Diagnostic approach: morphological patterns 781
Hepatocellular and hepatoid/epithelioid patterns 782
Glandular (biliary, papillary) pattern 783
Mixed epithelioid-glandular (hepatobiliary) pattern 783
Other predominant cell patterns 783
Cystic pattern 785
Benign hepatocellular tumours and tumour-like lesions 785
Malignant hepatocellular tumours and precursor lesions 795
Dysplastic lesions and small hepatocellular carcinoma 799
Hepatocellular carcinoma and variants 802
Combined hepatocellular-cholangiocarcinoma, classic type 820
Combined hepatocellular-cholangiocarcinoma with stem cell features 820
Biliary tumours and tumour-like lesions 823
Biliary hamartoma (von Meyenburg complex) 823
Bile duct adenoma 824
Biliary adenofibroma 825
Biliary and peribiliary gland cysts 825
Other benign biliary-like lesions 825
Intraductal papillary (biliary) neoplasms 826
Mucinous cystic neoplasm 827
Biliary intraepithelial neoplasia 828
Vascular tumours 835
Cavernous haemangioma 835
Infantile haemangioma 837
Epithelioid haemangioendothelioma 838
Other rare forms of vascular tumours 842
Nonvascular mesenchymal tumours and tumour-like lesions 843
Other tumours and tumour-like lesions 852
Leukaemia and myeloproliferative disease 853
Acute leukaemias 853
Hairy cell leukaemia 853
Chronic leukaemias 854
Chronic myeloproliferative disorders and myelodysplastic syndromes 854
Myelomatosis/multiple myeloma 855
Lymphomas and lymphoreticular neoplasms 856
Hodgkin lymphoma 856
Non-Hodgkin lymphoma 856
B-cell lymphomas 857
T-cell lymphomas 858
Primary hepatic lymphomas 858
Hepatosplenic T-cell lymphoma 859
Follicular dendritic cell tumours 859
Other primary hepatic lymphomas 859
Metastatic tumours 860
Because of its anatomical and functional complexity, the liver is involved in as many primary hepatic disorders as extrahepatic or systemic problems. It is always important to think ‘out of the organ’ when dealing with tumours and tumour-like lesions of the liver ( Table 13.1 ). Two major challenges exist. First, the chronically injured liver often becomes cirrhotic, giving rise to hepatocellular nodular lesions of variable biological status, all of which need to be distinguished from hepatocellular carcinoma (HCC). Second, the liver is a common site for metastases from all parts of the body, and these can mimic the two most important primary liver cancers, HCC and intrahepatic cholangiocarcinoma (ICC) with their respective variants, and vice versa. Thus, correlation with clinical and radiological findings is mandatory before rendering a definitive diagnosis.
|Epithelial tumours: hepatocellular |
Malignancy-associated and premalignant
Epithelial tumours: biliary
|Malignancies of mixed or uncertain origin |
Germ cell tumours
Patients with liver tumours fall into the following clinical settings:
Routine health screening uncovering abnormal liver function tests, elevated serum tumour markers or imaging abnormalities
Radiological surveillance for recurrence or metastatic disease in known cancer patients
Ultrasound screening of high-risk patients with known pre-existing liver disorders
Investigation for symptomatology referable to the liver or the primary disease
Investigation for abdominal mass or symptomatology in paediatric patients
Tissue samples come from fine-needle aspiration (FNA) biopsy, core biopsies, wedge biopsies or surgical resections, including explants (see Chapter 2 ). The necessity for preoperative tissue confirmation for classic HCC in the setting of cirrhosis is still debatable. However, one can increasingly foresee the need for small tissue sample procurement with the advent of personalized targeted molecular therapy and extended HCC recipient transplantation criteria. Also, with increasing recognition of primary hepatic carcinomas, differentiation is difficult to classify as either hepatocellular or cholangiolar, or there is dual differentiation within the same tumour, a feature that many now refer to as ‘biphenotypic’ differentiation.
Diagnostic approach: morphological patterns
The following algorithmic approach to the cytohistological diagnosis of tumours or tumour-like lesions serves as our recommended manner of segregating certain diagnostic possibilities when confronted with a tissue sample of a ‘mass’ lesion in the liver. This diagnostic algorithm is based on the recognition of basic patterns and cell profiles. The morphological categories are as follows:
Hepatocellular and hepatoid/epithelioid patterns, including polygonal, oncocytic, large-cell and transitional cell types
Glandular (ducts, glands and/or mucin) pattern, including biliary and papillary patterns
Mixed epithelioid-glandular (including hepatobiliary) pattern
Other predominant cell patterns: squamous, small/intermediate, clear, fat-containing, pleomorphic/giant, spindle, haematopoietic/inflammatory cell and other rare lesions
Hepatocellular and hepatoid/epithelioid patterns
This category encompasses tumour cells that are primarily polygonal and epithelioid in appearance ( Table 13.2 ). Epithelioid cells are generally loosely cohesive cells arranged in sheets and syncytial clusters. Despite the variability in shape and size, there is a general monomorphous appearance. Tumours in this category also may be composed of large cells, cells with oncocytic cytoplasm or cells deemed ‘transitional’ between hepatocellular and other hepatoid or epithelioid tumours (e.g. mixed hepatobiliary carcinoma, metastatic urothelial or neuroendocrine tumour).
|Patterns and cell profiles||Diagnostic considerations|
|Hepatobiliary lesions||Nonhepatobiliary lesions, primary or metastatic|
|Hepatocellular and epithelioid patterns |
Includes cell profiles:
Non-neoplastic (±fatty change):
Neoplastic (±fatty change):
|Glandular (ducts, glands and/or mucin) pattern |
Includes biliary and papillary patterns
|Mixed epithelioid-glandular pattern||Benign||Mesenchymal hamartoma|
Hepatocellular nodular lesions are characterized by benign or malignant hepatocytes forming cohesive trabeculae of variable cell thickness lined by sinusoidal endothelium. Well-differentiated hepatocellular nodular lesions must be distinguished from normal, hyperplastic, cirrhotic and steatotic liver parenchyma. Epithelioid-looking tumours of diverse origin often mimic HCCs, and vice versa. Although most lesions are of epithelial origin, some may be epithelioid variants of mesenchymal tumours. For example, hepatic angiomyolipoma is also a tumour with epithelioid and/or hepatoid features, and neuroendocrine tumours also can have abundant cytoplasm and present as large cells. Very rare tumours such as paraganglioma and phaeochromocytoma could also fall into this category.
Glandular (biliary, papillary) pattern
A glandular pattern, emanating from either ducts or acini, is the most common morphological pattern encountered in focal liver lesions ( Table 13.2 ). This consists of combinations and permutations of tubules, papillae and acini to signet ring-like cells, with variable amounts of intracellular/intraluminal-extracellular mucin (see Cystic pattern later). The benign biliary pattern is recognized on cytology by flat honeycomb sheets of uniform biliary-type cells with ‘bland’ nuclear features. In contrast, a disordered and irregular (so-called drunken) honeycomb appearance with increasing cellular pleomorphism indicates malignant transformation. Glandular lesions within or outside the liver may have similar histological and immunohistochemical profiles, and thus a diagnosis of primary versus secondary/metastatic carcinoma can be problematic. For example, metastatic pancreaticobiliary and gastric carcinomas have similar keratin profiles as cholangiocarcinomas, and therefore these lesions can be indistinguishable from each other. In addition, extensive ductular reaction, if not perceived in the appropriate context, can be mistaken for adenocarcinoma or other biliary tumours.
Mixed epithelioid-glandular (hepatobiliary) pattern
This pattern primarily includes the combined or mixed hepatobiliary carcinoma and its various subtypes ( Table 13.2 ). The most recognizable is combined hepatocellular-cholangiocarcinoma (HCC-CC) with an intimate admixture of HCC, adenocarcinoma and/or transitional components. Nonrepresentative sampling can be expected in small tissue samples.
Other predominant cell patterns
Most of the remaining tumours and tumour-like lesions fit into this approach if classified under cellular appearances, some of which may also indicate histogenesis ( Table 13.3 ). However, some of these categories can also occur in HCC or ICC.
|Predominant cell patterns||Diagnostic considerations|
|Hepatobiliary lesions||Nonhepatobiliary lesions, primary or metastatic|
|Squamous cell||Malignant||Intrahepatic cholangiocarcinoma (ICC) and variants|
|Fat-containing cell||Benign||Focal fatty change |
Hepatocellular nodular lesions with fatty change
|Malignant||HCC with fatty change|
|Pleomorphic and giant cell||Benign|
|Haematopoietic and inflammatory cell||Benign|
|Malignant||HCC with prominent inflammatory component|
|Other rare lesions||Germ cell tumours (e.g. teratoma, yolk sac tumour)|
Squamous cell carcinomas are typically not primary to the liver. These tumours often undergo necrosis with suppuration that mimics abscesses and obscures tumour cells. Keratin-induced granulomas can also be seen.
Both HCC and ICC may have small-cell features. Apart from the usual list of small round cell tumours, the pathologist must be aware of inflammatory/reactive haematological processes that also have small-cell features, although haematological malignancies tend to be diffuse when they involve the liver (see haematopoietic pattern later). Small round cell tumours are relatively common in the paediatric age-group.
The clear cell pattern is typically caused by glycogen deposits or fatty change of various types that can result in clear to bubbly cytoplasm. The classic prototypes of clear cell carcinomas are renal cell and adrenocortical carcinomas, but both HCC and ICC may also have clear cell features.
Hepatocellular nodular lesions with fatty change (steatosis) should be distinguished from steatosis of the background liver. It is also important to consider lipomatous lesions or other tumours that may contain fat cells (e.g. angiomyolipoma) and neoplastic/degenerative conditions that may contain cells mimicking intracytoplasmic ‘fat’ vacuoles.
Pleomorphic and giant cell
These tumours are typically poorly differentiated malignancies with high-grade pleomorphic and giant cell features (but not spindle cells). The giant cells are much larger and with multinucleation; tumour giant cells may exhibit extraordinary malignant atypia and mitotic activity, and these types are common in the more poorly differentiated HCCs. Histiocytic giant cells, as well as neoplastic giant cells of carcinomas and sarcomas (including undifferentiated/embryonal sarcoma of the liver), also could fall into this category.
The spindle cell pattern consists of elongated or fusiform cells, often arranged in interlacing fascicles. Differential diagnoses range from reactive fibroblastic stroma to mesenchymal tumours. Sarcomatoid carcinomas may also have this pattern and, when primary in the liver, may be of hepatocellular or biliary origin.
These tumours range from inflammatory/reactive to proliferative lesions. Intrahepatic and perihilar inflammatory masses may form as a result of distal obstructions, and such obstructions in turn may result from a malignant stricture or intraductal papillary neoplasm. Additionally, marked inflammatory reactions (both neutrophilic and lymphocytic) can be rarely seen in HCC or in tumours with extensive necrosis. Thus the clinician should consider the possibility of a mixture of reactive and malignant components.
Germ cell tumours, such as teratoma and yolk sac tumour, have characteristic features.
The pathologist should first and foremost ascertain whether the lesion is a true cyst with a specialized lining or a pseudocyst ( Table 13.4 ). It should be noted that some tumours may cavitate because of central infarction or locoregional ablational therapy. The nature of the cyst contents should be noted and triaged if clinically warranted. The demonstration of mucin is pathognomonic of a glandular neoplasm, but suppuration with or without superimposed infection can also occur even in an otherwise noninflammatory cyst.
|Cystic pattern||Diagnostic considerations|
|With epithelial lining||Without epithelial lining|
|Malignant||Cystic degeneration of malignant tumours (e.g. undifferentiated embryonal sarcoma of liver)|
Benign hepatocellular tumours and tumour-like lesions
The pattern of presentation of hepatocellular adenoma (HCA) has changed in the recent past such that the incidence is increasing, particularly in certain clinical and aetiological settings. As a result, we now recognize at least four subtypes of HCA: hepatocyte nuclear factor 1α-inactivated (HHCA), inflammatory (IHCA), β-catenin-activated (BHCA) and unclassified, each with different aetiological, histological and molecular characteristics ( Table 13.5 ). HCA is a benign neoplasm that typically arises in noncirrhotic liver; reports of HCAs in cirrhotic liver almost always represent macroregenerative nodules or extremely well-differentiated hepatocellular carcinomas, except in rare cases when definitive subtyping of an adenoma can be proven. HCA typically still occurs in women in the reproductive age-group (15–45 years). In the more remote past, its association with oral contraceptive (OC) use was well established, but the incidence of OC-related adenomas greatly decreased in the last few decades with the introduction of lower-dose OC preparations. HCA can also increase in size during pregnancy, when oestrogen levels are high. This association supports that sex hormones play a role in tumourigenesis by promoting cell proliferation, but are not the initiators of hepatocellular neoplasms. Higher risk of HCA is also associated with noncontraceptive oestrogens and anabolic-androgenic steroids.
|Lesion||Typical histological features||Key clinical associations and comments|
HCA has been reported in association with several hereditary disorders, and multiple adenomas can be seen in these patients. HCA can occur in type Ia glycogen storage diseases and less often in types III and IV. Patients with these disorders may develop multiple adenomas. Maturity-onset diabetes of the young, type 3 (MODY3), a form of autosomal dominant familial diabetes mellitus with germline mutations involving TCF1 gene, can be associated with HCAs (see subtypes later) and multiple adenomas as well. Other, less common associations are familial adenomatous polyposis, McCune–Albright syndrome, Fanconi anaemia, clomiphene use and Klinefelter syndrome.
More recently, a significant increase in HCA has been related to rising levels of obesity and metabolic syndrome. In addition, some specific forms of HCA associated with obesity now appear to be increasing in incidence, especially the inflammatory variants and the β-catenin-activated subtype. These obesity-related lesions present in a slightly older age-group in women, but also in men. Moreover, it is proposed that overweight/obesity could represent a major risk of malignant transformation of HCA, possibly via the IL-6 pathway. Chronic vascular disorders may also increase the risk for adenomas.
The mean age for patients with HCA was 30 years in older series, whereas the mean age has been 37–41 years in more recent series, reflecting changes in aetiology and risk factors for HCA. HCA occasionally occurs outside the reproductive age-group, but these are still uncommon. In the earlier series, patients usually came to medical attention with symptoms of abdominal pain, mass or haemorrhage. In more recent series, however, the majority of tumours were discovered incidentally on imaging, and about 20% of patients present with acute abdominal pain caused by haemorrhage into the tumour or into the peritoneal cavity, which may be life threatening.
HCAs demonstrate variable echogenicity on ultrasound (US) and cannot readily be distinguished from other lesions. On computed tomography (CT), HCAs are iso- or hypoattenuating and often heterogeneous with hyperattenuating areas corresponding to recent bleeding. On multiphasic CT, there is arterial enhancement and a persistent enhancement in the delayed phase. On magnetic resonance imaging (MRI), HCAs are typically hyper- or isointense relative to adjacent liver on T1-weighted images and slightly hyperintense on T2-weighted images. The presence of fat and telangiectasia lead to distinctive appearances on contrast-enhanced US and MRI and may be able to determine HCA subtypes with an accuracy approaching 85%.
By definition, HCA is a benign neoplasm composed of hepatocytes and almost always arises in a noncirrhotic liver. Multiple tumours may be present, and when ≥10 are present, this process has been designated as ‘adenomatosis’ (see later). The tumours often bulge from the surface of the liver and occasionally are pedunculated. The majority of tumours are 5–15 cm but can measure up to 30 cm in diameter. The colour varies from yellow or tan to brown ( Fig. 13.1 ), and there may be green areas of bile production as well as areas of necrosis or haemorrhage. Irregular areas of fibrosis may be present. Lesions may be well circumscribed or not, and in the latter case, the adenoma can be difficult to differentiate from the surrounding liver ( Fig. 13.2 ).
Microscopically, HCA is typically composed of hepatocytes arranged in thin cell plates (typically one or two cells thick) without cytological atypia ( Fig. 13.3 ). Tumour cells are typically uniform in size and shape, but some variation can be seen, which may result from regenerative and atrophic/ischaemic changes within the tumour. The reticulin framework is intact, outlining the cell plates in a pattern similar to that seen in normal liver without loss or fragmentation ( Fig. 13.4 ). In certain HCA subtypes, however, particularly the HHCA, there may often be near to complete encircling of small groups of tumour cells by reticulin fibres (‘packeting’). The nuclei of tumour cells are typically uniform and regular; the nuclear/cytoplasmic ratio is low; and mitoses are almost never seen (see Fig. 13.3 ). Nucleoli are seldom prominent. Occasional tumours, especially in patients with long exposure to contraceptive steroids, may have a few pleomorphic nuclei resembling large-cell change in cirrhosis.
