Supratentorial extraventricular ependymal neoplasms

A clinicopathologic study of 32 patients


  • Shanop Shuangshoti M.D.,

    1. Section of Neuropathology, Clinical Brain Disorders Branch, National Institutes of Health, Bethesda, Maryland
    Current affiliation:
    1. Department of Pathology, Faculty of Medicine, Chulalongkorn University, Bangkok, Thailand
    Search for more papers by this author
  • Elisabeth J. Rushing M.D.,

    1. Department of Neuropathology and Ophthalmic Pathology, Armed Forces Institute of Pathology, Washington, District of Columbia
    Search for more papers by this author
  • Hernando Mena M.D.,

    1. Department of Neuropathology and Ophthalmic Pathology, Armed Forces Institute of Pathology, Washington, District of Columbia
    Search for more papers by this author
  • Cara Olsen M.S.,

    1. Department of Preventive Medicine and Biometrics, Uniformed Services University of the Health Sciences, Bethesda, Maryland
    Search for more papers by this author
  • Glenn D. Sandberg M.D.

    Corresponding author
    1. Department of Neuropathology and Ophthalmic Pathology, Armed Forces Institute of Pathology, Washington, District of Columbia
    • Department of Neuropathology and Ophthalmic Pathology, The Armed Forces Institute of Pathology, 16th Street and Alaska Avenue NW, Building 54, Washington, DC, 20306-6000
    Search for more papers by this author
    • Fax: 202-782-4099

  • This article is a US Government work and, as such, is in the public domain in the United States of America



Published research on the clinicopathologic features of extraventricular ependymal neoplasms of the cerebral hemispheres has been scant.


Thirty-two archival cases were studied to investigate the prognostic impact of clinicopathologic parameters including flow cytometry, the proliferation (Ki-67) labeling index, and p53 expression.


Among these 32 cases were 2 subependymomas, 19 ependymomas, and 11 anaplastic ependymomas. No significant gender predilection was observed, and 45% of patients were in their second or third decade of life. The left cerebral hemisphere was 1.5 times more commonly involved. On available imaging studies, lesions were often cystic, with or without a mural nodule. Tumors expressed glial fibrillary acidic protein (87%), S-100 protein (77%), cytokeratin (43%), and epithelial membrane antigen (17%). Ki-67 proliferation index paralleled tumor grade. Immunoreactivity for p53 protein was observed in the 2 cases of subependymoma, in 10 of 11 anaplastic ependymomas, and in 6 of 17 ependymomas. Flow cytometry performed in 27 tumors revealed diploidy in 20 cases and aneuploidy in 4 cases (3 anaplastic and 1 classic ependymomas), with S-phase fraction ranging from 0.2–9.7. Eleven subjects were additionally treated with radiotherapy, and 3 with chemotherapy. Follow up was available in 25 (78%) patients.


The results of the current study suggest that there is no significant relation between histopathology, Ki-67 proliferation index, p53 immunolabeling, tumor ploidy, and biologic behavior. Cancer 2005. Published 2005 by the American Cancer Society.

Ependymomas are relatively uncommon nervous system tumors, constituting 1.2–7.8% of all intracranial neoplasms or 2–6% of all gliomas.1–2 In children, however, they represent the most frequent primary brain tumors after pilocytic astrocytomas and medulloblastomas.3 In patients < 20 years of age, approximately 60% of ependymomas arise in the infratentorium, or, more specifically, in the fourth ventricle.1 Conversely, spinal cord ependymomas are more frequent in the adult patient population; supratentorial ependymomas outside the ventricular system as a distinct location are infrequent and affect both pediatric and adult patients equally.1, 3–4

Since the first description of ependymoma as a specific entity by Bailey and Cushing in 1926, controversies have persisted regarding classification, treatment, and prognosis.2, 3, 5, 6 The World Health Organization (WHO) classification system7 recognizes four types of ependymal neoplasms, including subependymoma (Grade 1), myxopapillary ependymoma (Grade 1), ependymoma (Grade 2) and its variants (cellular, papillary, clear cell, and tanycytic subtypes), and anaplastic ependymoma (Grade 3).

