Ependymoblastoma: Dear, Damned, Distracting Diagnosis, Farewell!

Abstract

Ependymoblastoma is a diagnostic label that has been applied to a variety of rare central nervous system (CNS) tumors over the last eight decades. Consequently, there is uncertainty about whether such an entity exists and what its characteristic features might be. The current study, based on 14 cases from our institutional archives and identified by the search terms “ependymoblastoma,”“ependymoblastomatous,”“ependymoblastic” or “PNET with ependymal differentiation,” aimed to test the hypothesis that the ependymoblastoma is a distinct and recognizable entity. Ependymoblastic rosettes are a key diagnostic feature and were present in 11/14 (79%) tumors, eight (73%) of which were embryonal tumors with abundant areas of neuropil‐like differentiation. Three other cases showed rare ependymoblastic rosettes in the histopathological setting of a typical primitive neuroectodermal tumor (PNET), medulloblastoma (MB) or atypical teratoid/rhabdoid tumor (AT/RT). The remaining cases were all embryonal tumors with structures that mimicked ependymoblastic rosettes. Our results indicate that ependymoblastic rosettes are most frequently encountered in embryonal tumors with abundant neuropil and less frequently in other CNS embryonal neoplasms, including PNET, MB and AT/RT. We believe that ependymoblastoma as a diagnosis is neither precise nor specific and that it is time once and for all to retire this diagnosis from the lexicon of neuropathology.

INTRODUCTION

First described and then abandoned by Bailey and Cushing as a specific diagnostic entity, ependymoblastoma has enjoyed a long if somewhat ambiguous tenure in the diagnostic armamentarium of neuropathologists. In the 2007 World Health Organization (WHO) classification of brain tumors, ependymoblastoma is placed under central nervous system (CNS) primitive neuroectodermal tumors (PNETs), “a heterogeneous group of tumors occurring predominantly in children and adolescents”(10). Tumors within this category that produce ependymoblastic rosettes are termed ependymoblastomas. This reclassification of ependymoblastoma as a variant of CNS PNET is only the most recent attempt to find a suitable definition for this storied but ultimately problematic diagnosis.

In their initial classification of brain tumors, Bailey and Cushing distinguished two types of ependymal tumors: ependymoblastoma and ependymoma (1). Ependymoblastomas were characterized by the presence of “ependymal spongioblasts,” tumor cells with cytoplasmic processes that extend toward blood vessels (1). Tumors derived from these cells were distinguished by such perivascular processes, that is, perivascular anuclear zones (“pseudorosettes”). This was in contrast to ependymomas, which were derived from more mature cells, lacking processes. Bailey and Cushing subsequently recognized that this architectural distinction was not reproducible and correlated poorly with biological behavior. Therefore, they abandoned ependymoblastoma as a distinct diagnostic entity, retaining ependymoma as the preferred designation for these tumors (15).

The use of ependymoblastoma as a diagnostic term was reintroduced in 1938 by Kernohan, who, citing Bailey and Cushing, concluded that there were some ependymomas that must arise from more primitive ependymal cells (so‐called ependymoblasts) and that tumors arising from these cells should be so designated. Consequently, Kernohan proposed division of ependymomas into four categories, based on histologic type, and then “. . . further classified into two groups: ependymomas and ependymoblastomas, according to the degree of malignancy”(7). From a practical perspective this meant that the term “ependymoblastoma” was generally used for tumors that were felt to be malignant (11). Further refinement of the Mayo Clinic grading criteria for ependymomas by Mabon focused on grade of malignancy rather than the presumed histogenesis of these tumors. As the tumors that had been classified as ependymoblastomas demonstrated such wide variation in malignancy, this particular usage of ependymoblastoma was dropped (11).

Phoenix‐like, ependymoblastoma reappeared when Rubinstein appropriated it as the diagnostic label for a tumor that he proposed was distinct from the broad category of ependymomas on the one hand and malignant ependymomas on the other. In an initial report of two cases, he laid out his concept of this tumor as one that was composed predominately of primitive elements, but which focally exhibited evidence of ependymal differentiation, primarily through the presence of ependymoblastic rosettes (14). Rubinstein argued that these tumors were clearly examples of “malignant infiltrating glial neoplasms”(14). However, rather than showing evidence of the anaplastic features characteristic of a glioblastoma, these tumors were chiefly comprised of cells with primitive features, with an analogy in one case drawn to medulloblastoma. Believing that these tumors most likely represented a type of “primitive glioma” that arose from a glial precursor cell, Rubinstein proposed that these tumors should be designated ependymoblastomas (14). This designation was based on the (tenuous) assumption that “ependymoblast” was a reasonable term for the presumptive precursor cell.