Features that are variably present include fat, glycogen production, cytoplasmic lipofuscin granules, Dubin–Johnson pigment (pigmented liver cell adenoma) and rarely, Mallory–Denk bodies. The fat may be abundant, simulating fatty liver, particularly in some HHCA and IHCA subtypes. The sinusoids can be compressed, leading to a sheet-like appearance. Sometimes the sinusoids are dilated and can be mistaken for peliosis hepatis. Thin-walled vascular channels and small arteries are scattered throughout the tumours, whereas large arteries are seen at the periphery. The arteries are unpaired (lack corresponding bile ducts), because interlobular bile ducts are not found in HCA. However, bile ductules and progenitor cells may be present, particularly in the IHCA subtype. Kupffer cells are present but usually inconspicuous; stellate cells are occasionally seen. Haematopoietic elements may be noted in the sinusoids, and rare cases with noncaseating granulomas have been described. Areas of fresh haemorrhage and haemosiderin-laden macrophages from past haemorrhages may be seen, along with recent or previous infarcts. Immunohistochemically, the endothelial cells lining the plates can stain positively to variable degrees for CD34, similar to that seen in hepatocellular carcinoma (HCC). Other markers typically used for HCC, such as glypican 3 (GPC3) and alpha fetoprotein (AFP), are negative.
Cytologically, FNA aspirates are hypercellular, comprising small cohesive groups of almost normal-looking hepatocytes with transgressing endothelium in the absence of bile duct epithelium ( Fig. 13.5 ). The neoplastic hepatocytes may display fat vacuoles, pale glycogen-laden cytoplasm or pigments. They have relatively small, uniform central round nuclei with normal nuclear/cytoplasmic ratio (≤1 : 3 by estimating comparison of nuclear and cell diameters). Differential diagnoses include other well-differentiated hepatocellular nodules. Cell blocks may not prove sufficient for immunohistochemical (IHC) subtyping. An indeterminate cytodiagnosis may be proffered in the absence of appropriate clinical and radiological information.
Correlation of specific genetic mutations with clinical and histological features led to the recognition of subtypes of HCA ( Table 13.5 ), which formed the basis of the 2010 classification adopted by the World Health Organization (WHO).
Hepatocyte nuclear factor 1α-inactivated (HHCA)
Biallelic inactivation of TCF1 gene that codes for hepatocyte nuclear factor 1α (HNF1α) was observed in about 40% of sporadic HCAs in early studies of adenoma subtypes. The gene defect was found to be somatic in 90% but germline in 10%. These tumours tend to occur in young women, are associated with OC use as well as MODY3 and are often characterized by prominent intralesional steatosis ( Fig. 13.6 ). Both large and small droplets can be present, but in many cases, small-droplet fat can predominate and may be overlooked as an important diagnostic finding. Minor architectural irregularities such as small acinar structures can be seen ( Fig. 13.6 ). Inflammation and cytological atypia are typically absent, but the reticulin framework pattern often has a ‘packeted’ appearance in which reticulin fibres can partially encircle small clusters of hepatocytes ( Fig. 13.7 ). HNF1α positively regulates the FABP1 gene, which codes for liver fatty acid-binding protein (LFABP). LFABP is expressed in the cytoplasm in normal hepatocytes, as demonstrated by immunohistochemistry (IHC), but is low or absent in HNF1α-inactivated adenoma (HHCA) ( Fig. 13.8 ). This is one of the subtypes that may be more frequently associated with adenomatosis, so the liver parenchyma should be examined for the presence of small fatty nodules (also LFABP negative), which represent microadenomas ( Fig. 13.9 ). The risk for malignant transformation is likely very low.
Challenges in HHCA diagnosis
Steatosis is not a specific finding because it can be observed in other subtypes (particularly the inflammatory type), while rare HHCAs may lack steatosis. Loss of LFABP is useful to establish the diagnosis, but the IHC staining has to be carefully titrated because staining in normal hepatocytes can be weak. Hepatocellular carcinoma (HCC) arising in HHCA is rare (<5%) and can be associated with β-catenin activation. Since LFABP loss can be observed in HCC, it should be used for histological subtyping after the diagnosis of HCA has been established based on other histological features, such as cytological and architectural findings.
Inflammatory (formerly ‘telangiectatic’) (IHCA)
The IHCA subtype accounts for 40–50% of HCAs and is most common in women, although it also can occur in men with appropriate risk factors. IHCA is often associated with obesity and metabolic syndrome; steatosis and steatohepatitis can be present in the non-neoplastic liver. These tumours are characterized by activation of the interleukin (IL)-6 signalling pathway (IL6/JAK/STAT3 pathway) resulting from somatic gain-of-function mutations in the IL6ST gene that encodes the signalling co-receptor gp130. These mutations are observed in 60% of IHCAs, leading to activation of signal transducer and activator of transcription 3 (STAT3) in the absence of ligand. Less often, somatic mutations in other genes involved in the pathway, such as Fyn-related kinase ( FRK ), STAT3 , GNAS and Janus kinase 1 ( JAK1 ), have been implicated in IHCA, all of which promote the constitutive activation of STAT3.
The histological hallmarks of IHCA are sinusoidal dilation, ductular reaction, haemorrhage and inflammation ( Figs 13.10 and 13.11 ). Arterioles, often in clusters, and surrounded by a small amount of fibrous tissue with or without ductular reaction at the interface, can mimic portal tracts, but the arterial zones lack interlobular bile ducts (‘pseudo-portal tracts’) ( Fig. 13.11 ). Ductular reaction is seen in almost half the cases. Fibrous septa may be present, and therefore, in combination with ductular reaction, the findings can mimic focal nodular hyperplasia (FNH), although a nodular architecture typical of FNH is seen in <10% of cases. However, because of the similarities to FNH, many of these cases were previously termed ‘telangiectatic FNH’. Based on imaging characteristics, clonality and protein-clustering studies, it is now recognized that most of these lesions are IHCA. The distinction cannot be made by the presence of sinusoidal dilation alone; this feature can be present in a minority of FNH. C-reactive protein (CRP) and serum amyloid-associated (SAA) protein, both acute-phase reactants, are overexpressed in the neoplastic hepatocytes, so IHC staining for these markers helps to confirm the diagnosis ( Fig. 13.12 ).
Challenges in IHCA diagnosis
Sinusoidal dilation and/or inflammation may not be seen in up to 10% of IHCAs. Both SAA and CRP typically stain these lesions in a diffuse manner, although SAA can be negative or minimal in 5–10% of cases with otherwise typical morphology, while CRP will be diffusely positive in an estimated >80% of tumour cells in essentially all these lesions. Thus, CRP has higher sensitivity, and completely negative results are rare. Since both SAA and CRP can be positive in the adjacent non-neoplastic liver, careful correlation with morphology and CD34 staining is necessary to avoid interpreting this as a positive result in the tumour, especially in limited needle biopsies. It is important to note that activation of β-catenin is seen in approximately 10% of IHCAs, resulting in diffuse staining with glutamine synthetase (GS) with or without nuclear β-catenin staining; these tumours are thought to have a higher risk of progression to HCC. A combination of gp130 alterations and β-catenin mutation is likely in these cases.
The BHCA subtype occurs more often in men than do the other subtypes and often harbours atypical morphological features, such as cytological atypia, small-cell change and pseudoglandular/acinar architecture. Multifocal reticulin loss may occur ( Fig. 13.13 ), raising concerns for transition to HCC. Association with concurrent or subsequent HCC may occur in almost half the cases. These tumours are characterized by mutations in the exon 3 of CTTNB1 (β-catenin) gene, which is a critical component in the Wnt signalling pathway. In normal cells, β-catenin is located in the membranous region and can be demonstrated by IHC ( Fig. 13.14 ). Cytoplasmic β-catenin binds to a protein complex that targets it for proteasomal degradation. Deletions or point mutations in the exon 3 region interfere with the binding of β-catenin to the protein complex, which in turn interferes with the cytoplasmic degradation of the protein and leads to its nuclear translocation. Nuclear β-catenin activates T-cell factor/lymphoid enhancer factor (TCF/LEF) family of proteins, leading to expression of a variety of Wnt target genes that play an important role in cell proliferation. IHC demonstration of nuclear β-catenin can help to identify tumours with β-catenin-activation ( Fig. 13.14 ). Activated β-catenin leads to transcriptional upregulation of GLUL gene, which codes for GS, and diffuse GS staining can be demonstrated by IHC as evidence of β-catenin activation ( Fig. 13.15 ).
Challenges in diagnosis of BHCA
In view of the atypical histological findings, frequent reticulin loss, high risk of HCC and rare instance of metastases in these BHCAs, it may be better to regard these tumours as extremely well-differentiated variants of HCC rather than adenomas. Cytogenetic abnormalities such as gains of 1q and 8q are typical of HCC and are often seen in adenoma-like tumours with β-catenin activation, further supporting their classification as HCC. In some cases, adenoma-like tumours with β-catenin activation do not show features that are sufficient for definite diagnosis of HCC. It has been suggested that these tumours should be characterized as ‘atypical hepatocellular neoplasm’ or ‘hepatocellular neoplasms with uncertain malignant potential’ (HUMP). In addition to tumours with borderline histological features and those with β-catenin activation, this designation has also been recommended for all adenoma-like tumours that occur in men, women >50 years, rare aberrant IHC features (e.g. GPC3 or HSP70 positivity) and those occurring in unusual clinical settings (e.g. glycogen storage disease, Fanconi anaemia). In addition, it is possible that the β-catenin activation can occur in any other HCA subtype (HHCA, IHCA or unclassified) as a ‘second hit’ in the progression of HCA to HCC, and thus BHCA may not truly represent a subtype of HCA. In initial studies, the BHCA subtype accounted for 10–15% of total HCA, but the numbers in more recent studies are less than 5%, perhaps reflecting that many tumours initially considered as BHCAs are now being classified as ‘extremely well-differentiated HCC’ if enough cytoarchitectural atypia is present.
Nuclear β-catenin staining is present in less than half of the tumours with known β-catenin mutation or deletion (see Fig. 13.14 ), whereas diffuse GS staining has greater sensitivity and better correlation with β-catenin activation (see Fig. 13.15 ). Therefore, absence of nuclear β-catenin on IHC does not exclude β-catenin activation in a tumour. When present, nuclear β-catenin staining can be focal, and any unequivocal nuclear staining irrespective of number of positive tumour cells should be regarded as a positive result. Because of possible nonspecific background cytoplasmic staining, it is recommended that cytoplasmic β-catenin staining in the absence of nuclear staining not be considered positive.
The interpretation of GS immunohistochemistry plays a critical role in the classification of HCA subtypes. GS is an enzyme involved in ammonia detoxification by combining it with glutamate to produce the amino acid glutamine. In normal liver, GS expression is limited to a narrow rim of hepatocytes around the central vein. This zonation is thought to be the result of β-catenin activation in centrizonal hepatocytes, presumably as a result of Wnt signalling from the central vein. Activation of β-catenin in neoplastic hepatocytes leads to diffuse cytoplasmic expression of GS in the tumour cells. Due to the low sensitivity of β-catenin IHC, it has become standard clinical practice in diagnostic pathology to use diffuse GS immunostaining as a marker of β-catenin activation. In this context, diffuse staining has been defined as ‘moderate to strong’ cytoplasmic staining in >50% of tumour cells. Diffuse staining can be seen in almost all the tumour cells (diffuse homogeneous staining), or it can be seen in >50% but not all tumour cells (diffuse heterogeneous staining). The interpretation of diffuse homogeneous GS staining is often straightforward, but it can be difficult to discern whether staining is seen in >50% of tumour cells when a diffuse heterogeneous pattern is encountered. In contrast to the near-perfect correlation between diffuse GS staining and β-catenin activation in HCAs reported in some studies and in the WHO classification, recent studies have shown that GS is not a perfect marker of β-catenin activation, and many cases with diffuse GS staining (perhaps 20–30%) may not have evidence of β-catenin mutation or deletion. The correlation is likely to be higher for diffuse homogeneous GS staining. The reasons for the discrepancy are unclear but may be related to activation of other components of Wnt signalling pathway, such as AXIN1 mutation or mechanisms unrelated to mutations in genes in Wnt signalling, including alterations in vascular flow. It thus seems prudent to diagnose these tumours as HCA or HUMP with diffuse GS staining and convey that an underlying β-catenin ( CTTNB1 ) mutation or deletion is likely but not certain. GS staining in other subtypes of HCA tends to be patchy and, by definition, involves <50% of hepatocytes. This patchy staining can be perivenular and is often more prominent compared to normal liver (‘expanded perivascular staining’). Peripheral accentuation can be seen. In some instances, the patchy areas of staining at the periphery can connect with each other, creating a pattern reminiscent of map-like pattern seen in FNH (‘pseudo-map-like’).
Mutations in exons 7 and 8 of CTTNB1 have recently been described in 10% of HCAs, many of which were previously categorized as unclassified. These cases do not show atypical morphological characteristics or diffuse GS staining typical of tumours with exon 3 CTTNB1 mutations. Based on initial results, it is thought that mutations in exons 7 and 8 do not confer the malignant potential associated with exon 3 CTTNB1 mutations, but may result in diffuse heterogeneous staining in a small subset of cases.
HCAs that lack distinctive immunohistological and molecular abnormalities of the other three subtypes are currently considered ‘unclassified’. Some of these lesions, however, have been found to be associated with longstanding vascular flow abnormalities of the liver because of a variety of aetiologies, both congenital and acquired.
Specific clinicopathological settings
Hepatic adenomatosis is arbitrarily defined as the occurrence of ≥10 HCAs. This condition can be familial, in which case it usually occurs in association with a specific form of diabetes (MODY3) and is caused by germline mutation in the TCF1 gene that encodes HNF1α (see Fig. 13.9 ). These adenomas show loss of LFABP on IHC. Additional germline mutations in other genes such as CYP1B1 can increase the risk of HCAs in MODY patients. Young age, male gender and a history of diabetes can further increase this risk. Adenomatosis with IHCAs can rarely occur in McCune–Albright syndrome. Rare instances have been described in vascular conditions such as congenital absence of the portal vein and post-Fontan procedure.
Hepatic adenomatosis also has similar risk factors and clinicopathological settings as sporadic HCAs, including predilection for young women. Several reports have implicated obesity and fatty liver disease as risk factors for adenomatosis of the inflammatory subtype (see Fig. 13.2 ). In most cases, all HCAs in adenomatosis are of the same histological subtype, although a mix of IHCA and HHCA has been reported in the same case. Treatment is dictated by the size of the nodules and subtype of HCA. As with sporadic HCA, the risk for HCC is highest with larger tumours (>5 cm) and those with β-catenin activation. Therefore, resection is often performed in these patients as well as for symptomatic patients. Embolization and transplantation are options for unresectable cases.
Hepatocellular adenoma in glycogen storage diseases
HCAs may occur in about half the patients with type I and 25% of those with type III glycogen storage disease (GSD). IHCAs comprise 52–65% of adenomas, and most have mutations in the IL6/JAK/STAT3 pathway similar to sporadic IHCAs. Up to 20% of HCAs have been reported to have β-catenin activation, and HCC is well described in GSDs. Thus the diagnosis of HCA should be made with caution in this setting, especially when β-catenin activation is present and the designation of ‘atypical hepatocellular neoplasm’ or ‘HUMP’ can be used for cases with borderline features. Due to the risk of HCC, screening with serum AFP and annual US is recommended.
Adenoma-like tumours in other liver diseases
Nodular lesions that are histologically similar to HCA have been observed in the setting of vascular diseases, including Budd–Chiari syndrome, portal vein agenesis, hereditary haemorrhagic telangiectasia and tetralogy of Fallot. Loss of L-FABP and positivity for CRP has been described, although typical histological features of IHCA are not seen. Since these can be multiple and can be mixed with FNH-like nodules, it has been proposed that these HCA-like nodules are not neoplastic and represent regenerative nodules as a result of uneven blood perfusion throughout the liver. However, β-catenin activation and association with HCC have been reported, and careful review is necessary before concluding that the nodules are regenerative or benign.
Nodular lesions with histological features of IHCA and SAA/CRP positivity have been described in cirrhosis related to alcoholic as well as nonalcoholic steatohepatitis (NASH). Mutations in IL6/JAK/STAT pathway have been described, supporting the neoplastic nature of these nodules. Thus, rare cases of IHCA can occur in cirrhosis.
Hepatocellular carcinoma arising in hepatocellular adenoma
HCC arising in the setting of HCA is a rare phenomenon and has been reported in 4–8% of cases. Advanced age, male gender, use of anabolic steroids, metabolic syndrome and large tumour size have been cited as risk factors for HCC transformation in an adenoma. Both tumours occur concurrently in most reported cases; it has been assumed that this represents malignant transformation of an adenoma. BHCAs are more likely to be associated with concurrent or subsequent diagnosis of HCC. The adenoma portion often shows focal atypical morphological features and molecular features resembling HCC, suggesting that the adenoma-like portion in a subset of these tumours may represent an extremely well-differentiated form of HCC (see previous discussion).
