Utility of serum tumor markers, imaging, and biliary cytology for detecting cholangiocarcinoma in primary sclerosing cholangitis

Authors


  • Potential conflict of interest: K.H. is a coinventor on the patent for the UroVysion™ FISH probe set discussed in this study and receives royalties.

Abstract

There is limited information on test performance for detecting cholangiocarcinoma in primary sclerosing cholangitis (PSC), particularly when used sequentially. This study aimed to characterize diagnostic performance of serum carbohydrate antigen 19-9 (CA 19-9), ultrasonography, computed tomography, magnetic resonance imaging, cholangiography, and biliary cytologic techniques for detecting cholangiocarcinoma in PSC. All consecutive patients with PSC were screened and followed for development of cholangiocarcinoma from 2000 through 2006. Of 230 patients, 23 developed cytopathologically confirmed cholangiocarcinoma with an annual incidence of 1.2%. The optimal cutoff value for serum CA 19-9 was 20 U/mL, which yielded a sensitivity of 78%, specificity of 67%, positive predictive value (PPV) of 23%, and negative predictive value (NPV) of 96%. Serum CA 19-9 combined with either ultrasonography, computed tomography, or magnetic resonance imaging provided a sensitivity of 91%, 100%, and 96%, specificity of 62%, 38%, and 37%, PPV of 23%, 22%, and 24%, and NPV of 98%, 100%, and 98%, respectively, if at least one method was positive. Subsequent cholangiographic examinations in these patients increased specificity to 69% and PPV to 42% while maintaining sensitivity of 91% and NPV of 96%. Following this group, conventional cytology, aneuploidy detection by digital imaging analysis, and aneusomy detection by fluorescence in situ hybridization in brushing samples of biliary strictures had a sensitivity of 50%, 57%, and 86%, specificity of 97%, 94%, and 83%, PPV of 86%, 89%, and 80%, and NPV of 83%, 74%, and 88%, respectively, for detecting cholangiocarcinoma. Conclusion: Tumor serology combined with cross-sectional liver imaging may be useful as a screening strategy and cholangiography with cytologic examination is helpful for the diagnosis of cholangiocarcinoma in patients with PSC. (HEPATOLOGY 2008.)

Cholangiocarcinoma occurs with increased frequency in patients with primary sclerosing cholangitis (PSC) and is currently a leading cause of death in these patients.1–4 The development of cholangiocarcinoma is unpredictable on the basis of the duration, symptoms, and severity of PSC.2, 4 The cholangiographic appearance of benign strictures of PSC makes the diagnosis of cholangiocarcinoma difficult in this setting. Most patients are diagnosed with advanced unresectable disease, which imparts a dismal prognosis and is a contraindication of liver transplantation.3 However, innovative approaches using a combination of neoadjuvant chemoradiotherapy and liver transplantation provide excellent outcomes with 5-year survival of 82% in patients with PSC who have perihilar cholangiocarcinoma confined to the biliary tree.5

Currently, many physicians screen patients with PSC with a variety of diagnostic modalities, including: serum carbohydrate antigen 19-9 (CA 19-9) levels; ultrasonography (US), computed tomography (CT), or magnetic resonance imaging (MRI) of the liver; and magnetic resonance cholangiopancreatography (MRCP) or endoscopic retrograde cholangiopancreatography (ERCP), and subsequently perform cytopathologic testing to confirm the diagnosis of cholangiocarcinoma. However, there is limited information on the diagnostic performance of these tests for detecting cholangiocarcinoma in patients with PSC.

The overall objective of this retrospective study was to assess the diagnostic utility of serum CA 19-9; US, CT, and MRI of the liver; cholangiography; and biliary cytologic techniques for detecting cholangiocarcinoma in patients with PSC. Specifically, we reviewed our experience in screening and surveillance for complicating cholangiocarcinoma in a large series of patients with well-defined PSC to estimate the sensitivity, specificity, accuracy, positive predictive values (PPVs), and negative predictive values (NPVs) of each diagnostic test and the possible combinations of serologic, radiologic, and cytologic methods for detecting cholangiocarcinoma in patients with PSC. The information about diagnostic performance of these tests for detecting cholangiocarcinoma is important in the clinical management of patients with PSC.

Abbreviations

AUC, area under the curve; CA 19-9, carbohydrate antigen 19-9; CT, computed tomography; DIA, digital imaging analysis; ERCP, endoscopic retrograde cholangiopancreatography; FISH, fluorescence in situ hybridization; MRCP, magnetic resonance cholangiopancreatography; MRI, magnetic resonance imaging; NPV, negative predictive value; PPV, positive predictive value; PSC, primary sclerosing cholangitis; ROC curve, receiver operating characteristic curve; US, ultrasonography.

Patients and Methods

Patients.

Our patient population was composed of all consecutive patients with well-defined PSC who were identified using a computerized master diagnosis index. The diagnosis of PSC was based on cholangiographic findings of multifocal strictures and beading of the intrahepatic and/or extrahepatic bile ducts6 with compatible biochemical abnormalities and exclusion of secondary causes.3 These patients had their first medical evaluation for their liver disease at the Mayo Clinic, Rochester, MN, between January 2000 and December 2005. Patient follow-up for surveillance of cholangiocarcinoma was conducted until December 2006. In our institution, screening and surveillance for cholangiocarcinoma in PSC is usually performed using annual follow-up with serum CA19-9 level and abdominal US evaluation. The study was approved by the Mayo institutional review board and written informed consent was obtained from all patients for participation in medical research.

Serum Tumor Marker and Liver Biochemistries.

The serum concentrations of CA 19-9 were measured using the Bayer ADVIA Centaur Immunoassay System (laboratory reference range, 0-40 U/mL). Serum levels of CA 19-9 and liver biochemical indices were obtained at the latest time of surveillance for cholangiocarcinoma in patients without cancer or at the closest time to the diagnosis of cholangiocarcinoma in patients with cancer. The Mayo risk score was calculated at the latest time of surveillance or at the diagnosis of cholangiocarcinoma using the revised PSC Mayo risk score.7

Radiologic Evaluation.