The current study presents our experience with 32 supratentorial extraventricular ependymal tumors. Histologic and biologic parameters, including tumor grade, Ki-67 proliferation index, p53 immunolabeling, and flow cytometry were assessed to determine whether these parameters correlate with clinical outcome and their ventricular–periventricular counterparts.


Thirty-two cases of extraventricular ependymal neoplasms were retrieved from the Armed Forces Institute of Pathology database between the years 1986–1999. Ependymoblastomas and cases showing typical features of astroblastoma were excluded. Information regarding the location of the tumor, neuroradiologic appearance, patient demographics, and treatment (operative findings, adjuvant therapy, and followup) were retrieved from the accompanying patient records. In three cases, clinical data and pathologic material were available from two separate surgeries. In these three cases, pathologic studies (immunohistochemistry and flow cytometry) were performed on material from the second operation.

In all cases, tumor tissues were formalin-fixed, routinely processed, and paraffin-embedded surgical specimens. Five-micron sections stained with hematoxylin and eosin (H & E), and, occasionally, with phosphotungstic acid-hematoxylin (PTAH) methods were available for light microscopic examination.

Tumors were classified according to the WHO classification system.7 In addition, tumors were assessed for general architectural pattern, nuclear atypia, mitotic figures (1 = < 1/10 high power field [HPF]; 2 = 1–5/10 HPF; 3 = 5/10HPF), necrosis, hemorrhage, calcification, and vascular proliferation.

For immunohistochemical studies, the avidin-biotin peroxidase indirect technique with diaminobenzadine as the chromagen was applied, using antibodies-directed glial fibrillary acidic protein (GFAP) (DakoCytomation, Carpinteria, CA; polyclonal, dilution 1:4000), S-100 protein (DakoCytomation; monoclonal, dilution 1:600), cytokeratin AE1/AE3 (DakoCytomation; AKO, polyclonal, dilution 1:400), epithelial membrane antigen (EMA) (Ventana Medical Systems, Tucson, AZ; monoclonal, predilute), p53 protein (DakoCytomation;, monoclonal, dilution 1:1200), and the proliferation marker Ki-67 (DakoCytomation;, monoclonal, dilution 1: 80).

Immunostained sections were interpreted as positive if there was either focal or diffuse cytoplasmic reaction product with GFAP and cytokeratin; nuclear expression in the case of p53 and Ki-67; and membranous, ring-like or punctuate intracytoplasmic labeling with EMA.

To quantify the percentage of nuclei staining with Ki-67, a nuclear antigen, at least 500 tumor cells were manually counted three times in the region of tumor with highest nuclear staining, and an average percentage of positively stained nuclei was calculated.

For ultrastructural study, fresh tissue was fixed in phosphate-buffered 3% glutaraldehyde, Epon-embedded, cut at 1-micron thickness, and stained with uranyl acetate and lead citrate. Preparations were viewed on a Zeiss 109 electron microscope (Carl Zeiss, Oberkochen, Germany). Feulgen-stained sections were used for ploidy study by image analysis on a Coulter E system (Beckman Coulter, Fullerton, CA).

Survival probabilities were calculated according to the Kaplan–Meier method and were measured from the date of diagnosis until the date of last followup or until death. Bivariate associations between survival and prognostic factors were tested using the log-rank test. Multivariate associations between survival and prognostic factors were tested using the Cox proportional hazards regression. A two-sided probability level of 0.05 was chosen for statistical significance. Statistical analysis was performed using SPSS version 12.0 for Windows (SPSS, Chicago, IL).


Clinical, radiologic, and pathologic features are summarized in Table 1. Cases lost to followup were excluded. On the basis of available surgical reports or neuroimaging studies, all tumors were in the supratentorium and had no obvious connection to the ventricular system. Both genders were nearly equally affected with a female to male ratio of 1.1:1. The age at diagnosis ranged from 9 months to 75 years with a mean of 30.5 years for both genders, 28.4 years for males and 32.5 years for females. Forty-seven percent (15 patients) were in the second or third decade of life. The left cerebral hemisphere was more commonly involved than the right, with the left to right ratio of 3:2. The frontal lobe was the favored site (38%), followed by the parietal (22%), and parietooccipital (13%) regions. Imaging studies were available in only12 cases. Of these, a cystic lesion with or without a mural nodule was the most frequent radiologic finding (42%).

