Dynamic enhancement patterns of intrahepatic cholangiocarcinoma in cirrhosis on contrast-enhanced computed tomography: risk of misdiagnosis as hepatocellular carcinoma

This study aimed to assess the features of intrahepatic cholangiocarcinoma (ICC) at computerized tomography (CT) and verify the risk of misdiagnosis of ICC as hepatocellular carcinoma (HCC) in cirrhosis. CT appearances of 98 histologically confirmed ICC nodules from 84 cirrhotic patients were retrospectively reviewed, taking into consideration the pattern and dynamic contrast uptake during the arterial, portal venous and delayed phases. During the arterial phase, 53 nodules (54.1%) showed peripheral rim-like enhancement, 35 (35.7%) hyperenhancement, 9 (9.2%) hypoenhancement and 1 (1.0%) isoenhancement. The ICC nodules showed heterogeneous dynamic contrast patterns, being progressive enhancement in 35 nodules (35.7%), stable enhancement in 28 nodules (28.6%), wash-in and wash-out pattern in 15 nodules (15.3%) and all other enhancement patterns in 20 nodules (20.4%). There were no significant differences in the dynamic vascular patterns of ICC according to nodule size (p > 0.05). ICC in cirrhosis has varied enhancement patterns at contrast-enhanced multiphase multidetector CT. Though the majority of ICC did not display typical radiological hallmarks of HCC, if dynamic CT scan was used as the sole modality for the non-invasive diagnosis of nodules in cirrhosis, the risk of misdiagnosis of ICC for HCC is not negligible.

CT scan. CT-scans were performed with a multidetector-row helical quadruplephase (i.e., unenhanced, hepatic arterial, portal, and delayed phases) CT scanner (MDCT; Definition, Siemens Medical Systems, Erlangen, Germany). First, an unenhanced scan was obtained through the liver. Next, after intravenous infusion of 80-100 ml of a nonionic iodine-containing contrast agent (ultravist 370, Scherning AG, Berlin, Germany) using a power injector (Stellant CT Injection System, Medrad, Indianola, Pennsylvania ) at a rate of 4 ml/sec, contrast-enhanced scans were obtained in arterial phase with bolus test trigger for optimal characterization of focal hepatic lesions. Data acquisitions were obtained through the whole liver in a craniocaudal direction during a single breath-hold helical acquisition for 4-6 sec with 5 mm slice thickness and 0.5 s rotation time. The acquisition of the arterial phase was automatically started 4 s after the arrival of contrast agent in the aorta. The start of acquisition sequences was 60 s for the portal venous phase and 180 s for the delayed phase from the beginning of contrast agent injection.
CT findings were evaluated blindly by 2 radiologists with over 20 year experience of liver radiology (PC and SYD) who were blinded to the clinical and histological results. All imaging films were independently evaluated by both radiologists whereas discrepant diagnoses were jointly re-evaluated to reach a final consensus.
Categorization of enhancement patterns at CT. After intravenous contrast administration, the enhancement through each of the different phases was registered as follows: (1) hyperdense: increased density involving predominant parts (> 50%) of the lesion cross-section area compared to the surrounding liver parenchyma 13 ; (2) peripherally hyperdense: increased density limited to the periphery of the lesion, resembling a rim-like pattern; (3) isodense: same density as the surrounding liver parenchyma; (4) hypodense: lower density compared to the liver parenchyma involving predominant parts (> 50%) of the cross-sectional area of the tumor excluding peripheral rim-like enhancement. Dynamic pattern of enhancement was defined according to the combination of contrast enhancement in the different phases of the study (arterial, portal venous, delayed-venous), as follows: (1) stable enhancement: the nodule enhancement is unmodified from the arterial to the portal venous and delayed phases; (2) progressive contrast enhancement: the nodule enhances progressively over time, reaching maximal intensity in delayed phases; (3) "wash-in and wash-out" enhancement pattern: intense hyperdense during the arterial phase followed by hypodense in the portal and/or delayed venous phases; (4) all other patterns. This classification was modified from Iavarone et al. 14 . Statistical analysis. Quantitative variables, such as tumor size, age, serum levels of AFP, and Ca19-9, were expressed as median and range. Qualitative variables like etiology of liver disease, Child-Pugh score, and number of nodules, were expressed as count and proportions. Differences in signal intensity in baseline and post-contrast sequences and in dynamic enhancement pattern according to nodule size were evaluated by the Chi-squared test/Fisher's exact test for categorical variables. A P value of less than 0.05 was considered statistically significant. Statistical analysis was performed using the SPSS 12.0 software package (SPSS Inc, Chicago, IL). CT scan features. The size of ICC detected by CT at diagnosis was 5.6 ± 2.7 cm. Twenty one nodules (20.4%) were ≤ 3.0 cm while 77 (78.6%) nodules > 3.0 cm. Hepatic capsular retraction was revealed in 11cases (13.1%). Intrahepatic biliary dilatation was observed in 17 patients (20.2%). Nine patients had malignant portal veins thrombus (10.7%) and 12 patients had regional lymphadenopathy (14.3%). Hepatic heamangioma was observed in 1 patient (1.2%) and hepatic cyst noted in 5 patients (6.0%). Intrahepatic lithiasis was seen in 1 patient (1.2%). The contrast appearance of ICC during different vascular phases is summarized in Table 2.

