Article Text

Download PDFPDF

Cholangiolar pattern and albumin in situ hybridisation enable a diagnosis of intrahepatic cholangiocarcinoma
  1. Diane G Brackett1,
  2. Azfar Neyaz1,
  3. Kshitij Arora1,
  4. Ricard Masia1,
  5. Anthony Mattia1,
  6. Lawerence Zukerberg1,
  7. Joseph Misdraji1,
  8. Lipika Goyal2,
  9. Andrew X Zhu2,
  10. Cristina R Ferrone3,
  11. Omer H Yilmaz1,
  12. Vikram Deshpande1
  1. 1 Depatment of Pathology, Massachusetts General Hospital, Boston, Massachusetts, USA
  2. 2 Division of Hematology/Oncology, Department of Medicine, Massachusetts General Hospital, Boston, Massachusetts, USA
  3. 3 Depatment of General and Gastrointestinal Surgery, Massachusetts General Hospital, Boston, Massachusetts, USA
  1. Correspondence to Dr Vikram Deshpande, Department of Pathology, Massachusetts General Hospital, Boston, Massachusetts, USA; vdeshpande{at}


Aims The histological distinction of intrahepatic cholangiocarcinoma (ICC) from metastatic adenocarcinoma remains a challenge. The primary goal was to evaluate the diagnostic value of morphology and albumin expression in the diagnosis of ICC.

Methods We evaluated morphological patterns in 120 ICCs and 677 non-hepatic adenocarcinomas and performed in situ hybridisation (ISH) stain for albumin in the former cohort (retrospective cohort). We also identified 119 samples from primary and metastatic lesions, the validation cohort, in which albumin ISH was performed as part of the diagnostic workup. Targeted sequencing was performed on selected cases. We also mined existing expression profiling data including cases from The Cancer Genome Atlas (TCGA) (41 760 unique samples).

Results In the retrospective cohort, 45% of ICCs and <1% of non-hepatic adenocarcinomas showed a cholangiolar pattern; albumin ISH was positive in 93% of ICCs with significant intratumorous heterogeneity. In the validation cohort, 29% of ICCs showed a cholangiolar pattern and 88% expressed albumin, while all metastatic non-hepatic neoplasms were negative (n=37) (sensitivity 88% and specificity 100%). Targetable genetic alterations (IDH mutations and FGFR2 fusions) were identified in 31% of ICCs (10 of 32). An analysis of the TCGA data validated the specificity of the albumin assay.

Conclusions The cholangiolar pattern and albumin RNA ISH distinguishes ICC from metastatic adenocarcinoma with high specificity. Given the high prevalence of targetable mutations in ICC, albumin RNA ISH is an essential component in the workup of tumours of uncertain origin. A specific diagnosis of ICC could trigger molecular testing and uncover targetable genetic alterations.

  • Intrahepatic cholangiocarcinoma
  • cholangiolar
  • cholangiocellular
  • albumin
  • in situ hybridization

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


Targeting IDH mutations and FGFR2 translocations in intrahepatic cholangiocarcinomas (ICCs) could revolutionise current therapeutic paradigms. The dramatic response to FGFR2 inhibitors and increased progression-free survival with IDH1 inhibitor1–3 in cholangiocarcinoma necessitates high diagnostic accuracy. Although the diagnosis of ICC relies on histological evaluation, pathologists often sidestep a definitive diagnosis, instead choosing a relatively non-specific category, pancreaticobiliary-type adenocarcinoma, primarily because of the perceived non-specific histological appearance and immunohistochemical profile of ICC. Additionally, patients often present with non-hepatic metastases, often widespread, adding an additional layer of complexity.4 In this context, it is notable that metastatic carcinomas to the liver outnumber ICCs by a wide margin. In a prior study we showed that albumin, one of the best-characterised markers of hepatic progenitor cells, is ubiquitously expressed by ICCs and represents a novel biomarker for this neoplasm.5 6

The cholangiolar pattern ICC has received some attention over the last two decades, although almost exclusively in the Asian literature.7–9 It should be noted that intrahepatic gland-forming tumours dominated (>90%) by cholangiolar pattern are classified as cholangiocellular carcinoma.8–10

The goals of this study were to (1) identify histological patterns unique to cholangiocarcinoma that could aid in the distinction of ICC from metastatic adenocarcinoma, (2) re-evaluate the performance of an albumin RNA in situ hybridisation (ISH) stain and (3) validate these results in the clinic on a cohort of ICCs and mimics.