Table 1. Clinicopathologic Data on Extraventricular Ependymal Neoplasms with Follow-up
No.LocationCysticaTypeGFAPS-100CKEMAP53Ki-67Ploidy% S phaseRxCxRecurSurvival
  • Rt: right, Lt: left, F: frontal, CE: cerebrum, P: parietal, T: temporal, O: occipital, PO: parietooccipital, FP: frontoparietal, AE: anaplastic ependymoma, CE: cellular ependymoma, CC: clear cell ependymoma, CL: classic, GBM: glioblastoma multiforme, SE: subependymoma, PA: papillary; GFAP: glial fibrillary acidic protein, CK: cytokeratin, EMA: epithelial membrane antigen, D: diploid, A: aneuploid, UI: uninterpretable, ND: not done, Rx: radiotherapy, Cx: chemotherapy, Recur: recurrence, U: unkown.

  • a

    Cystic refers to the neuroimaging appearance.

1PSE++++D0+No3.5 yrs
2Rt CBCE+++1+D0.2UU3 yrs4 yrsa
3Lt PCC+++1+D9.4++No6 yrsa
5Lt PPA+++1+D0.3Nodays
6Lt FCE++++1+UI+U3 yrsa
7Rt PCE+++1+NDUUU6 mosa
8Rt OCL+++1+D3.6No5 yrsa
9Lt CBCE++2+D9.2UU4 yrs8.6 yrs
10Lt PCE++1+D1.018 yrs23 yrs
14Rt FCL+++1+D0.7+No6 yrsa
15Rt TCENDNDNDNDND2+ND+U22 yrs22 yrsa
16Lt FCE+1+D0.6UUU3.9 yrs
17Lt FCL+++1+A4.1+No13 yrsa
18Rt FPCL++ND+U6.8 yrs
19FCL+1+D2.2UUNo4 yrsa
21Lt P+CE++1+NDNo8 days
22Lt PAE+++3+UI++No18 yrsa
24Rt FAE++++2+D3.8UUU14 yrsa
25Lt P+AE+++3+A4.4+No2 yrsa
26Rt OAE++++1+D2.8No5 yrsa
27Lt F AE++2+D3.8UUU4.8 yrs
29Rt O+AE+2+D2.3+No5.5 yrsa
30FAE+++1+A1.0++3 yrs6 yrs
32Rt FAE++++3+D6.7UUU14 mos

There were 19 ependymomas (11 cellular, 6 classic, 1 clear cell, and 1 papillary variant), 11 anaplastic ependymomas, and 2 subependymomas. Tanycytic and myxopapillary variants were not observed. Necrosis and vascular proliferation were found in all anaplastic ependymomas (Fig 1), and 37% and 32%, respectively, in ependymomas. Hemorrhage, calcification, and microcysts were seen more frequently in anaplastic ependymomas, and so were marked nuclear atypia, increased mitotic activity, and high Ki-67 labeling index. None of the tumors contained melanin pigment.

Figure 1.

(A) Surgical specimen of an anaplastic ependymoma showing microvascular proliferation and mitotic figures. (B) Immunostaining with the Ki-67 proliferation marker reveals a high labeling index. (C) Many of the neoplastic cells express p53 reaction product.

Immunohistochemical studies demonstrated strong positivity for GFAP (27 of 31), S-100 protein (24 of 31), cytokeratin (13 of 30), and EMA (5 of 29). Immunostain for p53 protein was reactive in both of the subependymomas (100%), 91% of anaplastic ependymomas, and 35% of ependymomas. Electron microscopy was performed in Cases 12 and 15, and revealed intercellular junctions, cilia, basal bodies, and intracytoplasmic lumina. Flow cytometry in 27 tumors demonstrated diploidy in 20 cases and aneuploidy in 4 cases (3 anaplastic and 1 classical ependymomas). S-phase fraction ranged from 0.2–9.7 (mean, 3.5) in ependymomas, and 1.0–6.9 (mean, 3.9) in anaplastic ependymomas.

The extent of surgical resection was available in four patients, three with macroscopic total excision, and one with partial removal of the tumor. Four subjects with anaplastic ependymoma, six with ependymoma, and one with subependymoma received radiotherapy. Two patients with anaplastic ependymoma and one with ependymoma were treated with chemotherapy.