Discontinuation of OCs, hormone-containing intrauterine devices (IUDs) and anabolic steroids may lead to regression of HCA. HCAs associated with GSD may regress with dietary therapy. Resection is recommended for tumours ≥5 cm because of the risk of rupture, haemorrhagic complications and associated HCC. Nonsurgical options such as embolization can be considered for patients who are poor surgical candidates. Since some HCAs have been reported to increase in size despite the discontinuation of OCs or anabolic steroids, and the development of HCC has been reported despite regression in size, it is recommended that patients with smaller tumours have CT or MRI at 6-month intervals for 2 years, with subsequent follow-up dictated by the growth pattern and stability of the tumour over time.
Focal nodular hyperplasia can sometimes be difficult to distinguish from HCA. The presence of fibrous septa containing large arteries and ductular reaction favors FNH (see next section). However, these features can also be seen in IHCA, so the distinction can be challenging, especially in small biopsies. In these settings, the map-like GS staining in conjunction with histological features can confirm the diagnosis of FNH. Notes of caution, however, include that the periphery of HCAs may show a vague anastomosing pattern of staining reminiscent of the map-like pattern of FNH. In addition, SAA can be positive in 10–15% of FNH cases, typically in a focal manner, while CRP staining in the periseptal region is seen in most FNH cases without the diffuse staining typical of IHCA, with diffuse staining restricted to a small minority of FNH cases.
Distinguishing HCA from well-differentiated HCC can also be challenging and may be impossible on needle biopsies. The presence of wide cell plates, prominent pseudoglandular/acinar architecture, zones of small-cell change, cytological atypia and reticulin loss favour HCC (see Table 13.5 ). Diffuse sinusoidal staining for CD34 also favours HCC but is insufficient alone for the diagnosis. GPC3 and heat shock protein 70 (HSP70), if positive, strongly point to HCC. Similarly, the demonstration of cytogenetic abnormalities such as gains of chromosome 1 and 8 and TERT promoter mutation would indicate HCC, but these are not currently available for routine diagnostic use. If the histological changes are borderline, or if high-risk features such as male gender and β-catenin activation are present, it is prudent to designate the tumour as ‘atypical hepatocellular neoplasm’ or ‘HUMP’.
Focal nodular hyperplasia
Focal nodular hyperplasia (FNH) is a benign, non-neoplastic lesion that can occur in both genders and all ages, but is most common in adult women. In most cases, FNH is discovered incidentally at surgery or in imaging studies. FNH is thought to represent a hyperplastic response of hepatocytes to altered blood flow. Most women (70–80%) with FNH are taking OCs, but these are not thought to play an aetiological role. The hepatocellular component is polyclonal, which supports the conclusion that FNH represents a regenerative rather than a neoplastic process. Angiopoietin is involved in vasculogenesis and is upregulated in FNH.
Most FNH cases can be diagnosed by multiphasic CT scan. FNH shows hyperattenuation in the arterial phase of multiphase CT scan, and it becomes isoattenuating with normal liver in the delayed phase. The central scar is hypoattenuating and becomes hyperattenuating in the venous phase. On T1-weighted MR images, FNH is typically isointense to hypointense. On T2-weighted images, it is hyper- or isointense, and the scar is hyperintense in most cases. US is not a preferred modality because FNH is not well visualized. The central scar can be mistaken for fibrolamellar hepatocellular carcinoma (FLC) on imaging. If imaging shows typical features including a central scar, the patient may be followed without biopsy.
More than 85% of FNH lesions are <5 cm in diameter, but larger lesions can occur. FNH is usually well circumscribed but nonencapsulated. The colour is lighter than that of the adjacent normal liver. On cut section, the lesion is subdivided into smaller nodules by fibrous septa that often run into a stellate scar that may be central or eccentric ( Fig. 13.16 ). Several scars may be seen in large lesions. A central scar is absent in 20% of cases. Microscopically, FNH is characterized by hyperplastic nodules that are separated by fibrous septa, often radiating from a central scar. The fibrous septa and central scar contain dystrophic thick-walled arteries and to a lesser extent, veins. The arteries are often eccentrically thickened and show intimal and fibromuscular hyperplasia with poorly formed elastic lamina ( Figs 13.17 and 13.18 ). The scar and fibrous septa typically show a ductular reaction, most prominent at the interface between the nodules and fibrous stroma, and variable lymphocytic inflammation, while interlobular bile ducts are absent. The hepatocellular component is arranged in plates one to two cells thick, with intact reticulin framework, but thicker plates may be present focally, often near the center and periphery of the lesion ( Fig. 13.19 ). Complete circling of small groups of hepatocytes by reticulin fibres (packeting) is also a common feature. Fat accumulation can be seen, especially in the patient with fatty liver disease (see Fig. 13.18 ). Steatohepatitic changes such as ballooned hepatocytes and Mallory–Denk bodies may also be present and should not be confused with steatohepatitic variant of HCC. The periseptal hepatocytes adjacent to the fibrous septa of the lesion often show chronic cholestasis, with cholate stasis (‘pseudoxanthomatous transformation’) and copper accumulation. These features of chronic cholestasis are present in many cases of FNH and can be very helpful in differential diagnosis.
On IHC, FNH shows a highly characteristic geographical or map-like pattern on GS staining, with strong cytoplasmic staining of broad anastomosing groups of hepatocytes ( Fig. 13.20 ). The map-like pattern can be focal in needle biopsies and may not be well developed around the central scar or areas with fatty change. In normal liver, GS expression is limited to a narrow rim of hepatocytes around the central vein. This zonation most likely results from β-catenin activation in the centrizonal hepatocytes, which in turn may be the result of Wnt signalling from the central vein. Expansion of the β-catenin-activated centrizonal region leads to GS overexpression in the map-like pattern in FNH. β-catenin mutations are not observed. The ductular reaction is highlighted by cytokeratins 7 and 19 and neuronal cell adhesion molecule (NCAM). CD34 usually shows patchy sinusoidal staining, with diffuse staining in a small number of cases.
On FNA aspiration cytology, a polymorphous population of cells is seen, comprising hyperplastic hepatocytes with normal nuclear/cytoplasmic ratio (≤1 : 3, by estimating comparison of nuclear and cell diameters), biliary elements, endothelium and Kupffer cells. Non-neoplastic hepatocytes in FNH show some mild variation in cell and nuclear size but maintain the normal nuclear/cytoplasmic ratio; they are not monotonous or pleomorphic, similar to that seen within a family of siblings, unless they are identical twins, so-called sibling variation. Anisonucleosis is also a distinct characteristic of non-neoplastic hepatocytes. Another helpful benign feature is the presence of biliary ductular clusters. They represent ductular reaction at the hepatocellular-stromal interface of the scar and appear as curved, double-stranded rows of small cells with overlapping, oval nuclei and scanty or absent cytoplasm ( Fig. 13.21 ). The rows can be seen emanating from hepatocytes. Ductular clusters should be distinguished from bile duct epithelium, which appears as flat sheets of uniform equidistant cells with discernible cytoplasm. Ductular reaction can easily be mistaken for adenocarcinoma, such as cholangiolocellular carcinoma. Endothelial and Kupffer cells are difficult to separate cytologically and are recognized as spindle/comma-shaped nuclei and nuclear streaks. Stromal fragments are rare and may mimic leiomyosarcoma. Regenerative and dysplastic nodules are the primary differential diagnoses.
Telangiectatic FNH refers to a subset of cases with prominent sinusoidal dilation. Based on imaging features, clonality results and gene expression studies, most but not all ‘telangiectatic FNH’ cases are now classified as IHCA. ‘Focal nodular hyperplasia-like’ (FNH-like) lesion has been used to describe lesions morphologically similar to FNH, but that occur in the setting of liver disease (e.g. cirrhosis) and vascular disorders (e.g. idiopathic portal hypertension, hereditary haemorrhagic telangiectasia, haemangiomas, Budd–Chiari syndrome). These FNH-like lesions can be multiple and may not show the characteristic map-like pattern of GS staining seen in true FNH.
Differential diagnosis and treatment
FNH is a non-neoplastic lesion and does not require surgical resection even if large, unless the patient is symptomatic or the location of the lesion compromises liver function. Thus it is important to distinguish FNH from HCA and HCC (see Table 13.5 ). The nodular architecture, central stellate scar, fibrous septa, bile ductules and hepatocellular nodules distinguish FNH from HCA. However, these features can be seen in IHCA, and the distinction can be challenging on histological grounds alone, especially in needle biopsies. Map-like GS pattern is diagnostic of FNH and is not seen in IHCA. Staining with inflammatory markers such as SAA is typically negative or focal, while CRP staining is usually restricted to periseptal areas; diffuse staining typical of IHCA is uncommon in FNH. Sinusoidal dilation (telangiectasia) is more common in HCA but can be seen in 15–20% of FNH, and it cannot be relied on to distinguish these entities. Core needle biopsies can mimic ductopenic cirrhosis, but the clinical setting and presence of dystrophic arterioles suggest FNH.
Nodular regenerative hyperplasia
Nodular regenerative hyperplasia (NRH) is usually considered a reaction to heterogeneous or abnormal blood flow within the liver, which has many causes. The lesion consists of a nodular transformation of the liver that mimics nodules in cirrhosis but lacks the scarring typical of that condition, so NRH is not likely to be confused with a neoplasm (see Chapter 11 ). Occasionally, however, as in cirrhosis, there may be larger regenerative nodules among the smaller nodules, which may be identified on radiology and confused with other hepatocellular tumours, particularly HCA.
Malignant hepatocellular tumours and precursor lesions
Hepatoblastoma (HB) is the most frequent liver tumour in children, with an annual incidence of approximately 4 per 1 million in children under 5 years of age. HB mimics the developing fetal or embryonal liver histologically and may contain other heterologous epithelial or mesenchymal elements. A majority of cases occur before age 5 years, with rare reports of the tumour in older children and adults.
Aetiology and molecular aspects
Congenital anomalies are present in approximately 5% of patients with HB, and reported associations with other congenital diseases include Beckwith–Wiedemann syndrome, familial adenomatous polyposis coli, Li–Fraumeni syndrome, trisomy 18, GSD types I–IV, hemihypertrophy and Simpson–Golabi–Behmel syndrome. Rare cases of congenital HB have been reported.
Cytogenetic studies and comparative genomic hybridization have revealed that tumour cells have a relatively stable genome and usually harbour a limited number of chromosomal abnormalities (mean, three changes per tumour). Recurrent genetic alterations involve chromosomes 2, 20, 1, 8 and X. Frequent trisomies 2 and 20, often seen as the only abnormality, have also been observed in other embryonic tumours such as embryonal rhabdomyosarcoma. The DNA content is frequently diploid in the fetal type and aneuploid in half the embryonal and small, anaplastic cell types. Microsatellite analysis revealed regions with allelic loss of chromosomes 1 and 11.
Several studies revealed that genes encoding components of the wingless/Wnt signal transduction pathway, including the APC , AXIN1 and CTTNB1 genes are mutated in a large proportion of cases (>80% in total). These mutations lead to abnormal activation and induction of growth-promoting genes and other survival pathways (e.g. Notch pathway) that can be differentially expressed in different histological subtypes of HB. The central molecule of this pathway is β-catenin, which carries mutations, mostly in-frame deletions, in more than half of HB cases. An additional gene reported to be affected by recurrent mutations in HB is NFE2L2 , which encodes a transcription factor involved in the antioxidant response pathway.
Other typical molecular features include aberrant reprogramming of imprinted genes such as IGF2, H19, DLK1, GTL2, PEG3, PEG10, BEX1, MEG3 and NDN, which are abundantly expressed in fetal liver. Based on recent transcriptional data, HB can be classified into two distinct groups. The C1 subclass recapitulates liver features at late stages of intrauterine life, with a mostly fetal phenotype and the expression of markers of mature hepatocytes. The C2 subclass has a predominantly embryonal phenotype, with the expression of markers of hepatic progenitor cells (e.g. K19, EpCAM).
Studies of gene expression profiling have compared HB with HCC and identified upregulation of expression of MIG6, TGFβ1, DLK1 and IGF2 in HB. These genes were differentially expressed between HB and HCC, but they did not separate HB histological subtypes.
Approximately 75% of children with HB are male. They usually present with abdominal swelling and less frequently with weight loss, anorexia, nausea, vomiting, abdominal pain and jaundice. Thrombocytosis and marked elevation of serum AFP is present at diagnosis in most patients. Although biopsy is necessary to confirm the diagnosis, imaging plays an important role to determine the location and extent of tumour, assess feasibility of surgery and evaluate vascular invasion and extrahepatic extension.
Hepatoblastomas usually present as a single mass but can be multifocal and may measure up to 25 cm in diameter. The right lobe is more often involved. The non-neoplastic liver is usually normal. HB typically has a lobulated cut surface and may be yellow, brown, green or variegated, depending on the differentiation of the tumour and whether a mesenchymal component is present. In post-therapy resections, a heterogeneous appearance with areas of haemorrhage and necrosis is common. Calcification and ossification may be present. Specific attention should be paid to gross vascular invasion and surgical margins, and fresh tissue should be set aside for special studies whenever possible.
Histologically, HB may be classified as epithelial or mixed epithelial-mesenchymal, each with several recognized subtypes. The epithelial type is the most common and has several subtypes.The fetal epithelial pattern comprises one-third of the cases and is composed of sheets and thin trabeculae of polygonal cells resembling fetal hepatocytes with a small round nucleus, inconspicuous nucleolus, clear to finely granular cytoplasm with variable amounts of cytoplasmic glycogen and lipid giving a ‘light and dark’ pattern to the tumour when viewed at low-power magnification ( Fig. 13.22 ). Haematopoietic cells are almost always present, mimicking the typical extramedullary haematopoiesis of normal fetal liver. HBs that are composed exclusively of fetal pattern are referred to as pure fetal HB . By definition, the entire tumour is composed of uniform cells, and mitoses are <2 in 10 high power fields. In the crowded fetal subtype , the tumour cells are more closely packed, have higher nuclear/cytoplasmic ratios, decreased cytoplasmic glycogen, more eosinophilic cytoplasm and higher mitotic activity (≥2 in 10 high power fields) compared to the fetal pattern ( Fig. 13.23 ).
The fetal pattern is mixed with the embryonal pattern in 20% of cases, and the crowded fetal pattern may merge into the embryonal pattern (see Fig. 13.23 ). The embryonal pattern corresponds to the sixth to eighth week of the embryonic liver and consists of sheets, clusters or single small, angulated, hyperchromatic cells with a high nuclear/cytoplasmic ratio, prominent nucleolus and frequent mitoses. Embryonal cells often cluster into glandular, acinar or pseudorosette formations. Extramedullary haematopoiesis is often more prominent than the fetal pattern. The macrotrabecular pattern is seen in about 3% of cases and is composed of trabeculae >10 cells in thickness with either fetal or embryonal type cells ( Fig. 13.24 ). However, the tumour cells may rarely be larger and closely resemble HCC. In most cases, this pattern is mixed with other patterns of HB. The small-cell pattern is seen in about 3% of cases and is composed of noncohesive sheets of small cells similar to those of other ‘small blue cell’ paediatric neoplasms. The mitotic rate is often high, but some tumours do not show significant mitoses. Rhabdoid phenotype with eccentric nucleus and eosinophilic cytoplasmic globules can be seen. This pattern has been variably called small-cell anaplastic or small-cell undifferentiated HB. A tumour should contain at least 70% small-cell component to be considered a pure small-cell HB. Smaller components of this morphology should also be recognized because of a potential for adverse outcome.
Ductular differentiation can be seen in the epithelial variant as in fetal HB, often at the periphery of the tumour. When ductular differentiation is prominent, the tumour is referred to as cholangioblastic HB. Ductular reaction can occur at the periphery of the tumour, especially after chemotherapy, and should not be considered as the cholangioblastic variant.
The mixed epithelial-mesenchymal pattern contains areas of fetal and/or embryonal epithelial cells along with neoplastic mesenchymal elements ( Fig. 13.25 ). The mesenchymal component consists of immature and mature fibrous tissue, osteoid-like tissue and in some cases, hyaline cartilage. The cells in the mesenchyme and osteoid-like areas share staining characteristics with the epithelial component, being positive for keratin and showing nuclear β-catenin staining, indicating that they are metaplastic epithelial cells rather than truly of mesenchymal origin.
