Disruption of the pancreatic anastomosis with resultant sepsis is the cause of nearly 50% of deaths following pancreaticoduodenectomy (PD). Traditionally, the pancreatic remnant is anastomosed to the jejunum. Pancreaticogastrostomy (PG) was introduced as an alternative by Waugh and Clagett in 1946 and by Park, Mackie, and Rhoads in 1967. The purpose of this retrospective review was to assess the safety of PG at a single institution. Between 1986 and 1998 a total of 102 patients underwent PG following PD. The indications for PD were periampullary carcinoma (n = 89), pancreatitis (n = 7), and miscellaneous (n = 6). Altogether, 80 patients underwent the traditional Whipple procedure and 22 the pylorus-preserving Whipple (PPW) procedure. The PG was performed by a single-layer invagination technique to the posterior gastric wall using interrupted silk sutures. Leaks from the pancreatic anastomosis were detected by measuring amylase in fluid obtained from surgically placed drains. Operative mortality was 3.9% (4/102). The cause of death was uncontrolled upper gastrointestinal hemorrhage, sepsis, pulmonary embolus, and cardiac failure secondary to myocardial infarction. The mean operating time was 6.8 hours. Blood transfusion was given in 43 patients (42%), and the mean amount of the transfusion was 2.6 units. Nonfatal complications occurred in 35 patients (34%), and included leaks from the pancreatic anastomosis in 9 (8.8%), leaks from the biliary-enteric anastomosis in 4 (3.9%), and gastric paresis 7 (6.9%). Other complications included abscess, wound infection, colitis, delirium tremens, and hyperbilirubinemia. Discharge occurred 6 to 47 days (median 12 days) postoperatively and was prolonged in patients suffering from a complication. PD is associated with significant morbidity. PG is a safe alternative to pancreaticojejunostomy for managing the pancreatic remnant.