Abstract
Objectives:
Brush cytology has a low sensitivity for the diagnosis of cholangiocarcinoma. This study aimed to compare the standard approach (brush cytology) with a triple modality approach utilizing brush cytology, forceps biopsy and fluorescence in situ hybridization in terms of sensitivity and specificity for the diagnosis of cholangiocarcinoma.
Methods:
In a retrospective study at a single academic center, 50 patients underwent triple modality testing. Additionally, 61 patients underwent brush cytology alone. Intervention was endoscopic retrograde cholangiopancreatography with brush cytology, fluorescence in situ hybridization, and forceps biopsy. The main outcome measures included sensitivity, specificity, positive predictive value and negative predictive value.
Results:
Overall, 50 patients underwent triple tissue sampling, and 61 patients underwent brush cytology alone. Twenty-two patients were eventually diagnosed with cholangiocarcinoma. Brush cytology had a sensitivity of 42%, specificity of 100%, positive predictive value of 100% and negative predictive value of 88%. Triple tissue sampling had an overall sensitivity of 82%, specificity of 100%, positive predictive value of 100%, and negative predictive value of 87%. Within the triple test group, brush cytology had a sensitivity of 27%, forceps biopsy had a sensitivity of 50%, and fluorescence in situ hybridization analysis had a sensitivity of 59%.
Conclusions:
A triple modality approach results in a marked increase in sensitivity for the diagnosis of cholangiocarcinoma compared with single modality testing such as brush cytology and should be considered in the evaluation of indeterminate or suspicious biliary strictures.
Keywords
Background
Cholangiocarcinoma is a malignant transformation of epithelial cells lining the lumen of the intrahepatic or extrahepatic biliary tree [Aljiffry et al. 2009]. Symptoms arise when a biliary stricture progresses to obstruction. The ability to receive treatment, to provide accurate prognostic information, and to be considered for orthotopic liver transplantation under a cholangiocarcinoma protocol requires a tissue diagnosis. Conventional brush cytology specimens obtained during endoscopic retrograde cholangiopancreatography (ERCP) have a high specificity but a poor sensitivity for the diagnosis of cholangiocarcinoma [Tamada et al. 2011]. Brush cytology is generally considered to be the standard approach as it adds minimal time, expense, and risk; however, it may achieve a tissue diagnosis in as few as 20% of cases [Jailwala et al. 2000; Fogel et al. 2006]. The sensitivities of multiple modalities in trial settings have been low, and these results are generally higher than those seen in routine clinical practice. To improve sensitivity, sophisticated approaches have been employed; however, techniques such as intraductal endoscopic ultrasound (EUS) and biliary confocal endomicroscopy require specialized training and major equipment acquisitions resulting in availability in few academic centers. To increase the diagnostic yield, intraductal stricture forceps biopsy [Tamada et al. 2011] and fluorescence in situ hybridization (FISH) [Weber et al. 2008a] have been added to brush cytology, and a dual approach has increased sensitivity compared with brush cytology alone. No studies to date have examined a triple modality approach utilizing brush cytology, FISH, and forceps biopsy in the evaluation of biliary strictures. This study evaluated the impact of triple modality testing on sensitivity and specificity in the diagnosis of cholangiocarcinoma compared with brush cytology and single modality testing alone.
Methods
This retrospective study was approved by the Institutional Review Board at the Emory University School of Medicine. All patients in the study provided written informed consent for the ERCP. An interventional endoscopy database was queried to identify patients who underwent ERCP from December 2008 to November 2012 at Emory University Hospital and The Emory Clinic outpatient Ambulatory Surgery Center. Patients with a previous history or tissue diagnosis of cholangiocarcinoma were excluded. Baseline demographic characteristics and clinical symptoms at the initial presentation were recorded for each patient. The diagnostic workup prior to the ERCP was evaluated. Patients with biliary strictures evaluated for diagnosis at ERCP were included. The evaluation included patients’ biliary strictures, including patients with hilar, peri-hilar, and extrahepatic bile duct strictures.