Teratoid HB (HB with heterologous elements) refers to mixed HB with heterologous elements such as neural/neuroectodermal differentiation represented by mature brain, primitive neuroepithelial components forming tubules and rosettes, as well as melanin and retinal pigment. Rhabdomyosarcomatous elements can be present. Squamous and mucinous glands may be present, but these elements can also be seen without a neural component in epithelial HB. Rarely, glandular elements with subnuclear and supranuclear vacuoles reminiscent of yolk sac tumour can be present, but these are interspersed with other areas of typical HB. These tumours are not related to teratomas, and representations from all three germ layers typical of teratoma are not seen.
After chemotherapy, HB may show extensive necrosis and haemorrhage with nodules of fibrous tissue, and there may be no residual tumour. Osteoid-like tissue is frequently present and may be the only evidence of viable tumour. Viable epithelial areas may be seen grossly as pale friable areas within fibrotic nodules, while the mixed elements appear as tiny white areas of osteoid. The amount of residual tumour is variable, and it can be difficult to differentiate fetal pattern from entrapped benign hepatocytes. The histological subtype may be difficult to identify because of secondary changes such as marked fat accumulation and bizarre nuclei. Squamous epithelium with keratinization and foreign body giant cell reaction is often seen in post-therapy specimens.
Both fetal and embryonal elements in HBs express pancytokeratin, and hepatocellular markers such as hepatocyte paraffin 1 (HepPar 1), polyclonal carcinoembryonic antigen (pCEA) and AFP. GPC3 is positive in >90% of cases. Nuclear β-catenin is seen in the embryonal component but is absent or rare in most fetal HBs. However, rare cases of fetal HB may show prominent β-catenin nuclear staining. GPC3 staining in the fetal areas is slightly more intense in the eosinophilic cells compared with clear cells. GS shows diffuse cytoplasmic staining in both embryonal and fetal areas. The staining pattern in the crowded fetal pattern is similar to the fetal pattern, but GPC3 and GS staining can be stronger. SALL4, a marker of germ cell tumours, can be positive in embryonal HB. The ductular elements can be highlighted by cytokeratins 7 and 19. The tumour cells in small-cell variant are negative for HepPar 1, AFP, GS and GPC3. The tumour cells express cytokeratin and vimentin; INI1 may be lost in some tumours, especially those with rhabdoid phenotype. Cytokeratin 19 can be positive in scattered tumour cells. Strong and diffuse nuclear staining with β-catenin is uniformly present.
FNA cytology findings in epithelial HB depend on whether differentiated (fetal, embryonal and/or macrotrabecular subtypes) or undifferentiated elements are aspirated. Fetal elements resemble immature hepatocytes ( Fig. 13.26 ). Neoplastic cells are slightly smaller and have ample cytoplasm with small, central round nucleus, finely granular chromatin and normal nuclear/cytoplasmic ratio. Pleomorphism and mitoses are not seen. Embryonal subtype is characterized by cells with scant cytoplasm, angulated nucleoli, coarsely granular chromatin, high nuclear/cytoplasmic ratio and mitoses. The cells are usually arranged in acinar groupings, in a papillary pattern or in sheets. Extramedullary haematopoiesis may be discernible. Differentiated epithelial HB can resemble HCC cytologically. Cytological subtyping can be difficult without resorting to ancillary tests, including ultrastructural studies. The undifferentiated type consists of anaplastic, small round cells and is one of the differential diagnoses of small round cell tumours in childhood. In the mixed epithelial-mesenchymal type, mesenchymal tissue is rarely seen in FNA cytology.
The fetal pattern can be difficult to distinguish from normal infantile liver with extramedullary haematopoiesis. The presence of lobular configuration, portal tracts, central veins and a low nuclear/cytoplasmic ratio points indicate normal liver. Unlike HB, normal liver is negative for GPC3 and nuclear β-catenin, while GS is confined to the pericentral region. Alternating light and dark cell zones typical of fetal HB are not seen.
HCC can be mistaken for HB, especially with fetal and macrotrabecular patterns. HCC typically demonstrates more pronounced nuclear pleomorphism, intranuclear inclusions, atypical mitoses and globules of α1-antitrypsin. The presence of embryonal areas or mesenchymal elements helps in the diagnosis of HB. Compared to HCCs, HBs are more likely to show nuclear β-catenin and diffuse GS staining (70% versus 20%) as well as GPC3 staining (>90% versus <50% in well-differentiated HCC).
Small round cell tumours
The embryonal component and especially the small-cell subtype of HB can mimic metastatic Wilms tumour, and neural tumours such as neuroblastoma or primitive neuroectodermal tumour (PNET). The presence of admixed fetal areas, diffuse GPC3, diffuse GS and nuclear β-catenin staining and the absence of a renal mass favour embryonal HB. Neuroblastoma should be suspected in the presence of an adrenal mass, positive staining for PGP 9.5 and absence of GPC3 and nuclear β-catenin staining. PNETs can rarely present as primary liver tumours; CD99 positivity and fluorescence in situ hybridization (FISH) for the EWS break-apart probe can confirm the diagnosis. Malignant rhabdoid tumour (MRT) can occur in the liver and is characterized by abundant cytoplasm, variably prominent nucleoli and eosinophilic cytoplasmic globules. This is a highly aggressive tumour that is positive for cytokeratin, vimentin and epithelial membrane antigen; loss of nuclear staining with INI1 staining is a typical feature of MRT. Some small-cell HBs also show diffuse loss of INI expression and deletion of the INI1/BAF47 gene on chromosome 22, which is the same cytogenetic abnormality seen in MRT. These HBs often do not respond to conventional HB therapy and may benefit if treated as MRT.
Spindle cell tumour
Spindle cell tumours such as embryonal rhabdomyosarcoma can enter the differential diagnosis in HB with mesenchymal elements. Involvement of biliary tree rather than the liver parenchyma and staining positivity for desmin, muscle-specific actin and myogenin are typical of embryonal rhabdomyosarcoma. The presence of epithelial areas or other mesenchymal components (e.g. osteoid, bone, cartilage) suggests HB.
Calcifying nested stromal-epithelial tumour
Calcifying nested stromal-epithelial tumour of the liver is an extremely rare tumour that is composed of a spindle cell component along with epithelial, ductular and osteoid components; thus it could be confused with HB ( Fig. 13.27 ). This tumour affects children and young adults, some of whom may present with Cushing syndrome. Cytokeratin, vimentin and hormones (e.g. ACTH) are often positive, whereas GPC3 and nuclear β-catenin are negative.
Germ cell tumour
HB with teratoid features can mimic a teratoma. The hepatocellular component in teratomas usually resembles mature fetal liver, in contrast to a mix of fetal and embryonal patterns often seen in the epithelial component of mixed HB. It stains with GPC3 in a pattern similar to fetal HB. Typical features of yolk sac tumour, such as reticular growth pattern, marked myxoid stroma and ‘glomeruloid’ elements, are not observed in HB. GPC3 can be positive in the hepatocellular component of teratoma and in yolk sac tumours, but nuclear β-catenin would suggest HB. The teratomas have almost all been in infants, and the yolk sac tumours have been described in both children and adults.
Transitional liver cell tumour
This rare paediatric liver tumour presents in children and adolescents with clinical and histopathological features reminiscent of HCC. Molecular studies have shown genetic characteristics of both HB and HCC, characterized by a high number of somatic mutations, including genes such as CTTNB1 , RAD17, MSH6, TP53 and TERT promoter. Serum AFP is typically high. These tumours have morphological features intermediate between hepatoblasts and hepatocytes. Unlike the trabecular pattern seen in HCC, the tumour cells exhibit a diffuse pattern and can have a wide spectrum of appearances ranging from small to hepatoid cell. Multinucleated tumour giant cells can be present. As with HB, nuclear staining with β-catenin is present. These tumours are aggressive and typically do not respond to chemotherapy.
The most widely used staging systems are is that of the Children’s Oncology Group (COG) and PRETEXT schemes. The COG scheme classifies the tumours postoperatively by their resectability into four groups: stage I (complete resection), stage II (microscopic residual disease), stage III (macroscopic residual tumour) and stage IV (distant metastases). At presentation, approximately 38% of HBs are stage I, 9% stage II, 24% stage III and 29% stage IV. The majority of the 53% of unresectable (stages III and IV) cases can be rendered resectable with preoperative chemotherapy. Liver transplantation (LT) is an option for unresectable disease.
Developed by the International Childhood Liver Tumor Strategy Group (SIOPEL), the PRETEXT staging system is based on segmental involvement of the liver and is used to determine tumour extension before therapy: stage I (1 section involved, 3 adjoining sections are free), stage II (1 or 2 sections involved, 2 adjoining sections are free), stage III (2 or 3 sections involved, no two adjoining sections are free) and stage IV (all 4 sections involved). The sections in the PRETEXT staging are defined by grouping of liver segments: segments II and III (left lateral section), segment IV (left medial section), segments V and VIII (right anterior section) and segments VI and VII (right posterior section). In addition, the PRETEXT system also includes assessment of tumour involvement of inferior vena cava (IVC), hepatic veins (V), portal veins (P), caudate lobe, multifocality, rupture/intraperitoneal haemorrhage, lymph node involvement, extrahepatic abdominal disease (E) and distant metastases (M).
Treatment and prognosis
Surgery is the mainstay of treatment of HB, and complete resection is the only chance for cure. All patients with COG stage I tumours with pure fetal pattern are treated with surgery alone without chemotherapy because 5-year survival is almost 100%. The other histological variants as well as higher-stage tumours receive a combination of chemotherapy and surgical resection, the order determined by tumour resectability at presentation. For nonpure fetal histology, the disease-free survival is 97–100% for COG stage I and II, 70% for stage III and 40% for stage IV disease. Preoperative chemotherapy to shrink the tumour has resulted in improved resectability and improved the overall 5-year survival rate to 75% from 35% in the 1970s. Cisplatin/doxorubicin or cisplatin/5-FU/vincristine are the usual regimens. LT is an option for tumours that remain unresectable after chemotherapy.
The stage of the tumour at initial resection is the key prognostic factor in determining survival. Small-cell and macrotrabecular subtypes have an adverse outcome. The other histological patterns do not independently affect survival when adjusted for age, gender and stage. Other unfavorable prognostic factors include involvement of multiple lobes, vascular invasion, slow decline in AFP after therapy, aneuploidy, nuclear β-catenin staining, low p27 expression and high cyclin D1 expression.
Dysplastic lesions and small hepatocellular carcinoma
Large-cell change (originally ‘large-cell dysplasia’) refers to cellular enlargement with bizarre pleomorphic nuclei, hyperchromasia, prominent nucleoli and occasional multinucleation of hepatocytes ( Fig. 13.28 ). The cells have abundant cytoplasm, and thus the nuclear/cytoplasmic ratio is normal. This change can affect isolated or groups of hepatocytes and may occupy the entire cirrhotic nodule. Large-cell change is found in 20–30% of cirrhotic livers and in 60–70% of cirrhotic livers with HCC; it has been thought to convey an increased risk for development of HCC. However, hepatocytes with large-cell change have a low proliferation rate, greater apoptosis and increased markers of senescence, suggesting a derangement of the process of hepatocyte polyploidy, possibly caused by chronic inflammation-induced DNA damage or cholestasis. Other studies have found alterations in cell cycle, DNA damage and telomere shortening in large-cell change in hepatitis B cirrhosis, suggesting that large-cell change is a heterogeneous lesion and may be a precursor of HCC in certain patients.
Small-cell change (formerly ‘small-cell dysplasia’) is characterized by small hepatocytes with increased nuclear/cytoplasmic ratio, typically with cytoplasmic basophilia ( Fig. 13.29 ). When occurring in expansile nodules, the hepatocytes in small-cell change often have increased proliferative activity and molecular abnormalities, such as p53 overexpression, and are likely to be a precursor of HCC. In other patients, poorly defined or diffuse areas of small-cell change without nodular configuration may represent regenerative phenomenon or even ischaemic/degenerative change and are not preneoplastic. Small-cell regenerative foci are also common in chronic biliary disease.
The term dysplasia refers to abnormal hepatocellular proliferation that does not fulfill criteria of malignancy. A cluster of dysplastic hepatocytes <1 cm in diameter is referred to as dysplastic focus (see Fig. 13.29 ), whereas a cluster of dysplastic hepatocytes ≥1 cm in diameter is referred to as dysplastic nodule .
Large regenerative nodules (LRNs), also known as macroregenerative nodules and adenomatous hyperplasia, are by definition ≥1 cm. These typically arise in hepatitic cirrhosis ( Table 13.6 ) and are rare in biliary disease ( Fig. 13.30 ). LRNs can also occur in noncirrhotic liver in Budd–Chiari syndrome and portal vein thrombosis and frequently occur as regenerative nodules after necrosis, including necrosis from active and/or acute to subacute hepatitis, and in this setting, these nodules are not thought to be preneoplastic. On histology, LRN/macroregenerative nodules are usually composed of hepatocytes with normal or near-normal cytology, with plates one to two cells thick, although thicker plates can be present, particularly in the setting of regeneration following necrosis. Importantly, intact reticulin framework is present, as may areas of large-cell change ( Fig. 13.31 ). Ductular reaction may be prominent (see Fig. 13.30 ), and unpaired arterioles are rare or absent.
|Type of nodule||Cellular features||Reticulin framework||Other key features|
|LRN/MRN||Bland cytology, similar to normal liver or cirrhotic nodules with focal LCC||Intact reticulin framework similar to cirrhosis||Portal or portal-like zones; scattered LCC; iron, Mallory–Denk bodies, fat and bile may be present|
|LGDN||Bland cytology of uniform pattern||Similar to MRN||Similar to MRN|
|HGDN||Small-cell change, common (nuclear density <2× normal) *||Mostly intact reticulin, but focal fragmentation/loss may be seen; plate width more variable and acinar change more likely||No stromal invasion|
|WDHCC||Small-cell change most common (nuclear density >2× normal)||Multifocal loss, fragmentation or extensive irregularity of reticulin framework; thin trabeculae||Early HCC often shows fatty change and lacks ductular reaction at periphery; stromal invasion|
Low-grade dysplastic nodules (LGDNs) represent a clonal hepatocellular proliferation (a uniform population of hepatocytes) without features of high-grade dysplasia. LGDNs can be histologically indistinguishable from macroregenerative nodules (LRNs) in the absence of clonality studies ( Table 13.6 ). Both LRN and LGDN in cirrhosis may stand out from the background cirrhotic nodules by the presence of Mallory–Denk bodies (MDBs), bile, clear cell change, iron, copper or fat. In part, given that the histological diagnostic criteria overlap those of macroregenerative nodule, its likelihood of progression to HCC is not clearly defined.
High-grade dysplastic nodules (HGDNs) are characterized by hepatocellular proliferation with atypical cytological and/or architectural features that are not sufficient for an unequivocal diagnosis of HCC ( Table 13.6 ). By definition, stromal invasion is not seen. HGDNs typically show small-cell change and features that suggest increased cellular proliferation, such as plates more than two cells thick, cytoplasmic basophilia, high nuclear/cytoplasmic ratio, nuclear hyperchromasia or an irregular nuclear contour ( Fig. 13.32 ). These features may involve the entire nodules or may be confined to one or more foci within the nodule, giving the appearance of ‘nodule-in-nodule’ formation. Portal tracts are present within the lesion. Occasional unpaired arterioles, focal pseudoglandular/acinar architecture and focal reticulin loss can be seen, but these findings are not prominent. Steatosis and MDBs can be present, while iron is typically absent. IHC for CD34 shows patchy sinusoidal staining, usually at the edge; occasional nodules can show more diffuse expression. AFP is negative, and GPC3 expression is variable but typically absent or focal. HGDNs are generally considered high-risk lesions for HCC.
Fine-needle aspirates of dysplastic nodules are composed of a polymorphous cell population comprising hepatocytes occurring loosely or in one- to two-cell-thick cords, bile duct epithelium and ductular clusters. LGDNs contain hepatocytes often exhibiting large-cell change characterized by nuclear and cellular enlargement with normal nuclear/cytoplasmic ratio but minimal nuclear atypia ( Fig. 13.33 ). In contrast, HGDNs contain hepatocytes showing small-cell change consisting of small and monotonous cells with minimal nuclear atypia, a subtle increase in nuclear/cytoplasmic ratio and nuclear crowding ( Fig. 13.34 ). Distinguishing LGDN from LRN and HGDN from small early HCC can be challenging.
Small hepatocellular carcinoma refers to tumours up to 2 cm in diameter; these have a greater likelihood of long-term survival after surgery or other attempts at curative therapy. Similar to large regenerative and dysplastic nodules, small HCCs are larger than the surrounding cirrhotic nodules and may differ in colour (frequently green, yellow or mottled), or they may bulge from the cut surface. Small nodules are easily overlooked in explanted livers unless the specimen is thinly sectioned (at least 1 cm intervals) and examined in detail. Small HCCs can be divided into two subtypes: progressed and early HCC (early well-differentiated HCC, HCC with vague nodular pattern, HCC with indistinct borders). Progressed HCC has a capsule and histologically resembles classic HCC, whereas early HCC lacks a definite capsule and histologically can resemble HGDN. Early HCCs constitute 15−30% of small HCCs. Remnants of pre-existing early HCC in association with less well-differentiated HCC were reported in 74% of tumours <1 cm but in only 9% of tumours that were 4–5 cm, suggesting stepwise dedifferentiation as the mechanism of disease progression.
