Hepatocellular Carcinoma

Hepatocellular Carcinoma

Michael S. Torbenson, MD


Hepatocellular carcinoma is defined as a malignant epithelial tumor that (1) originates in the liver, (2) is differentiated toward hepatocytes as its primary line of differentiation, and (3) is not a hepatoblastoma. In addition to this basic definition, there are several commonly used terms that focus on small tumors. These terms are primarily used in research studies and in clinical management algorithms, but are worth knowing as a pathologist.

Small hepatocellular carcinoma

By definition, this refers to any hepatocellular carcinoma that is less than 2 cm in greatest dimension. In research studies, small hepatocellular carcinomas that arise in cirrhotic livers, and are either well or moderately differentiated, are further divided into early hepatocellular carcinomas versus progressed hepatocellular carcinomas.

Early hepatocellular carcinoma

By definition, these carcinomas are less than 2 cm in greatest dimension. They arise in cirrhotic livers and have ill-defined borders on gross examination. On histologic examination, the carcinoma is well differentiated and has residual portal tracts within the tumor, suggesting it arose out of a high-grade dysplastic nodule. Less commonly used terms that are synonyms include vaguely nodular hepatocellular carcinoma or small hepatocellular carcinoma with indistinct margins.

Progressed hepatocellular carcinoma

By definition, these carcinomas are less than 2 cm in greatest dimension. They arise in cirrhotic livers and have distinct margins on gross examination. These carcinomas are well to moderately differentiated and are more likely than early hepatocellular carcinomas to have angiolymphatic invasion.


Risk factors, noncirrhotic livers

About 20% of hepatocellular carcinomas develop in noncirrhotic livers. Of these, the background livers are clinically and histologically within normal limits in 50% of cases. However, in the remaining 50% of cases, there is evidence for chronic hepatitis, and many have mild fibrosis. Etiologies include chronic hepatitis B,1 chronic hepatitis C,2 genetic iron overload,3 fatty liver disease,4,5 chronic vascular disease,6, 7, 8 or malignant transformation of hepatic adenomas.9

Risk factors, cirrhotic livers

Most hepatocellular carcinomas (80%) develop in cirrhotic livers. The most frequent overall causes of cirrhosis (chronic hepatitis B, chronic hepatitis C, and alcohol abuse) are also the most frequent causes of hepatocellular carcinoma. Chronic hepatitis B is the most common risk for hepatocellular carcinoma in most countries in Asia. Chronic hepatitis C is the most common risk factor in the United States, much of Europe, and Japan. The new and potent antiviral drugs targeting hepatitis C and hepatitis B greatly reduce the risks for both cirrhosis and cancer. However, successful suppression or clearance of these viruses only reduces and does not eliminate the risk of hepatocellular carcinoma.10,11 Fatty liver disease is also a key risk factor, including both alcohol and metabolic syndrome driven fatty liver disease. Both of these are independent risk factors and can also be cofactors when present with other diseases such as chronic viral hepatitis. Aflatoxin B1 exposure is an important risk factor for hepatocellular carcinoma, particularly with chronic exposure. Aflatoxin B1 is produced by Aspergillus, a fungus that grows on food stored in warm, damp conditions.12 This risk factor is best documented in cases from Africa. Finally, hepatic iron accumulation is a cofactor for the development of hepatocellular carcinoma in cirrhotic livers, even when the underlying liver disease is not genetic hemochromatosis.13

Although cirrhosis from any cause is a key risk factor for hepatocellular carcinoma, the risk is not equitably distributed. For example, the 5-year cumulative incidences for hepatocellular carcinoma in one study was as follows: hereditary hemochromatosis (21%), chronic hepatitis C (17%), chronic hepatitis B virus (10%), and biliary cirrhosis (5%).14


Hepatocellular carcinoma can develop in the pediatric population, but almost all cases present after the age of 5. The most common risk factors are chronic hepatitis B, glycogen storage disease, and tyrosinemia; however, there are many others (Table 28.1).

In adults, hepatocellular carcinomas are generally divided into those that occur in cirrhotic livers (80%) and those that occur in noncirrhotic livers (20%). For hepatocellular carcinomas arising in cirrhotic livers, the incidence begins to rise around age 40 and peaks at approximately age 62 years. Men are at greater risk than women, about 8:1. In contrast, hepatocellular carcinomas in noncirrhotic livers develop later, with about 40% of individuals age 70 or older.15,20,21 There is still a male predominance, but it is much less striking at about 1.5:1. The prognosis is better overall, even though the hepatocellular carcinomas in noncirrhotic livers tend to be larger than in cirrhotic livers.21

Table 28.1 Etiologies for hepatocellular carcinoma in the pediatric population

Clinical setting

Comment and/or reference

Chronic viral hepatitis

Hepatitis B

Cirrhotic and noncirrhotic livers264

Hepatitis C

Very rare, cirrhotic livers265

Inherited diseases

Tyrosinemia type 1

Livers are typically cirrhotic266, 267, 268

Arginase deficiency

Noncirrhotic liver269

Wilson disease

Cirrhotic liver270

PFIC 1 (ATP8B1 mutation)


PFIC 2 (ABCB11 mutation; BSEP deficiency)

Cirrhotic and noncirrhotic livers267,271,272

PFIC 3 (ABCB4 mutation; MDR3 deficiency)


Tight-junction protein 2 deficiency

Cirrhotic livers274

Glycogen storage diseases

Types 1a,275 IV,267,276 and VI275


Alagille syndrome


Biliary Atresia


Cardiac cirrhosis


Congenital absence of portal vein


Prolonged TPN therapy

Noncirrhotic liver281

Radiation therapy for Wilm tumor

Noncirrhotic liver283

Turner syndrome


Table 28.2 Serum markers of hepatocellular carcinoma


Commonly used cutoff

Sensitivity (%)

Specificity (%)


400 ng/mL








7.5 ng/mL




There are several useful serum markers for hepatocellular carcinoma: α-fetoprotein (AFP), lens culinaris agglutinin-reactive AFP (AFP-L3), and des-γ-carboxyprothrombin (DCP) (Table 28.2). Although current management guidelines do not recommend their use in screening programs, they find considerable clinical use as diagnostic and follow-up markers. All of these markers have prognostic value, with positive testing indicating a worse prognosis.22


