br ACCEPTED MANUSCRIPT br Subgroup analysis br As
As the therapeutic indications for ERCP are more common in biliary strictures and published data show differences in diagnostic sensitivity according to location, we next analyzed the data separating biliary and pancreatic malignant samples. Sensitivity of brush T-5224 was higher in pancreatic strictures (66.7% vs. 42.9%) with a diagnostic accuracy of 81%, although the tendency did not reach statistical significance. Pancreatic brush samples from malignant strictures showed significantly higher expression levels of miR-16 (p=0.0098) and miR-196a (p=0.0254) compared to benign strictures, miR-221 expression was also higher in tumor samples, however, this tendency did not reach statistical significance (p=0.0973) (not shown). In contrast, all target microRNAs were significantly enriched in malignant biliary samples compared to normal pancreatobiliary samples as shown in Figure
2. MiR-196a proved to be the best marker studied, with highly different expression values separating cholangiocellular carcinoma from normal specimens (p=2.3x10-6).
Diagnostic correctness of single and combined microRNA markers
To assess the clinical usefulness of the highly up-regulated single microRNA markers, we next analyzed the data by receiver operating characteristic (ROC) curves. Since the specificity of brush-cytology alone was 100% in our series, we tried to choose optimal cut-off values for the investigated markers with the highest matching sensitivity value without sacrificing the good specificity of cytology. Using a cut-off value ∆Ct > -0.985 for miR-196a malignant and benign pancreatobiliary samples could be separated with a sensitivity of 69.2% and a specificity of 100%, these values could be improved further with the combination of routine cytology and miR-196a expression, resulting in a 84.6% true positivity count, maintaining the PPV at 100%. The combination of several microRNA markers did not improve these statistics (Table 2).
ROC analysis for the subset of pancreatic samples resulted in optimal cut-off values of ∆Ct > 7.975 (miR-16), ∆Ct > -1.68 (miR-196a) and ∆Ct > -0.345 (miR-221), which corresponded to 91.7% sensitivity and 88.9% specificity using miR-16 as the best single marker. Combined analysis (e.g. with cytology or other markers) did not improve the diagnostic performance of miR-16 marker, nor did miR-196a and miR-221 improve the diagnostic precision of routine cytology in pancreatic strictures (Table 3).
When analysing the subset of biliary specimens, the following cut-off values seemed to result in the best diagnostic values: ∆Ct > 7.975 (miR-16), ∆Ct > -0.985 (miR-196a) and ∆Ct > -0.345 (miR-221). MiR-196a as a single marker showed a sensitivity of 85.7% which corresponded to a specificity of 100% (Figure 3). Combined expression analysis of several markers did not increase diagnostic precision, however, miR-196a and cytology together reached a sensitivity of 92.9% with no false positives (Table 4).
Characterization of strictures in the pancreatobiliary tree poses a challenge to gastroenterology specialists knowing the shortcomings of imaging, the lack of specific plasma markers and the low diagnostic performance of routine brush cytology. Albeit its low sensitivity, brush cytology has a well -defined role in jaundiced tumour patients principally in the neoadjuvant and palliative setting, where a therapeutic ERCP needs to be carried out to normalize serum bilirubin levels and tissue confirmation is inevitable to initiate chemotherapy. During any therapeutic stenting procedure brush cytology can be performed with no reasonable extra effort, carries a very low complication risk , and a positive cytology result spares the patient from further demanding diagnostic procedures such as EUS/FNA or intraductal biopsies. Although it is a highly controversial issue in the EUS era, diagnostic ERCP-based sampling might still have a role in very carefully selected cases in non-mass forming strictures with clear imaging proof of ductal dilatation, or small lesions with failed FNA due to extensive perilesional inflammation. Intraductal imaging and sampling methods are even expected to help the early diagnosis of pancreatic cancer concomitant with intraductal papillary mucinous neoplasia (IPMN), growing silently in the pancreatic duct with no visible mass lesion available for early image diagnosis .
Fluorescent In Situ Hybridization (FISH) based cytology and more recently next-generation sequencing, have the potential to approach the desired sensitivity values in the diagnosis of various cancer types including pancreatobiliary malformations, however, the expertise needed and cost of the procedures hampers their widespread use in GI units worldwide . Knowing the often rapid progression and dismal prognosis of pancreatobiliary cancers and the insensitivity of brush cytology, we set out to find a cheap, reproducible and widely-available ancillary test that reliably distinguishes malignant from benign strictures. Although brush cytology has specificity values approaching 100%, the biggest drawback of the technique is the unacceptably low sensitivity for a test used in clinical decision making.