The differentiation of HGDN from early HCC is extremely difficult and may be impossible on a biopsy. This distinction has clinical implications, since patients with a definite diagnosis of HCC may be considered for resection or may receive priority for LT. On the other hand, most HGDNs are ablated, and these patients do not have priority for LT. The features that favour progressed HCC, such as uniformly thick cell plates, prominent pseudogland formation, multifocal loss of the reticulin framework and vascular invasion, are often absent in early HCC ( Table 13.6 ). Increased cellularity, increase in nuclear/cytoplasmic ratio and frequent pseudogland formation and fatty change are more common in early HCC than in HGDN, but these are not reliable for definite distinction in many cases. The feature that can best distinguish early HCC from HGDN is stromal invasion ( Fig. 13.35 ), which is characterized by invasion of abnormal hepatocytes into perinodular portal tracts, fibrous septa or adjacent parenchyma. Stromal invasion can also be intranodular, with invasion of portal tracts within the nodule. Stromal invasion is diagnostic of HCC but can be difficult to assess or sample on core biopsies.
Ancillary techniques that favour HCC include diffuse sinusoidal staining with CD34, nuclear staining with β-catenin, diffuse staining with GS, HSP70 overexpression and GPC3 positivity. Combined use of HSP70, GPC3 and GS has been recommended for this differential diagnosis, with high specificity for diagnosis of HCC when two of the three markers are positive. Areas of stromal invasion often lack ductular reaction at the interface between hepatocytes and portal tracts/fibrous septa, whereas this feature is present in LRN and most HGDNs. Use of cytokeratins 7 and 19 can be useful in the identification of stromal invasion by highlighting lack of ductular reaction.
Other, less common nodules can occur in cirrhotic liver. Lesions resembling FNH can occur in the cirrhotic liver and have been referred to as FNH-like lesions . In addition to typical morphological features of FNH, some of these lesions (but not all) show map-like staining with GS (see Table 13.5 ). In rare instances, nodules in cirrhosis show typical features of IHCA (see earlier), such as sinusoidal dilation, inflammation, dystrophic arterioles and strong staining with SAA/CRP. Mutations typical of IHCA may be present.
Hepatocellular carcinoma and variants
The clinical presentation of hepatocellular carcinoma (HCC) is related to effects of the mass or signs and symptoms of underlying chronic liver disease, including upper abdominal pain, weight loss, abdominal enlargement and hepatomegaly with or without a palpable mass. Signs of decompensated liver disease, such as jaundice, ascites, variceal haemorrhage, hepatic encephalopathy or obstructive jaundice, can be present. Rarely, patients may present with distant metastases or paraneoplastic syndromes such as hypoglycaemia, erythrocytosis, hypercholesterolaemia, hypercalcaemia, isosexual precocious puberty (in children), gynaecomastia (in the absence of cirrhosis), carcinoid syndrome, hypertrophic pulmonary osteoarthropathy, osteoporosis, hypertension, hyperthyroidism, dysfibrinogenaemias, porphyria cutanea tarda and a variety of other cutaneous changes.
High levels of serum alpha fetoprotein (AFP), an oncofetal antigen, can help in the screening, diagnosis and follow-up of HCC. The utility of AFP is limited by low sensitivity and specificity. Many other serum biomarkers, such as des-γ-carboxy prothrombin, are becoming available, but their usefulness remains to be proved. Serum AFP level and US of the liver are recommended every 6 months for screening of patients at risk of developing HCC.
CT and MRI are the most common modalities used for radiographic diagnosis of HCC. The criteria proposed by the American Association for the Study of Liver Diseases (AASLD) state that for tumours >1 cm in cirrhotic liver, biopsy is not necessary to confirm the diagnosis if classical features of HCC are seen on multiphasic contrast-enhanced CT or MRI. The classic features are enhancement of the lesion in the arterial phase and washout in the venous phase. Atypical imaging features are common in small (typically <2 cm) and well-differentiated tumours. Additional imaging (CT or MRI) is recommended for these patients, and biopsy is often required. Functional imaging of HCC using hepatocyte-specific gadolinium chelate agents is increasingly used and may increase sensitivity and specificity for HCC diagnosis compared to multiphasic MRI or CT. Of note, small tumours are increasingly being detected by surveillance in high-risk patients.
A system of standardized reporting of CT and MRI findings for HCC diagnosis in high-risk patients is called the Liver Imaging Reporting and Data System (LI-RADS). This system combines arterial enhancement with tumour size, venous washout, presence of capsule and growth compared to prior imaging to yield five diagnostic categories: LR-1 (definitely benign), LR-2 (probably benign), LR-3 (moderate probability of benign or malignant), LR-4 (probably malignant) and LR-5 (definitely malignant).
Epidemiology and aetiology
One of the most striking features of HCC is the wide variation in its incidence in different parts of the world. Overall, HCC represents the third leading cause of cancer-related mortality worldwide. It is the fifth most common cancer in men and the seventh in women worldwide, but East Asia and sub-Saharan Africa have by far the greatest number of cases, and in the countries of those regions, HCC is among the leading causes of death. In areas of low incidence, such as the United States, carcinoma of the liver (primarily HCC) accounts for only 2.3% of cancer deaths in recent years, with an annual incidence of approximately 4.1 cases per 100,000 population. However, the incidence of HCC in the United States has more than doubled over the past two decades and is anticipated to continue increasing over the next 20 years because of the growing number of patients with advanced hepatitis C virus (HCV) infection and NASH. At its current pace, HCC is projected to surpass breast and colorectal cancers to become the third leading cause of cancer-related death in the United States by 2030. In general, regions of the world with a high incidence of HCC are those that have a high prevalence of chronic hepatitis B virus (HBV) infection, but even within these areas there is geographical variability, e.g. in western Africa the country of Gambia has nearly five times the incidence of Nigeria, suggesting that environmental cofactors such as aflatoxin exposure may also be important. In contrast to the rising incidence in the United States, the incidence of HCC has been declining in some high-incidence areas, such as China and Hong Kong, partly related to HBV vaccination of children and reduction of aflatoxin exposure in grains. HCV infection is associated with many cases in countries such as Japan, where the prevalence of hepatitis B is intermediate but the incidence of HCC relatively high.
The incidence of HCC generally increases with age, although geographical differences exist. In Europe and the United States the peak age-specific incidence is in the seventh decade, while in Qidong province in China where the incidence is the highest in the world, the peak is in the fifth decade. In South Africa the average age of patients with HCC is 35 years, and 40% are 30 or younger, whereas in Taiwan (another area of high incidence), the majority of patients are 40–60 years old, with a peak incidence in the eighth decade. Nevertheless, HCC can occur in younger individuals and even young children. Regardless of geographical location, HCC occurs more frequently in men than women, with male/female ratios in various countries ranging from 2 : 1–5 : 1. The precise reason is not known, but it has been shown that many tumours have androgen receptors, suggesting that androgens may promote tumour development and growth. There is also a male predominance in risk factors, such as chronic viral hepatitis, alcoholism and smoking, which undoubtedly also play a role. Other authors, however, have attributed this gender bias to female sex hormones, because HCC incidence greatly increases in postmenopausal women who do not take hormone replacement therapy. Rare HCCs arise in association with HCA (see previous section).
The majority of patients who develop HCC have underlying cirrhosis, and consequently, cirrhosis is a major risk factor for HCC. Macronodular cirrhosis has been more strongly associated with HCC than micronodular cirrhosis, but cirrhosis of virtually any cause may be complicated by the development of HCC. Cirrhosis has geographical variation, so that in Japan and other Asian countries, approximately 90% of patients have underlying cirrhosis. In contrast, in some parts of Africa, only 50–63% of patients are cirrhotic, although most have precirrhotic forms of chronic viral hepatitis. Series from North America have reported 46–91% with cirrhosis. Nevertheless, every series includes some patients without cirrhosis and some with no evidence of any underlying liver disease.
The majority of cases of HCC in the world are caused by HBV, with the number of HCV-associated cases increasing in the Western world. HBV and HCV are the main causal agents of chronic hepatitis. Approximately 5–10% of HBV and 85% of HCV infections become chronic. The epidemiological association of chronic HBV or HCV infection with HCC is well established.
Globally, up to 80% of HCC is attributable to HBV or HCV. The risk of HCC is increased 5- to 15-fold in chronic HBV carriers and up to 17-fold in HCV-infected patients. The two viruses are different, and the precise mechanisms by which they can cause cancer are not well clarified. Besides a direct effect of the virus on the genome, HCC could also develop indirectly from the inflammation-necrosis-regeneration cycle that occurs in the setting of chronic hepatitis, suggesting that the pathogenesis of HCC is immune mediated.
HBV- and HCV-encoded proteins alter host gene expression and cellular phenotypes that are recognized as hallmarks of cancer. These changes promote growth factor-independent proliferation, resistance to growth inhibition, tissue invasion and metastasis, angiogenesis, reprogramming of energy metabolism and resistance to apoptosis in the face of persistent immune attack and during therapeutic intervention.
Of all risk factors, HBV infection has the strongest association with the development of HCC. The relative risk of HCC in patients serologically positive for hepatitis B surface antigen (HBsAg) is 98 times that of patients who are negative. Those who are also positive for e antigen (HBeAg), indicating active viral replication, have 3.6 times the risk of those who are HBsAg positive but HBeAg negative, suggesting that the activity of the disease plays a role in pathogenesis. Even patients with occult HBV infection (HBsAg negative with HBV DNA demonstrable in tissue by PCR) may develop HCC, although the incidence and relative risk are not known.
HBV is the prototypic member of the family Hepadnaviridae and consists of a partially double-stranded DNA genome of 3.2 kilobases (kb) enclosed by envelope proteins (HBsAg) (see Chapter 6 ). The genome is packaged with a core protein (HBcAg) and a DNA polymerase. After penetration of the virus in the cell, its genome becomes a covalently closed, totally double-stranded molecule that can eventually integrate into the host genome. Covalently closed circular DNA serves as a template for transcription of viral RNA, which is translated into viral proteins. The HBV polymerase, which reverse-transcribes and replicates HBV DNA, lacks proofreading ability and is therefore prone to generate mutations. Common mutations include the precore ( G1896A ) mutation, basal core promoter mutations ( A1762T/G1764A ) and deletion mutations of pre-S/S genes.
Integration of HBV in sites within the host genome has been seen as a possible carcinogenetic mechanism. It is present in >85–90% of HBV-related HCC and usually precedes the development of HCC. HBV integration induces a wide range of genetic alterations within the host genome, including chromosomal deletions, translocations, production of fusion transcripts, amplification of cellular DNA and generalized genomic instability. Many integrated events occur near or within fragile sites or other cancer-associated regions of the human genome that are prone to instability in tumour development and progression. Genetic instability associated with integration may alter the expression of oncogenes, tumour suppressor genes and microRNAs. A recent large-scale analysis of HBV DNA integration sites in cellular DNA found a preference at sites regulating cell signalling, proliferation and viability. Common gene targets of integration include the human cyclin A2 gene, the retinoic acid receptor gene, human telomerase reverse transcriptase, PDGF receptor, calcium signalling-related genes, mixed-lineage leukaemia and 60S ribosomal protein genes. A large proportion of HCCs have integrated HBV sequences encoding HBV X ( HBx ) and/or truncated-envelope pre-S2/S proteins, which both contribute to hepatocarcinogenesis.
Several HBV genes have been found in infected tissues more frequently than others, including truncated pre-S2/S, hepatitis B X gene ( HBx ) and a novel spliced transcript of HBV named hepatitis B spliced protein (HBSP). The proteins expressed from these integrated genes have been shown to have intracellular effects that may account for their association with HCC, including effects on cellular growth and apoptosis. Among these genes, HBx seems to play a pivotal role in hepatocarcinogenesis. HBx gene, indeed, harbours weak transcriptional transactivation activity. The 154-amino acid viral product ‘X’ ( HBx ) has been shown to be essential for HBV and woodchuck hepatitis virus (WHV) infection in vivo . It has been considered to be a prime candidate for mediating HBV pathological effects. With regard to oncogenesis, HBx can directly inactivate the tumour suppressor p53 and the negative growth regulator p55, both involved in the pathway of senescence, and can transcriptionally downregulate p21 and Sui1, both of which inhibit hepatocellular growth.
Hepatitis C rivals hepatitis B in importance as an aetiological factor in HCC, and it appears to be largely responsible for the rising incidence of liver cancer in the United States, where 21% of cases are associated with HCV. Hepatitis C is also believed to be responsible for the relatively high incidence of HCC in Japan and some other countries. Most cases of HCC associated with hepatitis C have occurred after the development of cirrhosis, and the risk is multiplied in the presence of other factors, such as male gender, advanced age, co-infection with other viruses (HBV, HIV), alcohol, diabetes and steatosis. Patients infected with both HBV and HCV have a much greater risk of HCC than those with either virus alone.
HCV is a positive-strand RNA virus belonging to the genus Hepacivirus (family Flaviviridae) (see Chapter 6 ). The genome is a 9-kb-long, single-stranded, linear RNA. Its replication occurs in the rough endoplasmic reticulum (RER), without reverse-transcriptase activity ; thus it is unable to integrate into the host genome. Instead, viral proteins and their evoked host responses contribute mostly to the viral oncogenic processes.
The mechanism of liver damage in HCV infection is the result of immune response and direct cytopathic effect. The continuous process of necrosis and regeneration may make the cells prone to the action of various procarcinogenic substances (possibly HCV proteins), with genetic instability and subsequent malignant transformation. Based largely on experimental in vitro results, an oncogenic role has been reported for three HCV proteins: core protein, NS3 and NS5A. Core protein intervenes in several cellular functions as apoptosis, signal transduction and transformation. It modulates p21WAF1 expression and physically interacts with p53, promoting apoptosis and cell proliferation. Moreover, core protein stimulates nuclear factor (NF) κB, increasing its levels and DNA binding activity and modifying the response to tumor necrosis factor (TNF) α. Microarray analysis has revealed threefold or more transcriptional changes in 372 of 12,500 known human genes in core-expressing Huh-7 cells, with most genes involved in cell growth or oncogenic signalling. Of particular interest is the marked upregulation of both Wnt-1 and its downstream target gene WISP2 . Small-interfering RNA against Wnt-1 blunted growth stimulation by HCV core, and conditioned medium from Wnt-1 transfected cells accelerated cell growth. HCV core downregulates E-cadherin expression at the transcriptional level. This effect is strongly correlated with hypermethylation of CpG islands of the E-cadherin promoter through concerted action of both DNA methyltransferase (DNMT) 1 and 3b and is abolished by a specific inhibitor of DNMT. HCV core also upregulates the expression of transforming growth factor (TGF) β. As HCV-infected livers progress from chronic hepatitis through cirrhosis to HCC, hepatocytic pSmad3L/PAI-1 increases with fibrotic stage and necroinflammatory grade, and pSmad3C/ p21WAF1 decreases. These results indicate that chronic inflammation associated with HCV infection shifts hepatocytic TGFβ signalling from tumour suppression to fibrogenesis, accelerating liver fibrosis and increasing the risk of HCC.
In summary, although HBV and HCV are the major risk factors leading to the development of HCC, the precise pathogenetic mechanisms linking viral infection and HCC remain uncertain. A number of transgenic mouse model studies have found that the expression of various viral proteins either alone or in combination lead to the development of HCC. Viral proteins also have been implicated in disrupting several cellular signal transduction pathways that affect cell survival, proliferation, migration and transformation. HBV and HCV also appear to have distinct pathways to cancer.
Aflatoxins, a family of mycotoxins produced by fungi of the Aspergillus genus, are powerful carcinogens in experimental animals. Contamination of food, particularly grains and peanuts, by these toxins is common in the very same parts of the world where HCC is most common, namely China and southern Africa, and indeed, before the association with hepatitis B was recognized, aflatoxin B1 was thought to be a principal cause of HCC. Aflatoxin is converted by cytochrome P-450 (CYP) to 8,9-exo-epoxide, which damages DNA. Aflatoxin B1 (AFB1) binding to guanine residues of DNA can produce G-to-T mutations that are not found in human HCC in areas with low AFB1 exposure. Such mutations in codon 249 of the p53 tumour suppressor gene are thought to play an important role in carcinogenesis in parts of the world with a high incidence of HCC. There is a synergistic cooperation between HBV and AFB1 in hepatocarcinogenesis, and the combination of chronic hepatitis B with dietary aflatoxin exposure more than triples the risk of developing HCC. Furthermore, aflatoxin exposure may also be associated with advanced liver disease in patients with chronic hepatitis C.