The findings from original radiological reports, including US with Doppler study, contrast-enhanced spiral CT, gadolinium-enhanced and/or ferumoxide-enhanced MRI of the liver, ERCP, and MRCP, were recorded to minimize any potential bias associated with radiographic reinterpretation for this study. CT examinations were performed on a variety of multislice CT scanners that were considered state of the art at the time with 3-mm and 5-mm slice thickness, 120 kVp, and mA settings that varied between 132 and 506. MRI examinations were performed using a General Electric Signa 1.5 Tesla MR scanner. Image acquisition was performed using a phase-array torso multicoil. Two acquisitions each were performed in the axial and coronal planes using a single-shot fast-spin-echo pulse sequence with a repetition time of infinity, average echo time equal to 90 ms, matrix size equal to 256/256, and slice thickness of 5 mm with 0-mm skip. Three additional acquisitions were performed in the coronal plane and in planes approximately 45° oblique to the porta hepatis. Findings were considered “definite” when imaging depicted mass lesions; “probable” when imaging revealed evidence of tumor infiltration of hepatobiliary parenchyma; and “possible” when imaging demonstrated the secondary sign of tumor obstruction of the bile duct; a fourth category was “no evidence of tumor.”8

Cross-sectional imaging studies of the liver were evaluated for the presence of the following radiologic features: presence and location of mass, marked bile duct wall thickening (greater than 4 mm, if measured), marked biliary dilatation, and focal biliary dilatation associated with the atrophy-hypertrophy complex. Definite features of the tumor were considered to be present if there was a well-defined mass exhibiting echogenicity different from that of the liver on US; a mass with discernible margins and typical enhancement consistent with that of neoplasm on CT (hypoattenuating relative to the liver parenchyma in the portovenous phase and hyperattenuating in the delayed phase); or a well-defined mass exhibiting abnormal signal intensity consistent with that of neoplasm on MRI (hypointense on spin lattice relaxation time [T1]-weighted images and hyperintense on spin-spin or transverse relaxation time [T2]-weighted images). A probable tumor was considered if there was a mass-like area with poorly defined margins, or a marked thickening of the bile duct wall without typical enhancement for neoplasm on contrast imaging studies; a possible tumor if there was marked biliary dilatation without mass, or focal intrahepatic biliary dilatation associated with the ipsilateral lobe atrophy; and no tumor if there was either minimal biliary dilatation or slight bile duct wall thickening compatible with PSC.

Cholangiograms by either ERCP or MRCP were evaluated for the presence of dominant biliary stricture (single stricture disproportionately severe relative to other strictures), marked biliary dilatation, biliary filling defects, location of filling defect or mass, and other findings associated with PSC. A diagnosis of tumor was considered to be definite if there was a biliary stenosis associated with a polypoid bile duct lesion; probable if there was a combination of a dominant stricture and marked proximal bile duct dilatation; possible if there was a dominant stricture without marked proximal bile duct dilatation, marked diffuse biliary dilatation alone, or a focal segmental biliary dilatation; and none if typical features of sclerosing cholangitis in the absence of a dominant stricture or marked biliary dilatation were present. Marked biliary dilatation was defined as the presence of one or more of the following findings: common bile duct 2 cm or greater, right hepatic duct 0.8 cm or greater, left hepatic duct 1 cm or greater, or intrahepatic ducts greater than 0.5 cm in diameter.6

Cytological Evaluation.

Current processing of bile duct brushings at our institution consists of collecting two separate samples from the biliary stricture as previously described.9 Biliary strictures were not routinely dilated unless necessary to gain access or were considered to be causing symptoms. Cytology, digital imaging analysis (DIA), and fluorescence in situ hybridization (FISH) analyses were performed by cytotechnologists who had no knowledge of the other test results or patient's clinical history. Cytologic diagnoses were classified as malignant, suspicious for malignancy, atypical, or benign using accepted criteria.10 DIA is a form of cytologic analysis that quantifies cellular constituents by using spectrophotometric principles, and the results were categorized as diploid (DNA index between 0.95 and 1.10), aneuploid (DNA index between 1.11 and 1.89), or tetraploid (DNA index between 1.90 and 2.10).11 Aneuploid and tetraploid results were considered positive for malignancy. FISH uses fluorescently-labeled DNA probes to chromosomal centromeres or unique loci to detect cells that have numeric or structural abnormalities indicative of malignancy.12 The UroVysion probe set (Abbott Molecular Inc., Des Plaines, IL) targets centromere regions of chromosome 3 (CEP 3), chromosome 7 (CEP 7), chromosome 17 (CEP 17), and band 9p21 (P16/CDKN2A gene). A patient's specimen was considered positive for malignancy if five or more cells showed gains of two or more of all four probes (polysomy) or if ≥10 cells showed three copies of chromosome 7 or 3 and two or fewer copies of the other probes.

Statistical Analysis.

Continuous variables were presented as mean ± standard deviation or median (range) and compared using standard parametric and nonparametric methods where appropriate. Frequency data were presented as number and percentage and compared using the chi-square test or Fisher's exact test where appropriate. The performance of tumor marker was assessed using the receiver operating characteristic (ROC) curve. Sensitivity, specificity, PPV, NPV, and accuracy, with their exact 95% confidence intervals (95% CI), were obtained for each testing method based on the binomial distribution. For each of the radiologic and cytologic tests, we made the evaluations based on the different possible outcomes indicative of malignancy. Survival curves were constructed using the Kaplan-Meier method and were compared by the log-rank test. All statistical testing was done at the conventional two-tailed α level of 0.05.

Results

After excluding patients who: were referred from elsewhere for management of their cholangiocarcinoma (n = 68); had no cholangiograms for review to confidently make a diagnosis of PSC (n = 16); had an unclear final diagnosis as to benign or malignant liver disease (n = 10); or had no diagnostic tests in surveillance for cholangiocarcinoma during follow-up (n = 4), a total of 230 patients with PSC qualified for the study. Patients had a mean age of 42.1 ± 16.5 years when first seen for their liver disease in our institution and were followed with a median duration of 4.1 years (range, 0.1-22.5 years) after the diagnosis of PSC. Of these, 29, 34, 28, 40, and 99 patients were followed for development of cholangiocarcinoma for 1 year, 2 years, 3 years, 4 years, and more than 5 years, respectively. Within the patients with cholangiocarcinoma, nine (39%) were detected by initial screening and in five (56%) of them the cancer was identified at an advanced stage disease (stage III-IV).13 The other 14 cholangiocarcinomas (61%) were detected by surveillance with an annual incidence of 1.2%, and 11 (79%) of them had cancers detected at an early stage (stage I-II).13 The diagnosis of cholangiocarcinoma was based on pathologic evidence from hepatobiliary specimens for 15 (65%) of the patients and by the presence of malignant cells on conventional cytologic examinations from biliary brush samples in eight patients (35%). Proof of the absence of malignancy in patients with PSC was based on benign pathologic findings of liver specimens (n = 15), or a 1 year or longer follow-up after the absence of malignant evidence in either biliary brushing specimens (n = 88), or radiologic evaluation of hepatobiliary tract (n = 104). Patients with cholangiocarcinoma tended to be older; had a significantly higher proportion with a history of smoking; the occurrence of symptoms, particularly jaundice; and more advanced liver disease than patients with PSC alone as shown in Table 1.

