Pancreatic and bile duct brushing cytology in 1000 cases

Abstract

Duct brushing cytology is an important tool in evaluation of the extrahepatic biliary tract and large pancreatic ducts. The emergence of neoadjuvant therapies underscores the importance of accurate preoperative diagnosis by noninvasive means. Liquid-based preparation methods, such as ThinPrep, have become popular for nongynecologic cytology specimens. Findings from bile and pancreatic duct brushings were reviewed over the 10-year period of 1994-2003. Cytologic material, imaging reports, and clinical data were reviewed and pathologic and clinical follow-up data were obtained. The slides were prepared by direct smear only (18.8%), direct smear plus cytospin (14.4%), or direct smear plus ThinPrep (66.8%). A total of 1118 specimens were identified (1008 bile duct, 110 pancreatic duct) from 864 patients. The cytologic findings were: 53.5% negative, 16.5% malignant, 18.2% suspicious for malignancy, 11% atypical/inconclusive, 0.8% inadequate. Follow-up in the form of either histology or at least 6 months clinical observation was available for 82.2% of cases (n = 971). Overall operating characteristics were: 52.6% sensitivity, 99.4% specificity, 98.9% positive predictive value, 67.1% negative predictive value, and 75.7% accuracy. Diagnostic agreement between cytology and follow-up was the main variable analyzed. Agreement was significantly affected by characteristics of the sampled lesion, with ductal narrowing having the lowest rate of malignancy. In addition, the ThinPrep method showed superior sensitivity and accuracy compared with other preparation methods (P = .02). Nonsignificant associations were noted for patient age and gender, site of lesion, and the presence of either stones or prior stent. In a large dataset from a single institution, brushing cytology showed modest sensitivity and high specificity. Diagnostic agreement was considerably better for benign cases. The combination of direct smear and the ThinPrep method showed superior sensitivity and accuracy.