Department of Laboratory Medicine and Pathology, Mayo Clinic and Foundation, Rochester, Minnesota
Corresponding author: Benjamin R. Kipp, PhD, Department of Laboratory Medicine and Pathology, Hilton 1072C, Mayo Clinic, 200 First St SW, Rochester, MN 55905; Fax: (507) 284-1599; email@example.com
Patients diagnosed with primary sclerosing cholangitis (PSC) and dominant strictures often undergo endoscopic retrograde cholangiopancreatography with brush cytology to exclude or confirm the development of malignancy. Equivocal (atypical or suspicious) routine cytologic results may confound patient management decisions, especially in the absence of a mass on imaging. The objective of the current study was to identify independent predictors of malignancy in patients with PSC with an equivocal cytology diagnosis.
Patients with PSC underwent brush cytology for routine cytology and fluorescence in situ hybridization (FISH) during endoscopy as per standard care. FISH slides were classified as polysomy if at least 5 cells displayed a gain of ≥ 2 probes. A retrospective search identified 102 patients without a mass lesion noted on initial imaging studies, an equivocal routine cytology, and ≥ 2 years of follow-up.
Of 102 patients, 30 (29%) with an equivocal cytology result developed cancer within 2 years. Serum carbohydrate antigen 19-9 (CA 19-9) levels ≥ 129 U/mL (hazard ratio [HR] 3.19; P = .001) and polysomy (HR 8.70; P < .001) were each found to be predictive of cancer. Of 10 patients who had elevated CA 19-9 levels and polysomy, all went on to develop cancer (9 within 2 years). Although only 10 patients were included in this subset, the combination of elevated CA 19-9 and polysomy was found to be predictive of cancer (HR 10.92; P < .001).
Primary sclerosing cholangitis (PSC) is a chronic liver disease that causes inflammation and damage to the biliary tract. Patients with PSC are at an increased risk of developing cholangiocarcinoma, with a lifetime risk of 10% to 20%. Nearly one-half of all patients with PSC will have a dominant stricture that impairs biliary drainage. Unfortunately, the fibrotic nature of these strictures makes it difficult to discern a benign from malignant stenosis by imaging studies, because patients with benign disease can have features suggestive of malignancy (eg, jaundice; abrupt changes in serum liver biochemical tests; and symptoms of weight loss, abdominal pain, etc.). Due to the location of these tumors, it can also be difficult to obtain endoscopic biopsies for a histopathologic diagnosis of malignancy. As a result, cytologic brushings are collected during endoscopic retrograde cholangiopancreatography (ERCP) to help diagnose pancreatobiliary tract malignancies.[3-7]
One limitation of routine cytology is its suboptimal clinical sensitivity. Numerous studies have reported that a positive cytology is highly predictive of the presence of malignancy.[3-7] Unfortunately, few patients (< 20% in our practice) with cholangiocarcinoma are unequivocally determined to be positive by cytology. Equivocal cytology results (ie, atypical and suspicious) are much more common (approximately 40% in our practice), but possess only modest clinical value to physicians due to their poor negative and positive predictive values. In a previous study that included 86 patients with PSC, only 67% of suspicious cytology results were found to be associated with a malignancy. This void in diagnostic accuracy highlights the need for additional testing methods for the detection of malignancy in patients with PSC.
Genomic instability, characterized by genomic gains and losses, is a trait of most human cancers including cholangiocarcinoma and pancreatic cancer.[10, 11] Therefore, tumor cells are likely to contain an abnormal complement of chromosomal material that can be detected by fluorescence in situ hybridization (FISH). FISH is a method that detects aneusomic cells using fluorescently labeled DNA probes targeting specific chromosomal regions. At the study institution, the UroVysion (Abbott Molecular Inc, Des Plaines, Ill) FISH probe set is used in conjunction with cytology to enhance the clinical sensitivity of biliary brushings. Because pancreatobiliary cancer cells are often aneusomic, this probe set successfully detects pancreatobiliary tumor cells even though it was originally developed to detect bladder cancer.[10, 12]
FISH has been shown to detect significantly more pancreatobiliary cancers than routine cytology alone, while maintaining high specificity. However, false-positive FISH results have also been reported in patients with PSC. Bangarulingam et al reported a specificity of 88% in a study of 235 patients with PSC and recommended FISH evaluation only when clinically indicated rather than as a screening tool. A subsequent study from our institution reported that patients with repeated polysomy results from serial examinations were at higher risk of cancer compared to patients with an initial polysomy followed by nonpolysomy results. Because of the challenges of managing patients with PSC with equivocal cytology results, especially within the context of negative imaging tests (ie, no mass at baseline); there is a need to assess the diagnostic power of various clinical and pathologic variables in this cohort of patients. The objective of the current study was to identify risk factors suggestive of malignancy in this clinical context for improved diagnosis and treatment.