Table 1. Characteristics and Laboratory of PSC Patients with and without Cholangiocarcinoma
 Patients without Cholangiocarcinoma (n = 207)Patients with Cholangiocarcinoma (n = 23)P Value
  • Data are expressed as median (range). The normal ranges are follows: alkaline phosphatase, 90–234 IU/L; aspartate aminotransferase, 12–31 U/L; alanine aminotransferase 12–45 U/L; total bilirubin, 0.1–1.1 mg/dL; albumin, 3.5–5 g/dL.

  • *

    Symptomatic includes abdominal pain, jaundice, weight loss, or cholangitis.

Age at PSC (years)40 (28-52)49 (34-61)0.06
Sex (F/M)87/1206/170.2
Smoking (current/former versus none)18 versus1897 versus160.006
Alcohol (current/former versus none)44 versus1633 versus200.4
Inflammatory bowel disease (%)157 (76%)18 (78%)1.0
 Ulcerative colitis131 (63%)13 (56%) 
 Crohn's disease24 (12%)5 (22%) 
 Unclassified colitis2 (1%)0 
Colonic dysplasia/cancer (%)14 (7%)1 (4%)1.0
Proctocolectomy (%)33 (16%)5 (22%)0.6
Symptoms/signs (%)   
 Symptomatic*103 (50%)18 (78%)0.01
 Abdominal pain63 (30%)6 (26%)0.8
 Jaundice51 (25%)11 (48%)0.02
 Weight loss35 (17%)6 (26%)0.3
 Cholangitis26 (13%)1 (4%)0.5
 Variceal bleeding8 (3%)01.0
 Ascites10 (5%)2 (9%)0.3
 Hepatomegaly33 (16%)3 (13%)1.0
 Splenomegaly22 (11%)2 (9%)1.0
Ursodeoxycholic acid therapy (%)116 (56%)10 (43%)0.3
Laboratory   
 Alkaline phosphatase (U/L)461 (231–800)328 (238–1,402)0.7
 Aspartate aminotransferase (U/L)70 (35–122)71 (37–130)0.7
 Alanine aminotransferase (U/L)117 (55–212)117 (43–192)0.9
 Total bilirubin (mg/dL)1.0 (0.6–2.3)2.4 (0.7–4.4)0.08
 Albumin (g/dL)4.0 (3.6–4.2)3.9 (3.0–4.2)0.1
 International normalized ratio1.0 (0.9–1.0)1.0 (0.9–1.2)0.09
Revised PSC risk score0.21 (−0.38–1.26)0.95 (−0.08–1.83)0.04

Characteristics and Survival of Cholangiocarcinomas

The median age of 23 patients with cholangiocarcinoma was 51 years (range, 19–68) at the time of cancer diagnosis. Patients with advanced cancer presented more often with jaundice and higher levels of serum bilirubin and alkaline phosphatase than those with early cancer as shown in Table 2. Of the 15 patients with early stage cancer, one patient with intrahepatic cholangiocarcinoma had a right hepatectomy, 10 patients with perihilar cholangiocarcinoma were treated with neoadjuvant chemoradiotherapy and exploratory laparotomy to rule out metastatic disease preceding orthotopic liver transplantation per the clinical protocol developed in our institution,5 and one patient is on the waiting list for liver transplantation after completing a course of chemoradiotherapy. All patients undergoing transplantation had a cancer diagnosis before transplantation. The other three cancers at early stage were not enrolled into our transplant protocol because of severe post-ERCP pancreatitis, repeated stroke, and waiting for transplantation at another center, respectively. The patients with advanced cancer were palliated with photodynamic therapy, chemoradiotherapy, biliary drainage, and analgesic medications. A total of 10 patients who have undergone transplantation are still alive with no evidence of recurrent tumor during a median follow-up of 13 months (range, 8.2–21.6) after transplantation. Observed survival among patients with an early-stage cancer was significantly higher than survival among patients with an advanced stage cancer (P < 0.0001).

Table 2. Clinical Features of the Patients with Cholangiocarcinoma and Comparison between Stage I–II and Stage III–IV Groups
FeaturePatients with Stage I–II Cancer (n = 15)Patients with Stage III–IV Cancer (n = 8)P Value
  • Data are expressed as median (range).

  • *

    Symptomatic includes abdominal pain, jaundice, weight loss, or cholangitis.

Age at diagnosis of cancer (years)52 (44–62)45 (35–65)0.7
Sex (F/M)5/101/70.4
IBD (%)12 (80%)6 (75%)1.0
Symptoms/signs (%)   
 Symptomatic*10 (67%)8 (100%)0.1
 Abdominal pain4 (27%)2 (25%)1.0
 Jaundice4 (27%)7 (88%)0.009
 Weight loss3 (20%)3 (38%)0.6
 Cholangitis1 (7%)01.0
 Ascites1 (7%)1 (13%)1.0
 Hepatomegaly1 (7%)2 (25%)0.3
 Splenomegaly02 (25%)0.1
Laboratory   
 Alkaline phosphatase (U/L)307 (136–1,068)1,233 (392–2,140)0.03
 Aspartate aminotransferase (U/L)48 (27–110)127 (77–176)0.06
 Alanine aminotransferase (U/L)77 (37–159)193 (138–258)0.07
 Total bilirubin (mg/dL)1.0 (0.5–2.5)9.5 (3.1–12.0)0.003
 Albumin (g/dL)3.9 (3.0–4.2)3.9 (3.1–4.2)0.8
 International normalized ratio1.0 (0.9–1.1)1.1 (0.9–1.4)0.2
Serum CA 19–9 (U/mL)23 (17–71)71 (45–1,609)0.07
Revised PSC risk score0.84 (−0.18–1.31)1.78 (0.51–3.40)0.09
Time elapsed since PSC to the diagnosis of cancer (months)40 (1–60)0 (0–11)0.01

Operative Performance of Individual Diagnostic Tests

Serum CA 19–9.

Over the 7-year follow-up period, 711 serum CA 19-9 tests were performed on 208 patients with a mean number of 3.4 tests per patient. At the time of cholangiocarcinoma diagnosis, the median level of serum CA 19–9 in cancer patients was 45 U/mL (range, 20–95) compared to the median serum CA 19-9 level of 15 U/mL (range, 10–27) for patients with PSC alone (P < 0.0001). By plotting sensitivity of serum CA 19-9 against 1 − specificity for each possible cutpoint, the area under the curve (AUC) was 0.79. The optimal cutoff value for serum CA 19-9 was 20 U/mL, which yielded sensitivity of 78% and specificity of 67% (Table 3).