Metastatic disease is found in 40% of individuals at the time of first presentation, with the most common sites being the lungs (50%), bones (30%), abdominal lymph nodes (30%), and adrenal glands (10%).23,24 Peritoneal spread (carcinomatosis) is uncommon at presentation but can develop in time, with autopsy studies showing focal (6%) or diffuse (3%) peritoneum spread.25

Needle track seeding

Hepatocellular carcinoma can rarely spread along needle tracks after biopsies or after ablation therapies. Subcapsular and high-grade hepatocellular carcinomas are the most likely to cause needle tract seeding and most tumor deposits develop within 6 months.26,27 The older literature demonstrated a frequency of tumor seeding of approximately 2.5%,28 leading some authors to advocate against using biopsies to diagnose hepatocellular carcinoma. However, tumor seeding is readily treated and does not reduce life expectancy.26 In addition, there is no evidence that pretransplant biopsy of a hepatocellular carcinoma reduces posttransplant survival.29 In addition to being readily treated, more recent and larger studies show the risk of seeding is actually 10× lower than previous thought, at approximately 0.25%.27 Finally, the increased use of more sophisticated biopsy methods, such as the coaxial technique, brings the risk of seeding down to essentially 0%.30

Table 28.3 AJCC staging system (8th Edition)





Primary tumor cannot be assessed


No primary tumor identified


Solitary tumor that is ≤ 2 cm

The tumor can be negative or positive for vascular invasion.


Solitary tumor > 2 cm but without vascular invasion


Solitary tumor that is > 2 cm with vascular invasion; or multiple tumors with no tumor > 5 cm


At least two tumors with at least one that is >5 cm

Tumors can be with or without vascular invasion


Direct invasion of any organ, except gallbladder; Or, perforation of the visceral peritoneum; Or tumor involving a major branch of the portal vein (including right or left branches, but not the smaller branches); Or tumor involving the right, left, or middle hepatic veins

Tumors can be single or multiple

From Amin MB, et al. AJCC Cancer Staging Manual. 8th ed. New York: Springer; 2017. © American Joint Committee on Cancer.


Hepatocellular carcinomas are staged for both clinical management decisions and for prognosis. After resection or transplantation, the AJCC system is commonly used to provide a traditional tumor stage based on tumor variables (Table 28.3). In addition to this tumor-based staging system, other systems are commonly used to decide if patients are candidates for surgery and to guide therapy in those patients who have unresectable tumors. To help guide therapy, patients are classified using the underlying liver function (e.g., Child-Pugh score) or physiologic reserve (e.g., performance status). The Barcelona Clinic Liver Cancer (BCLC) staging system incorporates tumor variables, underlying liver function and overall health status to classify patients into different stages, with each stage linked to recommended therapies. Stage 0 and stage A tumors are candidates for surgery with curative intent. Stage B tumors are not resectable and
instead are treated with transarterial chemoembolization (TACE) or ablation therapy. Stage C tumors are treated with sorafenib, whereas Stage D tumors are treated by best supportive care.



The preferred treatment for solitary or small numbers of tumors in individuals with well-preserved liver function is surgical resection with curative intent.31 Anatomical resections are performed whenever possible. In general, liver function is considered adequate for surgery when bilirubin levels are normal and the hepatic venous pressure gradient is less than or equal to 10 mmHg and/or the platelet count is greater than 100,000. In some individuals, surgery is not possible even with adequate liver reserve because of comorbid conditions such as heart disease.


Ablation therapy is used for small tumors (less than 3 cm) that are unresectable because of their location or because the patient has other serious comorbid conditions. Ablation methods include radiofrequency ablation (RFA) and ETOH injection. In general, radiofrequency ablation is preferred over ETOH injection because radiofrequency ablation has both a lower recurrence rate and better overall survival. However, ETOH injection is still used in cases where the anatomic location of the tumor precludes radiofrequency ablation therapy. Radiofrequency ablation can also be combined with transarterial chemoembolization for locoregional therapy.32,33 In addition to use as primary therapy for unresectable hepatocellular carcinomas, radiofrequency ablation is also used to downstage tumors >3 cm, either as bridge to transplantation or to surgery.34 The use of radiofrequency ablation as primary therapy in tumors larger than 3 cm is controversial.35

Transarterial chemoembolization

Transarterial chemoembolization or transarterial radioembolization therapy is used for tumors that are (1) unresectable and (2) in patients who are not candidates for ablative therapy and (3) in patients who do not have metastatic disease. Contraindications to transarterial chemoembolization are primarily Child-Pugh C cirrhosis or portal vein invasion, whereas relative contraindications are benign portal vein thrombosis or bilirubin levels greater than 3 mg/dL.35 Transarterial chemoembolization leads to improved survival compared to nontreatment groups,36 but is not curative. Transarterial chemoembolization is also used as a bridge to downstage tumors in hopes of making them resectable, or as a bridge to transplantation. In the transarterial chemoembolization procedure, microsphere beads are loaded with chemotherapy agents and injected into the branch(es) of the hepatic artery that are feeding the hepatocellular carcinoma. Radioembolization is quite similar, but the microspheres are loaded with Yttrium-90, which delivers high levels of local radiation. Of note, there is a relatively high frequency of complications following transarterial chemoembolization because the beads can spread outside the tumor, injuring the nonneoplastic liver. Transarterial chemoembolization is followed by acute hepatic decompensation in up to 20% of individuals, which is irreversible in about 3% of cases.35 Transarterial embolization beads can also spread to other organs, such as the stomach (Fig. 28.1), leading to ischemic injury.

Figure 28.1 Y90 beads. These beads escaped from the liver and lodged in the stomach.

Systemic chemotherapy

Systemic chemotherapy is used when other therapies are contraindicated, mostly in the setting of portal vein invasion or metastatic disease. Sorafenib is the only therapy widely used outside the setting of specific clinical trials. The SHARP trial37 is widely considered a landmark study in hepatocellular carcinoma therapy. The multicenter SHARP trial showed that sorafenib was superior to placebo for individuals with hepatocellular carcinoma and Child A cirrhosis, with a better median overall survival (11 versus 8 months). Although this improvement is clearly modest, it has led to renewed efforts to improve systemic chemotherapy for hepatocellular carcinoma.