Materials and methods

We examined two distinct cohorts: (1) primary cohort of patients with cholangiocarcinoma (n=120), bile duct carcinoma (n=20) composed predominantly of hepatic resections, and 677 non-ICC adenocarcinomas, and (2) a validation cohort consisting predominantly of needle biopsies stained with albumin to assess primary site.

Primary cohort

Definition of ICC and selection of cases

Intrahepatic tumours were distinguished from perihilar tumours and those involving the bile duct based on imaging and/or macroscopic evaluation. We excluded mixed hepatocellular cholangiocarcinoma. Using these criteria, we identified 120 ICCs and 20 bile duct carcinomas from the years 2000–2013.

Adenocarcinomas from other sites

We assessed cholangiolar pattern in a cohort of primary adenocarcinomas from the lung (n=114), pancreas (n=143), breast (n=92), colon (n=156), stomach (n=83) and periampullary region (n=26), as well as metastatic adenocarcinoma to the liver with the primary originating in the lung (n=12), colon (n=22), stomach (n=14) and breast (n=15).

Validation set

As a part of the diagnostic workup, we performed albumin ISH (Advanced Cell Diagnostics, see below) on 119 primary and metastatic carcinomas. Using our medical records, we excluded cases lacking clinical history and/or follow-up information. The cohort included patients with a diagnosis of cholangiocarcinoma, hepatocellular carcinoma, metastatic adenocarcinoma to liver and carcinoma of unknown origin. In patients originally diagnosed with a tumour of unknown origin, the primary site was determined by the treating oncologist. We also re-evaluated H&E stained slides to document the presence or absence of the cholangiolar pattern and/or bile duct pattern.

Histological features

Intrahepatic cholangiocarcinoma

The cholangiolar pattern is characterised by well-formed ducts with angular profiles. The individual ducts appear to connect, creating the ‘never-ending’ glandular pattern. The low power image resembles antler horns, a highly characteristic feature of this pattern (figures 1a–c and 2a-c). The nested pattern (figure 3, 1a and c) consists of monotonous cells arranged in an organoid pattern, an appearance resembling a neuroendocrine tumour, often prompting immunohistochemistry for neuroendocrine markers. Bile duct-type ICCs show abundant intracellular and luminal mucin. Undifferentiated and sarcomatoid ICCs are characterised by nests and sheets of undifferentiated malignant cells (figure 3d). Less common patterns included (1) ischaemia-like pattern, characterised by the loss of neoplastic cells with an intact connective tissue framework, (2) large duct pattern, comprised of large dilated glandular units and (3) cribriform pattern with comedo necrosis, often luminal. For statistical analysis, we recorded the most prevalent pattern, although minor elements were also documented.

Figure 1

Intrahepatic cholangiocarcinoma with cholangiolar pattern (A–C). All three cases were interpreted as showing a cholangiolar pattern by the three observers. Well-differentiated adenocarcinoma of the lung (D), pancreas (E) and ductal carcinoma of the breast (F), which were all interpreted as lacking cholangiolar pattern.

Figure 2

Intrahepatic cholangiocarcinoma (ICC) with cholangiolar pattern (A and B). ICC mimicking a Von Meyenburg complex (C). ICC with tubular pattern and clear cytoplasm (D), eosinophilic cytoplasm (E) and clear cell change (F).