Complete followup information was obtained in 25 (78%) patients. Two subjects succumbed to postoperative complications. In the ependymoma group, 10 patients were alive at the last followup period of 0.5–22 years (mean, 6.6 yrs), and 4 died at 3.9–23 years (mean, 10.6 yrs). Five patients with anaplastic ependymomas were doing well at the last follow up of 2–18 years (mean, 9 yrs), and three died of the disease at 1.2–6 years (mean, 4 yrs). One subject with subependymoma died at 3.5 years postoperatively. None of the patients developed cerebrospinal fluid spread of tumor cells or extraneural metastasis.

None of the prognostic factors considered were significantly related to survival time in the bivariate analysis or in the multivariate analysis (results not shown). The Kaplan–Meier survival curve is represented in Figure 2.

Figure 2.

Kaplan–Meier survival curve for 25 patients with extraventricular ependymomas. (+) indicates patient was still alive at last followup. Median survival time is 8.6 years (95% confidence interval, 5.8–11.4 yrs)


Extraventricular ependymomas are rare, and whereas they definitely have ependymal morphology, their histogenesis remains uncertain. Ependymal cell rests, which are frequently found at the angle of the ventricles where ependymal cells extend deep into the adjacent white matter, have been hypothesized as the cell of origin of these neoplasms.3, 8 However, random distribution of ependymomas around the periventricular region, rather than restriction to the angles of the ventricles, suggests an alternative mechanism.8

Our patients shared several clinical features with previously published cases.5, 6, 8–10 In the supratentorial compartment, we confirmed that there is a higher proportion of anaplastic (malignant) ependymomas compared with the proportion appearing in published data on their infratentorial counterparts.5 Although Afra et al.9 did not recognize the cerebral hemispheric predilection, the left hemisphere, particularly the frontal region, was more frequently involved in our patients and those of Molina et al.6 In keeping with previous studies, tumors were often well circumscribed; a cystic component may be observed.6, 8, 11, 12

From the standpoint of pathologists, ependymal neoplasms have to be distinguished from other glial and nonglial tumors, especially when they arise in the brain parenchyma with no connection to the ventricles. Among ependymal tumors, subependymomas seem to pose little diagnostic difficulty as they have a rather consistent histologic pattern. Typically, subependymoma is a paucicellular tumor, consisting of uniformly bland tumor cells with prominent meshwork of fibrillary cytoplasmic processes. Clustering of tumor cell nuclei and microcyst formation are common.7, 13 Although previously thought to occur almost exclusively in the cauda equina region, a case of myxopapillary ependymoma was recently reported in the temporal lobe; therefore, this benign variant should be considered outside of its usual site.14 Histologically, this variant is characterized by neoplastic cells arranged in a papillary configuration around vascularized mucoid stromal cores. Tanycytic ependymoma, an uncommon fibrillar variant not observed in our series, is characterized by streams of piloid, or hair-like, cells having “ependymal” nuclei. True ependymal rosettes are absent, and perivascular rosettes are usually inconspicuous.

Conversely, ependymoma and its malignant counterpart demonstrate variable histopathologic features. Although ependymal (true) rosettes or canals are pathognomonic of ependymal differentiation,13 they are commonly absent and, thus, not required for diagnosis. Care should also be taken not to misinterpret microcysts or round edematous spaces, incidentally surrounded by tumor cells, as evidence of true rosettes. Perivascular pseudorosettes, encountered in the majority of ependymomas, are by no means specific and must be differentiated from similar-looking structures found in other tumors, such as astrocytomas, astroblastomas, and papillary meningiomas.15

The widespread staining for AE1/AE3 found in many ependymomas, including to a lesser extent those in the present series, corresponded to the pattern of GFAP staining. Vege et al. documented a varied staining frequency for other keratins: cytokeratin (CK)7 (20%), CAM 5.2 (19%), CK903 (14%), and CK20 (8%). In general, significant staining for keratins other than AE1/AE3 is inconsistent with the diagnosis of ependymoma.16