Endoscopic procedure
ERCP was performed in the endoscopy suite with patients in a prone position under deep anesthesia administered sedation or general anesthesia. Selective bile duct cannulation was performed, and a wire was advanced to the intrahepatic ducts. A cholangiogram was obtained, and the biliary stricture was defined. The cannulation device (cannula or sphincterotome) was withdrawn over the wire and a brush was introduced. The brush (RX cytology brush, Boston Scientific, Natick, MA, USA or Infinity brush, US Endoscopy, Mentor, OH, USA) was passed across the stricture repeatedly at the discretion of the endoscopist. Typically 10–15 passes of the brush within the stricture were performed. The brush was withdrawn and sent for cytologic examination as described below. In the triple test group, another brush was advanced and another sampling was performed. This brush was withdrawn over the wire and sent for FISH analysis as described below. Standard or pediatric biopsy forceps (Boston Scientific) were then advanced into the bile duct alongside the wire and the jaws were advanced to the stricture. Several bites were obtained at the stricture and the biopsy was withdrawn. A repeat passage of the forceps was performed at the discretion of the endoscopist based on the presence or absence of material in the jaws of the forceps. The forceps were withdrawn, and the sample was sent to histology for analysis. One or more stents were then placed at the discretion of the endoscopists.
Bile duct brushing
The brush head from the bile duct brushing was clipped off into a vial of CytoLyt (Hologic, Inc., Marlborough, MA, USA) solution. The brush was retrieved from the CytoLyt solution and placed into a vial of PreservCyt (Hologic, Inc.) solution. It was capped and shaken until material no longer adhered to the brush and then left to stand for at least 5–10 min. It was subsequently processed in a ThinPrep (Hologic, Inc.) processor according to the manufacturer’s specifications and stained with a Papanicolaou stain. If residual specimen was available, it was submitted into a cell block. The sample including ThinPrep slides and hematoxylin and eosin-stained sections of the cell block were reviewed by a cytologist and a diagnosis was rendered.
Fluorescence in situ hybridization
Brush cytology specimens obtained for FISH during the ERCP were sent to the Mayo Clinic Cytology Laboratory for processing. Upon arrival, cell material was scraped from the endoscopic brush into the specimen fluid. The specimen was then split evenly for routine cytology and FISH processing. The cytology portion of the specimen was used to make two ThinPrep slides and Pap stained. A cytopathologist diagnosed each specimen as nondiagnostic, negative, atypical, suspicious, or positive for malignancy. Cells were harvested from the FISH portion of the specimen to create a cell pellet as previously described, including up to two washes with 3:1 methanol/acetic acid fixative [Kipp et al. 2004]. FISH slides were made by manually dropping 10 μL of cell pellet suspension at a time onto a microscope slide until the optimal cell intensity was reached. Slides were pretreated and hybridized with the UroVysion (Abbott Molecular, Des Plaines, IL, USA) FISH probe set as previous described [Kipp et al. 2004]. Slides were classified as polysomy if at least five cells had gain (more than two signals) of at least two probes (Figure 1), trisomy if at least 10 cells had gain of one probe, or negative if none of the aforementioned criteria were met.
Fluorescence in-situ hybridization using probes directed to chromosome 3 (red), chromosome 7 (green), chromosome 17 (aqua), and 9p21 (gold) of a cell obtained by biliary brushing at an endoscopic retrograde cholangiopancreatography. This cell is classified as polysomic (more than two copies of at least two probes). Polysomy in a biliary stricture is suggestive of cholangiocarcinoma.
Clinical course
Following the ERCP, a combination of radiography, pathology, and all subsequent clinical documentation (outpatient visits to a gastroenterologist, oncologist, primary care physician, as well as all hospital discharge summaries) over the span of approximately 4 years were reviewed for the eventual diagnosis of cholangiocarcinoma. Radiography included computerized tomography (CT) and magnetic resonance imaging (MRI). If, in all available documentation following ERCP, no diagnosis of cholangiocarcinoma was made or mentioned, the patient was considered to be disease negative. Morbidity and mortality data were determined. The subsequent management of cholangiocarcinoma was examined. The gold standard for a tissue diagnosis was considered to be anatomical pathology. Brush cytology reported as ‘atypia’ and FISH results reported as ‘equivocal’ were not considered positive for malignancy. Positive cytology and polysomic FISH results were considered test positive for statistical analysis. If all sampling techniques employed during an ERCP yielded no evidence of carcinoma, the ERCP was considered negative. If either brush cytology, forceps histology, or FISH were interpreted as positive, the triple modality was considered positive. Patient charts were reviewed to determine rates of post-ERCP complications, including pancreatitis, hemorrhage, perforation, and infection. Post-ERCP pancreatitis was defined as new abdominal pain with elevation in the pancreatic enzymes. Any clinical diagnosis of hemorrhage, perforation, or infection subsequent to the index procedure was sought in the medical record.