The clinical features of ependymoblastomas were better characterized in a subsequent series of 12 additional cases (12). These tumors occurred in young children (usually under the age of 2 years), were associated with an aggressive course (median survival of twelve months) and sometimes showed growth reminiscent of a medulloblastoma, including leptomeningeal dissemination. In this report, a further refinement was added to the histologic description; these tumors were characterized by both ependymal rosettes and multilayered rosettes that contained mitotic figures and a central lumen with an internal limiting membrane and associated cilia and blepharoplasts. These structures were termed “ependymoblastic rosettes”(12). The ependymoblastoma was thus situated as a primitive central neuroepithelial tumor arising from a presumed embryonal cell with a differentiation potential restricted to glial precursor cells and ependymal cells 12, 15. After Hart and Earle introduced PNET, subsequent authors, including both Rorke‐Adams and Becker, proposed the classification of ependymoblastoma as a type of PNET 2, 6, 13. In their schemes, as well as the 2000 WHO classification of nervous system tumors, the definition of ependymoblastoma shifted away from a tumor derived from a precursor cell with restricted differentiation along glial precursor and ependymal lines, as proposed by Rubinstein, to a CNS PNET with distinctive ependymoblastic rosettes (8).

A further complication has been the recent recognition of a pediatric embryonal CNS tumor (“pediatric tumor with abundant neuropil and true rosettes”) that contains multilayered rosettes with essentially the same features as those described in ependymoblastoma (4). In addition to such rosettes, this tumor demonstrates evidence of neuronal differentiation, neurocytes and ganglion cells against a neuropil‐like background, and has also been called “embryonal tumor with abundant neuropil and true rosettes”(3). Thus, not only has “ependymoblastoma” been used over the years as a diagnosis for various tumors—ependymomas, malignant ependymomas and embryonal CNS tumors containing ependymoblastic rosettes—but its defining feature can be present in PNETs showing extensive non‐ependymal (neuronal) differentiation. Curious to determine which pediatric CNS tumors might warrant a diagnosis of ependymoblastoma, we reviewed the pathology archives of two major pediatric centers (Children's Hospital of Philadelphia and St. Jude Children's Research Hospital) with the aim of testing the hypothesis that the ependymoblastoma is a distinct and recognizable entity.

PATIENTS AND METHODS

This study was conducted with appropriate Institutional Review Board approval for the use of human tissues. Given the varied nomenclature possibly used for cases of ependymoblastoma, we searched our archives (1987–2008) for reports that recorded the terms “ependymoblastoma,”“ependymoblastomatous,”“ependymoblastic” or “PNET with ependymal differentiation.” Sixteen cases, 14 primary tumors and two recurrences, were discovered according to these criteria. Routine hematoxylin and eosin staining and light microscopy, as well as immunohistochemical staining results, were used to establish the review histologic diagnosis of each tumor. Cases where ependymoblastoma had been used as the diagnosis for what was clearly an anaplastic ependymoma, and in which no ependymoblastic rosettes were identified, were excluded from this series.

RESULTS

The clinical and key immunohistochemical features for the 14 primary cases are shown in 1, 2, 3. For the two cases with recurrences, the histologic features in both primary and recurrent tumors were identical, so only data relating to the primary tumor have been included. Overall, there was a slight male preponderance (ratio 1.3:1). Patients ranged in age from nearly 8 to 64 months; mean age was 31 months. Five of the tumors were located in the posterior fossa, whereas the remaining nine were suptratentorial. One tumor (# 2) from the posterior fossa showed spinal metastases at presentation. Two tumors showed local recurrences, but no metastases.

Table 1.

Embryonal tumors with abundant neuropil and ependymoblastic rosettes (ETANER). Abbreviations: EMA = epithelial membrane antigen; GFAP = glial fibrillary acidic protein; ND = not done; NeuN = neuronal nuclei antigen.