Minimal evidence indicates that alcohol is directly carcinogenic or genotoxic, but alcoholic cirrhosis is the most common risk factor in the United States and many Western countries. It has been suggested that alcohol may act both as a primary cause of HCC and as a cofactor together with viral infection, but nongenotoxic mechanisms are most likely related to causation. Alcohol is metabolized to acetaldehyde by alcohol dehydrogenase (ADH) and CYP2E1, which is associated with microsomes. CYP2E1 produces reactive oxygen species (ROS) that can lead to DNA damage and transformation. Moreover, polymorphism in ADH and CYP2E1 may influence the risk of hepatocarcinogenesis. In alcohol-induced cirrhosis, the reserve of S -adenosylmethionine, the methyl donor for DNA methylation, is decreased. The effect could be hypomethylation of DNA, with consequent DNA instability, which could be associated with the development of cancer. Furthermore, a number of studies have suggested the importance of the overall intrahepatic and systemic events following alcohol consumption. Alcohol may be synergistic with other risk factors, increasing the likelihood of HCC in patients with hepatitis B or C, diabetes, obesity or smoking. Heavy alcohol consumption is associated with a sixfold increase in risk of HCC in cirrhotic patients.
Diabetes, obesity, metabolic syndrome and fatty liver disease
The constellation of problems associated with insulin resistance, including type 2 diabetes, obesity, hypertension and dyslipidaemia, has come to be known as the metabolic syndrome (see Chapter 5 ), and an increasing proportion of patients with HCC have associated features of metabolic syndrome. Diabetes mellitus is associated with a two- to threefold increase in risk of HCC, regardless of other risk factors, and obesity is associated with a two- to fourfold increased risk. Most patients with diabetes or obesity who develop HCC have cirrhosis, and since both diabetes and obesity can lead to cirrhosis through NASH, this presumably accounts for many of the cases. However, many also have occurred in non-cirrhotic patients with metabolic syndrome, suggesting that alternative pathways of hepatocarcinogenesis may be common in this setting. When viral hepatitis, alcohol or other risk factors are present, the fatty liver disease may be synergistic. Approximately one-third of HCC cases in the United States have no known aetiology, and non-alcoholic fatty liver disease (NAFLD)-related cirrhosis may be a predisposing factor in many of these. This hypothesis is supported not only by epidemiological data, but also by the suspected pathogenesis and molecular ‘lesions’ of NAFLD. Lipid peroxidation with production of free oxygen radicals is thought to be a central event in the pathogenesis of steatohepatitis. This oxidative stress may lead to proliferation of oval cells in humans as in experimental models. The ROS can induce mutations, for instance, in p53 gene. The common soil of insulin resistance and steatosis favors liver carcinogenesis by promoting adipose tissue-derived inflammation, hormonal changes, oxidative stress and lipotoxicity and stimulation of insulin-like growth factor (IGF)1 axis by hyperinsulinemia. The progression of neoplastic cells can also be facilitated by numerous alterations in growth factors, cytokines and activation of apoptosis. Other mechanisms involving diet, gut microbiome and genetic factors are increasingly important and clinically relevant.
Inherited metabolic diseases
Cases of HCC have occurred in a number of inheritable metabolic diseases, but the strongest associations are with hereditary haemochromatosis, tyrosinaemia and α1-antitrypsin (AAT) deficiency. Development of HCC is a frequent terminal event following the development of cirrhosis in hereditary haemochromatosis (~20%), hereditary tyrosinaemia (37%) and AAT deficiency (15%). Prevention of cirrhosis by phlebotomy in haemochromatosis and diet with NTBC (2-(nitro-4-trifluoromethylbenzoyl)-1,3-cyclohexanedione) therapy in tyrosinaemia can prevent HCC in all but a few cases. There is also an increased incidence of HCC in patients with porphyria cutanea tarda, but since that disease is frequently associated with alcohol use and chronic HCV infection, the role of the porphyria is unclear. A few cases of HCC have been reported in patients with glycogen storage disease (GSD) types I and III, in association with HCAs and presumably representing malignant degeneration of an adenoma. HCC has also rarely been reported in patients with other forms of porphyria, hypercitrullinaemia, fructosaemia, Wilson disease, Byler disease and Alagille syndrome.
Because of its relative frequency, hereditary haemochromatosis (HHC) is the metabolic disease that is the most important risk factor in development of HCC, and conversely, HCC is recognized as one of the most important complications of HHC. The risk of developing HCC in HHC individuals is significantly higher (up to 200 times) than the risk in the general population. The mortality rate from HCC in patients with HHC has been estimated at 8%. It is not completely clear if this is caused by the iron overload itself or the cirrhosis that frequently develops in untreated HHC patients. The prevalence of tumour formation in these patients is appreciably higher (18.5%) than in those with HHC in general (10.7%). However, HCC may also occur, although rarely, in HHC patients without cirrhosis, suggesting that hepatic iron per se might be directly hepatocarcinogenetic, in addition to its indirect effect through the development of cirrhosis. An important role of iron overload in hepatocarcinogenesis, however, has been also demonstrated in patients with iron excess without HHC as thalassaemia or so-called African iron overload. Several studies ( in vivo and in vitro ) point to a causative role for iron excess in hepatocarcinogenesis. Iron can act directly on cellular proliferation, as by inactivating p53. It can also act indirectly, for example, by influencing lipid peroxidation and production of reactive oxygen radicals. Iron is in fact a substrate for cell proliferation and could initiate the process of carcinogenesis. The effect of free iron could be caused by both direct damage to chromosomes and mutation through the increased production of ROS. Interestingly, in patients with HHC a particular type of mutation of p53 has been described ( A220G ) in the DNA-binding site of the protein. At the same time, it has been hypothesized that the formations of adduct due to lipid peroxidation may lead to DNA damage in patients with HHC. Evidence also suggests that excess hepatic iron induces immunological abnormalities that may decrease immune surveillance for malignant transformation.
Anabolic and contraceptive steroids
A number of hepatocellular tumours have been reported in patients taking anabolic steroids. Some of these have been classified as HCA and others as HCC, based on histological features. However, even those called HCC regress when the drug is discontinued, and none has been reported to metastasize. Contraceptive steroids are clearly associated with HCA. Several case-control series have reported an association with HCC, but others have found a protective effect of contraceptive steroid use as well as other factors that lead to high oestrogen exposure, including multiple pregnancies and early menarche. A meta-analysis of published studies concluded that there was insufficient evidence to establish a relation between OCs and the risk of HCC.
Molecular pathology of HCC is also discussed in Chapter 2 , so this section provides only a brief overview of molecular classifications of HCC with possible clinicopathological implications ( Table 13.7 ). As noted previously, HCC is a multistage and a heterogeneous disease linked to environmental, dietary/lifestyle factors, hereditary disorders and chronic liver disease or cirrhosis. In a small proportion of cases, HCC is observed in absence of liver disease. These facts, along with the different aetiologies responsible for liver damage (e.g. viral hepatitis, alcohol, iron overload, other causes of cirrhosis), account for the high molecular variability of HCC. During the last decade, several genomic alterations have been reported, including high-level DNA amplifications in chromosome 6p21 ( VEGFA ), observed in 6.7% of HCCs, and 11q13 ( FGF19/CNND1 ), as well as homozygous deletions in chromosome 9 ( CDNK2 ). Interestingly, improved survival of sorafenib-treated patients with VEGFA -amplified HCCs has been reported. Despite these research efforts, however, no molecular classification has yet been introduced in clinical practice.
|Aggressive HCC||Less aggressive HCC|
|DNA mutations||TP53||Exon 3 of CTNNB1|
|Signalling pathways||TGFβ, MET, canonical Wnt, E2F1||Notch, IGF2, AFP||Liver-specific Wnt|
|Cellular phenotype||Hepatoblast-like |
Cancer stem cell
|Histological phenotype||Less differentiated, larger tumours||More differentiated, smaller tumours|
|Steatohepatitic HCC||Clear cell |
Macrotrabecular/compact Vascular invasion
|Clinical features and outcome||High recurrence, poor survival, hepatitis B virus||Hepatitis C virus, alcohol, low recurrence, good survival|
|Prognostic gene signature||5-gene signature|
|Tumour marker (serum, IHC)||BIRC5||AFP, GPC3||GLUL, LGR5|
Several studies have attempted to establish molecular classifications of HCCs based on global gene expression profiles. Such classifications seek to group HCCs based on specific dysregulation of a limited set of biological pathways, allowing personalized treatment decisions targeting the ‘driver’ molecules responsible for cancer growth in each particular case. Four main molecular classifications of HCC have been published to date, reporting two, three, five or six discrete HCC subgroups, based on gene expression profiling in tumour samples. Recently, a transcriptome meta-analysis involving several hundred HCC tumours depicted a molecular classification onto which the subclasses, gene signatures and somatic DNA alterations can be mapped. According to this large meta-analysis, HCCs can be classified into two main subgroups ( Table 13.7 ). The first group is characterized by more aggressive biological and clinical features, showing increased genetic instability, cellular proliferation, ubiquitination, activation of prosurvival signals such as E2F1 and MET pathways, impairment of tumour suppressor TP53, gene signature of KRT19, larger and less differentiated tumour, higher incidence of tumour recurrence and poorer prognosis. The remaining tumours are characterized by less aggressive features, including preserved hepatocyte function, smaller and more differentiated tumour and better prognosis. The aggressive tumours are further subdivided into two groups: S1 subclass, characterized by relatively higher activation of TGFβ pathway and hepatocyte/cholangioma-like gene signature, and S2 subclass, characterized by positivity of stemness markers, epithelial cell adhesion molecule (EpCAM), high AFP and GPC3, activation of IGF2 pathway, relative suppression of interferon target genes and hepatoblastoma-like gene signature. The subset of less aggressive HCC tumours, S3 subclass, is characterized by somatic mutations accumulated in exon 3 of CTNNB1 (β-catenin gene) accompanied by induction of specific target genes, GLUL, LGR5 and SLC1A2, but not with canonical Wnt pathway target genes. Of note, slightly different sets of canonical Wnt pathway target genes are more preferentially induced in the S1 or S2 subclass, suggesting biological context-specific induction of WNT target genes in HCC. Interestingly, a recent transcriptome analysis has developed a molecular signature, based on combined expression level of five genes: HN1, RAN, RAMP3, KRT19 and TAF9. This molecular five-gene score is associated with outcomes of patients with HCC treated by resection in different clinical settings worldwide, independent of other clinicopathological features of HCC.
The tumour may form a single mass that replaces an entire lobe, or there may be multiple scattered discrete nodules. Numerous cirrhosis-like nodules can occur (‘cirrhosis-like HCC’) and can be missed on imaging and gross examination ( Fig. 13.36 ). Some tumours form an expanding mass well demarcated from the surrounding liver, with or without a capsule, whereas others appear to infiltrate the surrounding liver tissue at the tumour margin. Encapsulated tumours usually arise in a cirrhotic liver, with the expanding tumour causing compression atrophy of the surrounding cirrhotic nodules and incorporation of cirrhotic scars to form the fibrous capsule. Most HCCs are soft, often with areas of necrosis, with the exception of fibrolamellar and scirrhous variants, which can be grey-white and firm due to abundant stroma. HCCs can be variably tan or yellow, and green if they produce bile. Vascular invasion is common and the portal vein, hepatic veins as well as the vena cava may be involved. Invasion of major bile ducts is uncommon but may cause biliary obstruction.
Microscopically, HCC can have a highly variable appearance from tumour to tumour, even in different lesions within the same liver, as well as within the same lesion. Tumour cells in HCC resemble normal hepatocytes to a variable extent, depending on the degree of differentiation. HCC tumour cells are often smaller than normal hepatocytes (small-cell change; see earlier) ( Fig. 13.37 ), with high nuclear/cytoplasmic ratio (see cytology section later). Nucleoli and hyperchromasia can be variable, but prominent nucleoli can often be seen in less well-differentiated or pleomorphic tumour cells. Intranuclear inclusions (vacuoles) may occur ( Fig. 13.38 ). The cell membranes are often distinct ( Fig. 13.39 ), and tumour cells typically have a moderate amount of eosinophilic, finely granular cytoplasm. However, the cytoplasm can be variable on staining, from more basophilic than normal hepatocytes to highly eosinophilic. The basophilia is a feature often associated with small-cell change within HCC (see Fig. 13.37 ), whereas the eosinophilia can take the form of oncocytic change (see Fig. 13.39 ), typically caused by accumulation of abundant mitochrondria.
The tumour cells can grow in several architectural patterns ( Table 13.8 ). The trabecular pattern is the most common ( Fig. 13.40 ; see also Figs 13.37–13.39 ), in which the tumour cells grow in variably thick cell plates (typically three or more cells thick) separated by vascular sinusoids and lined by endothelial cells which phenotypically resemble capillary endothelium (CD34+) ( Fig. 13.41 ), rather than normal hepatic sinusoidal endothelium (CD34−). Dilation or expansion of canalicular forms at the centre of the trabeculae leads to a pseudoglandular or pseudoacinar pattern, and the prominence of this pattern can be highly variable, from scattered foci to near-complete involvement of large zones of tumour by this transformation ( Figs 13.42 and 13.43 ; see also Fig. 13.37 ). The trabeculae can compress the sinusoids, producing a compact pattern with sheets of tumour cells without obvious trabeculae formation ( Fig. 13.44 ). The peliotic pattern is characterized by tumour cells interspersed by large vascular lakes, mimicking peliosis hepatis. Bile pigment may be present in tumour cells or in dilated canaliculi in about 50% of tumours. Cytoplasmic fat is present in about two-thirds of tumours and abundant in about 10%; fat can be present as small or large droplets. It tends to be more prominent in smaller tumours (particularly notable in early HCC; see previous discussion) and often disappears in advanced cases. Large amounts of cytoplasmic fat and glycogen can cause the cytoplasm to appear white in routine sections, producing a clear cell appearance ( Figs 13.45 and 13.46 ). The term clear cell HCC has been used if the clear cytoplasm is a prominent finding. Cytoplasmic MDBs are present in approximately 20% of tumours and hyaline globules in another 20%, which may represent AAT or other proteins ( Figs 13.47 and 13.48 ). Ground-glass cytoplasmic inclusions similar to those observed in chronic hepatitis can be seen in the tumour cells. Pale eosinophilic cytoplasmic inclusions (‘pale bodies’) are most characteristic of the fibrolamellar variant but can be seen in classic HCC and represent accumulations of fibrinogen. Rare findings include bizarre multinucleated cells ( Fig. 13.49 ), osteoclast-like giant cells ( Fig. 13.50 ) and extramedullary haematopoiesis. Necrosis and inflammatory changes may also be present, particularly in more poorly differentiated tumours (see Fig. 13.38 ). When prominent, lymphocytic or neutrophilic inflammatory infiltrates in the tumour are recognized as distinct variants (see later). Connective tissue stroma is typically sparse and is a helpful diagnostic clue for HCC. Abundant fibrous stroma is present in scirrhous and fibrolamellar variants.
|HCC, WD to MD|
|CC, WD to MD|
The reticulin network of the hepatic plates is typically fragmented or lost in HCC ( Fig. 13.51 ). Thus, histochemical stain for reticulin fibres is one of the most useful techniques for the diagnosis of HCC. The reticulin network can be present in some cases, particularly in early HCC and in extremely well-differentiated to well-differentiated HCC, and thus does not exclude the diagnosis. However, in many of these cases, the framework in HCC has an abnormal pattern, such as increased deposition around single cells or small clusters of cells ( Fig. 13.52 ), accentuation of greatly widened trabeculae or acinar structures ( Fig. 13.53 ) or abnormal architecture admixed with loss or fragmentation of reticulin ( Fig. 13.54 ), not typical of benign hepatocellular lesions such as HCA. Of note, widened trabeculae (>3 cells) alone is not an indicator of HCC, because focal reticulin irregularities in plate width and architecture can also be seen in benign lesions, especially FNH (see Fig. 13.19 ) and at times, regenerative nodules. Also, the near-complete encircling, or ‘packeting,’ of small groups of tumour cells by the reticulin fibres is not specific for HCC and can also be seen in HCA (see Fig. 13.12 ) and FNH (see Fig. 13.19 ). Similarly, loss and fragmentation of reticulin staining are frequently seen with fat accumulation in hepatocytes, even in non-neoplastic or benign settings. Therefore the reticulin stain has limited utility in the diagnosis of HCC in the presence of prominent intralesional steatosis.