The median survival after a diagnosis of hepatocellular carcinoma is between 10 and 18 months.38, 39, 40 There are many clinical and pathologic factors that influence prognosis, but none are as important as whether the tumor can be resected, with successful resection providing a 5-year overall survival ranging from 25% to 80%.41 Other important prognostic findings include clinical factors such as patient age, gender, underlying liver disease, and overall physiologic status. Tumor variables that strongly influence survival include tumor size, tumor grade, and microvascular invasion. Portal vein invasion and metastatic disease indicate a poor prognosis.

Tumor subtypes also have prognostic significance. Hepatocellular carcinomas subtypes with a better prognosis, compared to conventional hepatocellular carcinoma, are clear cell hepatocellular carcinoma and lymphocyte-rich hepatocellular carcinoma. Subtypes with a worse prognosis are combined cholangiocarcinoma-hepatocellular carcinoma, combined neuroendocrine carcinoma-hepatocellular carcinoma, cirrhotomimetic hepatocellular carcinoma, sarcomatoid hepatocellular carcinoma, granulocyte colony stimulating factor (GCSF) producing hepatocellular carcinoma, and carcinosarcoma.

After successful surgery, approximately 60% of individuals will have tumor recurrences, with most recurrent disease located in the liver.42,43 The tumor recurrences represent the original tumor clone in two-thirds of cases, presumably from micrometastases that were not removed at the time of the original surgery. The remaining one-third of recurrent hepatocellular carcinomas result from a second independent primary, only seen in individuals with a background of severe chronic liver disease, usually with cirrhosis.44,45

Angiolymphatic invasion

Vascular invasion predicts tumor recurrence after TACE,46 after resection,47,48 and after liver transplantation.49 Vascular invasion is divided into macrovascular and microvascular invasion. For clinical care, macrovascular invasion is defined as tumor involving blood vessels that are large enough to be recognized by imaging or by gross examination. Macrovascular invasion almost exclusively involves either the portal venous system or the central veins/vena cava,50 whereas the hepatic arteries are almost never involved. The frequency of macrovascular invasion is between 5% and 30%. Macrovascular invasion can be a contraindication to resection so is more commonly identified in imaging-based studies or in autopsy studies than in surgical pathology specimens.

Microvascular invasion is evident only on histologic examination and has a frequency of about 30%, with a range of 15% to 60%.51 The wide range in the frequency results from small studies that can lead to skewed results and from varying sectioning protocols (finding vascular invasion is to some degree a function of the number of sections taken). Sections taken from the tumor—nontumor interface are generally the most productive for finding vascular invasion. Portal vein invasion is 10× more common than central vein invasion,52 whereas hepatic artery invasion is hardly ever found. When there is invasion of the hepatic arteries, the tumor is generally very aggressive and the prognosis is poor.

In most cases, the vascular invasion is fairly obvious, but there can diagnostic challenges. For clinical care, an endothelial lining should be present to make the diagnosis of vascular invasion. Retraction artifact can sometimes mimic lymphatic or venous invasion, but will lack an endothelial lining. In many cases, the vascular invasion leads to a distended vessel, with sides molded around the tumor thrombus. Fibrin or an organized blood clot can occasionally be associated with the vascular invasion, but they are not necessary for the diagnosis. In contrast, the presence of rare individual tumor cells, or even small clusters of tumor cells, are counted as sectioning artifacts and not vascular invasion when they are floating freely in an otherwise empty vascular lumen.



Putative precursor lesions to hepatocellular carcinoma can be microscopic or macroscopic findings, most of which occur in cirrhotic livers. No convincing precursor lesion has been found in noncirrhotic human livers, other than hepatic adenomas.

Glycogen storing foci

Although glycogen storing foci have been considered possible precursor lesions,53,54 the data on this point does not strongly support a major role for these lesions as precursors to hepatocellular carcinoma. Glycogen storing foci are composed of hepatocytes with more glycogen than the background liver. They are easily identified at low power, standing out from the background liver as well-circumscribed clusters of clear cells (Fig. 28.2). Although the genetic changes are not well studied, the data to date indicates they are polyclonal.55 In hepatocellular carcinoma resection specimens, they are found in the background liver
sections of cirrhotic (10%) and noncirrhotic (20%) livers, but the likelihood for any role as a precursor for hepatocellular carcinoma is diminished by their also being found in similar frequencies in the background liver sections of specimens resected for metastatic carcinoma (20%).56 In some cases, glycogen storing foci will have intermediate hepatocytes that stain positive for CK7.57

Small cell change

Small cell change, also known as small cell dysplasia, is defined as clusters of hepatocytes with reduced cytoplasm and mild nuclear hyperchromasia, but otherwise normal nuclear and cytoplasmic cytology (Fig. 28.3). Their increased N:C ratio causes them stand out at low-power magnification. Consistent with their role as a precursor for hepatocellular carcinomas, small cell change in cirrhotic livers can be clonal,55 have chromosomal damage, and telomere shortening.58, 59, 60

Figure 28.2 Glycogen storing foci. A distinct cluster of hepatocytes with clear cell change stand out at low power from the background liver.

Figure 28.3 Small cell change. The hepatocytes show small cell change, with an increased N:C ratio and mild nuclear hyperchromasia.

Large cell change

Hepatocytes with large cell change have large and hyperchromatic nuclei that stand out at low power, but retain a normal or nearly normal N:C ratio (Fig. 28.4). Large cell change was previously known as large cell dysplasia, with the named changed in order to reflect a more nuanced understanding of this lesion. Large cell change appears to be genetically heterogeneous, with data showing DNA damage, especially in cirrhotic livers.61 Large cell change is also associated with hepatitis B related cirrhosis62 and with cholestatic changes.63

The frequency of large cell change increases with the frequency of hepatocellular carcinoma in cirrhotic livers, but data suggests large cell change results from cellular senescence, designed to eliminate hepatocytes with DNA damage because hepatocytes with large cell change show increased apoptosis and decreased cell cycling.61,64 Even if large cell change is not a direct precursor to hepatocellular carcinoma, it does indicate liver parenchyma with widespread DNA damage and thus is associated with an increased risk for a subsequent diagnosis of hepatocellular carcinoma.65,66 Large cell change can rarely be found in noncirrhotic livers. This change has not been well studied, but is largely thought to be senescent in nature.

