br ACCEPTED MANUSCRIPT br Our results show that routine

ACCEPTED MANUSCRIPT
Our results show that routine K 252a had an average sensitivity of 53.8% with a specificity of 100%, as expected, in the subgroup analyses sensitivity for pancreatic strictures was higher compared to biliary strictures (66.7% vs. 42.9%) maintaining the perfect specificity (see Tables 2-4). The higher sensitivity value in pancreatic strictures was in line with available literature, which is probably due to the tightness of the strictures (intervention is needed at more advanced stages and the pancreatic duct is narrower to start with) [22]. Our average sensitivity values were higher compared to published data due to the fact that the results were based on a consensus blinded investigation by three expert cytopathologist, and more stringent analysis interestingly resulted in a somewhat higher malignancy detection rate [5]. We must also point out that published sensitivities for biliary sampling tend to be lower in papers where pancreatic head tumors are included, because pancreatic masses compress, but don’t necessarily invade the bile duct and therefore a truly negative cytology could erroneously be counted as a false negative one. False negativity could be attributed to well differentiated tumors, and erroneous sampling of the “normal” tissue adjacent to the tumor, on the other hand previous endoscopic manipulation (infection, mechanical causes) might result in “overstaging” cytology not observed in our set of samples (not shown). Although ERCP/brush cytology of parapapillary masses is important, we handled these tumors separately, as the organ of origin is hard to define, and in some cases even macroscopic pathology examination fails to conclude on the site of origin confusing statistical calculations. In parapapillary cases using endpoints based on brush cytology is not acceptable due to the possibility of false negatives, which would result in misclassification of the patient.
A prospective head-to-head comparison of EUS/FNA and ERCP mediated sampling resulted in comparable sensitivities for biliary masses (79% for both groups), underlining the utility of ERCP/brush cytology in indeterminate biliary strictures, however, the authors conclude to start with EUS first based on the higher incidence of pancreatic masses [23]. In the case of sampling pancreatic masses the superiority of EUS is evident, however, a more reliable comparison of the sampling methods per se would be possible by techniques minimizing technical obstacles, such as repeated sampling: one published sensitivity of EUS/FNA for pancreatic masses after a single needle pass is 17% (with gradual increase to 87% when more than seven passes were performed), which is well below the 66.7% sensitivity of ERCP/brush cytology after one sampling in our cohort [24], and repetitive brushings are known to increase the diagnostic yield of brush cytology as well [25]. Not to mention on-site cytopathology, which is known to improve EUS/FNA accuracy and hence could do the same for the accuracy of the ERCP
ACCEPTED MANUSCRIPT
brushings [26]. An advantage of ERCP guided sampling is the lack of tumor seeding, a known risk described in EUS/FNA patients [20, 27].
There is a growing body of literature about the epigenetic molecular markers, microRNAs (miRs), which can be utilized in clinical applications based on the stability and the disease-specific expression of these small RNA molecules. MicroRNAs have been reported to have important functions in the regulation of carcinogenesis and cancer progression as well as homeostasis. Deregulated microRNAs can give information on transcriptional regulation and most importantly may serve as biomarkers for survival and early detection of pancreatobiliary cancers [12]. The aim of the present study was to analyze whether determination of microRNA expression levels in intraductal brush cytology specimens is a viable molecular technique and if the tumor-associated microRNA expression in brush cytology specimens is a valuable tool in the diagnosis of pancreatobiliary cancers.
The isolation of microRNAs was successful from brush cytology specimens, RT-PCR expression analyses of miR-16, miR-196a and miR-221 showed a clear and reproducible statistical significance between malignant and benign pancreatobiliary specimens, indicating approximately 11.3-fold, 23.6-fold and 4.3-fold increased microRNA levels in tumor tissue. Interestingly miR-21, a well-established diagnostic and prognostic marker in other sample sources [28], was not useful as a separation marker on brush cytology isolates in our cohort. Target microRNAs were significantly enriched in the subgroup analysis of malignant biliary specimens compared to normal samples (Figure 2). MiR-196a proved to be the best marker for pancreatobiliary brushings in general, and even more so for the biliary subgroup, with highly different expression values separating cholangiocellular carcinoma from normal