Characteristics of patients.
Dynamic enhancement patterns at CT. The analysis of the vascular dynamic enhancement pattern throughout the different phases of the 98 ICC nodules is shown in Table 3. Twenty eight nodules (28.6%) showed a stable contrast enhancement pattern during the dynamic study including 20 peripherally hyperdens ( Fig. 1), 6 hypodense and 2 hyperdense during all the three vascular phases. Thirty five nodules (35.7%) demonstrated a progressive enhancement pattern: peripherally hyperdense in the arterial phase followed by centripetal progressive enhancement in the portal and the late phase in 27 nodules (Fig. 2), inhomogeneous hyperdense in the arterial phase followed by progressive enhancement in the portal and the late phase and reached maximal intensity in the delayed phase in 4 nodules, hypodense during the arterial and the portal phase followed by inhomogeneous hyperdense during the late phase in 2 nodules, globally hypodense during the arterial phase followed by peripherally hyperdense during the portal phase and the late phase in 2 nodules. Fifteen nodules (15.3%) displayed a "wash-in and washout" enhancement pattern: intense hyperdense during the arterial phase followed by hypodense during the portal and the late phase in 13 nodules (Fig. 3 and the portal phase followed by hypodense during the late phase in 2 nodules. Twenty nodules (20.4%) showed other enhancement patterns: peripherally hyperdense during the arterial phase and the portal phase followed by hypodense during the late phase in 5 nodules, peripherally hyperdense during the arterial phase followed by hypodense during the portal phase and the late phase in 3 nodules, inhomogeneous hyperdense during the arterial phase followed by peripherally hyperdense during the portal phase and inhomogeneous hyperdense during the late phase in 2 nodules, slightly inhomogeneous hyperdense during the arterial phase followed by hypodense during the portal phase and the late phase in 6 nodules, hyperdense during the arterial phase and the portal phase followed by isodense during the late phase in 1nodule, isodense during the arterial phase followed by peripherally hyperdense during portal phase and hypodense during the late phase in 1 nodule, slightly inhomogeneous hyperdense during the arterial phase followed by peripherally hyperdense during the portal phase and the late phase in 1 nodule, peripherally hyperdense during the arterial phase followed by isodense during the portal phase and hypodense during the late phase in 1 nodule. Wash-in and wash-out 4 3 8 Other patterns 7 5 8  Enhancement patterns of ICC nodules during dynamic contrast CT scan according to nodule size. There were no significant differences in the dynamic vascular patterns of ICC previously defined according to nodule size (≤ 3 cm, 3.1-5.0 cm, > 5 cm), including stable enhancement (p = 0.174, p = 0.429, p = 0.414, respectively), progressive enhancement (p = 0.724, p = 0.362 p = 0.565, respectively), wash-out enhancement (p = 0.718, p = 1.000, p = 0.807, respectively) and other enhancement patterns (p = 0.240, p = 0.189, p = 1.000, respectively).

Characterization of nodules according to enhancement pattern at CT. Thirty five nodules demon-
strating a progressive enhancement pattern and 28 nodules showing a stable contrast enhancement pattern were characterized as ICC (64.3%). Fifteen nodules displaying a "wash-in and washout" enhancement pattern were characterized as HCC ((15.3%)). Twenty nodules showing other enhancement patterns were indeterminate (20.4%).