Figure 3

Intrahepatic cholangiocarcinoma (ICC) with nested pattern of growth (A and C). The tumour is diffusely positive for albumin in situ hybridisation (B) and negative for synaptophysin and chromogranin (images not shown). Sarcomatoid ICC (D).

Twenty perihilar and bile duct carcinomas were also studied.

Interobserver study

To assess the ability of pathologists to recognise the cholangiolar pattern, we identified 17 tumours from the primary cohort, including 7 consecutive cholangiolar pattern ICCs and consecutive well-differentiated adenocarcinomas from the lung (3), pancreas (4) and breast (3) (figure 1). We concealed the non-neoplastic elements with masking tape to ensure that the adjacent normal tissue was undetectable. In a further attempt to confound observers, we included an ICC metastatic to lymph node. Three designated liver pathologists were blinded to the diagnoses and each recorded the presence or absence of cholangiolar pattern. Since none reported a prior knowledge of this pattern, they reviewed 10 images highlighting the cholangiolar pattern from three independent ICCs prior to the analysis.

In situ hybridisation

Branched-DNA technology ISH was performed on formalin-fixed paraffin-embedded (FFPE) tissue using an albumin ISH probe (Affymetrix, Santa Clara, California, USA) on an automated platform, as defined in prior publications.5 6 The lowest-magnification objective at which chromogen was visualised was recorded for each case (×2, ×4, ×10, ×20 and ×40). Albumin ISH was also performed on the validation cohort using Branched-DNA technology (Advanced Cell Diagnostics, Newark, California, USA).11

Targeted sequencing

Our internal tumour-profiling assay was performed on RNA extracted from FFPE specimens as part of routine clinical care. A tumour genotyping assay based on the SNaPshot multiplex platform system (Applied Biosystems, Carlsbad, California, USA) was used to simultaneously query more than 150 previously described hotspot mutations across 16 cancer genes.12

Fusion assay

The solid fusion assay is a targeted RNA-sequencing method of Anchored Multiplex PCR to detect multiple fusions including FGFR2 fusions. The methodology has been previously described.12


Primary cohort

We identified 120 ICCs including 75 hepatectomy specimens and 45 needle biopsies. The mean age of the patients was 63 years (range 17–85 years). There were 62 female and 58 male patients.

Pathology features

Macroscopic examination and/or imaging revealed mass-forming tumours with a mean tumour size of 6.7 cm (range 2–18). The neoplastic cells were embedded in abundant intratumorous stroma (97%). Most patients (60%, 72 cases) showed more than one histological pattern (table 1); however, predictably, a higher proportion of biopsies showed a single growth pattern compared with hepatectomy specimens: 60% vs 28%. Thirty-one (26%) tumours were well differentiated, 61 (51%) moderately differentiated and 28 (23%) poorly differentiated.

Table 1

Histological patterns in intrahepatic cholangiocarcinoma

Cholangiolar pattern

Forty-five per cent (54/120) of the ICCs showed a cholangiolar pattern, and in 39% this represented the major pattern (table 1). Nine cases (8%) showed a striking resemblance to Von Meyenburg complex, the so-called ductal plate malformation-type cholangiocarcinoma.13 The tubular pattern was seen in 55% of cases (n=67). The cholangiolar and tubular patterns tended to coexist, with 24% of cases (n=29) showing both patterns. The bile duct-type carcinoma pattern was rarely identified at an intrahepatic location (2%, n=3), but was seen in all 20 extrahepatic bile duct carcinomas.

Within the ICC cohort, 13 cases (11%) showed intraluminal mucin, while 9 cases (8%) displayed intracellular mucin. The latter was most often identified in bile duct-type ICCs (67%, two of three cases). Nineteen per cent of tumours (n=23) demonstrated clear cytoplasm.

Interobserver study and cholangiolar pattern in non-hepatic origin tumours

Among the 614 non-hepatic origin tumours, only a single case (0.2%), a pancreatic ductal adenocarcinoma, displayed a cholangiolar pattern. The other patterns did not distinguish ICCs from non-hepatic neoplasms.