Macroregenerative nodule

Macroregenerative nodules are defined as cirrhotic nodules that are at least 10 mm in diameter and
most stand out from the background liver because of differences in color or texture or because they bulge out from the cut surface more than other cirrhotic nodules. The vast majority of macroregenerative nodules develop in cirrhotic livers, being found in 15% to 30% of cases.71,72 Macroregenerative nodules are multifocal in about two-thirds of cases.72 Macroregenerative nodules average 12 mm in diameter and almost all are between 10 and 20 mm.71,72 As an exception, macroregenerative nodules in cases of biliary cirrhosis, especially extrahepatic biliary atresia, can get a lot bigger, often in the 5 to 7 cm range.

Figure 28.4 Large cell change. Scattered hepatocytes show large atypical nuclei, but overall have a preserved N:C ratio.

The core concept of a macroregenerative nodule is rather simple: a cirrhotic nodule that stands out on gross examination (see above), yet shows no or minimal cytologic atypia. The hepatocytes in the macroregenerative nodule should look cytologically like the hepatocytes in the rest of the liver. Patchy large cell change is acceptable when the background liver shows similar changes. There should be no small cell change or other cytologic atypia. Macroregenerative nodules also have scattered portal tracts (Fig. 28.5). They have no reticulin loss, are glypican 3-negative, and their proliferative rate is the same as the background liver.73 Imaging studies are limited in many cases by the lack of histologic confirmation, but do suggest a low risk of malignant transformation with long term follow-up.74 Macroregenerative nodules can be challenging to reliably separate from low-grade dysplastic nodules, especially on biopsy specimens.

Large regenerative nodules can rarely develop in noncirrhotic livers following a severe hepatic injury that leads to large areas of panacinar collapse,75 for example after fulminant autoimmune hepatitis. When the surviving hepatocytes regenerate, they can from large benign regenerative nodules that can suggest carcinoma on imaging studies and on gross examination.

Figure 28.5 Macroregenerative nodule. Portal tracts are present throughout the macroregenerative nodule.

Dysplastic nodules

Dysplastic nodules are found only in cirrhotic livers and are the most widely accepted precursor lesions for hepatocellular carcinoma. Based on histologic findings, they are further classified into low- and high-grade dysplastic nodules (Table 28.4).

The core concept of a dysplastic nodule is straightforward: dysplastic nodules show more cytologic and/or architectural atypia than cirrhotic nodules in the background liver, but do not have the full histologic features of malignancy. In practice, there are nodules that show only minimal and somewhat equivocal atypia, making them hard to classify as a large regenerative nodule versus a low-grade dysplastic nodules.

Natural history

Dysplastic nodules are associated with increased risk for hepatocellular carcinoma. Although many of the dysplastic nodules will disappear over time, others will progress to hepatocellular carcinoma. In addition, dysplastic nodules serve as a general marker of increased hepatocellular carcinoma risk for the liver as a whole. Overall, about 10% of low-grade dysplastic
nodules and 30% of high-grade dysplastic nodules will progress to hepatocellular carcinoma within 2 to 3 years.74,76,77 During a 2-year study interval, an additional 25% of individuals developed hepatocellular carcinoma outside of the dysplastic nodule in one study.76

Table 28.4 High-grade versus low-grade dysplastic nodules


Low-grade dysplastic nodule

High-grade dysplastic nodule


Large cell change

May be present

May be present

Small cell change


May be present

Nuclear hyperchromasia

Absent to mild

Usually mild

Overall nuclear atypia

Minimal to mild

Usually mild

Architectural changes

Portal tracts

Several to many

Few to none

Lobular arterioles





May be present



May be present

a Assuming a noncholestatic liver.

Histologic findings

In terms of cytologic atypia, dysplastic nodules can show either large cell or small cell change. In terms of architectural findings, dysplastic nodules retain portal tracts, which are found in essentially all low-grade dysplastic nodules and many high-grade dysplastic nodules. The architectural atypia in dysplastic nodules can include pseudogland formation (less helpful in cholestatic livers), vague nodule-in-nodule growth, or increased lobular arterioles. The finding of increased lobular arteries can be accompanied by strong and diffuse staining of the sinusoids with CD34. However, this pattern of CD34 staining is also found in some ordinary cirrhotic nodules. Dysplastic nodules will show a retained reticulin pattern. Mitotic figures are rare or absent and Ki-67 proliferation is often similar to the background liver or minimally elevated. Immunostains are negative for AFP and β-catenin nuclear accumulation. Most dysplastic nodules are either negative for glypican 3 or show patchy staining. Nodules that show strong and diffuse glypican 3 staining suggest hepatocellular carcinoma. Likewise, strong and diffuse staining for glutamine synthetase favors hepatocellular carcinoma. Immunostains for heat-shock protein 70 are not widely available, but positivity in a nodule also favors hepatocellular carcinoma.78 Dysplastic nodules are further divided into low- and high-grade dysplastic nodules (Figs. 28.6 and 28.7). Low-grade dysplastic nodules have mild dysplasia but clearly show more atypia than the background liver. High-grade dysplastic nodules have sufficient atypia that they would qualify in most cases for well-differentiated hepatocellular carcinoma if they had convincing reticulin loss (Table 28.4). Classifying dysplastic nodules as low grade or high grade can be challenging on needle biopsy specimens, but works reasonably well in resection specimens.


The differential for a low-grade dysplastic nodule is a macroregenerative nodule. In most cases, comparison to the background liver is sufficient to make a diagnosis because low-grade dysplastic nodules will have cytologic and architectural changes that are not present to the same degree in the background liver. Inevitably, there are nodules with just a smidgen more atypia than the background liver, making them hard to classify. One reasonable approach is show these cases to a colleague; if you’re both still uncertain if there is enough atypia for a dysplastic nodule, call it a macroregenerative nodule.

Figure 28.6 Dysplastic nodule, low grade. The hepatocytes show a slight increase in N:C ratio and patchy foci of psuedoglandular architecture.

Figure 28.7 Dysplastic nodule, high grade. This nodule shows significant nuclear pleomorphism and pseudogland formation.

