Methods
Collection of cases
A total of 75 cases treated between April 2010 and April 2020 were identified in the pathology archive system of the Department of Pathology at The First Affiliated Hospital of Zhejiang University. Hepatectomy had been performed in all patients. Written informed consent was obtained from the patients before the publication of this report to allow for description of their protected health information, including accompanying images.
Clinical data
Clinical data were retrospectively retrieved from the patients’ records, including age, sex, location, tumor size, clinical presentation, routine blood tests, liver function tests, hepatitis B virus surface antigen, hepatitis C virus (HCV) antigen, serum tumor markers [such as alpha-fetoprotein (AFP), cancer antigen (CA) 19–9, CA125, and carcinoembryonic antigen (CEA)], imaging presentation, and preoperative diagnosis. Follow-up data were obtained from clinical records or telephonically through October 2020.
Immunohistochemistry
Immunohistochemical staining for HMB45, Melan-A, S-100, smooth muscle actin (SMA), AFP, Hep, CD34, Ki67, CD163, CD68, CD117, TFE3, P53, Syn, CgA, and pancytokeratin (CK) was performed using the Envision method#. All immunohistochemical staining were performed using the BOND-III automated immunostainer (Leica Biosystems, Buffalo Grove, IL, USA).
Hepatic imaging
Patients were subjected to evaluation using imaging modalities. A total of 56 patients underwent computed tomography (CT), 47 underwent magnetic resonance imaging (MRI), and 67 underwent ultrasonography (US). We recorded the following characteristics: lesion morphology; blood vessel invasion; bile duct invasion; and lymphadenectasis.
Statistical analysis
Statistical analysis was performed using SPSS software version 19.0 (IBM Corp., Armonk, NY, USA). Variables were compared using analysis of variance and expressed as the mean ± standard deviation or median. All statistical p values were two-sided, with p values of <0.05 were considered to indicate statistical significance.
Results
Patients and clinical presentation
A total of 75 patients with hepatic PEComa, 52 women (69.3%) and 23 men (30.7%) were included; their median age was 48 years (range: 24–64). Fifty-five patients (73.3%) were asymptomatic, and lesions were found incidentally during regular physical examinations. Overall, 15 patients presented with abdominal discomfort (20%) and 5 patients presented with fatigue (6.7%). Levels of liver function and tumor markers, such as AFP, CEA, CA19-9, and CA125, were normal in the patients. None of the patients had tuberous sclerosis complex or HCV infection; seven patients (9.3%) had a history of hepatitis B virus (HBV) infection, with four having a history of HBV for >10 years and one having a history of hepatitis B that progressed to cirrhosis over a >20-year period. The tumors were located in the right liver lobe in 35 patients (46.7%), the left lobe in 34 (45.3%), both right and left lobes in 3 (4%), and the caudate lobe in 3 patients (4%). With respect to tumor size, the largest lesion measured 14 × 12 × 10 cm; four patients (5.3%) had a lesion diameter of >10 cm and seventeen patients (22.7%) had a lesion size between 5 cm and 9 cm. Twenty-three patients (30.7%) had lesion sizes between 3 cm and 5 cm and thirty-one patients (41.3%) had lesion sizes of <3 cm. Overall, 69 patients (92%) presented with a solitary, well-circumscribed heterogeneous mass and 6 patients (8%) presented with multiple masses. One male patient exhibited recurrence in the right liver and abdominal implant metastasis 3 years after surgery for left renal PEComa; one female patient showed recurrence in the left liver after right hepatic PEComa resection performed 10 years ago, and two patients received transcatheter arterial chemoembolization (TACE) in their caudate lobe before surgery (Table 1).
