Background
Primary hepatic sarcomatous carcinoma (PHSC) is a rare malignancy composed of both carcinomatous (either hepatocellular or cholangiocellular) and sarcomatous components [
1]. This entity is differentiated from hepatic carcinosarcoma (CS), which contains both hepatocellular carcinoma and a true heterogonous sarcoma component such as chondrosarcoma, malignant fibrous histiocytoma, osteosarcoma, leiomyosarcoma, fibrosarcoma, rhabdomyosarcoma, and other mesenchymal tumors arising in the liver [
1‐
4]. Various terms have been used to describe these biphasic tumors, including CS, sarcomatoid carcinoma, pseudosarcoma, and spindle cell carcinoma. Sarcomatous carcinoma (SC) and CS, however, are most commonly used and easily confused. Currently, pathologists are still faced with the dilemma of how to distinguish CS from SC when making pathological diagnoses. It is suggested that such mixed tumors should be diagnosed as SC when the sarcomatous component is predominantly composed of spindle cells, but the epithelial cells are still morphologically, immunohistochemically, and ultrastructurally identifiable [
1,
5‐
8].
PHSC is extremely rare, accounting for only 0.2% of the primary malignant liver tumors [
1]. The prevalence of sarcomatous hepatocellular carcinoma (S-HCC) and sarcomatous intrahepatic cholangiocarcinoma (S-ICC) is 1.8–9.4% of hepatocellular carcinoma (HCC) and 4.5% of intrahepatic cholangiocarcinoma (ICC) [
1,
9,
10], and the sarcomatous combined hepatocellular and cholangiocarcinoma (S-HCC–CC) is rarely published in the literature, with no more than 20 cases [
11‐
14]. It was reported in a previous study that the PHSC carried higher aggressiveness and poorer prognosis [
1,
5,
9‐
11,
15]. However, its prognostic significance remained unclear. Since the literature was restricted to either case reports or small case series [
7,
16‐
20], the cross-sectional imaging features of PHSC were largely ill-defined and clinical diagnosis was difficult. The purpose of our study is to further characterize these tumors by reporting the imaging findings and clinical features of a series of 10 patients, improving the understanding and diagnosis of tumors.
Discussion
In agreement with previous reports, most patients (70%) in this study were found to be positive for HBsAg and negative for hepatitis C virus antigen (HCVAg), suggesting that hepatitis B virus infection might be related to PHSC [
1,
5], especially to the S-HCC. Unlike S-HCCs, S-ICCs were reported to show a relatively high incidence in patients with HCV-related hepatitis as the “normal” ICCs [
24,
25]. Nevertheless, the two patients with S-ICCs in our study had history of alcoholic cirrhosis without any viral hepatitis or cirrhosis. The elevation of AFP was found in 57.1% of the patients with S-HCCs, which was slightly lower than that in “normal” HCC ones. As reported in some previous studies, S-HCC was characterized by lower serum AFP level [
9,
15]. Meanwhile, the opposite conclusion came from the other studies [
1,
5]. Therefore, further research about the relationship among AFP, hepatitis or cirrhosis and PHSC needs to be done with a larger sample in the future.
Our study demonstrated that PHSC generally presented hypovascularity seen as peripheral enhancement on the arterial phase imaging. The PHSC was characterized by the peripheral viable cancerous tissue with fibrous stroma and central necrosis or hemorrhage [
14,
26]. Similar to prior results [
6,
15], the necrosis was more frequently seen with a high frequency of 85.7% (six of seven) in the surgically resected tumors in our study. The poorly differentiated cells of the sarcomatoid component grew so rapidly that the neovasculature could not adequately supply the fast-growing malignant cells, resulting in the central necrosis. In additional to the peripheral ring enhancement, when the necrosis was accompanied by fibrous septum or was scattered
, the tumors might exhibit heterogeneous enhancement distribution such as a mix of rim and internal or internal only heterogeneous enhancement as shown in our study.
