions during the past 27 years were reviewed. These included 171 Whipple operations, 4 total pancreatectomies and 1 pylorus-preserving pancreaticoduodenectomy. Among them, 40 cases were complicated with anastomotic leakage following pancreaticoduodenectomy. Their management strategy and outcome were reevaluated. Furthermore, the factors suggested to affect pancreaticojejunostomy leakage were also analyzed.The overall operative mortality was 15.3%, which decreased to 6.7% in recent two years. However, a high complication rate of 50% remained. Among the complications of the whole series (46.6%), anastomotic leakage accounted for 22.7% (40/176). Leakage occurred in 28 pancreaticojejunostomies (16.3%), 9 hepaticojejunostomies (5.1%) and 6 gastrojejunostomies (3.4%). Twelve patients required reoperation for ongoing sepsis or bleeding. This experience disclosed that in most cases hepaticojejunostomy leakage (8/9) could be successfully managed without operation. While three of the six gastrojejunostomy leaks survived after conservative treatment, two of the remaining three patients operated died of sepsis. Among cases with pancreaticojejunostomy leakage, 12 survived after conservative treatment, whereas 6 died of sepsis. Among 10 operated patients, only 3 patients survived. Earlier reexploration for uncontrolled leakage, probably within the first eight postoperative days, seemed to be the only chance for life saving. As far as the risk factors of pancreaticojejunostomy leakage are concerned, there seemed to exert no significant influence in terms of intraoperative blood loss, type and sequence of anastomosis as well as pancreatic stenting. The only clue that may affect the surgical outcome is technical; more experienced (> or = 10 Whipple operations) surgeons tended to render less morbidity and mortality.The retrospective analysis of our experience in pancreaticoduodenectomy discloses a trend toward decreased mortality rates in the recent years but operative complications remain high. Among the possible complications, anastomotic leakage is still a troublesome concern. Although conservative treatment can benefit most patients, earlier reexploration for uncontrolled sepsis should be considered. If a good result is anticipated, this complicated procedure should only be performed by an experienced surgeon.