Abdominal Imaging

, Volume 39, Issue 6, pp 1228–1240

Mass-forming cholangiocarcinoma and adenocarcinoma of unknown primary: can they be distinguished on liver MRI?

Authors

  • Najwa Al Ansari
    • Department of RadiologyUniversity of North Carolina at Chapel Hill
  • Bong Soo Kim
    • Department of RadiologyUniversity of North Carolina at Chapel Hill
  • Saowanee Srirattanapong
    • Department of RadiologyUniversity of North Carolina at Chapel Hill
  • Charles T. A. Semelka
    • Department of Biomedical EngineeringUniversity of North Carolina at Chapel Hill
  • Miguel Ramalho
    • Department of RadiologyUniversity of North Carolina at Chapel Hill
  • Ersan Altun
    • Department of RadiologyUniversity of North Carolina at Chapel Hill
  • John T. Woosley
    • Department of Pathology and Laboratory MedicineUniversity of North Carolina at Chapel Hill
  • Benjamin Calvo
    • Department of SurgeryUniversity of North Carolina at Chapel Hill
    • Department of RadiologyUniversity of North Carolina at Chapel Hill
Article

DOI: 10.1007/s00261-014-0172-3

Cite this article as:
Al Ansari, N., Kim, B.S., Srirattanapong, S. et al. Abdom Imaging (2014) 39: 1228. doi:10.1007/s00261-014-0172-3

Abstract

Purpose

To determine MR features suggestive of mass-forming cholangiocarcinoma (CCA) or liver metastases of adenocarcinoma of unknown primary (AUP), and to compare the ability of two experienced radiologists to establish the correct diagnosis.

Materials and methods

61 patients with CCA or AUP, with MRIs were placed into two groups: population 1, 28 patients with certain diagnosis of either CCA or AUP; and population 2, 33 patients with uncertain diagnosis. Using population 1 with known diagnosis, two investigators formulated imaging criteria for CCA or AUP, which represented phase 1 of the study. In phase 2, two independent radiologists categorized the patients in populations 1 and 2 as CCA or AUP using the formulated criteria. This categorization was compared with the patient medical records and pathologist review. Findings were tested for statistical significance.

Results

In phase 1, solitary lesion, multifocal lesions with dominant lesion, capsule retraction, and porta hepatis lymphadenopathy were features of CCA; multifocal lesions with similar size, and ring enhancement were features of AUP. The number of lesions, capsule retraction, and early tumor enhancement pattern were observed to be significant features (P < 0.05). In phase 2, agreement between the two radiologists was good (k = 0.663). For population 1, the agreement was good (k = 0.659), and was fair for population 2 (k = 0.293). Concordance between the two radiologists, medical record, and the pathologist was found in 41/61 (67%) patients.

Conclusion

Distinctive features of CCA and AUP are identifiable on MRI images, which may aid the radiologist to establish the correct diagnosis.

Keywords

CholangiocarcinomaAbdominal MRIPrimary liver tumorLiver metastasesUnknown primary siteAdenocarcinoma

Primary cholangiocarcinoma of the liver accounts for approximately 15%–20% of all primary liver cancers and is the second most common primary hepatic malignant tumor worldwide [13] with a documented increase in incidence rates in developed countries [2, 3]. Intrahepatic cholangiocarcinomas can be classified into three types according to their morphology and growth pattern: the mass-forming type, the periductal infiltrative type, and the intraductal growing type [5]. Mass-forming cholangiocarcinoma (CCA) is defined as a rounded mass located in the liver parenchyma and is the most common intrahepatic cholangiocarcinoma, accounting for 60% of all these tumors [5, 6].

The majority of patients with cholangiocarcinoma develop this malignancy de novo. However, certain conditions and risk factors have been associated with an elevated risk of developing the disease. Primary sclerosing cholangitis (PSC) is the most well known of these conditions, associated with a prevalence of cholangiocarcinoma of 5%–15% [4]. Liver cirrhosis and viral hepatitis C and B are risk factors for mass-forming cholangiocarcinoma (CCA), and is stronger for hepatitis C than B [1]. These cancers are often difficult to diagnose, their pathogenesis is poorly understood, and their dismal prognosis has resulted in a conservative approach to their management.

The majority of malignancies in the liver are metastatic adenocarcinomas. In most of these cases, the patients have a known history of a primary tumor elsewhere. Unknown primary cancer represents a group of heterogeneous cancers and is defined by the presence of metastatic disease for which a primary site is undetectable on presentation. Approximately 5%–10% of all patients with a history of cancer are diagnosed with liver metastases of unknown primary adenocarcinoma (AUP) [7]. The inability to detect the primary site in AUP has been explained by two theories: metastases clinically manifest themselves first and the primary tumor remains occult due to a favorable metastatic ability over local tumor growth, or spontaneously disappears after seeding the metastasis [7]. Although, by definition, AUP does not have a recognized primary site, it has been reported that a primary site is eventually detected in 30% of cases, which may include at autopsy [8, 9].