The location of the tumor and its imaging characteristics are particularly important. Unlike fibrillary astrocytomas, which are diffusely infiltrating tumors, most extraventricular ependymomas are well circumscribed. Cystic lesions, with or without a mural nodule, are typical.6, 8, 11 In addition to extraventricular ependymoma, the differential diagnosis of supratentorial parenchymal cystic tumors with mural nodule includes gangioglioma, pleomorphic xanthoastrocytoma, and pilocytic astrocytoma. These clinical features should prompt the pathologist to consider the possibility of ependymoma because immunohistochemical preparations are of limited help in distinguishing astrocytic from ependymal neoplasms.13

Electron microscopy may be needed to identify ependymal differentiation, which includes zipper-like junctional complexes, microvilli, cilia with basal bodies, and intercellular microlumina containing variable numbers of microvilli and cilia.7, 13, 15

Astroblastomas are rare neuroepithelial tumors of uncertain origin that possess unquestionable clinical and pathologic similarities with ependymomas; astroblastoma are usually well demarcated masses, often with an associated cystic component, arising in the cerebral hemispheres of young patients.17 Perivascular pseudorosettes and prominent vascular sclerosis are important histologic features of this enigmatic tumor.13, 15 In contrast to the delicate tapering glial processes of ependymoma cells that point toward the central vascular lumen to form perivascular pseudorosettes, processes of tumor cells in astroblastoma are thicker and broaden as they reach targeted vascular cores.13, 15 Astroblastoma and ependymoma express GFAP, vimentin, S-100 protein, and occasionally EMA. These two tumors, therefore, cannot be differentiated on the basis of immunohistochemical methods, and definitive diagnosis may require ultrastructural evaluation.13, 15

The papillary growth pattern of papillary meningiomas may be confused with perivascular pseudorosettes of ependymomas.15 Most meningiomas, nonetheless, are dural-based, extraaxial lesions. Positive EMA and negative GFAP immunostains support the diagnosis, even though GFAP immunoreactivity has been observed in exceptional cases of papillary meningioma.17–19 Areas of typical meningothelial differentiation with a syncytial appearance, intranuclear vacuoles, and well formed cellular whorls should also be seen. Clear cell ependymoma is an uncommon variant of ependymoma deserving mention because of its predilection for extraventricular location (6 out of 8 cases).11, 13 In a recent review of 10 pediatric cases notes frequent extraneural metastases and early recurrence. In addition, many of these tumors are reported to have overtly anaplastic histologic features accompanied by losses of chromosome 18.21 Although these neoplasms may morphologically mimic neuronal or mixed glial tumors and oligodendrogliomas, they do not exhibit the infiltrating growth pattern of the latter on imaging studies.10 The lack of neuronal markers, such as synaptophysin, excludes clear cell ependymomas from neuronal or mixed neuronal-glial tumors.10, 12 Electron microscopy, again, may be necessary to establish the correct diagnosis.

The tumor suppressor gene p53 located on chromosome 17p13.1 is the most common gene involved in human cancers.22 Theoretically, unlike wild-type p53 protein (normal product of the gene) that has a short half-life, mutant p53 protein can be detected by immunohistochemical methods using specific antibodies.21 Although p53 mutations have only rarely been reported in ependymal neoplasms by molecular analysis,7, 22, 24 we identified p53 protein immunolabeling in 60% of ependymal neoplasms studied. As in the previous report by Rushing et al., demonstrating p53 immunoreactivity associated with increasing histologic malignancy in ependymal tumors,25 the anaplastic ependymomas in our series more commonly express p53 protein (Fig. 2A,C). Surprisingly, two subependymomas in our study were also immunoreactive for p53 protein. Whether the discrepancy between the incidence of p53 protein immunolabeling and p53 mutations identified by DNA methods represents expression of wild-type p53 gene in tumor cells, alternative mechanisms of p53 gene inactivation, or simply a cross-reaction of the antigen–antibody complex remains to be ascertained. Nevertheless, no correlation was found between p53 protein immunoreactivity and the clinical outcome of our patients. Alternatively, Zamecnik and colleagues found that p53 immunopositivity and/or MIB-1 LI of > 5% (after subtotal resection) or MIB-1 LI of > 15% (after complete resection) are the strongest indicators of aggressive tumor behavior and poor prognosis.26

In addition to mitotic count, we investigated the proliferative activity of tumor cells by measuring Ki-67 labeling and the percentage of cells in S phase.3 As observed in previous studies,25, 27, 28 we noted that an increased Ki-67 labeling index correlated well with increasing mitotic activity and histologic malignancy of ependymal tumors (Fig. 2B); however, there was no association between the proliferation markers and outcome in our subjects. This finding is in contrast to that of a study by Ritter et al. who found an increased Ki-67 labeling index to be linked with a poorer prognosis.29 The percentage of S-phase fraction of ependymal tumors varies considerably among different studies.3, 26, 29 However, similar to a report by Schwartz et al.,3 we found no significant relation between the percentage of tumor cells in S-phase and clinical outcome.