Table 1Demographic characteristics of patients with hepatic PEComa
| Cases (n=75) | Female (n=52) | Male (n=23) | P-value |
|---|
| Age (Mean ± SD) | 48.61±9.70 | 46.77±9.25 | 52.78±9.60 | 0.01 |
| Symptom, n (%) | | | | |
| No symptom | 55 (73.3) | 39 (75.0) | 16 (69.6) | 0.01 |
| fatigue | 5 (6.7) | 4 (7.7) | 1 (4.3) | 0.18 |
| Abdominal discomfort | 15 (20) | 9 (17.3) | 6 (26.1) | 0.44 |
| Location of tumors, n (%) | | | | |
| Right lobe | 35 (46.7) | 30 (57.7) | 9 (39.1) | 0.01 |
| Left lobe | 34 (45.3) | 20 (38.5) | 11 (47.8) | 0.11 |
| caudate lobe | 3 (4) | 0 | 3 (100) | 1 |
| Both side | 3 (4) | 3 (5.8) | 0 | 1 |
| Lesion number, n (%) | | | | |
| solitary | 69 (92) | 47 (90.4) | 22 (95.7) | 0.01 |
| multiple | 6 (8) | 5 (9.6) | 1 (4.3) | 0.10 |
| Tumor size, n (%) | | | | |
| ≥10cm | 4 (5.3) | 2 (3.8) | 2 (8.7) | 1 |
| 9–5 cm | 17 (22.7) | 12 (23.1) | 5 (21.7) | 0.09 |
| 3–5 cm | 23 (30.7) | 16 (30.8) | 7 (30.4) | 0.06 |
| <3 cm | 31 (41.3) | 20 (38.5) | 11 (47.8) | 0.11 |
| Blood test | | | | |
| WBC (Mean ± SD) | 5.49±1.71 | 5.47±1.86 | 5.53±1.36 | 0.91 |
| N% (Mean ± SD) | 60.57±9.62 | 59.90±10.57 | 62.09±7.08 | 0.39 |
| TB (Mean ± SD) | 12.54±7.50 | 11.99±7.40 | 13.75±7.76 | 0.38 |
| DB (Mean ± SD) | 4.34±2.62 | 4.16±2.63 | 4.74±2.59 | 0.40 |
| IB, median | 7.10 (5.0,10.0) | 7.00 (4.90,9.78) | 8 (5,12.15) | 0.27 |
| Ca199, median | 7.3 (4.2,13.90) | 9.7 (6.2,15.20) | 4.75 (2.08,11.95) | 0.01 |
| Ca125, median | 8.8 (5.6,15.70) | 9.35 (6.28,16.50) | 7.85 (4.85,12.80) | 0.29 |
| CEA, median | 1.6 (1.1,2.2) | 1.5 (1,2.2) | 1.75 (1.40,2.38) | 0.12 |
| HBsAg, n (%) | | | | |
| positive | 7 (9.3) | 3 (5.7) | 4 (17.4) | 0.71 |
| negative | 68 (90.7) | 49 (94.2) | 19 (82.6) | 0.01 |
Imaging presentation
Patients underwent imaging examination. All images demonstrated that the lesions were round or oval with clear borders. There was no blood vessel or bile duct invasion, except in one patient who had extensive abdominal metastasis and a lesion in the liver with no clear boundary. Most patients who underwent US had a hyperechoic or hypoechoic mass.
On CT examination, low density signal was usually observed in the scanning period, and the lesions were unevenly enhanced in the arterial phase and were still enhanced in the portal and delay phases. Additionally, the lesions often contained fat.
On MRI examination, most patients exhibited a slightly hypointense or isointense signal on T1-weighted imaging (T1WI), slight hyperintense signal on T2-weighted imaging (T2WI), and unevenly hyperintense signal on diffusion-weighted imaging (DWI). The enhanced images showed significant and homogeneous enhancement during the arterial phase, which indicated a relatively hypervascular, heterogeneous composition. This enhancement gradually weakened during the portal venous and equilibrium phases, and insufficient enhancement was observed in the late parenchymal phase.
Taken together, the diagnostic accuracy was 30.5%, 31.9%, and 10.4% within the groups of patients who underwent CT, MRI, and US examinations, respectively. According to the consistency of the combined diagnosis, the diagnostic accuracy with two imaging methods was 7.4%, 18.2%, and 31% within the US + CT, US + MRI, and CT + MRI groups, respectively. The combined diagnostic accuracy of these three imaging methods was 7.4%. The imaging characteristics were inaccurate for diagnosing these PEComas (Table 2).