Classically, CCA has been described as an irregular mass with an incomplete rim of enhancement during the hepatic arterial phase and the portal venous phase with late retention of contrast [6]. Unfortunately, these features are also commonly noted in liver metastases from adenocarcinoma.

Distinguishing CCA from AUP can be also challenging for surgical pathologists because no reliable histologic features are available to distinguish cholangiocarcinoma from metastatic adenocarcinoma. Additionally, the diagnosis of CCA and AUP both depends on the reasonable exclusion of an extrahepatic primary tumor [9, 10]. Progress has been attained using immunohistochemical stains to aid in such differential diagnosis [1115]. Although it is recognizably low, there is no precise data concerning the sensitivity and specificity in differentiating peripheral CCA and AUP [10, 11, 16] and the diagnosis of CCA is often made on the basis of clinico-radiologic studies and the final diagnosis is achieved by exclusion [9, 11].

The distinction between CCA and AUP is important from a management and prognosis standpoint [7, 12, 13]. In patients with CCA, surgical resection may offer long-term survival [13, 14], whereas chemotherapy offers survival benefit in AUP patients but the overall prognosis attributed to these patients is generally poor [11, 12].

To the best of our knowledge, no prior study has attempted to describe radiologic features to distinguish these malignancies, nor to compare the ability of experienced diagnosticians to discriminate these cancers. The purpose of this study was twofold: to determine MR features suggestive of CCA or AUP and to compare the ability of two experienced radiologists to establish one or the other diagnosis.

Materials and methods

Study population

Institutional review board approval was obtained, with waiver of informed consent for retrospective analysis. One investigator (NAA) with 2 years of experience in body MRI obtained a list of all patients with liver malignancies from the records of the Department of Pathology between January 2006 and December 2011. Hepatocellular carcinomas were excluded, while all other malignant liver lesions, including CCA and AUP, were incorporated. A total of 400 patients were derived from this data.

The same investigator reviewed the institutional computerized information system (CIS) on all these patients to exclude patients with a known primary extrahepatic malignancy (with known or presumed liver metastases from this primary), or histologically proven other primary hepatic malignancies. This generated a final population of 61 patients (25 men and 36 women, with a mean age of 61 years and a range of 38–97 years).

No treatment for malignancy had been administered at the time of the index MR study in any of the subjects. Seven patients possessed liver cirrhosis, 6 due to viral hepatitis [Hepatitis B (n = 2) and C (n = 4)] and 1 patient due to PSC.

Pathologic evaluation

Partial hepatectomy or wedge resection was performed in 26 patients and biopsy was performed in 35 patients in whom unresectable tumor was indicated. A liver pathologist with 20 years experience (JTW) retrospectively reviewed all the histologic specimens. The pathologist was instructed to interpret the pathology in the usual routine clinical fashion, which included review of all existing clinical information.

Criteria for categorization favoring CCA included (1) adenocarcinoma histomorphology; (2) positive CK7 staining [11, 15, 1720]; (3) the absence of immunohistochemical markers for adenocarcinoma arising from another anatomic site [11]; and (4) the absence of clinical data for a tumor from another anatomic site.

Criteria for the categorization favoring AUP included (1) adenocarcinoma histomorphology (for a cancer that could not have originated at the biopsy site); (2) negative CK7 staining; (3) the presence of immunohistochemical markers for adenocarcinoma arising from another anatomic site [11, 15]; (4) no primary tumor found after thorough medical history or physical examination (including breast and pelvic examination in women, colonoscopy and testicle and prostate examination in men); and (5) normal laboratory and imaging test results, including the results of a complete blood count, blood chemistry, or prostate-specific antigen (PSA) test, computed tomography (CT) of the chest, CT or MRI of the abdomen and pelvis, and/or PET/CT [11, 15, 20]. The pathologist categorized 49 patients as CCA and 12 patients as AUP (Fig. 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs00261-014-0172-3/MediaObjects/261_2014_172_Fig1_HTML.gif
Fig. 1

Study population flowchart.

Data collection

The initial investigator, in conjunction with another investigator, a radiologist with 15 years of experience in body MRI, (BSK), reviewed the medical record and pathologist’s retrospective study evaluation of these 61 patients to derive a population of patients, in which the diagnosis was considered “certain” (population 1) that included 28 patients, 18 patients with CCA, and 10 with AUP. Patients with CCA included 8 men and 10 women with a mean age of 66 years ranging from 45 to 85 years. Patients with AUP included 5 men and 5 women with a mean age of 65 years ranging from 49 to 97 years. Patients in whom the diagnosis was less certain (population 2) included 33 subjects, 12 men and 21 women with a mean age of 57 years ranging from 38 to 79 years). Patients with uncertain diagnosis were interpreted in agreement by the pathologist retrospective review and medical records as CCA (23 cases) and AUP (1 case). Disagreement occurred in 9 patients (8 patients were interpreted as AUP by the medical records and as CCA by the pathologist retrospective review; and 1 patient was interpreted as CCA by the medical records and as AUP by the pathologist retrospective review).

