Certain tumors can histologically mimic hepatocellular carcinoma, including epithelioid angiomyolipomas, renal cell carcinomas, adrenal cortical carcinomas, neuroendocrine tumors, and melanoma. These tumors can show a trabecular or acinar or solid growth pattern that can closely resemble hepatocellular carcinoma. In addition, the tumor cells in these cases can show cytological features that further suggest hepatocellular differentiation, including eosinophilic and granular or clear cytoplasm.

Most angiomyolipomas are distinguished from hepatocellular carcinoma by their triphasic nature, but epithelioid angiomyolipomas, especially in biopsy specimens, can be difficult to distinguish from a well-differentiated hepatocellular neoplasm. Immunohistochemical stains are helpful here, as angiomyolipomas are positive for smooth muscle actin, HMB-45 and Melan-A, but negative for cytokeratin, HepPar1, and arginase-1 (Figs. 15.4, 15.5, and 15.6).

Both clear cell and chromophobe renal cell carcinoma can be mistaken for hepatocellular neoplasms. A history of renal malignancy or the presence of a renal mass can help reach the correct diagnosis. Furthermore, some clear cell hepatocellular carcinomas have areas of more typical non-clear morphology, which can help distinguish it from metastatic clear cell renal cell carcinoma. Immunohistochemical stains are very helpful in the differential diagnosis; clear cell renal cell carcinomas are usually positive for vimentin while most hepatocellular carcinomas are either negative or show patchy staining (90 %), with the exception of poorly differentiated or spindle cell hepatocellular carcinomas which can show more diffuse reactivity for vimentin [2]. RCC antigen is negative in hepatocellular carcinoma, but RCC antigen has a lower sensitivity and specificity for metastatic renal cell carcinoma and should be interpreted cautiously [3]. On the other hand, the PAX8 immunostain is positive in the majority of metastatic clear cell renal cell carcinomas [4, 5]. However, a recent publication has shown that up to 20 % of clear cell hepatocellular carcinomas can also be positive for PAX8 [6], so other markers for hepatic differentiation also need to be examined. Clear cell renal cell carcinomas are negative for HepPar1, polyclonal CEA (canalicular pattern), and arginase-1 [7–9]. Chromophobe renal cell carcinoma, a close mimicker of hepatocellular neoplasms, is positive for PAX8 and CD117 and negative HepPar1 and arginase-1 (Figs. 15.7, 15.8, 15.9, and 15.10); this immunoprofile helps distinguish it from hepatocellular neoplasms. Of note, glypican-3 is not helpful in this differential diagnosis, as a subset of chromophobe renal cell carcinomas is positive.

Adrenal cortical carcinomas can involve the liver by metastasis or direct extension and can be difficult to differentiate morphologically from hepatocellular carcinoma (Fig. 15.11). Their immunoprofile, however, is distinct from hepatocellular carcinoma; adrenal cortical carcinomas are often positive for inhibin and Melan-A and are typically negative for HepPar1 (about 5–10 % of cases can be positive), polyclonal CEA (canalicular pattern), and arginase-1 [8–11].

Neuroendocrine tumors can resemble hepatocellular carcinoma. Some cases may have prominent clear cell morphology, while others will have ordinary cytology but mimic hepatocellular carcinoma with trabecular or acinar growth patterns. Neuroendocrine tumors, however, are strongly and diffusely positive for chromogranin and/or synaptophysin while they are negative for HepPar1 and arginase-1 (Figs. 15.12, 15.13, 15.14, and 15.15) [8, 9]. Hepatocellular carcinomas can show also focal neuroendocrine differentiation [12], especially when cholestatic, but will also be positive for markers of hepatic differentiation. Interestingly, neuroendocrine tumors metastatic to the liver that produce insulin can have a very distinctive feature, characterized by a marked steatosis involving only a rim of hepatocytes surrounding the neuroendocrine tumor (Figs. 15.16 and 15.17) [13].

Finally, melanomas can mimic hepatocellular carcinoma. A history of melanoma can often guide the pathologist to the correct diagnosis. Immunohistochemistry is invaluable in this setting, as melanoma is positive for S100, Melan-A, and HMB-45 and negative for HepPar1, polyclonal CEA, and arginase-1 (Figs. 15.18, 15.19, 15.20, and 15.21). Glypican-3, however, is usually not helpful in this setting as glypican-3 is frequently positive in melanoma [14].

There are several excellent potential panels of immunohistochemical stains, which can be useful in differentiating hepatocellular carcinoma from its metastatic morphological mimickers. If a well to moderately differentiated hepatocellular carcinoma is in the differential, than one useful approach would be to perform an immunostain for hepatic differentiation, such as HepPar-1 or arginase-1. If they are negative, then additional follow-up stains can be used to further classify the tumor. Depending on the morphology and the clinical findings, these additional stains might include PAX-8, inhibin, vimentin, S100, HMB-45, Melan-A, SMA, CD-117, polyclonal CEA, and CAM5.2 (Table 15.2). As discussed previously, an effective approach for poorly differentiated tumors is a first round of stains to determine lineage (carcinoma, sarcoma, melanoma, lymphoma), followed by additional stains to help identify the most likely site of origin.

Cholangiocarcinomas can assume many of the histological patterns of metastatic adenocarcinomas. However, if the bile ducts near the adenocarcinoma show high-grade biliary intraepithelial neoplasia (BilIN-3), this provides strong morphological evidence that the tumor is a primary cholangiocarcinoma rather than a metastasis. Other precursor lesions include mucinous cystic neoplasms and intraductal papillary neoplasms. However, precursor lesions are not identified in most cases. Furthermore, it is important to remember that some metastatic adenocarcinomas, such as colorectal adenocarcinomas, can have a prominent intrabiliary ductal growth pattern (“cancerization of the bile ducts”), which can closely mimic high-grade biliary intraductal neoplasia (BilIN-3).

A panel of immunostains is often required to distinguish cholangiocarcinoma from metastatic adenocarcinoma. Morphologically, adenocarcinomas arising from the upper gastrointestinal tract, pancreas, and biliary tract appear very similar to intrahepatic cholangiocarcinoma. Adenocarcinomas originating from all of these sites can be composed of small tubular or tubulopapillary structures embedded in a fibrous stroma. Therefore, distinction between intrahepatic carcinoma and metastases from upper gastrointestinal or pancreatobiliary sources is difficult, if not impossible, based on morphologic grounds alone. Furthermore, there are currently no immunohistochemical stains available in routine practice that can accurately distinguish between an intrahepatic cholangiocarcinoma and metastatic adenocarcinoma from upper gastrointestinal or pancreatic origin. In current practice, such distinctions rely first on ruling out other possible metastatic tumors and, when the morphology and immunostains are consistent with a biliary, pancreas, or upper gastrointestinal primary. When this is done, a final determination then relies on correlation with the imaging and endoscopic findings to identify the most likely site of origin. The lack of positive, affirmative immunostains or molecular assays that identifies biliary differentiation is a major weakness in our diagnostic armamentarium, one that will be hopefully rectified in the future.