Statistical analysis
Data were collected into a secure database behind the institution firewall. Data were stored in a Microsoft Excel database. Tabulations and summary data were calculated in Microsoft Excel. Sensitivity, specificity, positive predictive value and negative predictive value were calculated using two by two tables and statistical comparisons between groups were performed using χ2 testing with Yates’ correction (QuickCalcs Contingency Table, GraphPad Software, Inc., La Jolla, CA, USA). An α of 0.05 was considered statistically significant.
Results
Overall, 61 patients underwent brush cytology alone, and 50 patients underwent triple tissue sampling. Of the 50 patients undergoing triple tissue sampling, the average age was 58.9 years, with 40 men and 10 women. Specific indications (some overlapping) for ERCP included biliary stricture in 30 patients (60%), jaundice in 15 patients (30%), primary sclerosing cholangitis (PSC) in 13 patients (26%), and abnormal liver function tests (LFTs) in 14 patients (28%). Other indications included abnormal imaging and bile duct dilatation. The total calculated length of follow up ranged from 65 days to 1422 days, with an average of 799 days.
In the 50 patients undergoing triple modality testing, 22 patients had an eventual confirmatory diagnosis of cholangiocarcinoma and 28 patients were disease negative at the time of censure. In the 22 patients ultimately diagnosed with cholangiocarcinoma, there were a total of 18 positive tests by one or more of the three modalities. In these 18 patients, all had an eventual diagnosis of cholangiocarcinoma.
Triple tissue sampling had an overall sensitivity of 82% (p-value < 0.001), a specificity of 100%, a positive predictive value of 100%, and a negative predictive value of 87%. Complications related to ERCP included one patient with pancreatitis and one stent occlusion. One other patient developed Clostridium difficile infection 2 days post procedure. In the triple tissue group, brush cytology alone yielded a sensitivity of 27%, forceps biopsy had a sensitivity of 50%, and FISH analysis (i.e. polysomic result, Figure 1) had a sensitivity of 59% (Table 1, Figure 3).
In a group of 50 patients undergoing triple modality testing for the diagnosis of cholangiocarcinoma, the sensitivity, specificity, positive and negative predictive values are compared for single, dual and the triple test approaches.
Biopsy, forceps biopsy; Brush, standard brush cytology; FISH, fluorescence in situ hybridization (FISH).
In the group of 61 patients who underwent brush cytology alone, indications included jaundice in 25 patients (40%), abnormal imaging in 10 patients (16.3%), cholelithiasis/bile duct stone in 9 patients (14.7%), and biliary ductal dilatation in 6 patients (9.8%). Other indications included abnormal LFTs, strictures, stent manipulation, cholangitis, pancreatic head mass and PSC. Twelve patients eventually had a pathologic diagnosis of cholangiocarcinoma (Figure 2). Five of these patients had positive brush cytology results for malignancy, two patients had reactive changes with air-drying artifact, and the other five patients had negative brush cytology results. In this group, brush cytology had a sensitivity of 42%, specificity of 100%, positive predictive value of 100%, and negative predictive value of 88%.
A forceps biopsy specimen obtained via endoscopic retrograde cholangiopancreatography, 20× magnification. The normal cellular architecture is distorted, suggesting moderately differentiated adenocarcinoma. Signet ring cells can readily be identified.
Discussion
Though cholangiocarcinoma may be suspected based on typical imaging findings and clinical course, tissue diagnosis remains a major challenge. Multiple studies have attempted to improve on sensitivity by employing sophisticated technologies or by combining two modalities [Tamada et al. 2011] (Table 2). Intraductal ultrasound and intrabiliary probe-based confocal endomicroscopy are noteworthy technologies and may improve sensitivity; however, they require extensive training and equipment acquisition and are available in few academic centers. Dual modality testing has been of benefit; however, the diagnostic yield is still low with these two-prong approaches [Howell et al. 1992; Pugliese et al. 1995; Jailwala et al. 2000; Farrell et al. 2001; Kitajima et al. 2007; Weber et al. 2008b]. This study examined a triple modality approach in the evaluation of biliary strictures in patients with suspected cholangiocarcinoma. A triple modality approach utilizing brush cytology, forceps biopsy and FISH resulted in a marked increase in the sensitivity from 27–59% to 82% for the diagnosis of cholangiocarcinoma (Figure 3). While triple tissue sampling resulted in a significantly higher sensitivity than brush cytology, the increase in sensitivity was not associated with a decrease in the specificity (100% by both approaches).