Table 3. Diagnostic Performance of Serum CA 19-9 with Different Cutoff Values
 20 U/mL40 U/mL100 U/mL129 U/mL200 U/mL
True positive (n)1813533
Tumor location (n)     
 Intrahepatic43211
 Extrahepatic1410322
Tumor staging (n)     
 Stage I-II116200
 Stage III-IV77333
AUC0.790.710.600.570.57
Sensitivity (95% CI)78% (0.58–0.90)57% (0.37–0.74)22% (0.10–0.42)13% (0.05–0.32)13% (0.05–0.32)
Specificity (95% CI)67% (0.60–0.73)84% (0.78–0.88)99% (0.96–1)100% (0.99–1)100% (0.99–1)
PPV (95% CI)23% (0.15–0.33)30% (0.19–0.45)71% (0.36–0.92)100% (0.44–1)100% (0.44–1)
NPV (95% CI)96% (0.91–0.98)94% (0.89–0.97)91% (0.86–0.94)90% (0.85–0.94)90% (0.85–0.94)
Accuracy (95% CI)68% (0.62–0.74)81% (0.75–0.86)90% (0.86–0.94)90% (0.86–0.94)90% (0.86–0.94)

Based on our population, we also obtained and compared diagnostic performance of serum CA 19-9 using other cutoff values that are frequently reported in the literature (Table 3). The sensitivity of serum CA 19-9 decreased from 78% to 13% with increasing cutoff values from 20–200 U/mL, while the specificity increased from 67% to 100%. With increasing cutoff values of serum CA 19-9, post-test probability of having cholangiocarcinoma markedly increased from 23% to 100%, contrasting with the ability to predict the absence of cholangiocarcinoma, which decreased slightly from 96% to 90%. It is important to note that the cutoff values of either 129 or 200 U/mL identified only advanced cancer, whereas using lower cutoff values tended to detect cancer at an early stage, as shown in Table 3.

Cross-Sectional Liver Imaging.

During the follow-up period, 650 US examinations were performed on 210 patients, 198 CT examinations were performed on 117 patients, and 134 MRI examinations were performed on 105 patients. The mean number of US, CT, and MRI examinations was 3.1, 1.7, and 1.3 per patient, respectively. After combining the results of all cross-sectional imaging studies, a definite finding was considered to be present in eight patients with cancer (35%), a probable finding in five patients with cancer (22%), and a possible finding in eight patients with cancer (35%). In patients who had a liver mass, intrahepatic cholangiocarcinomas were confirmed with a needle biopsy of the lesion and exclusion of another primary source of adenocarcinoma.

The operative performances of each cross-sectional imaging study of the liver for detecting cholangiocarcinoma in PSC are shown in Table 4. As expected, US, CT, and MRI had a low sensitivity (10%–32%) if only a definite finding was considered positive for malignancy. None of the patients with a definite finding on US, CT, or MRI had a false-positive result, and, consequently, the specificity and PPV were 100%. The sensitivity of US, CT, and MRI increased to 57%–75% and the specificity dropped moderately to 79%–94%, if overall findings including definite, probable, or possible findings were considered positive. MRI had a higher sensitivity on definite findings but a lower sensitivity on probable or possible findings than that of either US or CT. Among the three methods studied and the four different subgroups of patients, US examination had the best accuracy and PPV for detecting cholangiocarcinoma in patients with PSC.

Table 4. Diagnostic Performance of Cross-Sectional Liver Imaging for Detecting Cholangiocarcinoma in PSC
 AUCSensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)Accuracy (95% CI)
  1. Definite finding if there was a well-defined mass; probable finding if a poorly defined mass-like area or marked bile duct wall thickening; possible finding if marked biliary dilatation without mass or focal intrahepatic biliary dilatation with atrophy of the ipsilateral lobe.

US      
 Overall finding0.7657% (0.37–0.76)94% (0.90–0.97)48% (0.29–0.67)95% (0.91–0.97)90% (0.86–0.94)
 Definite finding0.5510% (0.03–0.29)100% (0.98–1)100% (0.34–1)91% (0.86–0.94)91% (0.86–0.94)
 Probable finding0.6019% (0.08–0.40)97% (0.93–0.99)40% (0.17–0.69)92% (0.87–0.95)89% (0.84–0.93)
 Possible finding0.6329% (0.14–0.50)97% (0.94–0.99)55% (0.28–0.79)92% (0.88–0.95)90% (0.86–0.94)
CT      
 Overall finding0.7875% (0.51–0.90)80% (0.71–0.87)38% (0.23–0.55)95% (0.89–0.98)79% (0.71–0.86)
 Definite finding0.6325% (0.10–0.49)100% (0.96–1)100% (0.51–1)89% (0.82–0.94)90% (0.83–0.94)
 Probable finding0.5719% (0.07–0.43)95% (0.89–0.98)38% (0.14–0.69)88% (0.81–0.93)85% (0.77–0.90)
 Possible finding0.5831% (0.14–0.56)85% (0.77–0.91)25% (0.11–0.47)89% (0.81–0.94)78% (0.69–0.84)
MRI      
 Overall finding0.7163% (0.41–0.81)79% (0.69–0.86)40% (0.25–0.58)91% (0.82–0.95)76% (0.67–0.83)
 Definite finding0.6632% (0.15–0.54)100% (0.96–1)100% (0.61–1)87% (0.79–0.92)88% (0.80–0.93)
 Probable finding0.5211% (0.03–0.31)95% (0.89–0.98)33% (0.10–0.70)83% (0.74–0.89)80% (0.71–0.87)
 Possible finding0.5821% (0.09–0.43)84% (0.75–0.90)22% (0.09–0.45)83% (0.73–0.89)72% (0.63–0.80)

Cholangiography.

Over the 7-year period, 391 ERCP and 100 MRCP examinations were performed on 230 and 76 patients, respectively. The mean number of ERCP and MRCP examinations was 1.7 and 1.3 per patient, respectively. A polypoid bile duct mass at the hepatic duct confluence was present in three cancer patients, one of whom had complete obstruction of the left hepatic duct. Cholangiograms depicted dominant strictures involving the extrahepatic bile ducts in 26 patients, the hepatic duct confluence in 42 patients, the intrahepatic bile ducts in 15 patients, and both intrahepatic and extrahepatic bile ducts in three patients. Of these 86 patients, the biliary stricture was malignant in only 16 patients (19%). Interestingly, two patients with ductal cholangiocarcinoma had diffuse biliary dilatation without an obvious biliary stricture.

The operative performances of both cholangiographic techniques are described in Table 5. When all definite, probable, and possible findings were considered as positive for malignancy, the sensitivity of ERCP was higher but the specificity was lower when compared with MRCP alone. When only definite findings on ERCP or MRCP images were considered as positive for malignancy, the sensitivity dropped to 11%–13%, but the specificity and PPV increased to 100%. Concurrently performing a contrast-enhanced MRI of the liver can complement MRCP examination by adding other important tumor information. Thus, if a definite finding on MRCP and MRI was considered as positive, its sensitivity increased to 44%, which was higher than other imaging methods. When all definite, probable, and possible findings were considered as positive for malignancy, diagnostic performance of MRCP plus MRI was comparable with ERCP for detecting cholangiocarcinoma. The cholangiographic features of a dominant stricture and/or marked biliary dilatation had a higher sensitivity but a lower specificity and PPV than that of the presence of intraductal polypoid mass for predicting cholangiocarcinoma in patients with PSC.