Table 2Diagnostic accuracy of imaging modalities in patients with hepatic PEComa
| Imaging methods | Cases (n) | Simultaneous diagnostic accuracy (n/%) |
|---|
| US | 67 | 7 (10.4) |
| CT | 56 | 17 (30.5) |
| MRI | 47 | 15 (31.9) |
| Combined two imaging examinations |
| US+CT | 54 | 4 (7.4) |
| US+MRI | 44 | 8 (18.2) |
| CT+MRI | 29 | 9 (31) |
| Combined three imaging examinations |
| US+CT+MRI | 27 | 2 (7.4) |
Pathological findings
Thirty-eight patients had CAML and 37 patients had EAML; the tumors were circumscribed, either unencapsulated or encapsulated. The cut surface was soft, with color ranging from yellow to dark red, and areas of necrosis were present in four patients. Routine histopathological examination with hematoxylin and eosin staining was performed. Immunohistochemically, the positive expression rate of HMB45 was 98.4%, that of SMA was 93.2%, and that of Melan-A was 98.2%. For S100, the expression rate was 14.3% in general, and in hepatocytes was 4.3% (Fig. 1). The tumor cells were negative for AFP, CD117, CK, and TFE3 (Table 3).
Table 3Immunohistochemistry
| Cases
| Male (n=23)
| Female (n=52)
|
|---|
| Count (n/%) | Positive (n/%) | Count | Positive (n/%) | Count | Positive (n/%) |
|---|
| AFP | 24/33.8 | 0 | 8 | 0 | 16 | 0 |
| Hepatocyte | 49/69 | 2 (4.3) | 15 | 2 (13.3) | 34 | 0 |
| Ki 67 | 29/40.9 | 27 (93.1) | 7 | 6 (85.7) | 22 | 21 (95.5) |
| CD34 | 28/39.4 | 20 (71.4) | 6 | 2 (33.5) | 22 | 16 (72.7) |
| CD163 | 1/1.4 | 1 (100) | 1 | 1 (100) | 0 | 0 |
| CD68 | 2/2.8 | 1 (50) | 2 | 1 (50) | 0 | 0 |
| CD117 | 8/11.3 | 0 | 1 | 0 | 7 | 0 |
| Syn | 1/1.4 | 1 (100) | 1 | 1 (100) | 0 | 0 |
| CGA | 1/1.4 | 1 (100) | 1 | 1 (100) | 0 | 0 |
| CK | 38/53.5 | 0 | 9 | 0 | 29 | 0 |
| HMB45 | 61/85.9 | 60 (98.4) | 18 | 18 (100) | 43 | 42 (2.3) |
| Melan A | 56/78.9 | 55 (98.2) | 16 | 15 (93.8) | 40 | 40 (100) |
| SMA | 44/62 | 41 (93.2) | 11 | 9 (81.8) | 33 | 32 (97.) |
| S-100 | 28/39.4 | 4 (14.3) | 6 | 6 (100) | 22 | 4 (18.2) |
| TFE3 | 3/4.2 | 0 | 1 | 0 | 2 | 0 |
| P53 | 4/5.6 | 2 (50) | 0 | 0 | 4 | 2 (50) |
Microscopically, CAML comprised a mixture of thick-walled blood vessels, different spindle cells, and mature adipose tissue. EAML comprised epithelioid cells with pale, clear, eosinophilic, or foamy cytoplasm. Mitotic figures were relatively rare. The tumor cells were different in size and irregularly arranged. The cytoplasm was rich with red staining, the nucleus was large and oddly shaped, and the nucleolus was prominent (Fig. 2).
Treatment and follow-up
In our study, most patients underwent partial hepatectomy, and the follow-up study was completed for 47 patients. The follow-up continued until the end of the study. During the follow-up period, one patient whose liver tumor size was 12 × 11 × 8 cm with necrosis had a mediastinal metastasis 5 years after surgery; another patient whose liver tumor size was 1.7 × 1.5 cm had a lung metastasis 4 years after surgery, and two patients had EAML.
Discussion
PEComas are mesenchymal tumors with perivascular epithelioid cell differentiation comprising distinct cells that show a focal association with blood vessels and walls; they usually express melanocytic and smooth muscle markers.4 The “PEComas group” comprises AML, CCST, LAM, CCMMT, and other NOS. AML is the most common pathological type; it includes CAML and EAML, and other pathological types are relatively rare.2 PEComas are predominantly diagnosed in the uterus or kidney and rarely occur in the liver.5 Hepatic PEComa usually presents a solitary, well-circumscribed heterogeneous mass in patients with no potential liver diseases. A higher frequency of multiple masses can occasionally accompany tuberous sclerosis;6 however, we did not find tuberous sclerosis in our patients, and most patients (92%) presented with solitary lesions. Four patients with EAML and three with CAML had HBV infection; one patient with cirrhosis was diagnosed with HCC and received TACE treatment before hepatic surgery. Hepatic PEComa is mostly asymptomatic, and serological test results are normal. Owing to the highly variable histological composition of hepatic PEComa, these tumors often do not have typical imaging characteristics; thus, their diagnosis is difficult.