A wide armamentarium of immunohistohemical stains is available to confirm hepatocellular differentiation and to distinguish HCC from other primary or metastatic neoplasms (see Table 13.8 and later section on metastatic tumours ). Since limited tissue is often available for biopsies, judicious selection of markers is imperative for diagnosis. The strengths and limitations of various markers are outlined next.
Arg1 is a binuclear manganese metalloenzyme that catalyzes the hydrolysis of arginine to ornithine and urea; it is found only in the liver. Arg1 shows strong cytoplasmic staining in normal as well as neoplastic hepatocytes ( Fig. 13.55 ). Nuclear staining can be seen in some cases.
Arg1 is highly sensitive and specific marker for hepatocellular differentiation. These results are true across the spectrum of differentiation of HCC. Variants such as scirrhous HCC are usually positive. The staining tends to be diffuse in most cases, implying that this marker will yield similarly high sensitivity in biopsies. Most nonhepatocellular tumours are negative for Arg1.
Rare instances of focal positive staining in adenocarcinomas of the prostate, pancreas, colon, breast and biliary tree have been reported. Hepatoid adenocarcinomas that can occur in the gastrointestinal (GI) tract, pancreas and other sites are typically positive for Arg1. Occasional cases of HCC can be negative for Arg1, and thus a panel of markers is recommended. Since Arg1 is positive in both benign and malignant hepatocellular neoplasms, it is not helpful for distinguishing HCA from HCC.
Hepatocyte paraffin 1
HepPar 1 is a monoclonal antibody that was initially obtained using tissue from a formalin-fixed failed allograft liver. It is now known that it reacts with the urea cycle enzyme carbamoyl phosphate synthetase 1 of liver mitochondria, with a granular cytoplasmic staining in normal and neoplastic hepatocytes ( Fig. 13.56 ).
HepPar 1 has high sensitivity and specificity (both >80%) for the diagnosis of HCC. A majority of adenocarcinomas from most sites, including biliary tree, pancreas, colorectum, breast, urinary bladder and prostate, are negative or only focally positive. Other tumours that usually enter the differential diagnosis of HCC, such as neuroendocrine neoplasms, renal cell carcinoma, adrenocortical carcinoma, malignant melanoma and epithelioid angiomyolipoma, are typically negative.
HepPar 1 has low sensitivity (<50%) for poorly differentiated and scirrhous HCC. The staining is patchy in 20% of HCC cases and thus can be negative in the tumour on needle biopsies. Positive staining can be seen in pulmonary adenocarcinomas and less often in adenocarcinomas from other sites, such as stomach, oesophagus and biliary tree. Hepatoid carcinomas are positive for HepPar 1, and it is not useful for distinction of benign and malignant hepatocellular lesions.
GPC3 is a membrane-anchored heparin sulfate proteoglycan normally expressed in fetal liver and placenta. It is an oncofetal protein and is not expressed in normal or benign hepatocytes. GPC3 shows cytoplasmic staining in HCC, with membranous and Golgi pattern in some cases ( Fig. 13.57 ).
GPC3 has moderate to high sensitivity (65–80% in most studies) for HCC. Most cases of poorly differentiated and scirrhous HCC express GPC3. Since GPC3 is negative in normal liver, FNH and adenomas, it can be useful in distinguishing benign hepatocellular lesions from HCC.
GPC3 is not specific for HCC, and can be expressed in a wide variety of other tumours, including squamous cell carcinoma, melanoma, nonseminomatous germ cell tumours (e.g. yolk sac tumour, choriocarcinoma), gastric adenocarcinoma and rare cases of cholangiocarcinoma. The sensitivity is low (<50%) in well-differentiated HCC. GPC3 can occasionally be positive in cirrhotic nodules and areas of high necroinflammatory activity in hepatitic disease. High-grade dysplastic nodules can be variably positive for GPC3.
Polyclonal antibody to carcinoembryonic antigen
Polyclonal CEA cross-reacts with biliary glycoprotein I in the bile canaliculus, leading to a characteristic canalicular pattern of staining in normal and neoplastic hepatocytes. Most adenocarcinomas show cytoplasmic and/or luminal staining. Monoclonal CEA also shows cytoplasmic reactivity in adenocarcinomas, but the sensitivity is lower (~60%). HCC is nonreactive with monoclonal CEA.
The overall sensitivity for HCC is high (70–80%), and the canalicular pattern observed with pCEA is considered to be specific for hepatocellular differentiation ( Fig. 13.58 ).
The sensitivity of pCEA is low in poorly differentiated and scirrhous HCC. Cytoplasmic staining can be seen in some HCC cases. Occasionally, it can be difficult to distinguish canalicular from membranous or luminal staining pattern seen in adenocarcinomas.
Other hepatocellular markers
Alpha fetoprotein (AFP) is an oncofetal protein produced by the liver and yolk sac. Its expression in a tumour is specific for hepatocellular differentiation if germ cell tumours can be excluded. However, staining tends to be patchy, and sensitivity is low (30–50%) and almost always essentially absent in small HCCs. The low sensitivity and availability of better antibodies have rendered AFP as a less useful option for diagnosis. However, serum AFP levels remain helpful in the diagnosis and monitoring response to therapy for HCC.
The staining pattern with CD10, villin and bile salt excretory protein (BSEP) is similar to the canalicular pattern of pCEA and has the same drawbacks.
Thyroid transcription factor 1 (TTF-1) is expressed in nuclei of epithelial cells and tumours of the thyroid and lung. Cytoplasmic TTF-1 staining has been observed in 70% of HCCs. Since the staining often parallels the results obtained with HepPar 1, and the immunoreactivity varies with the clone and antigen retrieval technique, TTF-1 offers no benefit if HepPar 1 is part of the panel.
CD34 is a marker of endothelial cells which is negative or restricted to a few periportal sinusoids in normal hepatic sinusoidal endothelium. The sinusoid-like vasculature in HCC usually shows strong expression of CD34 (see Fig. 13.41 ), attributed to a change in the phenotype of endothelial cells (‘capillarization of sinusoids’) as a result of an arterialized blood supply. Increased sinusoidal staining is also observed in FNH and HCA but tends to be patchy compared with diffuse staining observed in most HCC cases. Staining in regenerative and dysplastic nodules is variable but is usually restricted to the periphery. Thus, CD34 is not a specific feature of HCC but can be helpful to confirm the presence of lesional tissue when dealing with limited liver biopsy specimens.
Staining with the commonly used anticytokeratin cocktail AE1/AE3 is typically patchy and tends to be more prominent in poorly differentiated HCC. Keratin 8 (K8) and K18 are expressed by normal and neoplastic hepatocytes, and thus most HCCs stain with CAM5.2 and 35βH11. The negative results in HCC with the AE1/AE3 cocktail are generally an artefact of the two components (AE1, AE3) not ‘balanced’ at the appropriate relative dilutions. Both K7 and K20 are negative in almost 75% of HCC. K7 positivity is seen in about 20% of cases and tends to be stronger in poorly differentiated HCC, similar to AE1/AE3. K20 staining in HCC is rare but can stain MDBs. Of note, K19 can be expressed in 10–20% of HCCs, but this finding alone is not enough to designate a tumour as combined HCC-CC. However, the expression of K19 is associated with an adverse outcome.
MOC31 is an antibody directed against a cell surface glycoprotein, EpCAM. It was initially described for its utility in distinguishing metastatic adenocarcinoma and mesothelioma. MOC31 shows a membranous pattern of staining in cholangiocarcinoma and metastatic adenocarcinoma from a variety of sites, such as colorectum, pancreas, stomach, lung, breast and ovary. The sensitivity is high (80–100%), and the membranous pattern is easy to interpret. Most neuroendocrine tumours, urothelial carcinoma and renal cell carcinoma are also positive. Most HCCs are negative or weakly positive for MOC31. Strong immunoreactivity has been described in 8–10% of HCCs.
Other useful antibodies and techniques
Albumin in situ hybridization (ISH) targeting the albumin messenger RNA is specific for hepatocellular differentiation as albumin is synthesized only in the liver. A new branched-chain assay may provide higher sensitivity for the diagnosis of HCC. However, the use of this test is limited by its restricted availability.
Nuclear staining with β-catenin and diffuse staining with GS is a marker of mutations in β-catenin or other components of the Wnt signalling pathway and, in most cases, favor HCC over benign processes such as adenoma.
Heat shock protein 70 (HSP70) is an antiapoptotic protein and is overexpressed in HCC. HSP70 staining is seen in up to 80% of early HCCs in resection specimens but in <50% of cases on biopsy, and only 5–10% of HGDNs are positive. Combined use of HSP70, GPC3 and GS can help in diagnosis of HCC over HGDN if two of these three markers are positive.
Gains of chromosomes 1, 7 and 8 are common in HCC and are seen in most cases of well-differentiated HCC, but these changes are not seen on typical HCA. These cytogenetic changes can be demonstrated by FISH or comparative genomic hybridization but are not currently available for routine clinical use.
Histological variants of hepatocellular carcinoma
Fibrolamellar hepatocellular carcinoma
Fibrolamellar HCC, or fibrolamellar carcinoma (FLC) per WHO (2010), typically occurs in young adults, with a mean age of 26 years at diagnosis. The tumour generally occurs in the absence of chronic liver disease or cirrhosis, and serum AFP level tends to be normal or with only mild elevation in 10–15% of cases. Serum B12 binding protein is often elevated. There are no known risk factors. Since FLC is a rare tumour (1–8% of HCCs), classic HCC is more common than FLC, even in children and young adults.
FLCs are distinct from typical HCCs at the histological, immunophenotypical and molecular levels. On gross examination, FLCs are tan-white tumours that are typically larger than classic HCC ( Fig. 13.59 ). A central scar can be present, often with calcifications, and can be detected radiographically. Microscopically, FLC is characterized by a trio of histological features, as follows:
Large polygonal tumour cells with abundant eosinophilic granular cytoplasm ( Fig. 13.60 ). The granular appearance is related to abundant mitochondria, lysosomes or endosomal cytoplasmic accumulations. Eosinophilic cytoplasmic inclusions paler than the tumour cell cytoplasm and with a ground-glass appearance (pale bodies) are often present and are thought to be caused by fibrinogen accumulation. However, pale bodies are not specific for FLC and can be seen in classic HCC.
Prominent, usually single, eosinophilic macronucleoli, generally associated with vesicular nuclear chromatin or margination of nuclear chromatin ( Fig. 13.61 ).
Lamellar fibrosis comprising parallel plate-like stacks of collagen (see Figs 13.60 and 13.61 ). The characteristic fibrous stroma can be patchy and may not be present in a needle biopsy or metastasis.
Additional variable findings within the tumour include degenerative atypia characterized by enlarged nuclei with smudgy chromatin as a common finding, occasional pseudoglandular/acinar structures with mucin, granulomas and focal calcifications (see Fig. 13.61 ). A clear cell variant of FLC has been reported. It is recommended that all three criteria listed above should be met for an unequivocal diagnosis, because it is important to use strict morphological criteria for the diagnosis of FLC. Of note, in retrospect, it is likely that stringent criteria were not used in the diagnosis of rare cases that have been described in cirrhotic liver.
Similar to classic HCC, hepatocellular markers such as HepPar 1, pCEA (canalicular pattern) and Arg1 are positive in FLC. GPC3 immunoreactivity is observed in 17–64% of cases, but positive results with AFP are rare, correlating with the often unremarkable serum AFP. However, K7 is positive in almost all cases, compared to only 20–30% of classic HCCs. CD68, a macrophage marker, is positive in almost all cases as well, but it is not specific for this variant since staining is also seen in 10–25% of classic HCC. It has been argued that the diagnosis of FLC should be made with caution if K7 or CD68 is negative. In some cases, neuroendocrine markers such as nonspecific enolase and chromogranin can be positive.
As noted, focal pseudoglandular differentiation, presence of mucin and strong K7 staining are not uncommon findings and thus should not be mistaken for cholangiocarcinoma (CC), metastatic adenocarcinoma or mixed FLC-CC. However, rare cases of true mixed FLC-CC have been described that demonstrate distinct, well-characterized FLC and CC components. Some FLC cases express EpCAM, NCAM, CD133 and CD44, which are thought to be progenitor cell-associated markers, suggesting that these tumours have stem cell-like characteristics. The plate-like fibrosis typical of FLC can be seen in a minority of scirrhous HCCs, but the typical cytological features of FLC are not present.
A remarkable recent molecular finding in many cases of FLC distinguishing this lesion from classical HCC is a consistent somatic deletion leading to DNAJB1-PRKACA fusion transcripts. Whole-genome sequencing has revealed that the fusion results from a 400-kb deletion on chromosome 19. The fusion transcript encodes a chimeric protein that couples a segment of the heat shock protein, DNAJB1, with the catalytic domain of protein kinase A (PKA) and exhibits full retention of PKA activity. Although the tumourigenic mechanism for this fusion protein is not yet proved, the fusion protein appears to drive overexpression of PRKACA, a key regulatory kinase with numerous downstream signalling targets. The fusion protein can be detected in paraffin-embedded tissue on a transcriptional and a genomic level using a reverse-transcriptase polymerase chain reaction (RT-PCR) and FISH assay, respectively. Of note, further data later found that some HCCs of non-FLC type may also have this deletion, so even though this deletion may be a highly consistent finding in FLC, this molecular marker may not be 100% specific.
Based on initial studies, FLC was thought to have a favourable outcome compared to classic HCC. However, FLC is an aggressive tumour, and recent data show similar outcomes in FLC and classic HCC occurring in noncirrhotic liver. The 5-year survival in FLC is 50–60% in most series, and surgical resectability is the most important prognostic factor. Adjuvant chemotherapy is not effective, and the median survival for unresectable cases is 1 year. LT has been used for unresectable cases. Peritoneal spread and lymph node metastasis are common, the latter occurring in up to 70% of cases (versus <5% for classic HCC). Therefore, regional lymph node dissection is often done along with lobectomy.
Scirrhous hepatocellular carcinoma
A small subset of otherwise classic HCC shows a prominent stroma, called scirrhous HCC. There are no standard criteria for this designation, but many studies have used stroma in ≥50% of the tumour to establish the diagnosis. Earlier publications have used the term ‘sclerosing hepatic carcinoma’ for liver tumour with prominent stroma. These descriptions likely include a mix of tumours, including CC, and this term should be avoided. Marked fibrosis can result from therapy such as transarterial chemoembolization (TACE); these cases are not regarded as scirrhous HCC. Clinical characteristics such as age, gender, cirrhosis and serum AFP levels in scirrhous HCC tend to be similar to classic HCC. Atypical radiographic characteristics on contrast imaging, such as peripheral ring enhancement in the arterial phase and delayed enhancement of the central region in the venous phase, are common in scirrhous HCC and can lead to an erroneous diagnosis of intrahepatic cholangiocarcinoma (ICC) or metastatic carcinoma.
Scirrhous HCC tend to occur in subcapsular location and often lack a capsule. The firm, white cut surface resembles ICC. Apart from the abundant fibrous stroma, scirrhous HCC often resembles conventional HCC. Some scirrhous HCC cases have small oval cells located at the periphery of the nests of tumour cells. These small cells may represent stem/progenitor cells (see later). Scirrhous HCCs often yield negative results on IHC with several common hepatocellular markers (e.g. HepPar 1, pCEA) in more than half the cases. On the other hand, markers such as K7, K19 and EpCAM are positive in almost two-thirds of cases. The abundant stroma along with this immunophenotype can be mistaken for CC. Positive staining for GPC3 and Arg1 is observed in a vast majority of scirrhous HCCs, and the use of both markers is recommended for diagnosis of tumours with abundant stroma.
Reports have been inconsistent about the prognosis in this scirrhous variant compared to classic HCC. Most recent studies indicate that scirrhous HCC is an aggressive variant. The abundant stroma and aberrant IHC profile along with the gene-profiling data suggest that these tumours show overlapping features of classic HCC and ICC.
Sarcomatoid hepatocellular carcinoma
Sarcomatoid HCC is characterized by sarcoma-like areas that have presumably evolved from classic HCC. It accounts for <5% of all HCC cases. Sarcomatoid changes in classic HCC can occur after chemotherapy or TACE. Some variation may be inherent in the use of ‘sarcomatoid HCC’ versus ‘carcinosarcoma’, but for most the general consensus is that the term carcinosarcoma in the liver includes both sarcomatoid HCC and sarcomatoid CC. The sarcomatous component comprises malignant spindle, epithelioid or pleomorphic cells, often with an associated component of classic HCC. Heterologous differentiation along rhabdoid, osteoid and chondroid lines can occur as part of the sarcomatous component. The sarcomatous component shows cytokeratin expression in up to two-thirds of cases, while hepatocellular markers are usually negative. In the absence of a classic HCC component, the distinction from a sarcoma (without HCC) can be difficult. History of chronic liver disease, epithelial differentiation by IHC or prior classic HCC would favour sarcomatoid HCC over a sarcoma in these patients. The prognosis is worse compared to classic HCC.