In adult astrocytic tumors, DNA content has been found to be a powerful prognostic indicator.30 However, few studies have examined ploidy in ependymomas and, in those that have, the results are rather inconclusive. Spaar et al. reported polyploidy to be more frequent in supratentorial malignant ependymomas but found no correlation with clinical outcome.29 Garcia et al. demonstrated frequent aneuploidy in high-grade ependymomas.27 Schwartz et al. found that the majority of supratentorial ependymomas were aneuploid and more likely to recur.3 A trend toward fewer recurrences in patients with diploid neoplasms was also observed.31 To the contrary, our study showed that most Grade 1 and 3 ependymomas are diploid, and no correlation was identified between DNA content and clinical behavior. Onguru reported that DNA ploidy status correlated with clinical outcome rather than overall survival.32 This study was based on a small sample size of 12 patients and was not statistically significant. An investigation using gene expression patterns appeared to validate unique molecular pathways in the pathogenesis of supratentorial ependymomas.33 In that study, the authors were able to classify supratentorial Grade 2 and 3 tumors with 100% accuracy.33

Unlike diffusely infiltrating astrocytomas, which extend well beyond their apparent margin, ependymomas are commonly well demarcated and amenable to surgical extirpation.4, 12 The survival benefit provided by complete, as opposed to partial, resection has been repeatedly emphasized in literature.4, 10, 12 Details of the extent of surgery in our patients were, unfortunately, too limited to provide prognostic assessment.

Postoperative irradiation plays a role in the management of ependymomas, but indications for this adjuvant therapy remain a subject of debate. In recent studies of supratentorial ependymomas, it has been suggested that radiotherapy should be given to patients with anaplastic ependymomas10, 12 and in cases where only partial resection of either benign or malignant tumors has been achieved.12 A postoperative MRI with contrast has also been recommended for further evaluation of the extent of resection,3, 10, 12 Early second-look surgery may be proposed to achieve total excision in selected patients with accessible residual tumor detected on postoperative MRI.10 In contrast to their infratentorial counterparts, supratentorial ependymomas tend to recur in regions amenable to surgery. Therefore, reoperation to attempt complete tumor resection before initiation of radiotherapy should be considered.10, 12 Prophylactic craniospinal irradiation is no longer advocated unless cerebrospinal (CSF) seeding is evident on imaging or cerebrospinal fluid studies.3, 12 In a multicenter study of adult patients with ependymoma, Reni reported that, on multivariate analysis, postoperative radiation therapy resulted in a trend of overall improved survival and was predictive of increased failure-free survival.34 Chemotherapeutic agents have been tried, particularly in children for whom radiotherapy is contraindicated or in previously irradiated subjects with an inoperable tumor, but their benefit is still questionable.10, 35

In conclusion, we have reported a series of 32 supratentorial, extraventricular, ependymal neoplasms. These findings parallel those of ventricular–periventricular ependymoma counterparts. No correlation was found between clinical outcome and histologic or biologic parameters, including tumor grade, p53 protein immunolabeling, proliferation markers, and tumor ploidy. The major limitation of the current study is that treatment effects could not be studied in detail because clinical information was not available in many of the cases. This reflects the referral nature of the material available for study. Therefore, future prospective studies with a larger patient population and long-term followup are needed to evaluate the role of therapeutic options. The study of novel biologic markers in this glioma subtype may generate additional prognostically relevant data. Finally, increased awareness among neurosurgeons and pathologists of the atypical presentation of ependymal neoplasms ensures the appropriate therapeutic intervention.


The authors thank C. J. Janosky, H. Makhlouf, R. L. Becker, Jr. and R. P. Turnicky for their excellent laboratory and technical support.