Table 5. Diagnostic Performance of Cholangiograms for Detecting Cholangiocarcinoma in PSC
 AUCSensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)Accuracy (95% CI)
  1. Cholangiogram was classified as definite if there is an intraductal mass, probable if a combination of a dominant stricture and marked biliary dilatation, possible if either a dominant stricture or marked biliary dilatation was present.

ERCP      
 Overall finding0.7991% (0.73–0.98)66% (0.59–0.72)23% (0.16–0.34)99% (0.95–1)69% (0.62–0.74)
 Definite finding0.5713% (0.05–0.32)100% (0.98–1)100% (0.44–1)91% (0.87–0.94)91% (0.87–0.94)
 Probable finding0.6643% (0.26–0.63)88% (0.83–0.92)29% (0.17–0.46)93% (0.89–0.96)84% (0.79–0.88)
 Possible finding0.5635% (0.19–0.55)78% (0.72–0.83)15% (0.08–0.27)91% (0.86–0.95)73% (0.67–0.79)
MRCP      
 Overall finding0.7778% (0.45–0.94)76% (0.68–0.83)21% (0.11–0.38)98% (0.92–0.99)76% (0.68–0.83)
 Definite finding0.5611% (0.02–0.44)100% (0.97–1)100% (0.21–1)93% (0.87–0.97)93% (0.87–0.97)
 Probable finding0.5833% (0.12–0.65)84% (0.76–0.89)14% (0.05–0.35)94% (0.87–0.97)80% (0.72–0.86)
 Possible finding0.6333% (0.12–0.65)93% (0.86–0.96)27% (0.10–0.57)94% (0.88–0.97)88% (0.81–0.93)
MRCP plus MRI      
 Overall finding0.8289% (0.57–0.98)75% (0.67–0.83)23% (0.12–0.39)99% (0.94–1)76% (0.68–0.83)
 Definite finding0.7244% (0.02–0.44)100% (0.97–1)100% (0.51–1)96% (0.90–0.98)96% (0.91–0.98)
 Probable finding0.5222% (0.06–0.55)82% (0.74–0.88)9% (0.03–0.28)93% (0.86–0.96)77% (0.69–0.84)
 Possible finding0.6022% (0.06–0.55)94% (0.87–0.97)22% (0.06–0.55)94% (0.87–0.97)88% (0.81–0.93)

Biliary Cytology.

A total of 216 brushing samples of the bile duct in 117 patients were obtained for cytology, DIA, and FISH testing during the 7–year period. Not all specimens obtained were adequate for DIA (n = 51) and FISH (n = 52) testing. The operative performances of conventional cytology were calculated based on only patients with pathologically-proven cholangiocarcinoma whereas those of DIA and FISH testing were calculated based on all patients with either pathologically-proved or cytologically-proven cholangiocarcinoma (Table 6). As expected, cytology had a very low sensitivity (8%) if only a positive cytology result was considered positive for malignancy. Sensitivity increased if both positive and suspicious results were considered positive (46%) and the specificity dropped slightly to 97%. Overall DIA (either aneuploid or tetraploid was considered as positive) had intermediate sensitivity (63%) but was higher than that of cytology, even when suspicious samples were considered as positive for the cytology analysis. When only tetraploid results were analyzed, the sensitivity was lower (19%) but the specificity was slightly increased to 92%. For the FISH results, when either polysomy or trisomy FISH findings were considered as positive, the sensitivity was 88% and specificity was only 73%. If only FISH polysomic results were considered positive, the sensitivity dropped to 38%; in contrast, the specificity increased to 98%. For the combinations of both DIA and FISH, when both DIA tetraploid and FISH polysomy were considered as positive, this combination had a very low sensitivity (6%), but the specificity was increased to 100%. This combination could be reviewed as equivalent to a positive result on conventional cytology for diagnosis of cholangiocarcinoma.

Table 6. Diagnostic Performance of Biliary Cytology, DIA, and FISH for Detecting Cholangiocarcinoma in PSC
 AUCSensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)Accuracy (95% CI)
  1. Diagnostic performance was calculated based on patients with histologically proven cholangiocarcinoma for cytology and on all patients with histologically or cytologically proven cholangiocarcinoma for DIA and FISH.

Cytology      
 Positive or suspicious0.7846% (0.23–0.71)97% (0.91–0.99)67% (0.35–0.88)93% (0.86–0.97)91% (0.84–0.95)
 Positive0.618% (0.01–0.33)100% (0.96–1)100% (0.21–1)89% (0.81–0.93)89% (0.82–0.94)
 Suspicious0.6738% (0.18–0.64)97% (0.91–0.99)63% (0.31–0.86)92% (0.85–0.96)90% (0.83–0.94)
DIA      
 Aneuploid or tetraploid0.7163% (0.39–0.82)80% (0.67–0.89)50% (0.30–0.70)87% (0.74–0.94)76% (0.64–0.84)
 Aneuploid0.6644% (0.23–0.67)88% (0.76–0.94)54% (0.29–0.77)83% (0.71–0.91)77% (0.66–0.86)
 Tetraploid0.5519% (0.07–0.43)92% (0.81–0.97)43% (0.16–0.75)78% (0.66–0.87)74% (0.63–0.83)
FISH      
 Polysomy or trisomy0.8088% (0.64–0.97)73% (0.60–0.84)52% (0.34–0.69)95% (0.83–0.99)77% (0.65–0.85)
 Polysomy0.6838% (0.18–0.61)98% (0.89–0.99)86% (0.49–0.97)83% (0.71–0.90)83% (0.72–0.90)
 Trisomy 7 or 30.6350% (0.28–0.72)76% (0.62–0.85)40% (0.22–0.61)82% (0.69–0.91)69% (0.57–0.79)
DIA and FISH      
 Aneuploid/tetraploid and polysomy/trisomy0.7663% (0.39–0.82)90% (0.79–0.96)67% (0.42–0.85)88% (0.77–0.94)83% (0.73–0.90)
 Tetraploid and polysomy0.536% (0.01–0.28)100% (0.93–1)100% (0.21–1)77% (0.65–0.85)77% (0.66–0.86)
DIA or FISH      
 Aneuploid/tetraploid or polysomy/trisomy0.7588% (0.64–0.97)63% (0.49–0.75)44% (0.28–0.61)94% (0.80–0.98)69% (0.57–0.79)
 Tetraploid or polysomy0.7050% (0.28–0.72)90% (0.78–0.96)62% (0.36–0.82)85% (0.72–0.92)80% (0.69–0.88)