Case	Age (months)	Sex	Location	Synaptophysin	Neurofilament	NeuN	GFAP	EMA	Ki‐67	Diagnosis
1	50	Female	Frontal lobe	+	+	ND	Scattered cells	+ Internal limiting membrane	Up to 50%	ETANER
2	15	Male	Posterior fossa	+	+	ND	−	+ Internal limiting membrane	Up to 30%	ETANER
3	19	Male	Frontal lobe	+	+	+	Scattered cells	+ Internal limiting membrane	Up to 70%	ETANER
4	38	Male	Frontal lobe	+	+	+	−	+ Internal limiting membrane	Up to 50%	ETANER
5	18	Female	Posterior fossa	+	+	ND	Scattered cells	−	Up to 70%	ETANER
6	29	Male	Frontoparietal	ND	+	ND	−	ND	Up to 80%	ETANER
7	5	Male	Suptratentorial	+	+	+	−	+ Internal limiting membrane	Up to 10%	ETANER
8	32	Female	Frontal lobe	+	+	+	−	−	Up to 15%	ETANER

Table 2.

Other tumors with ependymoblastic rosettes. Abbreviations: AT/RT = atypical teratoid/rhabdoid tumor; EMA = epithelial membrane antigen; GFAP = glial fibrillary acidic protein; MB = medulloblastoma; ND = not done; NeuN = neuronal nuclei antigen; PNET = primitive neuroectodermal tumor.

Case	Age (months)	Sex	Location	Synaptophysin	Neurofilament	NeuN	GFAP	EMA	Ki‐67	Diagnosis
9	28	Male	Posterior fossa	+	+	ND	+	ND	Up to 70%	MB
10	8	Male	Posterior fossa, fourth ventricle	+	Scattered cells	ND	Scattered cells	+ Internal limiting membrane	Up to 80%	AT/RT
11	64	Female	Frontal lobe	+	ND	ND	−	ND	Up to 60%	PNET

Table 3.

Tumors with structures that mimic ependymoblastic rosettes. Abbreviations: EMA = epithelial membrane antigen; GFAP = glial fibrillary acidic protein; ND = not done; NeuN = neuronal nuclei antigen; PNET = primitive neuroectodermal tumor.

Case	Age (months)	Sex	Location	Synaptophysin	Neurofilament	NeuN	GFAP	EMA	Ki‐67	Diagnosis
12	32	Male	Posterior fossa	ND	−	ND	+	Scattered + dot‐like cytoplasmic	Up to 12%	PNET
13	45	Female	Temporal lobe	−	Weakly +	ND	Scattered cells	−	ND	PNET
14	46	Female	Parieto‐occipital	+	ND	ND	Scattered cells	ND	Variable, up to 80%	PNET

The cases segregated into three distinct categories based on histologic features and immunophenotype. Ependymoblastic rosettes were present in 11/14 (79%) cases, which are summarized in 1, 2. The remaining cases were all embryonal tumors with structures that mimicked ependymoblastic rosettes, but where careful examination failed to confirm the presence of genuine examples (Table 3). Among cases containing ependymoblastic rosettes, the vast majority, 8/11 (73%), were tumors with the distinctive morphology previously described as “pediatric tumor with abundant neuropil and true rosettes” and which we refer to here as embryonal tumors with abundant neuropil and ependymoblastic rosettes (ETANER) (Table 1). Three embryonal tumors demonstrated at least rare or focal ependymoblastic rosettes, but lacked the full phenotype of ETANER and were appropriately classified as another type of pediatric embryonal CNS tumor (Table 2).

ETANER (Table 1)

Patients with these tumors showed a male preponderance, with a 1.7:1 (male : female) ratio, and were the youngest in the series with an average age of 26 months (range 5–50 months) at diagnosis. Three‐quarters (6/8) were supratentorial. These tumors were comprised of predominately undifferentiated embryonal cells arranged against a neuropil‐like background. Scattered cells demonstrated neurocytic or ganglionic differentiation, whereas the embryonal component formed hypercellular foci throughout the neuropil (Figure 1A,B). The hypercellular foci frequently contained distinctive multilayered rosettes that in well‐oriented sections opened into a small round central lumen bounded by an internal limiting membrane. No external limiting membrane was identified; neoplastic cells along the outer edge of these structures simply merged into the surrounding neuropil. Mitotic figures and karryorhectic debris were readily identified in the walls of these structures (Figure 1C). These structures displayed the typical features of ependymoblastic rosettes.

Figure 1.