The three pathologists distinguished all cholangiolar pattern ICCs (n=7) from non-hepatic well-differentiated adenocarcinomas (n=10). The high concordance between the three observers suggests that cholangiolar ICCs are reliably identified by an experienced pathologist aware of the features.

Albumin ISH

Albumin ISH was performed in ICC cases with available FFPE material (n=67). Most ICCs expressed albumin (62/67, 93%) with reactivity easily visualised at a ×2 or ×4 objective in 35 of 67 cases (52%) (table 2). Intratumorous heterogeneity was notable, with many tumour cells showing weak reactivity or were negative. Predictably, biopsy samples were more likely to be negative (3/12, 25%) than whole sections (2/55, 4%) (p=0.024). We did not observe an association between the intensity of reactivity and histological pattern, with one exception: ICCs with an undifferentiated component tended to express lower levels of albumin (table 3).

Table 2

Albumin reactivity in biopsy and resection samples

Table 3

Relationship between albumin expression and tumour patterns

Specificity and sensitivity of cholangiolar pattern and albumin ISH

The cholangiolar pattern identified ICCs with a sensitivity and specificity of 45% and 99%, respectively; however, the sensitivity in the retrospective cohort was lower for biopsy samples (40%). The sensitivity of albumin ISH was 93%; based on a prior study from this institution, the specificity was >95%.5

Validation cohort

This cohort was dominated by needle biopsies, and the majority (89%) of ICCs were diagnosed on these small biopsy samples. A cholangiolar pattern was identified in 29% of ICCs but none of the non-hepatic neoplasms. Bile duct-type carcinomas and hepatocellular carcinomas also lacked this pattern. Two of the 52 (4%) cases showed morphological features of bile duct-type ICC.

Albumin reactivity was noted in 89% of ICCs; in contrast, all metastatic carcinomas were negative (table 4). The reactivity ranged from focal to diffuse. Notably, and like the retrospective cohort, background reactivity was minimal, <1 dot per 100 cells. In cases with low reactivity, we required multiple dots (>5) within the cytoplasm of a single tumour cell (figure 4). Both cases with bile duct-type ICC were positive for albumin. Bile ducts, bile ductules and hepatocytes served as an internal control.

Figure 4

The abundant intratumorous stroma is compatible with intrahepatic cholangiocarcinoma; however, in the absence of cholangiolar pattern, the histological appearance is relatively non-specific (A). The results of an albumin in situ hybridisation stain are shown in (B) and (C). Note the focal expression of albumin (B) (arrow). Although most of the tumour is negative, a subset of cells is positive for albumin expression, manifested by multiple dots within the cytoplasm (arrows).

Table 4

Result from the prospective cohort

It is also notable that 19% (10 of 52) of ICC showed a targetable mutation, IDH1/2 or FGFR2 fusions. An example of a misdiagnosed ICC is illustrated in figure 5. Originally diagnosed a primary ovarian carcinoma, the detection of FGFR2 fusion leads to testing the sample with albumin ISH, which was diffusely positive.

Figure 5

A case of intrahepatic cholangiocarcinoma originally diagnosed as a primary ovarian carcinoma with hepatic metastasis (A). The fusion assay identified a FGFR2-CIT fusion. Following the identification of the fusion, albumin in situ hybridisation was performed, revealing strong and diffuse expression of albumin (B), supporting a hepatic primary. The patient was treated with an FGFR2 inhibitor.

Collectively, the sensitivity and specificity of albumin ISH for ICCs in the validation cohort was 89% and 100%, respectively.

The sensitivity of albumin for hepatocellular carcinoma (HCC) was 96%; notably, almost half of these tumours were poorly differentiated.

