Objective: Endoscopic ultrasound (EUS) and magnetic resonance imaging (MRI) have both been assessed individually as staging modalities for pancreatic cancer. The aim of our study was to assess whether tumor staging by both EUS and MRI in the same cohort of patients could predict resectability and unresectability in patients with pancreatic cancer.

Methods: A review of 63 patients evaluated preoperatively with both EUS and MRI for pancreatic adenocarcinoma between January 1995 and December 1998 was done. Patients were staged as resectable or unresectable by predefined criteria. Preoperative staging by both modalities was compared to surgical outcome and the sensitivity and predictive values of each modality for determining resectability and unresectability was determined.

Results: EUS did not allow for complete T- and N-staging in 10 patients; therefore, for EUS, the final analysis was done on 63 of 73 patients (86%). EUS correctly staged 22 of 36 patients with resectable tumors. The sensitivity of EUS for resectability was 61%, with a positive predictive value of 69%. All 73 patients had complete MRI examinations; therefore, the final analysis was done on all 73 patients. MRI correctly staged 30 of 41 patients with resectable tumors. The sensitivity of MRI for predicting resectability was 73% with a positive predictive value of 77%. MRI and EUS both predicted resectability in 18 patients, of whom 16 (89%) were found to be resectable on surgical exploration. MRI and EUS both predicted unresectability in 17 (27%) patients, of whom 4 (24%) were found to be resectable on surgical exploration. When both MRI and EUS agreed on resectability, the positive predictive value for resectability was 89%. When both MRI and EUS agreed on unresectability, the positive predictive value for unresectability was 76%.

Conclusions: Neither MRI nor EUS alone were highly sensitive or predictive of resectability. However, when both tests agreed on resectability, nearly all patients were found to be resectable on surgical exploration.