부산대학교 도서관

Keywords: cytology; grading; intraductal papillary mucinous neoplasm; mucin; mucinous cystic neoplasm; pancreas; PSC classification; serous cystadenoma Abstract Background Pancreatic cyst cytology evaluates for neoplastic mucin and epithelial grade. This study describes cytological features of low- and high-grade mucinous neoplasms (MNs) using gastrointestinal contaminants for comparison. Methods Histologically confirmed pancreatic cystic neoplasms were reviewed by a panel of cytopathologists to identify which, among 26 selected cytologic features, correlate significantly with low- and high-grade MN. A test for greater than or equal to four of eight high-grade features (three-dimensional architecture, high nuclear:cytoplasmic ratio, moderate nuclear membrane abnormalities, loss of nuclear polarity, hyperchromasia, >4:1 nuclear size variation in one cluster, karyorrhexis, and necrosis) was assessed for identifying a high-grade neoplasms. Additional characteristics of the cohort such as cyst fluid carcinoembryonic antigen results, molecular testing, Papanicolaou Society of Cytopathology classification, and select high-risk clinical features are described. Results Endoscopic ultrasound fine-needle aspirations from 134 MN and 17 serous cystadenomas containing gastrointestinal contaminants were included. The MN consisted of 112 (84%) intraductal papillary MNs (low-grade = 69, 62%; high-grade = 24, 21%; and invasive = 19, 17%) and mucinous cystic neoplasms (low-grade = 20, 90%; high-grade = 2, 10%). Half had greater than five clusters of epithelium for analysis. Compared with gastrointestinal contaminants, mucin from MN was thick and colloid-like (40% vs. 6%, p 20% of the smear area (32% vs. none, p < .01), and contained histiocytes (46% vs. 18%, p = .04). Greater than or equal to four of eight select high-grade features was present in 36% of high-grade MN with sensitivity 37% and 98% specificity. Conclusion Colloid-like features, >20% of smear, and histiocytes correlated with MN. Testing for greater than or equal to four high-grade features had low sensitivity and high specificity for high-grade MN. CAPTION(S): Table S1 Byline: Carlie Sigel, Xiao-Jun Wei, Narasimhan Agaram, Keith Sigel, Roshan Raza, Rebecca Andrade, Rema Rao, Pari Shah, Kevin Soares, Abha Goyal