Comparison of the sensitivity and specificity for the detection of cholangiocarcinoma by modality grouped by single, double and triple testing approaches in previously published studies and the current study.
FISH, fluorescence in situ hybridization; NR, not reported; pCLE, probe-based confocal microscopy.
Comparison of sensitivities within the group of patients undergoing triple modality testing. The sensitivity of each modality alone and each combination are charted. Biopsy, forceps biopsy; Brush, standard brush cytology; FISH, fluorescence in situ hybridization (Brush). Specificity and positive predictive value were 100% for all groups.
Cholangiocarcinoma is a highly desmoplastic tumor, composed of a small number of malignant cells surrounded by fibrous tissue [Lazaridis and Gores, 2005]. Sampling modalities may obtain fibrous tissue without neoplastic material, resulting in a low sensitivity for both forceps biopsy and brush cytology. Within the triple tissue sampling group, FISH had the highest sensitivity (59%), while brush cytology had the lowest (27%). In previous studies of brush cytology alone, there is wide variation in the sensitivity from 18% to 80% (Table 2). The sensitivity of brush cytology seen here in the triple testing group (27%) was in the lower range (Table 2) [Ryan, 1991; Ponchon et al. 1995; Dumonceau et al. 2008; Smoczynski et al. 2012]. The sensitivity was 42% in the group that had brush cytology alone, which is in line with prior studies. Forceps biopsy of a biliary stricture had a sensitivity almost twice that of brush cytology in our study, but was still moderate at 50%. Other studies have demonstrated that forceps biopsy, while more time consuming and technically difficult than brush cytology, has a somewhat higher sensitivity [Kubota et al. 1993a, 1993b; Ponchon et al. 1995; Pugliese et al. 1995; Sugiyama et al. 1996; Tamada et al. 2001; Higashizawa et al. 2002; Weber et al. 2008a]. The success of forceps biopsy is operator dependent, and requires a higher degree of skill than brush cytology as the forceps are manipulated into the duct alongside the wire and are not angled to improve the approach to the biliary orifice. Prospective studies have demonstrated the combination of forceps biopsy and brush cytology to have a sensitivity in the range of 50–75% (Table 2) [Pugliese et al. 1987; Howell et al. 1992; Jailwala et al. 2000; Farrell et al. 2001; Kitajima et al. 2007; Weber et al. 2008b]. Triple modality testing involves multiple services, brush being examined by cytology, biopsy by histology, and FISH by send out, which also likely contributes to the increased sensitivity seen with this approach.
FISH is a cytogenetic technique that has been shown to improve sensitivity in the diagnosis of malignant biliary duct strictures [Weber et al. 2008a]. FISH increases sensitivity when added to brush cytology, especially in patients with PSC [Kipp et al. 2004; Bamrungphon et al. 2007]. In a prospective study, when cytology was negative, FISH increased sensitivity (35–60%) compared with routine cytology, while preserving specificity [Moreno Luna et al. 2006]. In one study, in patients with negative forceps biopsy and brush cytology, FISH predicted malignancy in 62% of patients [Moreno Luna et al. 2006]. One prospective trial demonstrated FISH to have a sensitivity twice that of brush cytology (34% versus 15%) [Kipp et al. 2004]. FISH and forceps biopsy had a combined sensitivity of 39%, compared with 30% for combined brush cytology and biopsy [Kipp et al. 2004]. Our study found FISH to have a slightly higher sensitivity at 59% compared with other studies (30–50%) [Kipp et al. 2004; Fritcher et al. 2009; Smoczynski et al. 2012]. Our study utilized the UroVysion FISH probe set. This set, initially designed for the diagnosis of bladder carcinoma, is often used to aid in the diagnosis of biliary tract malignancy. It frequently identifies two different chromosomal abnormalities: polysomy and trisomy. A recent trial demonstrated that the false-positive rate of FISH increases significantly when trisomy 7 is included as a criterion for a positive result [Fritcher et al. 2009]. Polysomy is specific for malignancy, and when included as the only criterion for a positive result, the specificity approaches 100%, especially in patients without PSC. Trisomy is not specific for malignancy and has been observed in benign conditions [Broberg et al. 2001; Moreno Luna et al. 2006; Barr Fritcher et al. 2007]. In our study, the probe set identified both polysomy and trisomy, but only polysomy was considered a positive FISH result. There were no false-positive FISH results in our study.