Validation on the TCGA cohort

We queried 159 studies and 41 760 samples at the cBioPortal site for Cancer Genomics ( Albumin was consistently expressed in HCCs and ICCs (figure 6). Among the five cholangiocarcinomas with the lowest levels of albumin expression, three were classified as perihilar, mid and lower bile duct carcinomas (TCGA-W5-AA2H-01, TCGA-3X-AAVC-01 TCGA-3X-AAVC-01), while two cases were characterised as ICC (TCGA-3X-AAV9-01 and TCGA-W5-AA2R).

Figure 6

Analysis of The Cancer Genome Atlas (TCGA) data. Most intrahepatic cholangiocarcinomas (ICCs) (arrow) and hepatocellular carcinomas (HCCs) (arrow head) show albumin expression above the black bar. Also, note that some tumours with the highest expression of albumin showed histological features compatible with a hepatoid carcinoma. Additionally, a small cohort (0.18%) of breast carcinoma showed albumin expression like ICCs and HCCs.

A few other non-hepatic tumours, predominantly gastrointestinal in origin, also expressed high levels of albumin. A review of the histology (stomach, TCGA V9-A9DV; lung, TCGA-69–8255) showed histological features consistent with a hepatoid carcinoma: these tumours also showed high expression of alpha-fetoprotein. Notably, with non-hepatic neoplasms, albumin expression correlated positively with alpha-fetoprotein, arginase-1, glypican 3 and CPS1 (hepatocyte) expression (all, p<0.001, q=0.001), supporting the argument that these neoplasms represent hepatoid carcinomas or a variant thereof. We also identified a cohort of breast carcinomas (13 out of 7181, 0.18%) with high level expression of albumin.


In the primary cohort, the cholangiolar pattern identified ICCs with a sensitivity of 45% and a high specificity, 99%; although the sensitivity in the prospective cohort was lower (29%). The sensitivity of albumin ISH in the retrospective cohort was 93%. In a prior study, we evaluated a large cohort of 450 non-hepatic primaries neoplasms including tumours from the lung, pancreas, oesophagus and gastro-oesophageal junction, stomach, colon, bladder (urothelial carcinoma), ovary, endometrium, kidney (renal cell carcinoma) and breast; these tumours were uniformly negative for albumin.5 In the validation cohort, the sensitivity and specificity of albumin ISH was 89% and 100%, respectively.

The implementation of the assay in our clinical laboratory allowed us to evaluate a second cohort of patients and test the assay in a ‘real-world’ setting. Less than a third of ICCs showed a cholangiolar pattern, a significantly lower percentage than the primary cohort, a predictable outcome given the fact that this group was comprised predominantly of needle biopsy samples. The albumin stain was reviewed by a broad range of pathologists, including non-gastrointestinal pathologists, and maintained a high sensitivity, 89%. The assay also sustained its near-perfect specificity. The concurrent presence of cholangiolar pattern and albumin, two unrelated parameters, is virtually diagnostic of ICC.

A notable caveat to the albumin ISH assay is the heterogeneous expression of albumin, with 9% of ICCs in the primary cohort, particularly undifferentiated tumours, showing only rare transcripts, only visualised on a ×40 objective. The heterogeneity in expression is exaggerated in needle biopsy specimens. In this analysis, we classified tumours with rare cells harbouring multiple cytoplasmic transcripts as positive for albumin. This interpretation is justified by the extremely low non-specific reactivity of the branch chain ISH platform. Although non-specific reactivity is typically absent, we nonetheless would advocate a relatively stringent standard for labelling a biopsy positive for albumin: we require multiple dots (typically >5), specifically localised to the cytoplasm of multiple neoplastic cells. Finally, perhaps one of the most challenging aspects of interpreting the assay lies in distinguishing lesional cells from entrapped hepatocytes and bile ducts. Compared with ICC, entrapped hepatocytes show significantly higher expression of albumin. In contrast, entrapped bile ducts and ductules have lower levels of albumin expression, akin to ICC. Relative to entrapped bile ducts, the nuclei of ICC tend to be larger, a finding that can aid in the recognition of neoplastic cells.

