Premium
Risk stratification using cytomorphologic features in endoscopic ultrasonographic‐guided fine‐needle aspiration diagnosis of pancreatic ductal adenocarcinoma
Author(s) -
Huffman Brandon M.,
Esebua Magda,
Layfield Lester J.,
Witt Benjamin L.,
Schmidt Robert L.
Publication year - 2015
Publication title -
diagnostic cytopathology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.417
H-Index - 65
eISSN - 1097-0339
pISSN - 8755-1039
DOI - 10.1002/dc.23273
Subject(s) - medicine , receiver operating characteristic , malignancy , fine needle aspiration , radiology , pathology , kappa , pancreatic duct , pancreas , biopsy , linguistics , philosophy
Background Endoscopic ultrasonographic‐guided fine‐needle aspiration (EUS‐FNA) is the procedure of choice for the investigation of pancreatic lesions. It shows good sensitivity and excellent specificity. Diagnostic criteria have been published but not statistically validated for the diagnosis of malignancy and stratification of risk for malignancy. Methods A training set of 57 EUS‐FNAs and the validation set of 107 EUS‐FNAs were selected. Slides were independently evaluated by three pathologists. Sixteen morphologic features were evaluated in the training set. Average absolute agreement, kappa scores, and association with malignancy were statistically evaluated. Recursive partitioning and multivariant analyses were performed on the features tested in the training set. Agreement data, univariate‐odds ratios, and discriminatory power were calculated for the diagnostic features selected from the training set. The selected morphologic features formed a scoring rule that was then applied to the validation set. Results The average absolute agreement in the training set was 72%. Anisonucleosis, nuclear crowding, macro nucleoli, single atypical epithelial cells, and intracytoplasmic mucin showed the highest interrater reliability. Anisonucleosis, macronucleoli, single atypical epithelial cells, and intracytoplasmic mucin were most predictive of malignancy. A simple scoring rule was developed combining these morphologic features and applied to the validation set. Analysis of the area under the receiver operating characteristic (ROC) curve confirmed the statistical validity of the scoring rule. Conclusion A scoring system utilizing the presence or absence of anisonucleosis, macronucleoli, single atypical epithelial cells, and mucinous metaplasia yielded good discriminatory power (area under ROC curve = 0.87). Diagn. Cytopathol. 2015;43:613–621. © 2015 Wiley Periodicals, Inc.