The differential for a high-grade dysplastic nodule is primarily a well-differentiated hepatocellular carcinoma. High-grade dysplastic nodules can have similar levels of cytologic atypia to that of a well-differentiate hepatocellular carcinoma, but they will not have definite loss of reticulin, easily identified mitotic figures, or a significantly increased Ki-67 proliferative rate compared to the background liver. Almost all high-grade dysplastic nodules have retained portal structures, so their absence favors hepatocellular carcinoma. Dysplastic nodules are negative for nuclear accumulation of β-catenin and lack AFP expression. Unfortunately, neither marker is very sensitive for the diagnosis of hepatocellular carcinoma, but they can
be helpful when positive. As noted above, strong and diffuse staining for glypican 3 or glutamine synthetase also favors hepatocellular carcinoma.

Figure 28.8 Stromal invasion. The hepatocellular carcinoma invades into this portal tract.

Stromal invasion, when present, is consistent with hepatocellular carcinoma.79, 80, 81 Stromal invasion is defined as malignant cells invading the stroma of portal tracts or fibrous bands (Fig. 28.8). Stromal invasion is very helpful when present, but is very challenging to confidently identify in biopsies. Overall, stromal invasion is most frequently seen in moderately to poorly differentiated hepatocellular carcinomas, in which cases a dysplastic nodule is unlikely to be in your differential. The use of CK7 or CK19 immunostains can improve detection of stromal invasion, because a ductular reaction (highlighted by the keratin stains) is typically absent at the edges of hepatocellular carcinomas with stromal invasion, but present at the edges of high-grade dysplastic nodules.82 Nonetheless, CK7 and CK19 staining patterns do not map perfectly with stromal invasion and the criteria of stromal invasion remains sufficiently challenging to implement on needle biopsy that it is not widely employed to make this distinction. In essentially all cases, the cytologic and immunohistochemic findings are sufficient to make the appropriate diagnosis without stromal invasion.


Formal clinical management guidelines of small hepatocellular lesions are based more on lesion size and imaging findings because many are not biopsied, but in general dysplastic nodules are followed radiographically every 6 months until they either disappear or enlarge. If they enlarge, biopsies are helpful to determine if hepatocellular carcinoma has developed. Some centers treat high-grade dysplastic nodules with ablation therapies.83,84

Focal nodular-like hyperplasia

Cirrhotic livers have many vascular shunts and sometimes these can lead to nodules that look and stain-like focal nodular hyperplasias.67,68 To separate them from conventional focal nodular hyperplasias, they are called focal nodular hyperplasia-like lesions. They do not have neoplastic potential.69 In keeping with vascular shunting as their etiology, these lesions are more common after chemoembolization and in individuals with esophageal varices.69 However, this lesion does not appear to be identical to typical focal nodular hyperplasia based on transcriptome profiling.70


Gross findings

The gross findings of hepatocellular carcinoma are important for staging and for guiding sampling for sections. The gross description should comment on the number of tumors, their size(s), the margin status (distance to closest margin) the percent of tumor necrosis (estimated to the nearest 10%), and the presence or absence of gross vascular invasion.

Most hepatocellular carcinomas are soft and bulge out from the cut surface of the liver, a helpful finding when grossing in a cirrhotic liver explant. Hepatocellular carcinomas also frequently have different colors than the background liver. Some hepatocellular carcinomas have a capsule of inflamed fibrotic tissue.

Hepatocellular carcinomas can be further classified into one of four growth patterns on gross examination: (1) a single distinct tumor nodule; (2) a large dominant nodule and multiple smaller “satellite” nodules, usually within 2 cm of the dominant nodule, a pattern that results from local spread of the large dominant nodule; (3) multiple distinct tumor nodules that are sufficiently distant from each other that they would not fit for satellite tumors, a pattern that reflects a field effect, with foci of independent hepatocellular carcinomas; (4) numerous small tumor nodules that are about the same size as a cirrhotic nodule, usually with 30 or more nodules, a pattern called diffuse” or “cirrhotomimetic.” The tumor burden with this last pattern is always greater than recognized on imaging studies and often by gross examination, with the extent of tumor only evident with histologic examination.

As another gross finding, up to 4% of hepatocellular carcinomas are pedunculated, protruding from the surface of the liver.85 Pedunculated hepatocellular carcinomas can mimic metastatic disease to the adrenal gland on imaging studies,86 but the gross findings are otherwise typical of hepatocellular carcinoma. In addition, tumors with this growth pattern do not
have any unique histologic findings.87 Clinically, some studies have suggested pedunculated tumors have a better prognosis, but overall the data is limited and this growth pattern is not entirely convincing as an independent prognostic factor.85,88

The optimal number of sections needed to evaluate a hepatocellular carcinoma specimen has not been well defined, but the standard of care is generally regarded as at least one section per cm of tumor. Some pathologists put two or three or more mini-sections into a single cassette, cleverly counting them toward the total section count, an approach that does not always seem to be within the spirit of striving for the best possible patient care. Tumor grade and microscopic vascular invasion are key prognostic findings, ones that depend on reasonably thorough sampling. The tumor-nontumor interface should be well sampled, being high yield for identifying vascular invasion.

Sections from the background liver are used to examine for ongoing liver disease and to determine the fibrosis stage. Sections should be taken as far away from the hepatocellular carcinoma as possible because the liver near the edge of the tumor commonly shows significant inflammation and fibrosis. The surgical resection margin section is often taken separate from the background liver section because the surgical margin commonly shows cautery effect.

Microscopic findings

Architectural changes

Neither benign nor malignant hepatic neoplasms have normal portal tracts. Some hepatocellular carcinomas and fibrolamellar carcinomas will have entrapped normal portal tracts, mostly at the edge of the tumor, but the lack of true portal tracts is an important finding supporting a diagnosis of neoplasm. Some tumors can also have fibrous septae with vascular structures, but true portal tracts are absent. A second useful architectural feature is small arteries located in the hepatic lobules (Fig. 28.9). In the normal liver, arterioles are present only in the portal tracts, but hepatic neoplasms frequently show small arterioles in the hepatic lobules. As a caveat, aberrant lobular arteries are not diagnostic of a neoplasm, also being found when nontumor liver parenchyma shows vascular flow abnormalities, including portal and hepatic vein flow obstruction in nonfibrotic livers (Fig. 28.10) and cirrhosis from many different causes, especially when there is widespread central vein sclerosis, for example with alcohol related cirrhosis. A third useful architectural finding is abnormal hepatic cord organization. In the nonneoplastic liver, the hepatic cords or plates have a regular organization and are
one to two cells in thickness. In contrast, neoplasms will have abnormal growth patterns, such as plate thickening and pseudoacinar structures. The reticulin stain is discussed below and is a key method used to identify abnormal plate organization.

