Preoperative insulin secretion ability and pancreatic parenchymal thickness as useful parameters for predicting postoperative insulin secretion in patients undergoing pancreaticoduodenectomy

Abstract

Periampullary malignant neoplasms have been increasing in Japan, mainly in response to an increase in the incidences of pancreatic cancer, and glucose intolerance due to deterioration of insulin secretion is an important problem. We investigated preoperative parameters to predict postoperative insulin secretion and the need for insulin therapy in patients undergoing pancreaticoduodenectomy (PD). Thirty-six patients with malignant neoplasms of periampullary lesions were enrolled. Preoperative pancreatic parenchymal thickness was evaluated by computed tomography. Insulin secretion and glucose tolerance were evaluated by a 75-g oral glucose tolerance test and an intravenous glucagon loading test. The relationships between postoperative insulin secretion and preoperative parameters and the cut-off values for predicting the need for postoperative insulin therapy for glycemic control were investigated. Pancreatic parenchymal thickness and other preoperative parameters, including the increment of serum C-peptide (Δ C-peptide), fasting plasma C-peptide (F-CPR), insulinogenic index (I.I.) and fasting plasma glucose (FPG), were significantly associated with postoperative insulin secretion. Multiple regression analyses revealed that preoperative Δ C-peptide or F-CPR was the most significant determinant of postoperative insulin secretion, followed by pancreatic parenchymal thickness. In the receiver operating characteristic curve, the best preoperative cut-off values for predicting the need for postoperative insulin therapy were a Δ C-peptide of 0.65 ng/mL, a F-CPR of 0.85 ng/mL and a pancreatic parenchymal thickness of 6.0 mm. Both preoperative insulin secretion and pancreatic parenchymal thickness effectively predict postoperative insulin secretion and identify subjects who need postoperative insulin therapy for glycemic control.