Embryonal tumor with abundant neuropil and ependymoblastic rosettes. Ependymoblastic rosettes form conspicuous concentric cellular rings that appear as islands in neoplastic neuropil (A). Ependymoblastic rosettes forming distinctive multilayered structures with a small round central lumen bounded an internal limiting membrane. Cells along the outer edge of the rosette merge into the surrounding neuropil and embryonal tumor cells. Mitotic activity is prominent in the walls of the rosettes (B). Immunohistochemical staining for epithelial membrane antigen highlights the limiting membrane of the central lumen (C). Immunohistochemical staining for Ki‐67 highlights the proliferative activity within rosettes (D). Immunohistochemical staining for NFP highlights abundant intervening neoplastic neuropil as well as demonstrating cytoplasmic expression in many of the embryonal cells (E). Immunohistochemical staining for neuronal nuclei antigen highlights scattered cells with more advanced neurocytic differentiation within the neoplastic neuropil (F).

Immunohistochemical staining with antibodies to neurofilament and synaptophysin highlighted the abundant neoplastic neuropil. Varying proportions of embryonal cells showed thin rims of cytoplasmic expression of these neuronal antigens (Figure 1E). However, the cells within the ependymoblastic rosettes typically showed little expression of neuroepithelial markers. Neuronal nuclear antigen expression characterized scattered individual cells with more advanced neuronal differentiation (Figure 1F). Glial fibrillary acidic protein (GFAP) expression was limited to absent in these tumors. Epithelial membrane antigen (EMA) expression was generally restricted to the central lumen of ependymoblastic rosettes (Figure 1C). Ki‐67 immunolabeling showed wide variation within any tumor; the embryonal element displayed a high index, whereas this was low in the neuropil‐rich component; ependymoblastic rosettes showed strikingly high labeling, as summarized in Table 1 (Figure 1D).

Other embryonal tumors with ependymoblastic rosettes (Table 2)

In three cases, ependymoblastic rosettes were encountered in tumors that had features of other embryonal CNS neoplasms. Case 10 illustrates the characteristic combination of ependymoblastic rosettes with other features of idiosyncratic embryonal tumors encountered in this group. Situated in the posterior fossa, this tumor consisted of dense embryonal cells, among which were small foci of neuropil‐like differentiation distinguished by pale staining (Figure 2A). Scattered within this tumor were small multilayered rosettes with an EMA‐immunopositive internal limiting membranes (Figure 2A). In other areas of the tumor, epithelioid features were evident (Figure 2B). Finally, one portion of tumor showed histologic features resembling a malignant germ cell (yolk sac) tumor, an impression confirmed immunohistochemically by the expression of both cytokeratins and alpha‐fetoprotein (Figure 2C–E). No teratomatous elements were identified in this tumor; in particular, there were no features to suggest that the rosettes were part of an immature neuroepithelium. EMA was expressed by groups of neoplastic cells, and there was diffuse loss of nuclear expression of the SMARCB1 (INI1) gene product by immunohistochemical staining, supporting a diagnosis of atypical teratoid/rhabdoid tumor (AT/RT) (Figure 2F).

Figure 2.

Other embryonal tumors with ependymoblastic rosettes. Embryonal tumor with areas of neuropil‐like differentiation and small ependymoblastic rosettes (A). Focal epithelioid differentiation in neoplastic cells (B). Focal germ cell (yolk sac) differentiation (C). Immunohistochemical staining for cytokeratin (D), alpha‐fetoprotein (E) and INI1 (F).

Cases 9 and 11 were both embryonal tumors that contained variable amounts of neuropil‐like matrix, but not the extensive regions that characterized ETANER. In Case 9, a single ependymoblastic rosette was identified. However, throughout this cerebellar tumor the most striking feature was the presence of perivascular collars of embryonal cells with an anaplastic phenotype (Figure 3A). The tumor showed extensive immunolabeling with both neuronal and glial markers and a strikingly high Ki‐67 labeling index, supporting a diagnosis of medulloblastoma. Case 11 was a suptratentorial tumor that also combined undifferentiated embryonal cells with neuropil‐rich regions characterized by neuronal differentiation. Although ependymoblastic rosettes were identified, primarily within the more cellular portions of this tumor, the most striking architectural feature was the angiocentric arrangement of tumor cells similar to that described in Case 9. This was most evident in neuropil‐rich areas. Neoplastic cells demonstrated synaptophysin expression and an elevated Ki‐67 labeling index in the range of 60%, supporting the diagnosis of cerebral PNET.

Figure 3.