Operative Performance of Combined Diagnostic Tests

In our practice, the combination of serum CA 19-9 and radiologic testing is usually used in screening for cholangiocarcinoma in patients with PSC. Over the 7-year period, there were 230 combined examinations of serum CA 19-9 with different radiologic methods on the entire population. Measurements of serum CA 19-9 levels combined with US, CT, MRI, MRCP, or ERCP examinations were performed on 201, 142, 136, 146, 146, and 212 patients during the follow-up period, respectively. The results of elevated serum CA 19-9 (≥20 U/mL) combined with a radiologic method (overall finding) was considered positive when at least one method was positive and negative when both methods were negative (Table 7). As expected, combined tests improved sensitivity but with decreased specificity for the detection of cholangiocarcinoma when compared with individual tests. The combination of serum CA 19-9 and either CT, MRI, MRCP, or ERCP had the best sensitivity (∼100%) but with a low specificity (∼40%) whereas the combination of serum CA 19-9 and US had intermediate specificity (62%) with a good sensitivity (91%) for detecting cancer. The specificities of these combinations were increased to 89%–98% but the sensitivities dropped moderately to 42%–70% if both methods were positive (Table 7). The accuracy of the combination of CA 19-9 and cross-sectional imaging was greatly enhanced if both methods were positive compared to the accuracy if only one method was positive. Among the six different combinations, measurement of CA 19-9 level combined with US examination had the best accuracy in correctly identifying the patient group with cancer. All combinations of serum CA 19-9 and radiologic modalities had a very high NPV, which suggests that physicians can confidently use negative test results to rule out cancer.

Table 7. Diagnostic Performance of Combinations of Serum CA 19-9 and a Hepatobiliary Imaging for Detecting Cholangiocarcinoma in PSC
Combined Testing*AUCSensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)Accuracy (95% CI)
  • *

    (a) Considered positive on combined tests when at least one method was positive and negative when both methods were negative. (b) Considered positive on combined tests when both methods were positive and negative when either method was negative.

US plus CA 19-9      
 (a) Either positive0.7691% (0.72–0.97)62% (0.55–0.69)23% (0.15–0.33)98% (0.94–0.99)65% (0.58–0.71)
 (b) Both positive0.7245% (0.27–0.65)98% (0.95–0.99)71% (0.45–0.88)94% (0.90–0.97)93% (0.88–0.95)
CT plus CA 19–9      
 (a) Either positive0.69100% (0.85–1)38% (0.30–0.47)22% (0.15–0.31)100% (0.92–1)47% (0.39–0.55)
 (b) Both positive0.7350% (0.29–0.71)96% (0.92–0.98)60% (0.36–0.80)95% (0.90–0.97)92% (0.87–0.95)
MRI plus CA 19-9      
 (a) Either positive0.6696% (0.79–0.99)37% (0.29–0.46)24% (0.16–0.33)98% (0.88–0.99)47% (0.39–0.55)
 (b) Both positive0.6942% (0.23–0.64)95% (0.90–0.97)50% (0.28–0.72)93% (0.88–0.96)89% (0.84–0.93)
MRCP plus CA 19–9      
 (a) Either positive0.69100% (0.85–1)38% (0.30–0.47)21% (0.14–0.31)100% (0.93–1)47% (0.39–0.55)
 (b) Both positive0.6942% (0.23–0.64)95% (0.90–0.97)50% (0.28–0.72)93% (0.88–0.96)89% (0.84–0.93)
MRI/MRCP plus CA 19–9      
 (a) Either positive0.69100% (0.85–1)38% (0.30–0.47)21% (0.14–0.30)100% (0.93–1)47% (0.39–0.54)
 (b) Both positive0.7045% (0.21–0.72)94% (0.89–0.97)33% (0.15–0.58)96% (0.92–0.98)91% (0.86–0.94)
ERCP plus CA 19–9      
 (a) Either positive0.71100% (0.86–1)43% (0.36–0.50)18% (0.12–0.25)100% (0.95–1)49% (0.42–0.56)
 (b) Both positive0.7970% (0.49–0.84)89% (0.84–0.92)41% (0.27–0.57)96% (0.92–0.98)87% (0.82–0.91)

Subsequently, ERCP were performed on 88, 96, and 93 patients who had either serum CA 19-9 levels ≥20 U/mL or overall findings indicating tumor with US, CT, or MRI, respectively. Also, subsequent MRCP were performed on 46, 51, and 51 patients who had a positive result on combined examinations of serum CA 19-9 with US, CT, or MRI, respectively. In this setting, the operative performances of cholangiograms were described when definite, probable or possible findings were considered as positive for malignancy (Table 8). The sensitivity of ERCP was higher than that of MRCP but the specificity of both cholangiograms was relatively comparable regardless of type of a prior cross-sectional imaging. In these selected patients, the PPV of subsequent ERCP was approximately two-fold that of ERCP performed in the entire population.

Table 8. Diagnostic Performance of Subsequent Cholangiogram in Predicting for Cholangiocarcinoma in PSC Patients Who Had Either Serum CA 19-9 ≥20 U/mL or Suspicious Radiologic Findings on a Cross-Sectional Liver Imaging
Sequential TestingAUCSensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)Accuracy (95% CI)
  1. The result of combined measurement of serum tumor marker with a cross-sectional liver imaging and subsequent cholangiogram was considered positive when there was suspicious overall finding on cholangiogram in patients with either serum CA 19-9 ≥20 U/mL or a cross-sectional liver imaging showing suspicious overall finding.

US plus CA 19-9      
 ERCP0.7790% (0.70–0.97)63% (0.51–0.74)42% (0.28–0.57)96% (0.85–0.99)69% (0.59–0.78)
 MRCP0.7071% (0.36–0.92)69% (0.54–0.81)29% (0.13–0.53)93% (0.78–0.98)70% (0.55–0.81)
CT plus CA 19–9      
 ERCP0.7490% (0.71–0.97)57% (0.46–0.68)37% (0.25–0.51)96% (0.85–0.99)65% (0.55–0.73)
 MRCP0.6971% (0.36–0.92)66% (0.51–0.78)25% (0.11–0.47)94% (0.79–0.98)67% (0.53–0.78)
MRI plus CA 19–9      
 ERCP0.7691% (0.72–0.97)61% (0.49–0.71)42% (0.29–0.56)96% (0.85–0.99)68% (0.58–0.76)
 MRCP0.7075% (0.41–0.93)65% (0.50–0.78)29% (0.14–0.50)93% (0.79–0.98)67% (0.53–0.78)

In the clinical context of patients who have either serum CA 19-9 levels ≥20 U/mL or suspicious findings on cross-sectional images of the liver, a total of 51 biliary brushing samples of the bile duct lesions in 51 patients were obtained for conventional cytology, DIA, and FISH testing during the study period. Not all specimens obtained were adequate for DIA (n = 19) and FISH (n = 19) testing. In these patients, the sensitivity of cytology slightly increased to 50% while maintaining the high specificity when both positive and suspicious cytology results were considered as positive for malignancy (Table 9). For overall DIA, specificity slightly increased to 92%–94% while maintaining intermediate sensitivity when either aneuploid or tetraploid DIA samples were considered positive for malignancy. For the FISH results, the specificity increased slightly to 80%–83% while maintaining high sensitivity when either FISH polysomy or trisomy was categorized as positive for malignancy. Interestingly, the PPV in these highly selected patients increased to 83%–100% for cytology, 86%–89% for DIA, and 77%–83% for FISH. When FISH and DIA were combined, the sensitivity moderately decreased to 50%–64%, but the specificity and PPV were 100% when both tests showed aneusomy on FISH and aneuploidy on DIA.

