Comparing accuracy of high-risk features for detecting advanced neoplasia in pancreatic cystic lesions: a systematic review and meta-analysis

Authors Abhiram Duvvuri, Harikrishna bandla, Vivek Chandrasekar Thoguluva, Chandra Dasari, Madhav Desai, Venkat Nutalapati, Vishnu Moole, Narimiti Anvesh, Patel Harsh, Frank Gress, Prateek Sharma, Divyanshoo Rai Kohli.


Background The American Gastroenterological Association recommends endoscopic ultrasound (EUS) for evaluating pancreatic cystic lesions (PCL) with ≥2 high-risk features (HRF), whereas the American College of Gastroenterology recommends EUS for ≥1 HRF. This systematic review and meta-analysis compared the diagnostic accuracy of using ≥1 vs. ≥2 HRF for assessing the risk of advanced neoplasia (AN) and performing EUS in PCL.

Methods An electronic database search was performed for eligible studies. AN was defined as pancreatic adenocarcinoma, intraductal papillary mucinous neoplasm or mucinous cystadenoma with high-grade dysplasia, pancreatic intraepithelial neoplasia and pancreatic neuroendocrine tumors. HRF included cyst size ≥3 cm, solid component, and dilated pancreatic duct ≥5 mm. The primary outcome was the sensitivity and specificity of using ≥1 vs. ≥2 HRF as an indication for EUS to detect AN in PCL.

Results Of 38 studies initially screened, 8 were included in the final analysis. Seven studies assessed the accuracy of ≥2 HRF and 4 studies assessed ≥1 HRF. The pooled sensitivity, specificity, positive and negative predictive values of EUS for detecting AN were 41.7% (95% confidence interval 19.5-67.8%), 90.8% (81.9-95.5%), 30.4% (19.4-44.2%) and 94.3% (89.6-97.0%) with ≥2HRFs, and 77.1% (66.1- 85.3%), 72.7% (50.4-87.5%), 17.95% (10.3-29.4%), 98.1% (90.8-99.6%), respectively, with ≥1 HRF.

Conclusion Performing EUS for PCL with ≥1 HRF could offer greater sensitivity in detecting AN compared to ≥2 HRF, with a similar negative predictive value.

Keywords Pancreatic cystic lesions, pancreatic cancer, endoscopic ultrasound, meta-analysis

Ann Gastroenterol 2021; 34 (5): 743-750

Original Articles