In the present study, the most common dynamic pattern of enhancement was progressive enhancement with persistent enhancement included in S-ICCs and S-HCCs. Additionally, the washout could also be found in S-HCCs. It was concluded in a previous radiologic–pathologic correlation study of S-HCC-CC that areas of arterial phase enhancement and later phase wash out were suggestive of HCC, progressive enhancements suggestive of CC, persistent and slight hypoenhancement in the subcapsular region suspicious for sarcomatoid transformation and hypo-density with little or no enhancement in keeping with the tumor necrosis [
12]. Histologically, pleomorphic spindle shaped cells having loose mutual contact and fibrous stroma showed persistent or progressive enhancement, viable cells displayed a trabecular pattern with little or without fibrous stroma in the periphery of the tumor exhibited typical HCC enhancement washout, and a definite glandular pattern with fibrous stroma also showed progressive enhancement [
11,
26]. Similar to the result, the S-HCC-CC in our study also presented as a lobulated multinodular confluent tumor with different dynamic enhancement patterns (Figs.
6 and
7). It was supposed that the diverse tissue composition might determine the various enhancement patterns. However, even the S-HCC exhibited different dynamic enhancement characters. So we inferred that the diverse imaging findings of the sarcomatous carcinoma not only depended on the tissue composition but also the proportion. The manifestations varied when certain histopathological components ranged from focal to prominent. Therefore, it may be necessary to further sub-classify PHSC not only using morphological criteria that define biologically distinct subgroups but also the amount of certain component, which may be related not only to the imaging findings but sometimes even the biologic behavior of these cancers [
11].
On the MR imaging of five patients, bright signal intensity similar to that of cyst or hemangioma on T2WI might be explained by necrosis [
14] and the signal might be attributed to hemorrhage seen as hypointensity or hyperintensity on T1WI and hypointensity on T2WI not only inside the center of the tumor but also in the subcapsular area. The other S-HCCs showed inhomogeneous high signals on DWI and T2WI, similar to the “normal” type. Five tumor capsules were observed on imaging, and only two of them were confirmed in pathology. It was reported that a high incidence was correlated with well differentiated HCC and the tumor capsules were much more common in ordinary HCC when compared with the S-HCC [
5]. Similar to the result, we did not find capsules in S-HCCs, except for a complete capsule in the S-ICC and a partial one in the S-HCC-CC pathologically.
All of the PHSCs in our study located near the liver subcapsular area where the liver capsules were frequently involved (five of seven) and sometimes subcapsular metastasis or peritoneal seeding (three of 10) occurred. The invasion of the liver capsular in sarcomatous carcinoma was more common than that in the “normal” type, which might be explained by the sarcomatoid component [
20]. It was found that one of mass-forming S-ICCs with a subcapsular rupture in our study protruded out of the liver contours, involve diaphragm and resulted in multifocal tumor seeding. To the best of our knowledge, four of the 30 S-ICCs reported in the English literature presented with spontaneous rupture thus far [
14,
27]. The spontaneous rupture of hepatic tumor was the result of a complex interaction of various factors such as location, composition or pressure and so on [
28‐
30]. The “normal” cholangiocarcinoma seldom ruptured spontaneously as a hard tumor with abundant fibrous stroma. Nevertheless, the S-ICC in our study presented as multiple cysts indicating more necrosis and less fibrous stroma, resulting in a fragile tumor. In comparison to the “normal” cholangiocarcinoma, the S-ICC was prone to rupture.
In additional to the liver capsular involvement found in our study, we also observed that the vascular invasion or thrombosis (50%), intrahepatic metastasis or satellite nodule (40%) were relatively common, similar to prior results [
9,
16]. As for the treatment of the tumor, it was reported that radical resection at an early stage may contribute to a relatively favorable prognosis. Nevertheless, its treatment protocols and effects were still controversial. Sometimes surgical removal of the tumor alone seemed to be insufficient. The sarcomatous carcinoma was thought to be associated with aggressive tumor biology, frequent metastasis, low resectability and frequent recurrence after curative resection, segmentectomy and even liver transplantation [
1,
5,
31]. The TNM stage was revealed to be one of independent risk factors for overall survival [
5], which had been partially proved in our study that all deaths occurred in PHSC patients with stage III-IV between 2 and 5 months and no tumor recurrence or progression happened in the patients with stage I between 16 and 24 months.
We acknowledge several limitations to our study. The major one was its retrospective nature make the complete section-by-section matching between imaging and pathologic findings technically unfeasible. Thus, to some extent our explanations for the imaging findings might be considered speculative. Due to the retrospective nature, we were unable to categorize the amount of the certain component such as sarcomatous component and evaluated its radiological findings and prognostic significance. The second limitation was relatively small sample size, which had inherent shortcomings but was unavoidable because of the rare incidence of these tumors. The last one was that our study was only descriptive for a selected case without control group and no statistical data on differential features between them were obtained.