Structures that mimic ependymoblastic rosettes. Tumors with neuropil differentiation and prominent perivascular cellular collars that resemble ependymoblastic rosettes (A,B). Highly cellular tumor with myxoid vascular degeneration giving rise to structures that resemble ependymoblastic rosettes but which lack internal limiting membrane (C). Immunohistochemical staining for collagen type IV of same tumor highlights the blood vessels giving rise to this appearance (D).

Embryonal tumors with features that mimic ependymoblastic rosettes (Table 3)

Three cases were originally noted to have ependymal‐type differentiation in the setting of an embryonal tumor. Upon review, these tumors lacked ependymoblastic rosettes, but contained structures that mimicked them. One previously described mimic of the ependymoblastic rosette is an arrangement of neoplastic cells in perivascular collars. Encountered in cases both with and without ependymoblastic rosettes, these perivascular collections can be mistaken for ependymoblastic rosettes during a cursory examination (Figure 3A). Usually the absence of an internal limiting membrane is sufficient to resolve any doubt, but in some cases the problem can be more challenging. Case 14 is illustrative in this respect. This superficially situated supratentorial PNET contained extensive myxoid vascular degeneration that gave rise to structures resembling the multilayered rosettes described above (Figure 3B). However, upon careful examination, these structures lacked an internal limiting membrane and, in better preserved areas, a central vascular structure could frequently be identified (Figure 3C). No typical ependymoblastic rosettes were identified. Immunohistochemical staining for EMA failed to demonstrate an internal limiting membrane, whereas vascular markers highlighted the blood vessels that gave rise to this appearance. Although neurocytic and ganglionic differentiation were identified morphologically and by immunohistochemical staining, this tumor lacked the neuropil‐like regions described earlier.

DISCUSSION

Originally proposed by Bailey and Cushing as a subtype of ependymoma, ependymoblastoma has had a long and varied tenure as a diagnostic entity. Beginning with Kernohan, ependymoblastoma was used as a designation of malignancy in ependymomas. Having fallen into disuse, ependymoblastoma was rehabilitated by Rubinstein, who used it to describe a putative embryonal tumor with ependymal differentiation. In its most recent iteration, as a subtype of PNET, ependymoblastoma has become synonymous with a distinctive histopathologic structure, the ependymoblastic rosette. Given the diverse application of ependymoblastoma over the years, we sought to establish whether or not there existed any single distinctive entity that warrants this diagnosis. Using the broad criteria described earlier, our search yielded a diverse group of pediatric embryonal CNS tumors characterized by the ependymoblastic rosette or its mimics. Our study of these tumors established three patterns. First, ependymoblastic rosettes are most likely to be encountered in a distinctive tumor characterized by abundant neuropil‐like regions and prominent neurocytic differentiation (ETANER). Second, ependymoblastic rosettes are not specific and may be encountered in medulloblastoma, PNET and AT/RT. Third, perivascular collars of neoplastic cells may mimic ependymoblastic rosettes, particularly in the setting of myxoid degeneration.

In the vast majority (73%) of cases, ependymoblastic rosettes were encountered in embryonal tumors with abundant areas of neuropil‐like differentiation. This combination was first described by Eberhart et al as the “pediatric neuroblastic tumor containing abundant neuropil and true rosettes”(4). Histologically, this distinctive tumor combines features that have been associated historically with both cerebral neuroblastoma and ependymoblastoma and is characterized by the combination of neuropil‐like regions, often containing neurocytic and ganglion cells, and undifferentiated embryonal cells, among which there are rosettes with well‐formed central lumens. Variably well‐formed internal limiting membranes are present within these rosettes, whereas along their outer aspect the cells of the rosette merge into groups of embryonal cells or surrounding neuropil. Rosette‐forming cells create a multilayered wall, in which both mitoses and karryorhectic debris are readily identified.

Immunohistochemically, these tumors show strong and consistent expression of synaptophysin and neurofilament proteins, both in neoplastic cells and neuropil‐like areas. Proliferative activity in these tumors tends to be biphasic. Although the more differentiated areas appear relatively quiescent, significant elevation is encountered in embryonal areas and ependymoblastic rosettes. Expression of GFAP, initially reported in just one of the original tumors described by Eberhart, has been confirmed as variably present in subsequent descriptions 3, 5, 9.

In the 13 cases reported to date, several clinical features have been described. This is clearly a pediatric tumor, all cases occurring in children ≤4 years of age. There is no apparent predilection for either gender. Although these are tumors that may be encountered in both infra‐ and supratentorial locations, they are more common in the latter where they tend to be large and cystic. These are aggressive tumors that have a poor prognosis despite intensive therapy; just over half (7/13) of the reported cases were dead because of the disease between 6 and 14 months (3).