MATERIALS AND METHODS
After approval by the Mayo Clinic Institutional Review Board, a retrospective query of our pathology reporting database was performed to identify patients with equivocal cytology results (atypical or suspicious) reported between October 2003 and April 2010. If a patient underwent multiple ERCP procedures with equivocal cytology results during this time frame, the earliest equivocal cytology brushing was included. If brushings from multiple anatomic sites were collected during a single ERCP procedure and classified as equivocal, then the most severe equivocal cytologic diagnosis was included for the study (ie, suspicious over atypical). However, a patient was not included in the study if one of the multiple brushings was diagnosed as positive.
Patient medical records were evaluated to ascertain age, sex, serum carbohydrate antigen 19-9 (CA 19-9) level, total bilirubin, alkaline phosphatase, inflammatory bowel disease (IBD) status, and receipt of ursodeoxycholic acid therapy. If a serum CA 19-9 level was not evaluated during the clinical visit associated with the equivocal brushing, then the closest serum CA 19-9 level taken before the brushing was used for statistical analysis. The cancer status at 2 years after the equivocal brushing result was also recorded. Pathologic results (positive biopsy, positive fine-needle aspiration [FNA] specimen, or subsequent positive ERCP brushing) or a mass lesion identified by imaging on follow-up were considered to be diagnostic of malignancy. A clinical course of at least 2 years without cancer after the equivocal brushing was required for patients to be classified as benign in the current study. The target population for this study was patients with PSC who were suspected to have cancer at the time of an equivocal cytology result and therefore those with a history of pancreatobiliary malignancy, those referred to the Mayo Clinic for the management of pancreatobiliary malignancy, and those who had a mass lesion identified on cross-sectional imaging at baseline were excluded.
Brushing Collection and Routine Analysis
As per standard practice, a cytology brush was passed through the stricture during ERCP and placed in PreservCyt fixative solution (Hologic Inc, Marlborough, Mass). In the laboratory, cellular material was scraped from the brush into the fixative solution. The sample was split equally. One half was used to make two ThinPrep (Hologic Inc) slides, which were both Pap-stained. A cytopathologist evaluated each case as nondiagnostic, negative, atypical, suspicious, or positive for malignancy (Fig. 1) using established criteria.[16, 17]
A FISH slide was produced from the other one-half of the sample using the UroVysion probe set as previously described.[18, 19] In summary, cells were harvested from the fluid to form a cell pellet. Cell suspension was incrementally applied (10 μL) within a 2-cm circle on a charged slide until optimal cellularity was achieved. Slides were pretreated through a series of reagents using a VP2000 Automated Processor (Abbott Molecular Inc) to prepare the cells for hybridization. UroVysion, which contains centromere enumeration probes (CEP) to chromosomes 3, 7, and 17 and a locus-specific probe directed to 9p21, was applied to the circle area containing cell suspension and coverslipped. Slides were heated for 3 minutes at 73°C to denature the DNA, followed by overnight hybridization of probe DNA to specimen DNA at 37°C. After hybridization, slides were briefly washed in 0.1/%NP40 of 2X saline sodium citrate at 74°C for the removal of nonspecific probe signals. A nuclear counterstain was applied and slides were coverslipped.
