Selected MicroRNAs May Aid the Preoperative Diagnosis of Pancreatic Adenocarcinoma
NC Panarelli, RK Yantiss, XK Zhou, N Kitabayashi, Y-T Chen. Weill Cornell Medical College, New York, NY
Background: Pancreatic carcinomas show altered expression of microRNAs (miRNAs). They often overexpress miR-21, -181b, -221, and -196a, but show decreased levels of miR-217. In this study, we investigated the expression of these five miRNAs in cytologic preparations obtained from fine needle aspirations to evaluate their utility in distinguishing adenocarcinoma from benign pancreatic lesions in limited samples.
Design: Cell block specimens from 38 clinically, or surgically, confirmed pancreatic adenocarcinomas and 11 benign pancreatic lesions, including chronic pancreatitis and pseudocysts, were analyzed for miRNA expression and normalized with small RNA U6 endogenous controls using quantitative RT-PCR. Mean miRNA expression levels, both individually and in combinations, were statistically compared between the cytologically malignant and benign groups, and fold differences were calculated. A logistic regression model that best distinguished these two groups was used to classify cases originally interpreted as suspicious for malignancy.
Results: Successful miRNA amplification was achieved in all 11 benign lesions and 35/38 cancers. The preoperative cytologic diagnoses of these 35 cases were carcinoma (26) and suspicious for carcinoma (9). The 26 cases diagnosed as carcinoma showed overexpression of miR-21 (p<0.001), -221 (p<0.01), and -196a (p<0.001) compared to the 11 benign lesions [fold change range: 9.45 (miR-196a) to 3.73 (miR-221)]. Combined evaluation of these three miRNAs significantly discriminated between carcinoma and benign lesions (p<0.001) with a 180-fold difference. Expression of miR-181b and -217 was statistically similar among benign and malignant groups. Based on these results, a logistic regression model [miR-221 + 2 x miR-196a] was generated, which would accurately predict 24/26 (92%) carcinomas as malignant and 8/11 (73%) benign cases as non-malignant. When applied to cases suspicious for carcinoma, the model correctly classified 8 of 9 (89%) as malignant.
Conclusions: Most pancreatic aspirates yield sufficient miRNA for analysis. Carcinomas show significant overexpression of miR-21, -221, and -196a compared to benign pancreatic lesions. Our logistical regression model accurately predicts the presence of carcinoma in non-diagnostic samples, suggesting that miRNA analysis may be a useful adjunct to the preoperative evaluation of pancreatic lesions.
Category: Liver & Pancreas
Tuesday, March 23, 2010 1:30 PM
Platform Session: Section C, Tuesday Afternoon