By definition, these tumours lack morphological evidence of epithelial differentiation, and the diagnosed is based on IHC. Undifferentiated tumours are thought to have a poorer prognosis than conventional HCC. With the wide availability of IHC markers, this diagnosis is seldom made.
Lymphoepithelioma-like hepatocellular carcinoma
Lymphoepithelioma-like HCC is characterized by a prominent lymphoid infiltrate. These tumours are rare, so definite criteria for diagnosis and clinicopathological correlations are not well established. Some cases can morphologically resemble nasopharyngeal carcinoma, and rare instances of Epstein–Barr virus (EBV) positivity have been described. In some cases the lymphocytic infiltrate may obscure the tumour cells and mimic a lymphoma.
Other histological subtypes
A variety of other histological subtypes, although not recognized as HCC variants in the WHO 2010 classification, are sufficiently distinctive to merit discussion.
Steatohepatitic hepatocellular carcinoma
In this variant the tumour shows typical morphological features of steatohepatitis in an otherwise classic HCC ( Figs 13.62 and 13.63 ). These features include steatosis, ballooning, MDBs and pericellular fibrosis. Most affected patients have steatohepatitis in the non-neoplastic liver, and risk factors associated with metabolic syndrome are often present. This variant can also occur in the absence of metabolic risk factors and steatohepatitis in non-neoplastic liver. The outcome is similar to classic HCC.
Granulocyte colony-stimulating factor-producing hepatocellular carcinoma
This unusual variant is characterized by a striking neutrophilic infiltrate, which can obscure the tumour cells and mimic an infectious/inflammatory process. Fever and leukocytosis can accompany the tumour. Serum levels of granulocyte colony-stimulating factor (G-CSF) are high, and G-CSF has been demonstrated by IHC in the tumour cells.
Clear cell hepatocellular carcinoma
Clear cell HCC is not a distinct histological subtype (see Fig. 13.46 ) because clear cells can be seen in any subtype of HCC. Thus the importance of recognizing this variant pattern is its distinction from other clear cell tumours, such as clear cell renal cell carcinoma. The distinction is usually straightforward because hepatocellular markers (e.g. Arg1, HepPar 1) are negative in renal cell carcinoma, whereas staining for renal transcription factors PAX-2 and PAX-8 is negative in HCC.
Diffuse cirrhosis-like hepatocellular carcinoma
Diffuse cirrhosis-like HCC is also not a distinct histological subtype, but it represents a characteristic growth pattern comprising multiple small nodules admixed with regenerative nodules in cirrhotic liver (see Fig. 13.36 ), with no evidence of a single, dominant, larger HCC within the liver. The tumour nodules may mimic cirrhotic nodules on imaging and gross examination. AFP is normal or mildly elevated in most cases. Histologically, the tumour nodules show features of well- or moderately differentiated HCC and typically show a similar morphology among all the nodules throughout the liver.
HCCs are notably heterogeneous in differentiation and histological patterns within the same lesion. Aspirating different parts of the tumour may yield more classic features, obviating the need for costly ancillary tests. On the other hand, the aspirate may not be representative. An impression of tumour architecture can also be gleaned from smears. Grading of HCC is based on nuclear features and follows the criteria outlined for histological grading.
Gross inspection of the hypercellular smears reveals trails of cohesive arborizing aggregates of malignant hepatocytes imparting a granular pattern of spread ( Fig. 13.64 ). The broad, tongue-like cords may be wrapped by peripheral endothelium ( Figs 13.65 and 13.66 ). Transgressing endothelium with basement membrane material resembles pink ‘tramlines’ on May–Grünwald Giemsa (MGG) smears ( Fig. 13.67 ). Pseudoglands are surrounded by neoplastic hepatocytes, similar to adjacent sibling cells ( Fig. 13.68 ). In the better-differentiated lesions, the hepatocytic characteristics with increased nuclear/cytoplasmic ratio (>1 : 3) are easily recognizable ( Fig. 13.69 ). Atypical bare or stripped hepatocytic nuclei may be identified in all grades of HCC ( Fig. 13.70 ). Multinucleated tumour giant cells, which may be of ‘osteoclastic’ or pleomorphic type, are encountered with increasing frequency in higher-grade lesions ( Fig. 13.71 ). Bile pigment can best be seen in MGG smears as blackish blue clumps in hepatocytic cytoplasm, ‘ropy cords/broken twigs’ within canaliculi or bile plugs within pseudoacini ( Fig. 13.72 ). Intracytoplasmic fat and glycogen vacuoles are also best observed in MGG smears. HCC cells with fatty change exhibit large cytoplasmic vacuoles with eccentric nuclei, accompanied by free fat globules in the background ( Fig. 13.73 ). HCC cells loaded with glycogen exhibit clear cell change ( Fig. 13.74 ). Intracytoplasmic inclusions (hyaline globules, MDBs, pale bodies) and intranuclear cytoplasmic inclusions may be seen ( Fig. 13.75 ). Bile duct epithelial cells, if present, are few and far apart. The background may be haemorrhagic or necrotic, especially after locoregional ablational therapies.
Cytological features of malignancy are lacking at the well-differentiated HCC (WDHCC) end of the spectrum, whereas resemblance to hepatocytes is lacking at the poorly differentiated HCC (PDHCC) end. Cohesion is the rule in HCC. However, a tendency to dissociation is noted in highly WDHCC with cords ≤2 cells thick; and in PDHCC with virtually absent reticulin. WDHCC cells tend to be conspicuous by their small size, monotony, subtle increase in nuclear/cytoplasmic ratio and nuclear crowding (see Table 13.8 ). Another pitfall is the ‘nodule-in-nodule’ lesion, where an HCC subnodule arises within a ‘parent’ nodule, whether dysplastic or regenerative ( Fig. 13.75 ). This raises the issue of adequacy of fine-needle aspiration (FNA) evaluation, given the focality of proliferative clones within such nodules. Small HCC, as with other well-differentiated hepatocellular nodules, are prone to fatty change. Lesional fatty change in the absence of steatosis is a useful parameter indicating representative sampling. Cytodifferentiation of highly WDHCC from other well-differentiated hepatocellular nodules is extremely challenging, often with indeterminate reports. On the other hand, PDHCC cells are highly pleomorphic with thin nuclear membranes, irregular nuclear contours and even absent nucleolus ( Fig. 13.76 ). Immunohistochemistry is necessary for separating PDHCC from ICC and metastatic carcinoma.
On cell block/microbiopsies, representative HCC tissue may yield invaluable architectural details, unpaired arteries, diffuse sinusoidal capillarization (CD34), canalicular formation (pCEA, CD10), abnormal/deficient reticulin and stromal invasion. Ductular reaction (K7 or K19) should be lacking within or at the invasive front of HCC. A panel of GPC3, Arg1, HepPar 1, pCEA, CD10, CD34, MOC31 and K7 is helpful in diagnosing and differentiating HCC from metastatic adenocarcinoma on FNA biopsies.
Variant histological patterns of HCC have special features in FNA. FNA cytology of the fibrolamellar variant is characterized by discohesive, large polygonal cells with abundant oncocytic granular cytoplasm, large nucleus and nucleolus, and low nuclear/cytoplasmic ratio ( Fig. 13.77 ). Intracytoplasmic hyaline globules and well-defined pale bodies are common. The other distinctive feature is lamellar fibrosis, seen as pink parallel bands of fibrous tissue within tumour fragments in MGG smears. Trabecular arrangement is not observed.
Other variations in histology may include steatosis, as well as a prominence of clear, small, spindle or giant cells. HCC with fatty change can be focal or diffuse (see Fig. 13.73 ). Highly WDHCC with fatty change can easily be misinterpreted as non-neoplastic hepatocytes from steatosis or focal fatty change. PDHCC with fat vacuoles can mimic lipoblasts or signet-ring cell adenocarcinoma. In HCC, clear cell type, the tumour cells exhibit pale, finely vacuolated and fragile cytoplasm surrounding a small, central round nucleus (see Fig. 13.74 ). Mimics include metastatic clear cell tumours, especially renal cell and adrenocortical carcinomas. In HCC of small-cell type, the tumour cells are small with scanty cytoplasm, round to oval nuclei with granular chromatin, high nuclear/cytoplasmic ratio and tendency to dissociation and microacinar formation ( Fig. 13.78 ). They mimic neuroendocrine tumours, but the chromatin is not of the ‘salt and pepper’ type. Closer scrutiny may reveal classic HCC features. Other small, round cell malignancies need to be excluded. Spindle cell HCC, or purely sarcomatoid HCC, is rare and often seen in conjunction with tumour giant cells. The spindle cells exhibit various degrees of pleomorphism. Ancillary tests are required to distinguish them from sarcomas and other types of sarcomatoid carcinomas. HCC can also contain, or can be composed entirely of, giant cells. The pure giant cell variant is rare and must be distinguished from giant cell sarcomas. There is an abundance of multinucleated tumour giant cells of the pleomorphic type. Discohesion, bizarre nuclei and abundant mitoses, including abnormal forms, are features ( Fig. 13.71 ).
Lastly, in undifferentiated HCC, the tumour cells are pleomorphic, undifferentiated and larger than the small-cell category ( Fig. 13.79 ). Because the tumour cells bear no obvious resemblance to hepatocytes, it is difficult to differentiate them from other types of undifferentiated carcinomas cytologically.
Grading and staging
A histological grading system for HCC was first described by Edmondson and Steiner in 1954. The tumours were graded on a scale of I–IV, with increasing nuclear irregularity, hyperchromasia and nuclear/cytoplasmic ratio associated with decreasing cytological differentiation for each successively higher grade, and a greater emphasis on the amount and appearance of the cytoplasm and the nuclear/cytoplasmic ratio, so that grade IV tumours had scant cytoplasm, even though the nuclei might be minimally anaplastic. This system is not widely used in practice, and the system proposed by WHO in 2010 based on a combination of cytological features and differentiation is generally employed for grading ( Table 13.9 ). A correlation between the grade and prognosis has been reported. PDHCC are associated with higher recurrence after LT.
|Well differentiated||Moderately differentiated|
The staging system proposed by the American Joint Committee on Cancer (AJCC) is based on a combination of tumour characteristics (size, number, vascular invasion, direct extrahepatic extension), nodal status and distant metastasis ( Table 13.10 ). The clinical utility of this system is limited because most HCCs arise in the setting of underlying liver disease, which complicates therapy and leaves the patient at risk for developing new tumours, even if the current HCC is successfully eradicated. Therefore, several other staging systems have been proposed that combine tumour characteristics with clinical and laboratory features, including the Barcelona Clinic Liver Cancer (BCLC) staging ( Table 13.11 ), Cancer of the Liver Italian Program (CLIP) score, Japan Integrated Staging (JIS) and Hong Kong Liver Cancer (HKLC) staging system. The BCLC system has been validated in many studies and incorporated into the AASLD guidelines for treatment of HCC. However, the system that allows for the best prognostic stratification continues to be debated.
|Primary tumour (T)|
|T0||No evidence of primary tumour|
|T1||Solitary tumour without vascular invasion|
|T2a||Solitary tumour with vascular invasion or multiple tumours, none >5 cm|
|T2b||Multiple tumours >5 cm|
|T3||Single tumour or multiple tumours of any size involving a major branch of portal vein or hepatic vein|
|T4||Tumour(s) with direct invasion of adjacent organs other than gallbladder or with perforation of visceral peritoneum|
|Regional lymph nodes (N)|
|NX||Cannot be assessed|
|N0||No regional lymph node metastasis|
|N1||Regional lymph node metastasis|
|Distant metastasis (M)|
|Stage 0||Tis N0 M0|
|Stage 1||T1 N0 M0|
|Stage 2||T2 N0 M0|
|Stage 3||T3 N0 M0|
|Stage IVA||T4 N0 M0|
|Any T N1 M0|
|Stage IVB||Any T Any N M1|
|Stage||ECOG performance status||Tumour||Okuda stage||Liver functional status|
|0 (very early HCC)||0||Single, <2 cm||I||CTP class A|
|No portal hypertension|
|A (early HCC)|
|A1||0||Single, any size||I||No portal hypertension|
|A2||0||Single, any size||I||Portal hypertension present|
|A3||0||Single, any size||I||Portal hypertension|
|A4||0||2 or 3 tumours, all <3 cm||I–II||CTP class A or B|
|B (intermediate)||0||Large, multinodular||I–II||CTP class A or B|
|C (advanced HCC)||1–2 *||Vascular invasion * or extrahepatic spread||I–II||CTP class A or B|
|D (end-stage HCC)||3–4 †||Any||III †||CTP class C †|
Spread and metastases
The average doubling time of small, untreated tumours is estimated to be 5–6 months. From an encapsulated mass in the early stages, the tumours infiltrate the surrounding liver as they enlarge and can form satellite nodules. Microvascular invasion is common, found in approximately 50% of tumours, and intrahepatic metastases, presumably by haematogenous spread, are present in 60% of tumours <5 cm in diameter and >95% of those >5 cm. Nodules growing from intrahepatic venous involvement or those adjacent to the main tumour are likely to be intrahepatic metastasis, whereas other nodules likely represent multicentric tumours. Presence of concurrent multifocal early HCC, a well-differentiated component at the periphery or distinct histologies in tumour nodules would favour multicentricity over intrahepatic metastasis. Thrombosis of the portal vein or its branches occurs in 65–75% and that of the hepatic veins in 20–25% of advanced tumours. Tumours with large-vessel vascular invasion involving vessels with a muscular wall or >1 cm are twice as likely to recur after resection.
Invasion of large bile ducts with obstructive jaundice occurs in approximately 5% of HCC patients. Advanced tumours often invade the diaphragm and gallbladder. Breach of Glisson capsule is rare, so peritoneal dissemination is uncommon. Extrahepatic metastases are common, being found at autopsy in more than half of cases. Lymph node metastasis is uncommon in classic HCC and is observed in <5% of patients at presentation. Haematogenous spread is observed in 40–60% of cases, with lung being the most common site. Other potential sites for metastases include adrenal glands, bone, stomach, heart, pancreas, kidney, spleen and ovary.
Treatment and prognosis
Surgical resection offers the only potential for cure for HCC and works best for resectable tumours in noncirrhotic liver or in cirrhotic patients with adequate liver reserve. Liver transplantation (LT) offers the possibility of cure for cirrhotic patients with HCC. Patients selected based on Milan criteria (solitary tumour ≤5 cm, or up to three tumours and none >3 cm) have a 5-year survival exceeding 70%. It has been argued that Milan criteria are restrictive, and expanded criteria (e.g. UCSF criteria: solitary tumour ≤6.5 cm, or up to three nodules with the largest ≤4.5 cm, and total tumour diameter ≤8 cm) have been successfully used to obtain 5-year survival of >60%. If LT is not an option, US-guided percutaneous ablation can be done using alcohol or thermal techniques such as radiofrequency waves, microwaves, laser or cryoablation. Ethanol injection and radiofrequency ablation have become the most widely used methods to cure small lesions (usually <3 cm) or as a bridge measure for patients awaiting definitive therapy by resection or LT (for lesions 3–5 cm). Since HCC receives its blood supply from the hepatic artery, transarterial embolization (TAE) can lead to tumour necrosis and prolong survival and is recommended for unresectable tumours >5 cm. The embolic material may be combined with lipoidol and antineoplastic drugs such as doxorubicin (TACE). Transarterial injection of yttrium-90 microspheres has been used for unresectable tumours (radioembolization). Systemic radiation, chemotherapy and hormonal therapy have largely proved unsuccessful in HCC. However, stereotactic body radiation therapy (SBRT) can deliver large doses of radiation and can be helpful to ablate the tumour. Sorafenib, a multikinase inhibitor, has been shown to prolong survival in patients with advanced HCC. National Comprehensive Cancer Network (NCCN) guidelines state that sorafenib can be considered after ablative or arterial therapy if there is persistent tumour and liver function is adequate.
Favourable prognostic factors for HCC are age <50 years, female gender, resectable tumour, better-differentiated tumour, low mitotic index, absence of vascular invasion, encapsulated tumour and absence of cirrhosis. The overall 5-year survival in HCC is 15–20% but can be as high as 75% for tumours <5 cm. For patients with advanced disease, survival beyond 1 year is unusual.
Combined hepatocellular-cholangiocarcinoma, classic type
The WHO 2010 classification defines combined hepatocellular-cholangiocarcinoma (HCC-CC) as a tumour containing unequivocal, intimately mixed elements of both HCC and CC ( Fig. 13.80 ). ‘Collision tumours’ comprising separate but adjacent nodules of HCC and CC are not considered HCC-CC. Similarly, fibrolamellar HCC with focal glandular differentiation or mucin is not regarded as a combined HCC-CC, as had been suggested in the older literature.