Figure 28.9 Hepatocellular carcinoma, abnormal lobular artery. This finding is also called a naked artery.

Figure 28.10 Aberrant arteriole in nontumor liver. A small lobular arteriole in seen in this case with a long-standing portal vein thrombosis.

Cytologic changes

Cytologic atypia includes nuclear and cytoplasmic changes. Cytologic atypia varies a lot between hepatocellular carcinomas, forming the basis for determining tumor grade. Nuclear changes consist largely of prominent nucleoli, variable amounts of hyperchromasia, membrane irregularities (Fig. 28.11), nuclear size variation, and rarely multinucleation (Fig. 28.12). One area with some uncertainty are cases with focal cytologic atypia similar to large cell change (larger hyperchromatic nuclei with a normal N:C ratio) in a tumor that would otherwise qualify for a hepatic adenoma. Some pathologists consider these to be “degenerative changes” acceptable within a hepatic adenoma, but many pathologists do not. At this point, data on this specific point is limited, but such tumors are best considered as atypical adenomas at a minimum.

Figure 28.11 Hepatocellular carcinoma, nuclear abnormalities. The tumor cells show nuclear pleomorphism with prominent nucleoli.

Figure 28.12 Hepatocellular carcinoma, multinucleation. Many of the tumor cells are multinucleated.

Cytoplasmic changes include reduced volume, often accompanied by increased basophilia. Although an increased N:C ratio is common in hepatocellular carcinoma, this finding is not universal and some carcinomas retain abundant pink cytoplasm. Other cytoplasmic changes include heavy lipofuschin deposits (Fig. 28.13), hyaline bodies (Fig. 28.14), pale
bodies (Fig. 28.15), or Mallory-Denk bodies (Fig. 28.16). Hepatocellular carcinomas with hyaline bodies appear to have a worse prognosis, but there are no known clinical correlates for Mallory-Denk bodies or pale bodies. Hepatocellular carcinoma can also show abundant glycogen accumulation, steatosis, or cholestasis, making them stand out from the background liver.

Figure 28.13 Hepatocellular carcinoma, lipofuschin. The tumor cells have abundant lipofuschin.

Figure 28.14 Hepatocellular carcinoma, hyaline bodies. This image is from a subtumor in a clear cell hepatocellular carcinoma that had high-grade cytology and numerous hyaline bodies, whereas the background tumor did not.

Figure 28.15 Pale bodies in a conventional hepatocellular carcinoma.

Growth patterns

These are the four major histologic growth patterns found in hepatocellular carcinomas (Figs. 28.17, 28.18, 28.19, and 28.20)89: trabecular (70%), solid (also known as compact, 20%), pseudoglandular (also known as pseudoacinar, 10%), and macrotrabecular (1%). All are defined by the H&E findings without use of special stains. Growth patterns are not the same as hepatocellular carcinoma subtypes,
which are described in their own section below. In fact, any of the growth patterns can be found in any of the hepatocellular carcinoma subtypes.

Figure 28.16 Hepatocellular carcinoma, Mallory-Denk bodies.

Figure 28.17 Hepatocellular carcinoma, trabecular growth pattern.

Figure 28.18 Hepatocellular carcinoma, solid growth pattern.

Figure 28.19 Hepatocellular carcinoma, psuedoglandular growth pattern.

Figure 28.20 Hepatocellular carcinoma, macrotrabecular growth pattern.

The solid growth pattern is just like it sounds: solid sheets of cells with no definite trabecular, pseudoacinar, or macrotrabecular growth. The trabecular variant has trabeculae of variable thickness, but less than 10 cells. In contrast, the macrotrabecular pattern is defined by trabeculae at least 10 cells in thickness on average. In the pseudoglandular pattern, the tumor cells form small gland- or rosette-like structures. In some cases, the psuedoglands are considerably large, often being filled with thin, granular material, a pattern of growth sometimes called acinar or pseudocyst of follicle-like (Fig. 28.21). Additional changes can accompany any of the above growth patterns. These include bile accumulation (Fig. 28.22), peliosis-like areas (Fig. 28.23), or clusters of benign macrophages in the tumor sinusoids (Fig. 28.24).

Figure 28.21 Hepatocellular carcinoma, large pseudocyst.

Figure 28.22 Hepatocellular carcinoma with bile.

Figure 28.23 Hepatocellular carcinoma, peliosis-like changes.

Figure 28.24 Hepatocellular carcinoma, macrophages in sinusoids. Benign macrophages are seen within the tumor sinusoids.

Figure 28.25 Hepatocellular carcinoma, nodule within nodule. This hepatocellular carcinoma (top) had a distinct nodule within it that shows higher grade cytology (bottom).

The above patterns are straightforwardly defined, but not always so straightforward to apply. In some cases, compressed trabeculae can resemble a solid growth pattern. Likewise, the macrotrabecular pattern can also show areas of more compressed growth that resembles a solid growth pattern. In addition, about 50% of resection specimens have mixed patterns, most commonly trabecular plus one or two others. Finally, the results will vary, sometimes considerably, depending on the cutoff used to score a growth pattern as being present. Most studies use the reasonable cutoff of 5% but others use higher or lower cutoffs.

These growth patterns are significant for several reasons. First, familiarity with them can be useful when making a diagnosis of hepatocellular carcinoma. For example, recognizing the pseudoacinar growth pattern can help avoid a misdiagnosis of cholangiocarcinoma or combined hepatocellular-cholangiocarcinoma. Secondly, the macrotrabecular pattern has a worse prognosis when compared to the solid growth pattern.90 In particular, macrotrabecular hepatocellular carcinomas with small basophilic tumor cells are typically AFP-positive and have extensive angiolymphatic invasion.