Clinically, the differential diagnosis of a hepatic lesion before surgery primarily depends on imaging examinations, and how to make a correct preoperative diagnosis with imaging examinations is worth investigating. Another study has described the morphological imaging characteristics of hepatic PEComa on CT, MRI, or US.5 It has been suggested that dynamic imaging is the most reliable method for differentiating AML from HCC7,8 because AML can be easily distinguished from hepatocytes with adipose tissue. Previous reports have suggested that contrast-enhanced CT or MRI is a useful diagnostic modality for hypervascularity and arteriovenous characteristics in PEComas.9–11 Researchers have now defined PEComas as neoplasms without adipocytes, thereby distinguishing them from classical AML.12 The diagnostic accuracy of liver PEComa using imaging modalities is low and that of CT combined with MRI is only 20%.13,14 In our study, the diagnostic accuracy of CT combined with MRI was 31%. Contrast-enhanced CT and MRI of PEComa showed that the lesions were significantly and heterogeneously enhanced in the arterial phase, continually enhanced in the portal venous phase, and slightly hypodense in the delayed phase, which exhibits a broad spectrum of imaging findings.
Hepatic PEComa can be diverse based on echogenicity of blood flow in or around lesions. For cases with fat, particularly in lesions with fat comprising >50% of the tumor, AML can be easily diagnosed. Diagnosis of tumors with no or little fat is challenging.15 Most tumors show low signal intensity on T1WI and high signal intensity on fat-suppressed T2WI and DWI images on MRI. Unfortunately, these findings are not specific in several tumors.16 Previous reports have suggested hypervascularity and arteriovenous connections as the features of PEComa. The current study demonstrated the utility of MRI in the early diagnosis of liver PEComa. Combining our cases, we created a form to summarize the characteristics of different hepatocyte tumors to distinguish PEComas (Table 4).
Table 4Imaging characteristics of different tumors
| Pathological features | Enhancement reason | Blood supply | Plain MRI | Arterial phase | Portal phase | Venous phase | Enhanced form |
|---|
| AML | Central stellate scar, no liver lobules, bile duct fibrosis | Composed of blood vessels and fibrous tissue | Small arteries | T2W1 is hyperintense or slightly hyperintense; T1W1 is isointense or slightly hypointense | Heterogeneity strengthening | Fades away, strengthened area continues | Isointense or hypointense | Fast in and slow out |
| Hemangioma | Blood-filled vascular cavity and fibrous separation | Blood pool filling | Hepatic artery | T2W1 is hyperintense; T1W1 is slightly hypointense | Peripheral nodular enhancement | Extending toward the center | Filling reinforcement | Fast in and slow out |
| FNH | Abnormal arrangement of normal stem cells, central stellate scar, tubule structure | Scars are composed of blood vessels and bile ducts | Small arteries; blood is eccentric | T2W1 is isointense or slightly hyperintense; T1W1 is isointense or slightly hypointense (center high) | Clearly uniform strengthening, low center | Quick removal, low center | Isointense, strengthened at the center | Fast in and out, center delay strengthened |
| Hepatocellular adenoma | Hepatocytes are hypertrophied and lipid-bearing, bile ducts are absent and can be malignant, and 1/3 have pseudocapsule | Absence of portal vein and bile duct, easy bleeding | Subcapsular arteriole blood flow is concentric | Easy to bleed, signal diversification, envelope is hypointense | Significantly strengthened | Isointense or hypointense | Isointense or hypointense | Fast in and out |
| Lamellar fibrocarcinoma | There is a central scar and fibrous separation, which can be accompanied by calcification and necrosis | Contains many fiber separations | Both arteries and portal veins exist | T2W1 is hyperintense; T1W1 is slightly hypointense. (center low) | Early periphery strengthening | Further strengthening | Low periphery, strengthened center | Fast in and slow out |
| Metastasis carcinoma | Multiple blood spread | Tumor angiogenesis | Small arteries | T2W1 is hyperintense; T1W1 is slightly hypointense | Early edge enhancement | Quick removal | Isointense or hypointense | Fast in and out |
PEComas are positive for melanocytic and muscle markers, including HMB45, Melan-A, and SMA,4 but negative for other markers, including AFP, CD117, CK, and TFE3. Detection of TFE3 protein expression is recommended for PEComas because TFE3-positive PEComas are associated with poor prognosis.17,18 In our study, TFE3 protein was detected in only three patients and these patients had a good prognosis.
