Endoscopic Ultrasound and Computer Tomography Are Inaccurate Methods of Classifying Cystic Pancreatic Lesions

Abstract

Background and aims: Despite advances in imaging modalities, preoperative diagnosis of pancreatic cystic lesions remains difficult. The aim of the study was to assess the accuracy of endoscopic ultrasound (EUS) and computer tomography (CT) to preoperatively distinguish benign (BE) from potentially malignant and malignant pancreatic cystic lesions (M/PM). Methods: Photograph series obtained from EUS examinations of 66 patients with cystic pancreatic lesions were blindly reviewed by 3 endoscopic ultrasonographers. 41 of those 66 pts also underwent a CT scan at our institution, which was blindly reviewed by a single radiologist. CT and EUS classification into BE and M/PM was correlated with the final diagnosis, which was established by surgical pathology, diagnostic fine needle aspiration or follow-up imaging. Interobserver agreement was measured using kappa statistics. Results: The final diagnosis was BE in 35 pts (pseudocyst 24, true cyst 6, serous cystadenoma 4, anomalous cystic duct 1) and M/PM in 31 pts (adenocarcinoma 11, IPMT 3, mucinous cystadenoma 16, neuroendocrine tumor 2, papillary and solid epithelium neoplasm 1). EUS classification by the three examiners was correct in 65% to 67%. Interobserver agreement was 50%. Kappa values for pairs of EUS examiners were 0.16, 0.43 and 0.53. CT classification was correct in 71% and in agreement with the EUS classification in 56% to 61% (kappa 0.12 to 0.27). Ultrasonographic features of chronic pancreatitis (at least 3 of 8 criteria) were more common in patients with pseudocysts than in other cystic lesions (p<0.05 for all 3 EUS examiners). Solid areas were associated with adenocarcinoma. No significant correlations with BE, M/PM and final diagnosis were found for other sonographic cyst characteristics (contour, wall thickness, content, septation). Conclusions: EUS and CT cannot accurately distinguish between benign pancreatic cystic lesions (BE) and malignant or potentially malignant ones (M/PM). There is poor interobserver agreement in classifying these lesions.