“Certainty” for the diagnosis of either CCA or AUP was considered when the pathologist’s review contained all the above criteria and matched the patient medical records, which included radiologic reports and the original pathology interpretation. Uncertainty arose when there was discordant diagnosis or despite eventual concordance there was a low level of confidence to achieve the final diagnosis because the retrospective pathologist’s review did not contain all the above criteria and/or the information present in the medical records was unclear (Fig. 1).

MRI technique

MR imaging of the abdomen was performed on 1.5 Tesla (T) (Vision, Sonata, or Avanto, Siemens Medical Systems, Malvern, PA) or 3.0 T (Trio, Siemens Medical Systems, Malvern, PA) MR systems using a phased-array torso coil in all patients. All MR imaging examinations were performed with the following protocol: for pre-contrast images transverse and coronal T1-weighted in-phase spoiled gradient echo (SGE); transverse T1-weighted out-of-phase SGE; T2-weighted sequences that included a fat-suppressed turbo spin-echo sequence and/or a fat-suppressed half-Fourier rapid acquisition with relaxation enhancement (RARE) sequence in the transverse and coronal planes. Pre- and serial post-contrast fat-suppressed 3D gradient-echo images (3D-GRE) were performed. Intravenous gadodiamide (Omniscan; GE Healthcare, Oakville, Ontario, Canada) or gadobenate dimeglumine (MultiHance, Bracco Diagnostics, Princeton, NJ) was administered by means of a power-injected (Medrad, Pittsburgh, PA) bolus of 0.1 mmol/kg and 0.05 mmol/kg, respectively, at 2 ml/s in all patients followed by a bolus 20 ml of saline flush. Gadobenate dimeglumine has replaced gadodiamide in all gadolinium-enhanced MR studies of adult patients beginning in June 2007. Contrast-enhanced images were acquired with 3D-GRE sequences in the hepatic arterial (approximately 18 s); portal venous (1.5 min); and interstitial (2.5 min) phases in the axial plane. At 3 min, post-contrast 3D-GRE was acquired in the coronal plane. The details of the sequence parameters used at 1.5 and 3.0 T MRI scanners are displayed in Table 1.
Table 1

Parameters of sequences used at 1.5 and 3.0 T MRI scanners

Parameter

Pre-contrast sequences

Post-contrast sequences

T2-weighted half-Fourier RARE

T1-weighted 2D-SGE in-phase

T1-weighted 2D-SGE Out-of-phase

T1-weighted 2D-SGEa

Dynamic T1-weighted 3D-GREb

3.0 T

1.5 T

3.0 T

1.5 T

3.0 T

1.5 T

1.5 T

3.0 T

1.5 T

TR (ms)

2000

1500

169

142

169

142

142

3.07

4.3

TE (ms)

95

90

2.5

4.4

1.58

2.2

4.4

1.32

1.6

Flip angle (°)

150

180

57

70

57

70

70

13

10

Echo train length

179

156

BW/pixel (Hz)

781

651

400

490

400

490

490

500

350

Matrix (phase × frequency)

204 × 256

192 × 256

204 × 256

192 × 256

204 × 256

192 × 256

204 × 256

192 × 256

160 × 320

FOV (mm)

350 × 350

400 × 400

400 × 400

350 × 350

400 × 400

350 × 350

350 × 350

400 × 400

360 × 360

Rectangular FOV

87.50%

81.30%

100%

75%

100%

75%

75%

93.80%

80%

No. of section

30

20

28

19

23

25

19

72

72

Section thickness (mm)

8

8

8

8

8

8

8

3

3.5

Intersectional gap (mm)

1.6

1.6

1.6

1.6

1.6

1.6

1.6

0.6

0.7

No. of signal acquisition

1

1

1

1

1

1

1

1

1

Fat suppression

Fat sat

Fat sat

None

None

None

None

None

Fat sat

Fat sat

Respiratory control

BI

BI

BH

BH

BH

BH

BH

BH

BH

a, T1-weighted 2D-SGE was used to acquire images in the hepatic arterial dominant phase at older 1.5 T scanners. In the portal venous phase and interstitial phase, a dynamic 3D-GRE sequence was used; b, Dynamic 3D-GRE sequences were used to acquire all three post-contrast phases at 3.0 T and at most recent 1.5 T scanners; TR, repetition time; TE, echo time; Hz, Hertz; FOV, field of view; RARE, rapid acquisition with relaxation enhancement; SGE, spoiled gradient echo; 3D-GRE, three-dimensional gradient echo; BI, Breathing independent; BH, Breath hold

Image analysis

In phase 1 of the study, the previously mentioned two investigators (NAA, BSK) evaluated in consensus the population 1 patients to determine imaging characteristics for CCA and AUP. Features examined for included number, size, location, signal intensity, hypovascular or hypervascular pattern of enhancement on arterial-phase images, progressive enhancement on venous-phase images, washout on delayed images, late capsular enhancement, peritumoral enhancement, the presence of porta hepatis lymphadenopathy, capsular retraction overlying the tumor (CROT), generalized involved segment parenchymal volume loss (GISPVL), segmental biliary dilatation directly related to the mass, and vascular encasement. The size of each tumor mass was measured at its greatest diameter in the axial plane using non-contrast T1-weighted images.