The entire cellular area of each FISH slide was evaluated with a fluorescence microscope by a cytotechnologist screening for cells with abnormal FISH patterns. Abnormal FISH cell types included trisomy (> 2 signals of a single locus [eg, CEP 7]), tetrasomy (exactly 4 copies of each of the 4 loci), or polysomy (gain of ≥ 2 loci, unless each locus had 4 copies, which was classified as a tetrasomic cell) (Fig. 2). As previously published, cutoff values for each abnormality included 10 cells with trisomy, 10 cells with tetrasomy, or 5 cells with polysomy.[9, 13, 15, 19, 20] If none of the aforementioned thresholds were reached, a specimen was classified as negative. It is important to note that a cell with a gain of multiple probes was classified as polysomic unless every probe exhibited tetrasomy (4 copies), in which case it was classified as a tetrasomic cell. It is important to distinguish tetrasomic cells from other cells with multiple gains (ie, polysomic cells) because tetrasomic cells are less specific for malignancy. Although some tumors are tetraploid, a cell with normal DNA content (2n) in the G2 phase of the cell cycle (4n) will likewise have 4 copies of each chromosome (ie, tetrasomy). In our experience, only approximately 40% of patients with a tetrasomic FISH result have cancer, and therefore differentiation from polysomy is important to preserve the specificity of a polysomic FISH result. Of all FISH abnormalities, polysomic cells are most likely to be associated with malignancy. Therefore, it is standard practice in our laboratory to screen the entire FISH slide in search of polysomic cells, even if a cutoff value for another FISH abnormality (eg, trisomy) has been reached. Screening was concluded if the cutoff value for polysomy was reached (5 cells) or the cellular area was exhausted.
Two-year cancer status was determined based on whether the patient developed cancer at any point between baseline and 2 years after baseline. Clinical characteristics initially were compared with 2-year cancer status using chi-square tests for categorical data (sex, cytology, FISH, serum CA 19-9 level, receipt of ursodeoxycholic acid therapy, and IBD status) and Kruskal-Wallis tests for continuous data (age, serum alkaline phosphatase level, and total bilirubin). The sensitivity, specificity, and positive predictive value were calculated for serum CA 19-9 versus 2-year cancer status using a cutoff value of 129 U/mL.
Associations between overall cancer risk (incorporating time to malignancy) and clinical characteristics were assessed with Cox proportional hazards regression models, and univariate associations were graphically depicted with Kaplan-Meier plots. The assumption of proportional hazards for each predictor was assessed graphically and tested by way of interaction with time (no deviation from this assumption was detected). The concordance index (C-statistic) was calculated for each regression model. This is a measure of predictive accuracy whereby a value of 1.0 indicates that the model perfectly discriminates patients with different outcomes, and a value of 0.5 indicates a model with no predictive information. The median time to malignancy was estimated along with 95% confidence intervals (95% CIs) using the Kaplan-Meier method. Patients who were found to have cancer immediately after their baseline brushing/tests were included in these analyses, and their time to cancer was set at 0.5 days (3 patients). We evaluated the serum CA 19-9 level as a binary predictor and as a continuous variable. These analyses yielded similar associations with cancer, and therefore it was ultimately included as a binary predictor (< 129 U/mL vs ≥ 129 U/mL) in accordance with clinical practice and the literature. All analyses were conducted using SAS statistical software (version 9; SAS Institute Inc, Cary, NC). P values < .05 were considered to be statistically significant.
There were 102 patients with PSC (73 men and 29 women) who fulfilled study criteria. Descriptive statistics are outlined in Table 1. The mean age of the patients with an equivocal cytology diagnosis was 47.8 years (range, 20 years-76 years). Seventy patients (69%) had IBD and 48 patients (47%) were receiving ursodeoxycholic acid therapy at the time of the equivocal brushing. The mean alkaline phosphatase level was 415 U/L (data were available for 100 patients). The mean total bilirubin level was 3.1 mg/dL (data were available for 99 patients). A total of 66 cytology brushing specimens were interpreted as atypical (65%) and 36 brushings were termed suspicious (35%) by a cytopathologist. The mean serum CA 19-9 level was 667 U/mL (median, 31 U/mL [range, 1 U/mL-26,200 U/mL]).