Intraductal ultrasound (IDUS) is a novel technique that can be performed during ERCP and has been studied in cholangiocarcinoma [Tamada et al. 2011]. Using a transpapillary approach, a probe is inserted into the bile duct to produce images of the bile duct wall. IDUS requires proficiency in ERCP as well as EUS, and requires specialized training in the interpretation of intraductal echoendosonographic imaging. Most problematically, the approach produces additional imaging data but does not provide a tissue diagnosis. Wire-guided probe-based EUS equipment is expensive and has limited availability. Thus patients would not be able to proceed to chemotherapy or orthotopic liver transplantation on the basis of the results, and diagnostic uncertainty could still persist despite the additional imaging findings. IDUS may have an additional role in staging [Tamada et al. 2002; Vazquez-Sequeiros et al. 2002; Stavropoulos et al. 2005; Levy et al. 2008; Weber et al. 2008a]. In our center, EUS fine-needle aspiration (FNA) is frequently performed for the evaluation of lymph nodes in patients with cholangiocarcinoma; however, FNA of the primary tumor is not performed which is consistent with our current cholangiocarcinoma protocols for liver transplant consideration.
There are several limitations in this trial. The study reports on a single center experience. Nevertheless, our data for each modality are similar to those previously published in single or dual modality studies. This was also a retrospective study. A retrospective design was helpful in allowing us to examine the outcome for patients following diagnosis. Additionally, we believe that a triple modality approach is superior to a single modality approach and has become our standard for patients in whom cholangiocarcinoma is suspected on the basis of imaging and clinical presentation. One question arises if the triple modality approach was selected in patients with a higher pretest probability of cholangiocarcinoma, thus contributing to the higher diagnostic rate in this group. To look at this question, we also analyzed the sensitivity of brush cytology alone within the group undergoing the triple modality testing. In this group, brush cytology alone had a sensitivity of 27%, suggesting that it was the additional modalities rather than an increased proportion of patients with malignancy in the triple modality group leading to the increased sensitivity. The sample size was moderate but similar to other trials and was sufficient to see a major difference in the sensitivity between the approaches.
Biliary strictures remain a major diagnostic challenge. This study demonstrates a marked increase in the sensitivity when forceps biopsy and FISH are added to standard brush cytology; however, 82% is still lower than desired in clinical practice. The approach described here does not require major additional procedural time and does not add significantly to the procedural risk or cost of an ERCP. From a technical standpoint, the triple modality approach would be within the standard skill set for providers routinely performing ERCP. New developments in imaging technologies such as probe-based ultrasound and probe-based confocal laser endomicroscopy (pCLE) will likely continue to improve the detection and targeting of intraductal lesions. Confocal endomicroscopy is an emerging advanced imaging technique that allows for real-time microscopic evaluation of bile duct mucosa, and can incorporate information such as blood flow and cellular architecture [Wani and Shah, 2013]. Recent studies have suggested that pCLE has a high sensitivity and negative predictive value to exclude malignancy [Meining et al. 2012; Heif et al. 2013]. However, this technique is expensive and, like IDUS, requires specialized training in interpretation of endoscopic microscopy. Ongoing studies are examining this approach. Molecular markers may have a role in improving the detection of intraductal malignancy and is an area of active study. Overall, this study demonstrates a substantially higher sensitivity with the triple modality approach which could be adoptable in a straightforward manner due to the relatively modest burden in additional cost, time, and procedural complexity.
Conclusion
This study demonstrates a substantially higher sensitivity for triple modality testing (brush cytology, FISH, and forceps biopsy) of 82% compared with brush cytology or single modality testing alone (27–59%) in the diagnosis of cholangiocarcinoma. Unlike more complex technologies for evaluating biliary strictures, the additions in cost, time, and procedural complexity with the triple modality approach are relatively modest.
Footnotes
Acknowledgements
AN, JB and SB participated in data collection and were major contributors to manuscript preparation. ML, EBF and GG participated in specimen analysis and contributed to manuscript preparation. SK, KW and QC participated in study design, data collection and manuscript revision. FW conceived of the study and participated in its design, data collection, coordination, manuscript drafting and final revisions of the manuscript. All authors read and approved the final manuscript.
We would like to thank the authors listed above for their diligence and commitment to this study and to the completion of the manuscript
Conflict of interest statement
The authors declare that they do not have anything to disclose regarding conflicts of interest with respect to this manuscript.
Funding
This study was funded in part by a development grant from the Department of Medicine at Emory University School of Medicine.