The categorization of tumor signal intensity on T1- and T2-weighted images, as well as the analysis of the enhancement pattern on hepatic dominant images, were performed using criteria previously published [7]. Lesion enhancement was analyzed on arterial, venous, and interstitial phase images in patients who underwent standard MR protocol. Patterns of lesion enhancement were classified as either lesional enhancement (ring enhancement, diffuse homogeneous enhancement, diffuse heterogeneous enhancement, and mixed type of enhancement including ring and heterogeneous), perilesional enhancement, or negligible enhancement.

Enhancement was considered lesional when the lesion size on pre-contrast images and the images obtained after administration of contrast material was identical. Perilesional enhancement was determined by noting enhancement beyond the visualized margins of the tumor with extension of contrast enhancement into the surrounding hepatic parenchyma. Negligible enhancement was considered when no appreciable enhancement was demonstrable.

The lesions were considered hypervascular when the lesions exhibited intense early enhancement comparable to pancreas or renal cortex, hypovascular when the lesions showed negligible or minimal enhancement on hepatic arterial dominant and portal venous-phase images. Nearly isovascular metastases were considered when the lesions exhibited comparable signal intensity to liver on arterial-phase images [7, 20].

The ability to retain contrast was evaluated comparing pre-contrast and arterial-phase images with the liver parenchyma on post-contrast images obtained at portal and interstitial phases. Washout appearance was considered when a hypervascular lesion showed decrease of signal intensity from the arterial phase to later phases that rendered lesion hypointense relative to the liver; fading was considered when a hypervascular lesion showed decrease of signal intensity in later phases that rendered the lesion isointense to liver; and retention was defined when there was accumulation of contrast within the tumor over time, irrespective of the early vascular enhancement, rendering the lesion hyperintense relative to the liver. Lymphadenopathy in porta hepatis was considered when the short axis of the lymph nodes was greater than 1 cm.

Criteria were divided into “strict criteria” for the diagnosis, which was derived from the literature; and “expanded criteria,” criteria that the two investigators (NAA, BSK) established from review of population 1 images. Strict criteria for CCA included CROT, solitary tumor, and diffuse heterogeneous enhancement with no washout and no late capsule enhancement [6, 21, 22]; for AUP, multifocal lesions with no dominant lesion pattern and ring enhancement [7]. Both groups did not have a history of a primary malignancy elsewhere, which was a circumstance considered critical for the diagnosis of both types of lesion.

Expanded criteria were criteria included for CCA: the presence of porta hepatis lymphadenopathy, vascular encasement, GISPVL, multifocal lesions with dominant lesion pattern, and lesion size greater than 5 cm. Expanded criteria for AUP included the absence of porta hepatis lymphadenopathy, the absence of vascular encasement, and the absence of CROT or GISPVL. The feature of multifocal lesions with dominant lesion pattern was defined as the circumstance, where one mass had a diameter 4 times greater than the next largest sized lesions.

In phase 2 of the study, two independent radiologists with 20 and 7 years experience (RCS, MR) of interpreting body MRI were provided with information from criteria determined from phase 1 of the study. These investigators were asked to analyze and categorize lesions as CCA or AUP on MR studies of patients using this information, both the strict and expanded criteria, examining these studies on a PACS workstation. This analysis was performed in a blinded fashion with no provision of any information other than that patients were part of this study. This analysis was done for patient populations 1 and 2 separately. The diagnosis identified by the radiologist was compared with the medical records and the pathologist’s diagnosis of populations 1 and 2. The level of agreement between the two radiologists’ blinded review was evaluated.

Statistical analysis

Significance for the presence of individual features for strict and expanded criteria in phase 1 of the study was tested using Fisher’s exact for categorical data, and independent t tests for numerical data. Comparison of the level of agreement for the diagnosis by the radiologist review, pathologist’s retrospective review, and medical records was made using McNemar’s test. In all cases, values of P < 0.05 were considered to be a statistically significant difference.

The level of agreement between the two radiologists for the presence of individual features for strict and expanded criteria and for the categorization of lesions (phase two) was tested with kappa statistics. All the statistical analyses were performed with MedCalc for Windows, version 11.3.0.0 (MedCalc Software, Mariakerke, Belgium).