Table 1. Descriptive Statistics at Baseline/Equivocal Cytology Result by Cancer Status After 2 Years
Within 2 years of the equivocal brushing result, cancer was diagnosed in 30 patients (29%), including cholangiocarcinoma (27 patients), pancreatic adenocarcinoma (2 patients), and gallbladder adenocarcinoma (1 patient). Approximately 73% of patients (22 of 30 patients) had pathologic evidence of cancer as determined by bile duct biopsy or resection (5 patients), biopsy/FNA demonstrating metastatic disease in a lymph node or the peritoneum (6 patients), subsequent positive routine cytology of a biliary brushing (5 patients), orthotopic liver transplant (4 patients), Whipple resection (1 patient), or pancreatic FNA (1 patient). The remaining patients with cancer (8 of 30 patients; 27%) were diagnosed by a combination of clinical features including the detection of a new, malignant-appearing mass on cross-sectional imaging studies (subsequent to baseline evaluation) and/or death from metastatic disease. The median follow-up time for patients without malignancy was 4.3 years (interquartile range, 2.3 years [3.0 years-5.3 years]).
Patients with polysomy by FISH analysis were found to have a significantly (P < .001) higher likelihood (76%) of developing a pancreatobiliary tract malignancy within 2 years compared with patients with a negative FISH result (16%) or other type of FISH abnormality (12%) (Table 1). The serum CA 19-9 level was also found to be significantly associated with the risk of developing cancer within 2 years. The 21 patients with elevated serum CA 19-9 levels (≥129 U/mL) were more likely to be diagnosed with cancer compared with the 76 patients with serum levels below this cutoff value (52% vs 24%; P = .011). Using a cutoff value of 129 U/mL, we observed a sensitivity of 37%, a specificity of 85%, and a positive predictive value of 50% for the detection of malignancy. Age, sex, receipt of ursodeoxycholic acid therapy, IBD status, serum alkaline phosphatase level, total bilirubin level, and routine cytology result were not found to be significantly associated with 2-year cancer status.
Laboratory values at baseline versus those at the time of cancer diagnosis were compared in those cancer patients for whom data points were available (26 patients with available data for serum alkaline phosphatase level and total bilirubin, and 23 patients with data available for serum CA 19-9 level). There was no significant difference noted in the distribution of alkaline phosphatase (median of 439 U/L vs 350 U/L; P = .121), total bilirubin (median of 2.1 mg/dL vs 1.9 mg/dL; P = .098), or serum CA 19-9 (median of 58 U/L vs 45 U/L; P = .837) at baseline versus the time of cancer diagnosis, respectively.
We next analyzed the data with Cox proportional hazards regression models to take into account the time to cancer detection. Using univariate analysis, we found that polysomy by FISH analysis (hazard ratio [HR] 8.70; P < .001), a serum CA 19-9 level ≥ 129 U/mL (HR 3.19; P = .001), and polysomy with a serum CA 19-9 level ≥ 129 U/mL (HR 10.92; P < .001) were significant predictors of a patient developing malignancy (Table 2) (Fig. 3). Among those patients with polysomy FISH, the median time to malignancy was estimated to be 71 days (95% CI, 23 days-365 days), and among those with a serum CA 19-9 level ≥ 129 U/mL, the median time was estimated to be 105 days (95% CI, lower limit of 23 days, upper limit could not be calculated). The median time to malignancy was not reached among the reference groups (Fig. 3). The unadjusted HR for polysomy by FISH with a serum CA 19-9 level ≥ 129 U/mL versus nonpolysomy by FISH and a serum CA 19-9 level < 129 U/mL was 20.44 (95% CI, 7.94-52.63).
Table 2. Results of Time to Cancer Analysis
HR (95% CI)
HR (95% CI)
Abbreviations: 95% CI, 95% confidence interval; CA 19-9, carbohydrate antigen 19-9; FISH, fluorescence in situ hybridization; HR, hazard ratio; NA, not applicable.
Multivariable model c-statistic = 0.804.
Data regarding the alkaline phosphatase level were not available for 2 patients.
Data regarding bilirubin were not available for 3 patients.
Data regarding the CA 19-9 level were not available for 5 patients.