The heterogeneous distribution also impacts assessment of the cholangiolar pattern. The cholangiolar pattern may not be captured on a biopsy, representing another limitation of this algorithm. Nevertheless, the pattern is readily recognisable, once observers are aware of its features, as evidenced by the universal agreement between the three reviewers. Despite tumour heterogeneity in both cholangiolar pattern and albumin expression, given the high specificity of these two features, when present, should still be sought out when considering a diagnosis of ICC.

Although the specificity of albumin for ICC is high, several additional caveats deserve mention, the first of which is the near-ubiquitous expression of albumin in HCC. At our institution, albumin is often used to define arginase-1 negative poorly differentiated HCCs: 61% of HCCs in our study were negative for arginase—1% and 46% were poorly differentiated.14 Nevertheless, we detected albumin in 96% of HCCs, validating this assay as the most sensitive and specific marker of HCC.6 Albumin may also be expressed in pancreatic acinar cell carcinoma (ACC), a tumour that frequently expresses other markers of hepatocellular differentiation and thus may present a diagnostic pitfall.14 In a previous study, albumin ISH was positive in 25% of cases of ACC (5 of 20 cases). In a smaller series, a significant proportion of cases were positive for albumin expression, four of seven cases.15 However, given the rarity of ACC and the low likelihood of it presenting as a tumour of unknown origin, this neoplasm does not represent a major pitfall in routine practice. Furthermore, trypsin and chymotrypsin immunohistochemical stains can be used to reliably detect ACCs, effectively distinguishing it from HCCs.

Two independent studies have validated the use of albumin as a marker of hepatic origin. In one study, albumin was positive in 71% of small duct cholangiocarcinomas (most ICCs) and 18% of large duct cholangiocarcinoma,16 a lower sensitivity than the current series. It should be noted that unlike the current study, this analysis was performed on tissue microarrays composed of 0.1 cm cores of tissue, and given the heterogeneity in expression, these samples may not have been adequate to assess albumin expression.16 In another study, albumin reactivity was noted in 36 of 37 (97%) HCCs and 14 of 22 (64%) ICCs.17 The authors also evaluated breast carcinomas, and adenocarcinomas from the lung, oesophagus, pancreas, bladder, endometrium, endocervix, ovary, kidney, as well as mesotheliomas, and papillary thyroid carcinomas. With the exception of a case of lung adenocarcinoma, the 454 non-hepatic tumours were negative for albumin.17 This study was performed using a manual platform and not an automated assay as performed in this study.17 Additionally, the authors did not specify the age of the paraffin blocks in these studies.16 17 We have noticed significant degradation in RNA quantity in blocks older than a decade, as evidenced by the significantly lower sensitivity (25%) in our retrospective biopsy cohort.

For pathologists unfamiliar with albumin ISH staining, it is important to reiterate why it is much preferred over immunohistochemistry. As discussed previously in Ferrone et al,5 ,5 there is strong non-specific background staining of albumin due to its ubiquitous presence in serum. This severely limits the use of albumin immunohistochemistry in specifically targeting the cells responsible for protein expression.

The ability to mine existing data represents a powerful means of validating current immunohistochemical and ISH assays. In this analysis, we queried the expression of albumin in 41 760 samples across 159 studies. Apart from ICCs and HCCs, a small cohort of breast carcinomas (<1%) also expressed albumin. In the appropriate context, it would thus be advisable to assess for markers of mammary differentiation in tumours that express albumin. Gastrointestinal tumours with high expression of albumin, for the most part, appear to represent examples of undiagnosed hepatoid carcinoma, as demonstrated by the morphological appearance and high expression of AFP. Finally, tumours arising from the perihilar region and the common bile duct express significantly lower levels of albumin; notably, only one of six cases in this series was positive for albumin.