Results

Phase 1

6/18 (33%) patients with CCA had all strict criteria: CROT, the most distinctive feature of CCA, was seen in 10/18 (56%) patients; solitary lesion in 9/18 (50%); and diffuse heterogeneous enhancement on hepatic arterial dominant phase images was seen in all 18/18 (100 %) patients (Table 2; Fig. 2). 5/18 (28%) patients of population 1 presented with all strict and expanded criteria for CCA: multifocal disease dominant lesion pattern was seen in 9/18 (50%) and porta hepatis lymph nodes was seen in 12/18 (67%) of patients. Regarding enhancement, hypovascular enhancement was observed in 13/18 (72%) of patients (Fig. 3), and hypervascular enhancement in 5/18 (28%) (Fig. 4). Moderately high signal intensity on T2-weighted images was seen in 9/18 (50%), GISPVL in 8/18 (44%), peritumoral enhancement in 14/18 (78%) (Fig. 5), bile duct dilatation in 7/18 patients (39%), and vascular encasement was seen in 11/18 (61%). Only 3/18 patients had lesions <5 cm (Table 3 and 4).
Table 2

The presence of strict and expanded criteria in population 1

CCA

AUP

Strict criteria

 Capsule retraction

10/18 (56%)

Ring enhancement

9/10 (90%)

 Diffuse heterogeneous enhancement

18/18 (100%)

Multifocal lesions with similar size

7/10 (70%)

 Solitary lesions

9/18 (50%)

  

Expanded criteria

 Multifocal lesions with dominant lesion pattern

9/18 (50%)

Lack of porta hepatis lymphadenopathy

10/10 (100%)

 Porta hepatis lymph nodes

12/18 (67%)

Lack of CROT

10/10 (100%)

 GISPVL

8/18 (44%)

Lack of vascular encasement

10/10 (100%)

 Vascular encasement

11/18 (61%)

  

Data are numbers of patients, with percentages in parentheses

CCA, mass-forming cholangiocarcinoma; AUP, liver metastases of adenocarcinoma of unknown primary

https://static-content.springer.com/image/art%3A10.1007%2Fs00261-014-0172-3/MediaObjects/261_2014_172_Fig2_HTML.jpg
Fig. 2

Cholangiocarcinoma (CCA) in a 74-year-old man. A Axial T2-weighted image shows heterogeneous moderately high signal intensity of 7.9 cm mass. Capsule retraction (arrow) is noted. T1-weighted images obtained in the hepatic arterial dominant phase (B) and at 2.5 min (C) demonstrate early mild to moderate diffuse heterogeneous enhancement (B) with progressive enhancement over time (C). Pathologic specimen of the tumor with H&E (D). Note that the presence of enlarged porta hepatis lymph nodes (arrow, A and B) is also nicely depicted on the later post-contrast image (C). The features of CROT, early diffuse heterogeneous enhancement, progressive enhancement, and porta hepatis lymph nodes are features of CCA. Both the blinded radiologists correctly interpreted the lesion.

https://static-content.springer.com/image/art%3A10.1007%2Fs00261-014-0172-3/MediaObjects/261_2014_172_Fig3_HTML.jpg
Fig. 3

Hypovascular cholangiocarcinoma (CCA) in a 68-year-old man. Capsule retraction is noted. T2-weighted image A shows a heterogeneous moderately high signal intensity mass in left hepatic lobe. T1-weighted images obtained in the hepatic arterial dominant phase (C) and at 2.5 min (D) demonstrate early mild to moderate diffuse heterogeneous enhancement (B) with progressive enhancement over time (C). Peritumoral enhancement is seen to resemble ring enhancement. Care must be taken in interpreting studies to distinguish peritumoral enhancement from simple ring enhancement. This distinction is facilitated by recognizing that perilesional enhancement occurs beyond the confines of the tumor as established on non-contrast (often pre-contrast T1-weighted (B)) images. Both blinded radiologists correctly interpreted the lesion.

https://static-content.springer.com/image/art%3A10.1007%2Fs00261-014-0172-3/MediaObjects/261_2014_172_Fig4_HTML.jpg
Fig. 4

Hypervascular cholangiocarcinoma in a 77-year-old man. T1-weighted images obtained in the hepatic arterial dominant phase (A) and delayed phase images (B, C) demonstrate an 8.2 cm hypervascular tumor with progressive heterogeneous enhancement over time and a rim of peripheral enhancement. Both the blinded radiologists correctly interpreted the lesion.

https://static-content.springer.com/image/art%3A10.1007%2Fs00261-014-0172-3/MediaObjects/261_2014_172_Fig5_HTML.jpg
Fig.5

Hypervascular ring-enhancing lesion in a 54-year-old man. A Axial T2-weighted image shows a heterogeneous moderately high signal intensity mass. T1 weighted images obtained in the hepatic arterial dominant phase (B) shows a hypervascular thick enhancing ring with subtle heterogeneous central enhancement. Delayed phase images (C) shows fading of the outer portion to near isointensity and subtle retention of contrast. This patient received a final diagnosis of AUP. There was no agreement between the two radiologists.