Table 9. Diagnostic Performance of Cytology, DIA, and FISH for Diagnosis of Cholangiocarcinoma from Biliary Tract Lesions during ERCP in Patients Who Had Either Serum CA 19-9 ≥20 U/mL or Overall Findings on Cross-Sectional Liver Imaging
 AUCSensitivity (95% CI)Specificity (95% CI)PPV (95% CI)NPV (95% CI)Accuracy (95% CI)
  1. Diagnostic performance was calculated based on patients with histologically proven cholangiocarcinoma for cytology and on all patients with histologically or cytologically proven cholangiocarcinoma for DIA and FISH.

Cholangiographic lesions in patients with either serum CA 19-9 ≥20 U/mL or overall US finding
 Cytology0.7950% (0.25–0.75)100% (0.85–1)100% (0.61–1)79% (0.60–0.90)82% (0.67–0.92)
 DIA0.7864% (0.35–0.85)92% (0.65–0.99)88% (0.53–0.98)73% (0.48–0.89)78% (0.58–0.90)
 FISH0.8791% (0.62–0.98)83% (0.55–0.95)83% (0.55–0.95)91% (0.62–0.98)87% (0.68–0.95)
 FISH or DIA0.8391% (0.62–0.98)75% (0.47–0.91)77% (0.50–0.92)90% (0.60–0.98)83% (0.63–0.93)
 FISH and DIA0.8264% (0.35–0.85)100% (0.76–1)100% (0.65–1)75% (0.51–0.90)83% (0.63–0.93)
Cholangiographic lesions in patients with either serum CA 19–9 ≥20 U/mL or overall CT finding
 Cytology0.7645% (0.21–0.72)97% (0.83–0.99)83% (0.44–0.97)82% (0.66–0.92)83% (0.68–0.91)
 DIA0.7250% (0.25–0.75)94% (0.72–0.99)86% (0.49–0.97)71% (0.50–0.86)75% (0.57–0.87)
 FISH0.8283% (0.55–0.95)81% (0.57–0.93)77% (0.50–0.92)87% (0.62–0.96)82% (0.64–0.92)
 FISH or DIA0.7983% (0.55–0.95)75% (0.51–0.90)71% (0.45–0.88)86% (0.60–0.96)78% (0.60–0.90)
 FISH and DIA0.7550% (0.25–0.75)100% (0.81–1)100% (0.61–1)73% (0.52–0.87)79% (0.60–0.90)
Cholangiographic lesions in patients with either serum CA 19-9 ≥20 U/mL or overall MRI finding
 Cytology0.8050% (0.25–0.75)96% (0.80–0.99)86% (0.49–0.97)80% (0.63–0.90)81% (0.66–0.91)
 DIA0.7762% (0.36–0.82)93% (0.70–0.99)89% (0.57–0.98)74% (0.51–0.88)79% (0.60–0.90)
 FISH0.8285% (0.58–0.96)80% (0.55–0.93)79% (0.52–0.92)86% (0.60–0.96)82% (0.64–0.92)
 FISH or DIA0.7985% (0.58–0.96)73% (0.48–0.89)73% (0.48–0.89)85% (0.58–0.96)79% (0.60–0.90)
 FISH and DIA0.8162% (0.36–0.82)100% (0.80–1)100% (0.68–1)75% (0.53–0.89)82% (0.64–0.92)

Discussion

Clinically, cholangiocarcinoma should be suspected when patients with PSC show rapid, progressive liver disease with worsening jaundice, weight loss, and abdominal pain. However, our data showed that the clinical presentation of patients with early cancer did not differ from that of patients with PSC alone, whereas patients with advanced cancer often present with cholestatic features. This suggests that a screening and surveillance strategy to identify cancer early should not depend on clinical manifestation or liver biochemistries. On this basis and with no relevant risk factors for cancer identified, we therefore recommend our patients with PSC be screened and regularly followed-up for developing cholangiocarcinoma. With this recommendation, cholangiocarcinomas in our patients with PSC were detected during follow-up, with an annual incidence of 1.2%, which falls within the range (0.5%–2.5%) reported in published natural history studies.2, 4, 14 Screening and surveillance for cholangiocarcinoma in PSC has the greatest impact on patient survival, because two-thirds of cancers were detected at an early stage of disease and the majority of them have been treated with a potentially curative option: liver transplantation after neoadjuvant chemoradiotherapy.5 Hence, there is an urgent need for optimal tools of diagnosis and monitoring to detect tumor early in this patient population.

Previously, serum CA 19-9 testing was introduced as a surrogate marker for the screening of pancreatic and colorectal carcinoma in asymptomatic people. There is no available information of serum CA 19-9 level for monitoring biliary tract cancer in patients with PSC. The normal range of serum CA 19-9 level in patients with PSC should be defined as the value below the ninety-fifth percentile of the crude distribution of CA 19-9 in large numbers of populations at low risk for subclinical biliary tract cancer. Unfortunately, we could not estimate the normal range from our population due to non-normal distribution of serum CA 19-9 level. However, our data suggest that a cutoff value of 20 U/L is superior to the manufacturer's claimed value of 40 U/L for identification of patients at risk for cholangiocarcinoma. This is in part due to undiagnosed early cancers using the manufacturer's reference range. As shown in Table 3, if we applied serum CA 19-9 at a cutoff value of 40 U/ml for screening/surveillance cholangiocarcinoma in PSC patients, we would identify only one-third of patients with an early stage cancer, whereas the adoption of a lower limit of 20 U/L for CA 19-9 would increase the effective case finding for early cancers confined in the biliary tree. However, the cost-effectiveness and long-term effect of a cutoff value of 20 U/L for serum CA 19-9 on the natural course of PSC need to be prospectively assessed.

