Introduction: Endoscopic ultrasound (EUS) and pancreas protocol computed tomography (CT) are the principal imaging modalities for the diagnosis and staging of pancreatic cancer. Factors associated with poor survival after surgical resection include positive margins and node-positive disease. The purpose of this study is to investigate the ability of EUS and CT to predict a margin negative (R0) and lymph node negative (LN-) resection as well as the need for venous resection in patients undergoing pancreaticoduodenectomy (PD).
Methods: Patients with pancreatic head adenocarcinoma undergoing palliative or curative resection during the last five years were identified from a prospectively collected database. EUS data collected included abutment or invasion of the mesenteric vasculature while CT data included a blinded evaluation of the circumferential degree of vascular involvement as well as the presence of metastases, ascites, local organ invasion, and adenopathy. Patients with incomplete EUS or CT data were excluded. Preoperative imaging was then compared to intraoperative findings and final pathology. Logistic regression was used to identify which EUS and CT features were most important determining surgical results.
Results: One hundred seven patients met study criteria. Sixty-one patients (57%) underwent potentially curative PD, 17 (28%) of whom required mesenteric vein resection. Fifteen (25%) patients had an R0, LN- resection; 26 (43%) an R0, lymph-node positive (LN+) resection; 3 (5%) had an R1, LN- resection; and 17 (28%) an R1, LN+ resection. Of the 46 unresectable patients, 26 were thought resectable by preoperative imaging. Five had gastric outlet obstruction requiring operation, but the others could not be resected due to undiagnosed liver cirrhosis (1), carcinomatosis (3), liver metastases (5), and locally advanced disease (12). Multivariate regression demonstrated superior mesenteric vein (SMV) involvement on CT as the most significant predictor of unsuccessful R0 resection (OR 0.29 per 90 degrees encasement, P=.02). Periportal adenopathy by CT was the only significant predictor of unresectability (OR 3.42, P=.02). No imaging findings reached statistical significance predicting node-positivity or vein resection. Addition of EUS to CT did not improve the sensitivity or specificity of CT for predicting R0 resection.
Conclusions: Pancreas protocol CT imaging appears to be the most valuable modality in predicting resectability of pancreatic adenocarcinoma as compared to EUS. Periportal adenopathy and SMV involvement best predict resectability, and such findings may help limit the number of non-therapeutic laparotomies.