Table 3

MR features of intrahepatic mass-forming cholangiocarcinomas and liver metastases of unknown primary in population 1

MR features

Total (n = 28)

CCA (n = 18)

AUP (n = 10)

P-Value

Number of lesions

107

39

68

.002

Diameter (cm)*

7.8 ± 3.5

8.9 ± 3.8

5.8 ± 4.4

.598

 ≤5

7 (25)

3 (17)

4 (40)

.001

 >5

21 (75)

15 (83)

6 (60)

 

Feature

.679

 Multiple with dominant lesion

10 (36)

9 (50)

1 (10)

 

 Multiple with similar size

7 (25)

0

7 (70)

 

 Solitary lesion

11 (39)

9 (50)

2 (20)

 

Location

.435

 Periphery

8 (28)

4 (22)

4 (40)

 

 Central

12 (43)

11 (61)

1 (10)

 

 Replacing the entire liver parenchyma

5 (18)

0

5 (50)

 

 Replacing one entire lobe

3 (11)

3 (17)

0

 

T1-weighted signal intensity

 Low

 

18 (100)

10 (100)

 

T2-weighted signal intensity

.240

 High markedly

2 (7)

0

2 (20)

 

 High moderately

16 (57)

9 (50)

7 (70)

 

 High mildly

10 (36)

9 (50)

1 (10)

 

Vascular encasement

.673

 

14 (50)

11 (61)

0

 

Capsular retraction

.004

 

10 (36)

10 (56)

0

 

Bile duct dilatation

.443

 

9 (32)

7 (39)

2(20)

 

Volume loss

.752

 

8 (29)

8 (44)

0

 

Porta Hepatis LN

.343

 

12 (43)

12 (67)

0

 

Data are numbers of patients, with percentages in parentheses

CCA, mass-forming cholangiocarcinoma; AUP, liver metastases of adenocarcinoma of unknown primary

† Data are statistically significant

Table 4

Enhancement pattern of intrahepatic mass-forming cholangiocarcinomas and liver metastases of unknown primary in population 1

 

Total (n = 28)

CCA (n = 18)

AUP (n = 10)

P value

Vascularity

1.00

 Hypovascular

18 (64)

13 (72)

5 (50)

 

 Hypervascular

10 (36)

5 (28)

5 (50)

 

Enhancement pattern

.027

 Diffuse heterogeneous

18 (64)

18 (100)

0

.001

 Ring enhancement

8 (29)

0

8 (80)

 

 Mixed

2 (7)

0

2 (20)

 

Retention of contrast

.752

 

22 (79)

18 (100)

4 (40)

 

Peritumoral enhancement

.863

 

17 (61)

14 (78)

3 (30)

 

Data are numbers of patients, with percentages in parentheses

CCA, mass-forming cholangiocarcinoma; AUP, liver metastases of adenocarcinoma of unknown primary

† Data are statistically significant

For AUP, 6 of 10 patients (60%) had all strict criteria: ring enhancement was observed in 9 of 10 patients and multifocal lesions with dominant lesion in 7 of 10 patients (Table 2; Fig. 6). No patients with AUP possessed enlarged porta hepatis lymphadenopathy, CROT, GISPVL, or vascular encasement; the absence of these findings represents the expanded criteria. The presence of all expanded criteria for AUP in population 1 was observed in 10/10 (100%) patients (Table 2). Other features for AUP that were observed included hypervascular lesions in 5/10 (50%) of patients, progressive enhancement on delayed phase in 4/10 (40%), and moderate hyperintensity on T2 of lesions in 7/10 (70%). Multifocal lesions with dominant lesion pattern were present in only 1 patient (10%) and solitary pattern in 2 patients (20%). Peritumoral enhancement was seen in 3 patients (30%) (Tables 3 and 4), and bile duct dilatation was seen in 2/10 patients (10%).
https://static-content.springer.com/image/art%3A10.1007%2Fs00261-014-0172-3/MediaObjects/261_2014_172_Fig6_HTML.jpg
Fig. 6

Adenocarcinoma of unknown primary (AUP) metastases in a 56-year-old woman. A Axial T2-weighted image shows multiple high signal intensity lesions. Axial gadolinium-enhanced T1-weighted images (at a slightly inferior location) in the hepatic arterial dominant phase B show hypervascular lesions with ring enhancement. Axial T1-weighted at 2.5 min C shows fading of the enhancing ring. No capsule retraction was evident and no enlarged porta hepatic lymph nodes were observed. There was an agreement between the two radiologists. Sections of tumor D show a pleomorphic cell population consistent with metastatic, poorly differentiated carcinoma. Scattered benign hepatocytes are entrapped within this focus of tumor ×200, H&E.

Significant differences between CCA and AUP were observed in the number of lesions, significantly greater for AUP (P = 0.002) (Table 3); lesion size, >5 cm for AUP; CROT in CCA; and tumor enhancement pattern (P = 0.026) and in particular diffuse heterogeneous enhancement in CCA (P < 0.001) (Table 4).