Classical AMLs are universally regarded as slow-growing benign tumors. However, it is becoming increasingly clear that some PEComas should be considered tumors of uncertain malignant potential, such as EAML. Malignant PEComas often show the following characteristics: epithelioid cells ≥70%; tumor size of >5 cm; invasion of blood vessels; ≥2 mitoses per 10 high-powered fields; pathological mitosis and necrosis; and tumor aggressiveness.19
Among our patients with EAML, three patients had necrosis, two patients were admitted to the hospital for the second time owing to a relapse, one had undergone surgery for left renal PEComa 3 years earlier, and one had undergone right hepatic PEComa resection 10 years earlier.
Most authors agree that the criteria for malignancy and the biological behavior of PEComas have been insufficiently established. Indeed, PEComas feature several biological behaviors, including benign as well as malignant behavior and uncertain malignant potential. Folpe et al.20 performed statistical analysis and found that the behavior of PEComas in the gynecologic tract and soft tissue correlated with a tumor size of >5 cm, infiltration, high nuclear grade and cellularity, mitosis in >1/50 high-power fields, necrosis, and vascular invasion. PEComas showing two or more worrisome histological features should be classified as aggressive. In our cases, most of the patients (41.3%) had tumor sizes of <3 cm. We should take into account that tumors of >10 cm with necrosis tend to become malignant.
Because this disease is rare, it is difficult to perform a clinical, therapeutic trial, and therefore the treatment for hepatic PEComa remains controversial. The variable proportions of the different components of hepatic PEComa make diagnosis difficult before surgery, and the majority of patients undergo surgical resection. Furthermore, postoperative complications or recurrence have rarely been reported.21 In our study, all patients underwent surgical resection, including laparotomy and laparoscopic partial hepatectomy. Two patients had received TACE before the surgery. Two patients relapsed after surgery, including one male patient whose liver tumor size was 12 × 11 × 8 cm with necrosis, who had mediastinal metastasis 5 years after surgery and pathological necrosis, and one other male patient, whose tumor presented as a solitary mass with a size of 1.7 × 1.5 cm and who had lung metastasis 4 years after the surgery. Pathological examination showed that the tumor had vascular invasion. Both patients received chemotherapy in other hospitals. Yang et al.14,22 recommended the performance of fine-needle aspiration (FNAB) combined with HMB45 and Melan-A staining in all asymptomatic patients with lesions of <5 cm and without serological abnormalities, when hepatic PEComa is suspected. If FNAB is performed and the pathology indicates a benign pattern, then observation and imaging follow-up are recommended.22 For patients with unresectable malignant PEComas, O’Malley et al.23 recommended that applying targeted therapy, such as mTORC1 inhibitors, may have a specific effect.
Most of these patients were found incidentally during a regular physical examination. In our study, some patients presented with abdominal discomfort, and four had a history of a liver mass for >2 years. They were diagnosed with AML and then followed by observation. We found that the tumors had gradually increased to >5 cm at 2-year follow-up and that the patients visited the hospital upon experiencing abdominal discomfort. The patients had a good prognosis after surgery. This indicates that patients with PEComas of <5 cm can be administered conservative treatment. However, we found one patient with a solitary mass of <5 cm who had a lung metastasis 4 years after the surgery. Pathological examination showed that one worrisome histological feature was the invasion of blood vessels. Further research must allow accurate prediction of the biobehavior of this lesion and establish criteria for discrimination between malignant and benign tumors. In our study, most patients who underwent surgical resection had a good outcome; therefore, surgical resection may be the best choice for hepatic PEComa therapy.
Our study also had some limitations. The efficacy and prognosis were not compared with different treatments, such as conservative therapy, TACE, or surgery. Although we sorted different liver tumor characteristics according to MRI, owing to PEComa heterogeneity, the application of the characteristics described in Table 4 may have some limitations. When imaging diagnosis is challenging, FNAB should be used to make an auxiliary diagnosis.
In conclusion, AML is the most common type of PEComa in the hepatic tissue. There are no specific symptoms of hepatic PEComa, and serological examinations and imaging modalities for accurate preoperative diagnosis are lacking. EAML should be considered a tumor of uncertain malignant potential. However, the prognosis of PEComa after resection is promising.