Some hepatocellular carcinomas show a nodule-in-nodule growth pattern, with a main tumor mass showing one morphology and within it morphologically distinct nodule(s) with higher grade cytology that represent emergence of a more aggressive tumor clone (Fig. 28.25). These higher grade nodules often show some morphologic similarities to the larger, lower grade nodule. In addition to this pattern, other hepatocellular carcinomas don’t show a dominant nodule with emergence of a higher grade nodule, but instead show multiple adjacent nodules that have distinctly different morphologies (Fig. 28.26) often with the same overall cytologic grade, a finding of uncertain molecular genesis, but one that possibly reflects genomic instability.

Figure 28.26 Hepatocellular carcinoma, multiple morphologies. The nodules are adjacent to each other and show distinctly different morphologies.

Histologic grading

Tumor grade is a strong predictor of overall patient survival and disease free survival. This prognostic power is seen after resections in cirrhotic livers,47,91 noncirrhotic livers,48 and after liver transplantation.47,49 Many (one-third) of hepatocellular carcinomas have two or more nuclear grades. In these cases, the predominant tumor grade and the worse tumor grade should be reported. The worse grade tends to drive prognosis.92 Tumor grade correlates with tumor size and with angiolymphatic invasion, but has independent prognostic value in most multivariate studies.

The modified Edmondson-Steiner grading system (Table 28.5) is frequently used in research studies.89
In clinical practice, and many research studies, a threetier system is used: well differentiated, moderately differentiated, and poorly differentiated (Table 28.6). In addition, some pathologists will use additional categories of very well differentiated and undifferentiated (Table 28.6 and Figs. 28.27, 28.28, 28.29, 28.30, and 28.31).

Table 28.5 Modified Edmondson-Steiner grading system for hepatocellular carcinoma




Abundant cytoplasm; minimal nuclear atypia


Mild nuclear atypia with prominent nucleoli, hyperchromasia, and nuclear irregularity


Moderate nuclear atypia with greater hyperchromasia and nuclear irregularity


Marked nuclear pleomorphism, marked hyperchromasia, and anaplastic giant cells

Table 28.6 Grading hepatocellular carcinomas



Very well differentiated

  • Adenoma-like on H&E

  • Minimal to no nuclear atypia

  • N:C ratio is normal

  • Diagnosis based on patchy reticulin loss or abnormal immunostain findings


  • Hepatic differentiation is readily evident on H&E

  • Mild nuclear atypia

  • N:C ratio is normal or increased

  • Additional stains needed to prove malignancy in most cases

Moderately differentiated

  • Clearly malignant on H&E

  • Overall morphology strongly suggests hepatocellular origin

  • Moderate nuclear atypia

  • N:C ratio increased almost always

  • Stains are helpful to confirm hepatocyte differentiation

Poorly differentiated

  • Clearly malignant on H&E

  • H&E morphology consistent with a range of poorly differentiated carcinomas

  • Marked nuclear atypia, or moderate nuclear atypia but minimal tumor cytoplasm with very high N:C ratio

  • Immunostains needed to confirm hepatocellular differentiation


  • No morphologic features of hepatic differentiation or glandular differentiation on the H&E

  • Stains are needed to prove the neoplasm is epithelial

  • H&E differential includes lymphoma, melanoma, or mesenchymal tumors

  • Keratin-positive, but keratin-negative for hepatic markers such as HepPar1, Arginase, and glypican 3

  • Metastatic disease excluded by immunostains, imaging, and clinical findings

Figure 28.27 Hepatocellular carcinoma, very well differentiated. This hepatocellular carcinoma (right, with normal liver on the left) cannot be distinguished from an adenoma without special stains.

Figure 28.28 Hepatocellular carcinoma, well differentiated. The tumor cells have plenty of cytoplasm, and there is only mild nuclear atypia. Note, however, that the tumor nuclei lack the regular, orderly spacing of benign proliferations.

Changes after chemoembolization therapy

Chemoembolization or ablation therapy is used to treat hepatocellular carcinomas localized to the liver, both as primary therapy in unresectable tumors and to shrink tumors, allowing surgery in previously unresectable tumors. Not surprisingly, chemoembolization or ablation therapy can affect both the gross and microscopic findings.

Treated tumors are staged in the same fashion as nontreated tumors. The percent necrosis should be estimated to the nearest 10% using the gross and/or histologic findings. The percent necrosis after TACE
ranges from 0% to 100%, with an average of 50% to 70% necrosis in those tumors with necrosis.46,93,94 About 30% of tumors are completely necrotic.46 Hepatocellular carcinomas with strong and diffuse CD34 staining and negative VEGF staining appear to be more resistant to TACE.95 There is no reliable way to distinguish necrosis that resulted from treatment versus spontaneous tumor necrosis and all necrosis is included in the estimate.

Figure 28.29 Hepatocellular carcinoma, moderately differentiated. The tumor morphology strongly suggests hepatic differentiation and the tumor is clearly malignant.

Figure 28.30 Hepatocellular carcinoma, poorly differentiated. There is striking nuclear atypia. Immunostains demonstrated hepatic differentiation. Imaging studies reported no other tumor sites.

Treatment can lead to other histologic changes, such as intratumoral inflammation and intratumoral fibrosis. Treated tumors are also more likely to express CK19 and have areas with biliary and/or spindled cell morphology.96,97 The potential effect of treatment on tumor grade has not been well studied, but one study found a higher tumor grade in TACE-treated hepatocellular carcinomas.98 This same study also found an increase in multinucleated tumor cells and cytoplasmic hyaline inclusions following treatment.98 Embolic beads can be found in the tumor and in the adjacent nontumor liver. Embolic beads can also escape the liver, leading to damage to other organs, in particular the stomach (Fig. 28.1).

Figure 28.31 Hepatocellular carcinoma, poorly differentiated. This tumor has scant basophilic cytoplasm with large pleomorphic nuclei.

Immunohistochemical stains and special stains

Immunohistochemical stains are used to (1) decide if a well-differentiated hepatic tumor is benign or malignant (Table 28.7) and (2) decide if a clearly malignant tumor is hepatocellular carcinoma (Table 28.8).