Discussion
Ninety eight ICC consecutively identified in cirrhotic patients between January 2005 and July 2015 in our hospital exhibited varied enhancement patterns at contrast-enhanced multiphase multidetector CT. Over half of the ICC nodules (54.1%) had a peripheral rim-like enhancement during the arterial phase, whereas during the portal and delayed phase, 26 nodules (49.1%) showed a centripetal progressive enhancement,20 nodules (37.7%) a stable enhancement and 7 nodules (13.2%) other enhancement pattern. The proportion of ICC nodules showing peripheral rim-like enhancement during the arterial phase in our patients was similar to previous studies by Kim et al. 11 and Lavarone et al. ( 42.8%, 50% respectively). We consider that the presence of a central necrosis and fibrosis may account for the rim-like arterial contrast uptake at the periphery of the nodules 13 14 . We favor to interpret these slight variances in the light of the difference of sample size. Twenty two ICC nodules (15.3%) displayed a "wash-in and wash-out" enhancement pattern in our study, resembling the radiological hallmarks of HCC, namely, arterial hypervascularity followed by venous or delayed phase washout 10 , and in fact, these nodules were independently characterized as HCC by radiologists with over 20 year experience of liver radiology in our hospital (PC and SYD). The rate of ICC nodules showing "wash-in and wash-out" enhancement pattern in our study is lower than that reported by Kim et al. (21.4%) 11 14 . However, this definition may not be suitable for characterizing nodules that show predominantly hyperdense (i.e., involving 60∼ 90% of the nodule cross section area) during the arterial phase followed by hypodense in the portal and/or delayed phase. The majority of HCC in cirrhosis displayed inhomogeneous enhancement during the arterial phase 11,15,16 , therefore, it is not reasonable to correspond arterial hypervascularity to only global intense contrast enhancement during the arterial phase. In our opinion, to define predominantly hyperdense of a nodule during arterial phase as arterial hypervascularity, regardless of global or predominantly partial 13,15 is more reasonable. The median tumor size was 3.0 cm in the study of lavarone et al. 14 , while 5.3 cm in our study. However, after stratification, the ICC nodules ≤ 3 cm enrolled in our study (21 nodules) is comparable to that of lavarone et al. (23 nodules) because much more ICC nodules were included in our series. More importantly, 4 ICC nodules ≤ 3 cm (19.0%) showed "wash-in and wash-out" enhancement pattern and there were no significant differences in the dynamic vascular patterns of ICC according to nodule size (≤ 3 cm, 3.1-5.0 cm, > 5 cm) in our study. Patients with cirrhosis who undergo surveillance may have an earlier stage of HCC at diagnosis 17 . Unfortunately, surveillance for HCC in patients with cirrhosis is recommended but may not be successfully performed. Less than 20% of patients with cirrhosis who developed HCC received regular surveillance in the United States 18 . HCCs were detected during surveillance in the minority of patients even at major referral center 19 . Surveillance for HCC is still not a consolidated practice as it should be 20 . Therefore, differentiation of ICC from HCC in large nodules is still a common clinical situation worldwide for most liver tumors in cirrhosis are found not at very early stage. (3) Forty ICC nodules were enrolled and 18.8% of cirrhosis related to hepatitis B virus infection in the study of lavarone et al., while 98 ICC nodules enrolled and 85.7% of cirrhosis related to hepatitis B virus infection in our study. Additionally, in the paper of lavarone et al., one ICC nodule showed hyperdense during both the arterial and portal phases, followed by hypodense in the delayed venous phase (patient No. 33, Table 3), should be judged as "wash-in and wash-out" pattern according to the definition, though wash-out occurred in the delayed phase and be subsequently characterized as HCC. Because radiological diagnosis of HCC should be based on imaging techniques (4-phase MDCT/dynamic contrast enhanced MRI) showing arterial hypervascularity and venous or delayed phase washout according to the practice guidelines of AASLD and EASL 9,10 . This discrepancy reflects the lack of internationally agreed standard for the review of radiological films aimed at optimizing the diagnosis of ICC in cirrhosis. Our results indicate if dynamic contrast CT scan was used as the sole modality for the non-invasive diagnosis of HCC in cirrhosis, about 15% of ICC would be misdiagnosed for HCC, however, this estimate needs to be clarified prospectively in the future.
There were some limitations to our study. First, it was limited by its retrospective nature and thus by our limited control regarding patient selection. Therefore, we are unable to calculate the sensitivity and specificity of dynamic contrast CT in differentiation between ICC and HCC in cirrhosis. Second, though we provided a largest series of pathologically proven ICC in cirrhosis seen on contrast-enhanced multiphasic multidetector CT up to now, the sample size of ICC nodules ≤ 3 cm is still limited and just comparable to that of the previous study 14 . This reflects the low incidence of ICC in patients with cirrhosis, accounting for about 1-2% of all new nodules in cirrhosis 21,22 and surveillance for hepatic tumor in patients with cirrhosis is not widely performed in our country. Prospective multiple center study including more small ICCs in cirrhosis is expected Third, it is impossible for us to compare the enhancement patterns of CT and MRI because only 11 out of 84 patients of ICC underwent both MRI and CT scan in our series, though the enhancement appearance of ICC at MRI was reported much different from that of HCC 23 and this reflects the fact that in many countries worldwide, with a high incidence of HCC, MRI is still not as readily available as CT. The usefulness of our present results is restricted to centers which use CT for characterization of hepatic nodules in cirrhosis.
In conclusion, ICC in cirrhosis has varied enhancement patterns at contrast-enhanced multiphase multidetector CT. Though the majority of ICC did not display typical radiological hallmarks of HCC, if dynamic CT scan was used as the sole diagnostic modality for the non-invasive diagnosis of nodules in cirrhosis, the risk of misdiagnosis of ICC for HCC is not negligible.