Phase 2

For population 1, the extent of diagnostic agreement between radiologist 1, pathologist review, and the medical records was 100%: lesions were correctly classified as CCA and AUP in 28/28 cases. For population 2, there was no significant difference between the radiologist 1 review and the medical records (P = 0.38): the radiologist correctly classified lesions as CCA and AUP, in accordance with medical records, in 21/33 (64%) cases (15 CCA and 6 AUP); there was no significant difference between the radiologist 1’s review and the pathologist’s retrospective review (P = 0.35): the radiologist correctly classified lesions as CCA and AUP, in accordance with pathologist’s retrospective review, in 20/33 (61%) cases (18 CCA and 2 AUP);

For population 1, radiologist 2 correctly classified lesions as CCA and AUP in 25/28 cases (89%) (P = 0.12), in comparison to pathologist review and the patient medical records (15 CCA and 10 AUP). For population 2, there was no significant difference between radiologist 2’s review and the medical records (P = 0.28): lesions were correctly classified as CCA and AUP, in accordance with medical records, in 25/33 cases (75.8%) (19 CCA and 6 AUP). There was no significant difference between radiologist 2’s review and pathologist’s retrospective review (P = 0.2): lesions were correctly classified as CCA and AUP, in accordance with pathologist’s retrospective review, in 22/33 cases (66%) (21 CCA and 1 AUP) (Table 5).
Table 5

Results of Phase 2 (Population 1 and Population 2)

Concordance

Population 1

Population 2

Pathology/medical records

Pathology

Medical records

CCA

AUP

CCA

AUP

CCA

AUP

Radiologist 1

18/18 (100)

10/10 (100)

18/31 (58)

2/2 (100)

15/27 (56)

6/9 (67)

Radiologist 2

15/18 (83)

10/10 (100)

21/31 (68)

1/2 (50)

19/27 (70)

6/9 (67)

Data are numbers of patients, with percentages in parentheses

CCA, mass-forming cholangiocarcinoma; AUP, liver metastases of adenocarcinoma of unknown primary

The overall agreement between the two radiologists’ review was good (k = 0.663); for population 1, the level of agreement was good (k = 0.659) and for population 2, it was fair (k = 0.293). In contrast, concordance between two radiologists, medical record, and the pathologist was found in 41/61 (67%) patients.

Discussion

In our study, we have attempted to evaluate the ability to distinguish CCA from AUP. Although the MR imaging features of intrahepatic CCA have been reported, CCA is sometimes a diagnosis of exclusion. It may present late as a single large peripheral mass, but multiple tumor nodules are not uncommon [21]. The histomorphologic spectrum is broad, preventing an unequivocal distinction between CCA and metastatic tumors arising from another primary site [20, 24]. Consequently, reasonable exclusion of an extrahepatic primary tumor is necessary [5]. AUP, especially from foregut-derived tissues (lung, esophagus, stomach, and pancreas), is histologically indistinguishable from CCA, and must be differentiated on clinical grounds [20, 2325]. AUP is an imperfect diagnosis. The accuracy of distinguishing metastatic adenocarcinomas can be improved by immunoprofiling with a combination of cytokeratins, primarily with CK 7 and CK 20 immunohistochemical staining [1015, 17, 20]. The expressions of CK 7 and CK 20 of metastatic gastric and colorectal adenocarcinoma, which represent a large portion for the source of liver metastases, are different from that of CCA [11, 15]; however, some metastatic liver tumors, especially from pancreatic and gallbladder cancers, are still challenging to diagnose, as the CK expression is similar to CCA [15].

In earlier years, primary tumors were more difficult to detect because imaging techniques had lesser spatial and contrast resolution, the diagnosis of AUP may have included CCAs, and also because histopathologic differences were not well established [20, 2629]. Improvement in the detection of the primary site and metastases in patients with adenocarcinoma liver lesion has occurred in recent years [20, 2527]. The incidence of AUP has decreased, perhaps, as a reflection that primary site detection has improved [11, 20, 24]. Although challenging, the distinction between CCA and AUP is important as they may require different therapies [7, 11, 2832]. Complete surgical resection remains the only potentially curative option for patients with CCA and routine lymphadenectomy at the time of surgical resection is recommended [33]. On the other hand, chemotherapy has been the basis of AUP treatment.

Criteria for the imaging diagnosis of CCA that are well established in the literature, and that we have termed strict criteria, include solitary lesion, CROT, and diffuse heterogeneous arterial-phase imaging with progressive lesion enhancement [5, 6, 21, 22]. Porta hepatis lymphadenopathy is a feature that we have described in our study and this expanded criteria feature was commonly observed (67% in CCA vs. 0% in AUP). Despite our findings, a previous study reported that re-metastasis from liver metastases via lymphatic flow can occur in patients with liver metastasis from colorectal cancer [34]. The absence of porta hepatis lymphadenopathy in AUP may be related to the small number of patients, the different biologic behavior of AUP, or that our study only included patients with untreated liver lesions. Multifocal lesion with dominant lesion type is a novel description in our study and was observed in 50% of CCA patients; this finding would need to be observed in other studies to validate its importance. The presence of CROT has been reported in previous reports [21, 29] as a distinctive feature of CCA. Kim et al [6] reported that approximately 69% of their cases showed retraction of liver surface contour. This finding is in close agreement with our study, in which 50% of CCA patients in phase 1 showed CROT. It should be noted that rarely contour retraction may be observed in metastases with desmoplastic matrix [35].