The differential for a well-differentiated hepatic tumor depends in part on the background liver. In noncirrhotic livers, the differential is focal nodular hyperplasia, hepatic adenoma, and hepatocellular carcinoma. In a cirrhotic liver, the differential is focal nodular hyperplasia-like lesion, macroregenerative nodule, dysplastic nodule, and hepatocellular carcinoma. Of note, some nodules in cirrhotic livers can express CRP or SAA,99,100 but at this point the term hepatic adenoma for these nodules is hard to justify, both because it leads to unnecessary confusion for the clinical team and because there is no evidence that such lesions behave clinically like an adenoma.

A wide variety of immunostains can be used when evaluating poorly differentiated tumors. Some of the more commonly used stains are shown in Table 28.9 along with the frequency of positive staining in hepatocellular carcinomas.

Table 28.7 Stains used to distinguish benign from malignant hepatic lesions




  • Rarely (<1%), well-differentiated hepatocellular carcinomas will have no obvious reticulin loss


  • To be helpful, staining should be significantly higher than background liver

  • Of note, not all hepatocellular carcinomas will have a higher proliferative rate


  • Not widely used as there are better stains

  • Strong diffuse staining is commonly but not always found in hepatocellular carcinoma. Also, some hepatic adenomas can show strong and diffuse CD34 staining

Glypican 3

  • Positive in about 50% of well-differentiated hepatocellular carcinomas

  • Also stains lipofuschin and benign hepatocytes with significant inflammation

  • Focal staining also found in some dysplastic nodules


  • Positive in about one-third of all hepatocellular carcinomas

Glutamine synthetase

  • Diffuse and strong positive in about 50% of hepatocellular carcinomas

  • β-Catenin inactivated adenoma also diffuse positive

Universal rules governing the use of immunohistochemistry

Four fundamental, immutable laws govern the use of immunohistochemistry in tumors.

Law 1. All special stains should be interpreted in conjunction with the H&E findings. For example, if a tumor’s morphology is inconsistent with hepatocellular carcinoma, immunostain findings will not change that. Cross checking positive immunostains with the H&E is also important to avoid mistaking the positive staining of entrapped hepatocytes for positive staining of a tumor.

Law 2. The sensitivity and specificity of immunostains invariably get worse as more studies are published. The first one is (almost) always the best, but the sensitivity and specificity will fall as more data accumulates. This is because the performances of stains depend on a variety of factors including those specific to laboratory methods and those that reflect biology, such as the strong correlation that can be seen between stain sensitivity and tumor grade, underlying liver disease, and/or tumor subtype. For these

reasons, most expert liver pathologists will choose from a panel of top performing stains because a wisely chosen panel of stains will help mitigate individual weaknesses of different immunostains.

Table 28.8 Stains used to demonstrate hepatocellular differentiation



Arginase 1

  • Positive in 90% of hepatocellular carcinomas

  • Negative tumors cluster at both ends of differentiation: well and poorly differentiated can both be negative


  • Positive in 90% of hepatocellular carcinomas

  • Negative cases are usually poorly differentiated

Glypican 3

  • Positive in 80%-90% of hepatocellular carcinomas

  • Performs poorly in well-differentiated hepatocellular carcinomas, where only 50% are positive

In situ hybridization for albumin

  • Positive in >95% of hepatocellular carcinomas

  • Positive in 80%-95% of cholangiocarcinomas

Polyclonal CEA (canalicular pattern)

  • Positive in 60%-80% of hepatocellular carcinomas

  • Canalicular staining indicates hepatic differentiation

  • Membranous or cytoplasmic staining is compatible with hepatocellular carcinoma but not specific

  • There are better stains as this one tends to be challenging to interpret in poorly differentiated carcinomas

CD10 (canalicular pattern)

  • Performance is similar to polyclonal CEA


  • Positive in about one-third of hepatocellular carcinomas

  • Can be useful in poorly differentiated tumors, which may lose expression of other markers such as HepPar and Arginase

None of these stains are 100% specific and they are best used in combination with clinical findings and H&E morphology

Table 28.9 Frequency of positive staining for nonhepatocellular markers in hepatocellular carcinoma


Approximate frequency


Epithelial markers



Cytokeratin AE1/AE3




Positive tumors have a worse prognosis. Positive cases typically coexpress CK7



Hepatocellular carcinomas that are cholestatic or occur in younger individuals (<40) are positive in the majority of cases



Positive cases typically coexpress CK7





Stains for site of origin



Staining is often focal in hepatocellular carcinoma. Also, 20%-30% of cholangiocarcinomas are CDX2-positive



TTF1 can show cytoplasmic staining with some antibodies, but not nuclear staining




Limited data



May be a marker of “stem cellness”



Staining is typically strong and cytoplasmic and can have a membranous accentuation





Mesothelioma markers



Usually patchy. Strong diffuse staining is seen in many cholangiocarcinomas





Limited data



Limited data

Stains used to subtype hepatic adenomas



Stain is positive in normal liver; lost staining is informative



Should be strong and fairly diffuse to be informative; some clones show considerable background staining



β-Catenin (nuclear)


Can be very patchy

Glutamine synthetase


Should be strong and diffuse to be informative

Other markers

CD117 (Kit)




Nonneoplastic hepatocytes are also positive. Membranous staining is seen in both tumor and nontumor



About 10% of conventional hepatocellular carcinomas are positive. Sarcomatoid hepatocellular carcinomas are positive

Law 3. If there is a discrepancy between the morphology and immunohistochemical findings, additional studies must be performed. Examples of additional studies include submitting more sections on resection specimens, repeating discrepant stains, and performing an additional round of immunostains that focuses on clarify the differential(s) raised by the discrepant findings.

Law 4. A difficult case is the wrong time to first use a stain with which you’re not familiar. The temptation can be very strong to use an unfamiliar stain in the hopes of getting yourself out of a diagnostic jam. However, this often leads to diagnostic misadventures. If you cannot resist temptation, at a minimum, the results of unfamiliar stains should be reviewed (not just verbally, but actually reviewed) by somebody with skill in interpreting the stain.

Only gold members can continue reading. Log In or Register to continue

Stay updated, free articles. Join our Telegram channel

Nov 24, 2019 | Posted by in GASTROENTEROLOGY | Comments Off on Hepatocellular Carcinoma

Full access? Get Clinical Tree

Get Clinical Tree app for offline access