As described earlier, the histopathologic features of the eight cases reported in our series are consistent with this diagnosis. They show the characteristic abundant neuropil‐like areas, undifferentiated embryonal cells, and ependymoblastic rosettes, as well as the predominantly neuronal immunophenotype reported in these tumors. Clinically, our cases showed a male preponderance, and one patient was 50 months at the time of resection, extending slightly the reported age range for this disease. Like those previously reported, our cases also showed a distinct predilection for supratentorial sites. Although survival data were not available for our cases, local recurrence in two and metastasis at presentation in one are consistent with the aggressive behavior described previously.

The appropriate nomenclature for this tumor remains to be decided, as it is yet to be included in the WHO classification. However, considering the dominant neuronal differentiation it shows, we believe that the appropriate designation should not be ependymoblastoma. In their original report, Eberhart et al proposed that this tumor fell within the general category of PNET and offered a descriptive designation “pediatric neuroblastic brain tumor containing abundant neuropil and true rosettes”(4). In a subsequent report, the nomenclature was revised slightly to “embryonal tumor with abundant neuropil and true rosettes”(3). Molecular characterization of these tumors has thus far provided limited data to guide the refinement of nomenclature or classification of this histologically distinctive tumor 3, 5. However, for the moment, we propose “ETANER.”

Our series of cases clearly demonstrates that ependymoblastic rosettes may also be encountered in other embryonal CNS neoplasms including medulloblastoma, PNET and AT/RT. Current understanding of the molecular and genetic features that distinguish these three embryonal CNS neoplasms makes it clear that it is no longer justifiable to lump together such cases on the basis of a single histologic feature, that is, an ependymoblastic rosette (10). It is well recognized that embryonal tumors can show varying patterns of histologic differentiation; our Case 10, illustrated in Figure 2, is an excellent example. This embryonal tumor demonstrated a complex histopathology, including a focus of germ cell differentiation, as well as loss of SMARCB1 (INI1) expression. The presence of ependymoblastic rosettes in this tumor illustrates the lack of diagnostic specificity conferred by identifying these structures outside the specific context of the biphasic tumor described above. Finally, in our series we encountered PNETs that contained structures mimicking ependymoblastic rosettes. Generally, these were formed by perivascular collars of tumor cells in the setting of myxoid degeneration, suggesting the need for caution where limited sampling or poor preservation preclude definitive assessment.

Between 1926 and 2007 the term ependymoblastoma had been inconsistently applied to a variety of tumors, including ependymomas, malignant ependymomas, presumed primitive gliomas arising from a glial precursor and differentiating along ependymal lines, and PNETs. In the case of PNETs, the presence of ependymoblastic rosettes has become a key diagnostic feature. In the 2007 WHO Classification of nervous system tumors, ependymoblastoma was placed within the spectrum of PNET and explicitly defined as a tumor “with ependymoblastic rosettes”(10). This means that for all practical purposes the diagnosis of ependymoblastoma could be extended to include any CNS embryonal tumor with ependymoblastic rosettes. However, as our series makes clear, this structure is most often encountered in what is clearly a distinct entity dominated by a neuronal phenotype: the ETANER. Futhermore, as we illustrate with three cases in this series, other CNS embryonal neoplasms, including medulloblastoma and AT/RT, which have molecular genetic features distinct from PNET, can also contain ependymoblastic rosettes.

The strong association of the ependymoblastic rosette, on the one hand, with a clinically and pathologically distinct tumor entity, and its occasional occurrence, on the other hand, among a variety of embryonal CNS neoplasms suggest that, just as in times past, the use of the term ependymoblastoma is both confusing and imprecise. On the basis of the present and other published series, neuropathologists encountering ependymoblastic rosettes would be well advised to search for other distinctive features of an embryonal tumor with abundant neuropil and neurocytic differentiation, as this is the most likely context in which to encounter these rosettes. Failing that, they can proceed with confidence to work up the embryonal tumor and diagnose it for what it truly is most likely to be, a medulloblastoma, PNET or other embryonal tumor exhibiting, as part of its diverse differentiation, ependymoblastic rosettes. Although these structures are distinctive, they have not been illuminating for successive generations of neuropathologists. The evidence of this case series would suggest that it is time to retire, once and for all, the diagnosis of ependymoblastoma.