Angiosarcoma predominantly arises in the skin or superficial soft tissue. Less frequently, it can occur in various organs and has been reported in ovary , oral cavity , small intestine , thyroid , pleura , lung , testis [10, 11] and parotid [2, 3], even in fibroadenoma  and phyllodes tumour . Occasionally, it also may be part of dedifferentiated liposarcoma  or mixed malignant tumors [15, 16]. Epithelioid angiosarcoma most commonly involves deep soft tissue. Histologically, epithelioid angiosarcoma predominantly consists of sheets of highly atypical round cells with prominent nuclei, resembling poorly differentiated carcinoma, so, it may cause a great diagnostic confusion . The diagnosis hint of angiosarcoma may be the existence of sinusoid-like spaces or the remarkable necrosis or hemorrhage. Unfortunately, there is a variant in squamous cell carcinoma, namely, pseudoangiosarcomatous cell carcinoma which also shows anastomosing vascular and gland-like spaces mimicking angiosarcoma . So, in addition to histological findings, immunostaining must be used for distinguishing the two entities.
The endothelial cells show reactivity for several markers, including CD31, CD34 and von Willebrand factor (factor VIII). Among them, CD31 is considered the most sensitive and most specific endothelial cell marker. Epithelioid angiosarcoma is positive for CD31, but it is classically negative for CD34. In our case, there was strong and diffuse expression for CD31 in almost all tumor cells, and CD34 expression was absent. So we made the diagnosis of epithelioid angiosarcoma.
Some papers have documented that angiosarcoma or epithelioid angiosarcoma can express cytokeratin . Miettinen M et al. found that epithelioid angiosarcomas were often positive for CK8 and CK18 (approximately 50%) . Therefore, CK may be not helpful in distinguishing epithelioid angiosarcoma from squamous cell carcinoma. In our case, although CK showed a diffuse expression, but contrary to the diffuse and strong cytokeratin immunoreactivity in most of the carcinoma, the staining seemed to be relatively weak. This may represent an important clue to the possibility of non-epithelial lesions.
P63 is a p53 homolog that is expressed in various normal epithelial tissues and epithelial malignancies [19, 20]. Many pathologists believed that immunohistochemical staining for p63 protein was a useful marker for distinguishing spindle cell carcinoma from sarcoma [21, 22]. Jo VY et al. also demonstrated that p63 immunohistochemical staining is limited in soft tissue tumors, and failed to find expression of P63 in 20 cases of angiosarcomas . On the contrary, in our case we found diffuse p63 expression in the tumor cells. So far, there is only one literature that described the positive expression of p63 in angiosarcoma, that is, Kallen ME et al. reported nuclear p63 expression might been seen in 20%–30% of malignant vascular tumors, including angiosarcoma, epithelioid angiosarcoma and epithelioid hemangioendothelioma . Senoo M et al. found that p63 can affect expression of vascular endothelial growth factor via interactions with hypoxia inducible factor-1α . Consequently, it is reasonable that p63 can express in vascular tumors, and p63 may not be as specific for the epithelial differentiation as the literatures reported [21–23]. Moreover, in contrast to described by Kallen ME et al. that only a minority of tumor cell nuclei were immunoreactive in most p63-positive vascular tumors , in our case over 75% of the tumor cells was positive. The p63 expression in angiosarcoma represents another diagnostic pitfall.