Using multivariable analysis, we included the significant variables from the univariate analysis (FISH and serum CA 19-9 level) and routine cytology rather than all predictor variables to avoid model overfitting. Polysomy by FISH (P < .001) remained a statistically significant predictor of malignancy after adjusting for cytology result and serum CA 19-9 level. There was no statistically significant interaction noted between polysomy and serum CA 19-9 level, indicating that the effect of one does not depend significantly on the other (or vice versa). The adjusted HR for polysomy versus negative FISH result was 6.96 (95% CI, 2.97-16.34) (Table 2) (Fig. 4). The C-statistic for this model was 0.804.
The results of the current study indicate that polysomy by FISH analysis can identify patients with PSC without a mass lesion apparent at baseline who are at highest risk of developing a pancreatobiliary malignancy within the setting of equivocal routine cytology results. Multiple studies have described the difficulties of evaluating biliary cytology specimens, with clinical sensitivities reported to be < 50%.[5, 23-25] Intraobserver and interobserver reproducibility of atypical and suspicious categories have been reported as variable.[5, 26] Because of the difficulty in analyzing these specimens, more objective molecular tests such as FISH may be used to aid cytopathologists in the diagnosis of pancreatobiliary tract malignancies.
In the current study, the findings on univariate and multivariable analyses indicated that polysomy by FISH is the best predictor of malignancy in patients with an equivocal cytology result in the absence of a periductal mass on cross-sectional imaging. The strong association between polysomic FISH results and pancreatobiliary malignancy is consistent with previous studies assessing both PSC and non-PSC populations.[9, 27] This is also consistent with studies assessing FISH testing on cytology specimens from other anatomic sites (eg, bladder, esophagus, and lung). Although the main benefit of performing FISH is to increase diagnostic sensitivity, specificity has been shown to be excellent, especially in non-PSC patients.[13, 28, 29] However, there have been some reports of false-positive polysomy results with FISH analysis in patients with PSC.[14, 15, 30] As a result, previous reports suggest that polysomy by FISH analysis in patients with PSC must be corroborated with clinical findings and verified over time.[14, 31]
In the current study, 25 patients with equivocal cytology were found to have polysomy by FISH. Six of these patients were not found to have cancer on clinicopathologic follow-up within 2 years of the equivocal cytology result (ie, positive predictive value of 76%). False-positive FISH results can occur due to the misinterpretation of cells on the FISH slide, clinical treatment after brush cytology that reduces tumor burden before it can be pathologically or clinically confirmed by an alternative test (gold standard for this study), and/or the ability of FISH to detect premalignant chromosomal changes in cells (ie, dysplasia) that have not progressed to cancer. DeHaan et al reported that 3 of 7 dysplastic lesions from PSC patients (43%) harbored polysomic cells by FISH. It is interesting to note that 1 of the “false-positive” patients in the current study was found to have a positive cytology and a mass that was diagnosed as cholangiocarcinoma 2.7 years after the initial equivocal diagnosis. This suggests that the positive predictive value of polysomy FISH and equivocal cytology is even higher than 76%, because our statistical cutoff value was set at 2 years of follow-up. It is also important to note that not all FISH abnormalities carry the same clinical significance. Data from the current study suggest that separating polysomy from other abnormal FISH findings (ie, trisomy and tetrasomy) is useful for predicting whether patients will have cancer. This is consistent with a report from Bangarulingam et al that concluded that patients with other FISH abnormalities had clinical outcomes comparable to those of patients with negative FISH results and should be managed accordingly.