Early heterogeneous enhancement is a feature of primary malignant tumors arising from the liver and may reflect that the feeding vessels are in continuity with background liver [20]. In contrast, metastases migrate to the liver and parasitize the surrounding blood vessels, creating the ring appearance of blood supplying the most vascularized outer portion of the tumors [30]. In our study, all CCA in population 1 had diffuse heterogeneous enhancement, and 28% of these were hypervascular lesions. Nanashima et al [31] and Kim et al [6] reported that 46% and 29%, respectively, of the intrahepatic cholangiocarcinomas in their studies showed hypervascular enhancement pattern. This appearance is well recognized and might carry a better prognosis with longer disease-free survival [6, 31]. Although the ring enhancement pattern is most common in AUP [7], Danet et al [30] reported occasional (17%) heterogeneous enhancement in metastatic lesions with diameters larger than 3.0 cm. In our study, ring enhancement was observed consistently in AUP and was significantly different between AUP and CCA.

Peritumoral enhancement has been reported in less than 50% of patients with liver metastases [30] and in 25%–45% patients with CCA [6]. Based on our results, as could be expected from the literature, peritumoral enhancement did not provide useful distinction.

Retention of contrast is correlated with the presence of fibrosis, which is observed in CCA [21, 23]. This retention is not usually observed in hypervascular metastatic lesions, although this could be observed in hypovascular liver metastases [7]. In population 1, 90% of CCA and 40% of AUP showed retention of contrast. Hepatobiliary contrast agents do not seem to provide additional information in the distinction between CCA and AUP. A recent study has shown that mass-forming intrahepatic cholangiocarcinomas present as a hypointense lesion on gadoxetate disodium enhanced in late venous-phase MR images [36]. Pseudo-washout pattern on hepatobiliary phase may appear because of progressive background liver enhancement [36]. Vascular compression is common with CCA, and is not commonly observed in AUP. In our study, 56% of patients with CCA had this feature, and we did not observe vascular compression in any case of AUP in population 1. Minimal T2 hyperintensity and biliary duct dilatation were more common in CCA vs. AUP, but this did not result in useful distinction.

Despite the use of modern MRI and expanded criteria for diagnosis, we found that an inability to characterize lesions still occurred: the overall agreement between the 2 radiologists’ review was good, in particular for population 1, although concordance for all patients between the two radiologists, medical record, and the pathologist was found in only 41/61 (67%) patients overall.

In current clinical practice at our institution, tumors that remain uncertain as to whether they represent CCA or AUP are generally managed as if they are CCA. This categorization despite uncertainty may also be present in the literature in studies that describe the appearance of CCA [14, 20, 22, 25].

Limitations of our study include the following. First, it was limited by its retrospective nature and thus by our limited control regarding patient selection. Second, at the present time these lesions may be histopathologically indistinguishable, and pathologists rely heavily on clinical and imaging findings to establish the diagnosis of CCA and AUP. Despite our best efforts to select only clearly agreed upon CCA and AUP in phase 1 of our study, upon which we based our imaging criteria, there is uncertainty whether our classification was definitively correct. To a large extent this reflected that uncertainty still exists with our proposed “reference standard” of diagnosis, as the pathology literature persists to this day describing difficulty distinguishing CCA from malignancies of gastrointestinal and pancreatic origin, and hence AUP from these, emphasizing the challenging nature of histopathologic evaluation of those tumors [20]. Third, we had only 17 patients with AUP, which may reflect that only patients with at least 6 months of follow-up without discovering a primary site were included. Primary sites are currently more often detected than in former years, which is likely the principle reason why the occurrence of AUP is decreasing. Lastly, we only included patients with untreated liver lesions. Therefore, applying these criteria greatly limited the number of patients in our study. Although this circumstance decreased the number of patients, it achieved considerable benefit by showing the native appearance of these tumors [30]. The small number of patients with AUP might have also introduced a selection bias. It is conceivable that with a larger number of patients, features such as biliary dilatation or capsule retraction may be identified in a small number of AUPs, as these features have been reported in some colorectal metastases [35, 37]; despite the relatively small number of patients, our results support that these particular features are at least considerably more common in CCA.

In conclusion, our study has reported on several patterns and combinations of MR imaging features that may correlate with CCA or AUP. We have also shown that uncertainty in diagnosis remains, even with histopathology. Recommendations include that the final diagnosis may rest with the combination of clinical/laboratory information, liver MRI, and histopathologic evaluation.

Copyright information

© Springer Science+Business Media New York 2014