Our observations, confirmed in a previous study by Campbell et al.,8 showed that by using more than one cross-sectional form of liver imaging, radiologic abnormalities corresponding to tumors could be depicted in most cases. Furthermore, the visualization of tumor lesions was a more reliable sign (100% PPV) than the secondary sign of marked bile duct wall thickening, marked biliary dilatation, or atrophy-hypertrophy complex (PPV of 22%–55%). However, the primary sign of tumor could be found in only 35% of cancers. The use of CT and MRI yielded higher sensitivity than US in detecting cholangiocarcinoma but US examination provided better accuracy than CT and MRI in distinguishing cholangiocarcinoma from underlying PSC. The better sensitivity of CT and MRI derives from their ability to depict the infiltrative lesions with contrast-enhanced techniques. Nevertheless, the specificity of findings on contrast-enhanced imaging seem to be impaired by the fibrosing inflammatory nature of PSC, which disrupts blood flow, leading to difficulty in distinguishing abnormal perfusion of hepatobiliary parenchyma in PSC from malignant lesions. Although probable and possible radiologic findings are an imperfect sign of cholangiocarcinoma in PSC, their detection should prompt careful evaluation with definitive confirmation techniques.

Approximately one-third of our patients with PSC developed prominent biliary strictures during the follow-up; however, only 19% of biliary strictures were caused by cholangiocarcinoma. This finding confirms that cholangiography alone is imprecise for distinguishing between benign processes of PSC and malignant biliary strictures.8, 15, 16 Our results showed that ERCP and MRCP plus MRI performed equally well in predicting cholangiocarcinoma complicating PSC. However, when assessing the test properties of MRCP alone in identifying primary and secondary signs of tumor, its sensitivity was lower than that of ERCP. This may be due to less spatial resolution and inadequate depiction of the contour of strictures by the MR technique. Additionally, the biliary tree has often collapsed because of a preceding biliary drainage when patients are referred for MRCP examination. Thus, if the suspicion of cancer is high, a negative MRCP does not sufficiently exclude the diagnosis.

Although it is too early to draw any definite conclusions from this extensive experience, our observations suggest that there may be a role for the test strategy of initially combining measurement of serum CA 19-9 level with cross-sectional examination of the liver for screening and surveillance of cholangiocarcinoma in patients in PSC. However, it should be considered that four of five patients with positive results on this combination did not have cancer. To deal with the high false-positive rate of this strategy, subsequent testing with cholangiography provides additional information to maximize specificity and PPV in predicting cholangiocarcinoma and seems to be more appropriate than to routinely offer ERCP to all patients. If there are suspicious cholangiographic findings, the individual estimated risks of having cancer increased from 1.0 in 4.0 to 1.0 in 2.4 based on the type of cholangiography. Further, these patients should be considered for an aggressive approach with cytologic examinations for exploring suspicious bile duct lesions. However, conventional cytology can confirm the presence of malignancy in only one-half of the cases. This finding reinforces the need for better cytologic techniques to enhance sensitivity in the diagnosis of cancer. DIA and FISH testing have been increasingly used to determine the chromosomal alterations of individual cells in malignant pancreatobiliary strictures.9 These cytologic techniques are more applicable for biliary brush samples with limited cellularity, and may detect malignancy before it is detectable by other means. The DIA and FISH techniques used as an adjunct to biliary cytology can improve sensitivity and PPV for detecting cancer while maintaining an acceptable specificity.9 Some false-negative DIA and FISH results could be due to a lack of the diagnostic cells in the specimen rather than an inability of the DIA or FISH assay to detect the cells. One argument for using this serial testing for detecting cholangiocarcinoma is that this strategy will detect 9 of every 10 cases. Furthermore, it will largely reduce the cost and the number of patients undergoing invasive procedures with a consequent reduction in unnecessary morbidity related to the procedure.

Some limitations should be considered for interpreting the results in our study. First, participants were followed for different lengths of time and had a different number of tests performed. Thus, it is difficult to rigorously compare the test properties between each diagnostic modality. Second, in this tertiary center, tests performing with higher PPV in detecting cholangiocarcinoma may not be reproducible in populations with a lower prevalence of cholangiocarcinoma. Finally, improvements in radiologist experience and technology have occurred continuously over time and might have some influence on the interpretation of images. Although these factors seem to influence the reported performance to some degree, it would be impossible to control for all of these in a retrospective analysis or even in a prospective study.

The principal findings of this study relate to the clinical utility of a variety of tests for the accurate diagnosis of cholangiocarcinoma complicating the course of patients with PSC. Tumor markers and a variety of radiologic modalities are individually imperfect tools for the detection of cholangiocarcinoma in this population. However, initially combining serum CA 19-9 with cross-sectional liver imaging has become the preferred testing strategy for triggering further investigations in diagnosing cholangiocarcinoma. On the basis of test properties, cost, and availability, serum CA 19-9 (using a cutoff value of 20 U/mL) and abdominal US at 12-month intervals appears be a useful strategy for screening/surveillance of cholangiocarcinoma in PSC, as proposed in Fig. 1. Given that this strategy is based on 23 cancers, of which nine were found on initial screening and only 14 were detected during surveillance, there is a need for prospective trials to validate our results. Following this strategy, if a mass lesion is depicted, the diagnosis of cancer should be straightforward. If imaging reveals evidence indicating biliary tract lesions or obstruction, subsequent cholangiography in these patients can increase the possibility of detecting ductal cholangiocarcinoma. Finally, a diagnosis of cancer will be confirmed with cytologic techniques of biliary sampling. This strategy has the potential to identify early cancer and should be important in decision making for management of PSC.

Figure 1.

Proposed algorithm for screening and surveillance for cholangiocarcinoma in patients with PSC. Patients should be followed-up annually and a serum CA19-9 level and an abdominal ultrasound evaluation obtained. If patient has only elevation of serum CA19-9 levels more than 20 U/mL or only presence of ultrasonographic evidence of biliary tract dilatation/bile duct wall thickening, we recommend MRCP as the initial investigation and then proceed to ERCP if additional information or intervention such as tissue sampling is needed. If patients have both elevation of serum CA19-9 levels more than 20 U/mL and ultrasonographic evidences of biliary tract dilatation/bile duct wall thickening, or have deterioration in the clinical condition, patients should be referred for ERCP with appropriate endoscopic intervention and biliary sampling of suspicious lesions. If patients have a suspicious liver mass on US, patients should have either contrast-enhanced CT or MRI of the liver to define the extent of the lesion and then the diagnosis of cholangiocarcinoma should be established with endobiliary biopsy during ERCP or US-guided needle biopsy of the lesion. Fine needle biopsy is ideally avoided until resectability has been assessed by a specialist surgeon. For patients with questionable cytopathologic findings, we recommend scheduled surveillance that combines the traditional tests with cholangiography and tissue sampling every 3-6 months.

Ancillary