The expense and invasive nature of ERCP with epithelial sampling precludes its use as a regular screening method in patients with PSC . In addition, statistical principle and a previous report have demonstrated that the use of FISH in a population with a low cancer prevalence reduces the positive predictive power of the assay. In our practice, patients with PSC undergo ERCP only if it is clinically indicated, including patients with jaundice, cholangitis, abdominal pain, new/progressive hilar stricturing, or elevated laboratory values. Laboratory values, including serum CA 19-9, are assessed every 3 months to 4 months. This tumor marker is generally elevated in patients with PSC with pancreatobiliary malignancy,[33, 34] but is often considered unreliable because of the potential for false-positive results.[35, 36] The use of serum CA 19-9 testing for the detection of malignancy in patients with PSC is controversial, but some studies have reported that its use in conjunction with other modalities is beneficial.[3, 30] We included serum CA 19-9 with a cutoff level of 129 U/mL as a variable in the current multivariable analysis to assess its diagnostic power in comparison with other variables. Although the serum CA 19-9 level was found to be associated with malignancy on univariate analysis, it was not found to be an independent predictor of malignancy in patients with PSC in the presence of equivocal cytology. Unfortunately, serum CA 19-9 cannot be used independently as a triage mechanism for ERCP because of its low negative predictive power. In the current study, 18 of 76 patients with a serum CA 19-9 level < 129 U/mL (24%) were diagnosed with a malignancy within 2 years of the equivocal cytology result. It is important to note that an elevated serum CA 19-9 level in conjunction with polysomy by FISH analysis was found to be diagnostic of malignancy on extended follow-up (Fig. 3D). Ten patients in the current study cohort fulfilled these criteria and 9 of 10 were diagnosed cancer within 2 years of the equivocal cytology. The remaining patient was previously discussed and was diagnosed with cholangiocarcinoma 2.7 years after the equivocal cytology result.
Although the development of cancer was heralded by an abrupt increase in serum total bilirubin and/or CA 19-9 values in an occasional patient, this was not the usual scenario. Indeed, the distribution of serum laboratory values (serum alkaline phosphatase, total bilirubin, and serum CA 19-9) among patients who developed malignancy was not found to be significantly higher at the time of cancer diagnosis compared with the baseline reading. This finding suggests that surveillance of alkaline phosphatase, total bilirubin, and serum CA 19-9 levels does not reliably indicate the presence of malignancy in this cohort of patients.
One concern for clinicians is how to properly manage patients with PSC with atypical versus suspicious cytology results. It has been reported that the positive predictive value of suspicious cytology is lower in patients with PSC compared with non-PSC patients (67% vs 87%-100%, respectively), but how to translate this difference into risk stratification in patients with PSC is unclear. In the current study, the likelihood of developing cancer was greater for patients with suspicious versus atypical cytology results (42% vs 23%; P = .045). However, considering time to malignancy, there was not a significant difference noted between patients with atypical and suspicious cytology (P = .086) (Fig. 3A). Multiple studies have reported the misinterpretation of pancreatobiliary cytology brushings from patients with PSC because these specimens tend to contain reactive cells that are difficult to differentiate from neoplastic cells.[5, 23-25] An advantage of FISH in this setting is that it does not depend entirely on cell morphology. For example, abnormally large nuclei on a FISH slide will likewise be recognized as having increased size but will not be classified as abnormal unless the corresponding signal patterns are abnormal.
Patients with PSC who have an equivocal routine cytologic brushing result but lack a mass lesion on imaging studies pose a challenging clinical dilemma for gastroenterologists. Although this clinical scenario is relatively uncommon and, consequently, the current study is limited to a small number of patients, its findings remain important for the management of patients with PSC in clinical practice. We assessed laboratory and clinical variables based on nearly a decade of clinicopathologic data in this subset of patients with PSC and found that those with polysomy by FISH analysis were at an increased risk of harboring or developing malignancy compared with those without polysomy. Patients with both an elevated serum CA 19-9 level and polysomy by FISH are at a very high risk, because all patients in this cohort were found to have cancer on extended clinicopathologic follow-up.
No specific funding was disclosed.
CONFLICT OF INTEREST DISCLOSURES
Sarah Jenkins has received fees for participation in review activities such as data monitoring boards, statistical analysis, endpoint committees, and the like. The Health Sciences Research Department at the Mayo Clinic works on a charge-out basis to support this project, including work for data management, analysis, and manuscript assistance and review. Dr. Halling and the Mayo Clinic have a patent on, and receive royalties from, the sale of the FISH probe set (UroVysion) discussed in this article. Drs. Halling and Kipp receive grant funding from Abbott Molecular Inc, the manufacturer of the FISH UroVysion probe set discussed herein.