The cholangiolar pattern is also a highly specific means of diagnosing an ICC. The pattern, in its most overt form, mimics the hepatic duct plate and recapitulates the embryonic biliary system. However, there are significant differences between this North American series and those from Asia.7 Bile duct-type ICCs, characterised by larger ducts and abundant intracellular and extracellular mucin, constitute a higher proportion of ICCs in the Asian population: 59% of cases in one series compared with only 2%–4% of ICCs in this series.7 Notably, bile duct-type ICCs tend to be negative for albumin.16 The sensitivity of the albumin assay will thus be significantly lower at an institution with a higher incidence of bile duct-type ICC.

Gemcitabine and cisplatin currently represent the standard of care for the treatment of ICC, although the outcome of these patients continues to be dismal.18 The discovery of two potential molecular targets, IDH1 mutations and FGFR2 fusions, has the potential to change current therapeutic algorithms. Currently in phase III trials, inhibitors of IDH1 and FGFR2 fusion in phase I and II have shown significant promise. Of note, the randomised phase III trial of AG120, an IDH1 inhibitor, in cholangiocarcinoma achieved its primary endpoint and demonstrated a statistically significant improvement in progression-free survival compared with placebo. As highlighted in figure 5, the diagnosis of targetable ICC in the metastatic setting represents a major challenge, particularly since these biopsies may be seen by non-gastrointestinal pathologists and could mimic a wide range of primary tumours. We would thus advocate an albumin stain for patients who present with a primary of unknown origin, since a precise diagnosis may offer these patients a range of specific and often targeted therapies. It should also be noted that FGFR2 fusions have been identified in a wide range of other tumours including pancreatic adenocarcinoma, breast uterine endometrial adenocarcinomas, papillary thyroid carcinomas and ampullary adenocarcinomas, among others. Chondrosarcoma, glioblastoma and acute myeloid leukaemia also harbour IDH1/2 mutations. Thus, defining the cell of origin is critical in deployment of targeted therapy, further underlining the impact of albumin ISH. While we attempt to present the relevance of a diagnosis of ICC in the era of targeted therapy, the study design did not directly address the overall therapeutic impact of the albumin ISH, specifically on disease survival.

The cholangiolar pattern and an albumin stain represent robust tools to a definitive diagnosis of ICC. By assessing the stain in the clinical laboratory, we present a real-world study of ICC and their mimics. Nevertheless, while the specificity of albumin ISH remains high, the sensitivity has ranged from as low as 64% to as high as 99%, and thus a negative result cannot be viewed as an exclusionary criterion for ICC. Furthermore, this study under-represents bile duct-type ICCs, a subgroup that may tend to lack albumin expression.16 Given its high sensitivity and specificity, we recommend evaluating all tumours of uncertain origin for albumin ISH. Finally, as with any other tumour marker, the results must be viewed in the overall clinical context, and continued reliance on other relatively specific markers of origin such as TTF-1 for lung and GATA-3 for breast is recommended.

Take home messages

  • Cholangiolar pattern and albumin RNA in situ hybridisation (ISH) distinguishes intrahepatic cholangiocarcinoma (ICC) from metastatic adenocarcinoma with high specificity.

  • Given the high prevalence of targetable mutations in ICC, albumin RNA ISH is an essential component in the workup of tumours of uncertain origin.

  • A specific diagnosis of ICC could trigger molecular testing and uncover targetable genetic alterations.



  • Handling editor Runjan Chetty.

  • DGB and AN contributed equally.

  • Contributors DGB, AN, OHY, CRF and VD were involved in study concept and design and drafting the manuscript. DGB, AN, KA, RM, AM, CRF, and LZ were involved in collection of sample retrieval, laboratory work and data collection. DGB, AN, CRF, VD and JM performed analysis and interpretation of data. LG, AXZ and VD were involved in critical revision of the manuscript for important intellectual content and study supervision.

  • Funding The study was supported by a research grant from Affymetrix, CA, USA. Although this specific study was not supported by Advanced cell diagnostic, the authors institution received support from this institution.

  • Competing interests None declared.

  • Patient consent for publication Not required.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.