Fax: (011) 31 104087905
Identification of relevant prognostic histopathologic features in 69 intracranial ependymomas, excluding myxopapillary ependymomas and subependymomas
Version of Record online: 9 DEC 2005
Copyright © 2005 American Cancer Society
Volume 106, Issue 2, pages 388–395, 15 January 2006
How to Cite
Kurt, E., Zheng, P.-P., Hop, W. C. J., van der Weiden, M., Bol, M., van den Bent, M. J., Avezaat, C. J. J. and Kros, J. M. (2006), Identification of relevant prognostic histopathologic features in 69 intracranial ependymomas, excluding myxopapillary ependymomas and subependymomas. Cancer, 106: 388–395. doi: 10.1002/cncr.21608
- Issue online: 5 JAN 2006
- Version of Record online: 9 DEC 2005
- Manuscript Accepted: 10 AUG 2005
- Manuscript Revised: 9 AUG 2005
- Manuscript Received: 17 MAY 2005
- mitotic index;
- Ki-67 labeling index;
- prognostic factors
The results of attempts to identify histopathologic parameters that contribute to the clinical outcome of patients with ependymomas have been controversial. This may be due to the relative rareness of ependymomas. Furthermore, in many investigations, myxopapillary ependymomas and subependymomas were included and may have confounded results, because those tumors should be considered clinicopathologic entities distinct from the other ependymomas.
In this retrospective study, the influence of the histologic subtype of ependymoma and of individual histologic features on the outcome of 69 patients with ependymomas was investigated. Myxopapillary ependymomas, subependymomas, and ependymomas with spinal localizations were excluded from the analysis. The ependymomas were subdivided into cellular, papillary, clear cell, and tanycytic subtypes. The study extended over a period of 30 years.
No differences in clinical outcome between the four histologic subtypes of ependymomas were revealed. Neither tumor localization (either infratentorial or supratentorial), patient age, nor gender affected survival. The survival of patients who underwent complete tumor resection differed significantly from that of patients who underwent partial resection. In univariate analysis, the features of nuclear atypia, the mitotic index, and the MIB-1 labeling index (LI) significantly influenced survival. With regard to survival, the presence of microcysts, blood vessel density, and the feature of vascular hyalinization demonstrated a trend but did not reach significance. In multivariate analysis, only the mitotic index and the MIB-1 LI were identified as factors with independent prognostic significance (P = 0.027 and P = 0.023, respectively). Both proliferation indices were correlated strongly with each other.
The results of the univariate analysis indicated that, for patients with intracranial ependymoma, nuclear atypia, the mitotic index, and the MIB-1 LI significantly influenced survival. In the multivariate analysis, the mitotic index and the MIB-1 LI were the only features that had independent prognostic significance. Because both showed strong correlations, only one of them should be included in a grading scheme for intracranial ependymomas. Cancer 2006. © 2005 American Cancer Society.
Over the last decades, attempts have been made to isolate factors of prognostic significance for patients with ependymal neoplasms. In virtually all studies, the tumor location was identified as the most important prognosticator, irrespective of the histology of the tumor.1, 2 The histopathologic parameters to be used in a grading scheme have been subject to controversy, and some authors have disputed the usefulness of grading of ependymomas.3–7 To some extent, this controversy may have been caused by the relatively small series of ependymomas published—this glioma subtype accounts for only 9% of all neuroepithelial tumors.8 Furthermore, many studies included both subependymomas and myxopapillary ependymomas. Subependymomas typically appear in the walls of the lateral and third ventricles and have an invariable histology, with low cell-density and proliferation parameters. In addition, myxopapillary ependymomas, which arise most often in the filum terminale, display a constant histology with low indices of proliferation. The recent finding of a particular genetic background of myxopapillary ependymomas has strengthened the notion that these lesions form a distinct entity.9 For these reasons, both subependymomas and the myxopapillary variant of ependymoma should be excluded from grading studies. According to World Health Organization criteria, the other ependymomas are subdivided into cellular, clear cell, papillary, and tanycytic variants, as determined by their predominant histology.8 Despite the results from previous studies, the relevance of histopathology to the clinical outcome of patients was reappraised in a few recently published articles on intracranial ependymomas.4, 5 Unfortunately, in one study, an undetermined number of myxopapillary ependymomas were included; whereas, in the other study, the histologic parameters investigated were not evaluated for their individual prognostic power.
In the current retrospective analysis, we evaluated 13 histopathologic features for their influence on the clinical outcome in a group of 69 patients with ependymomas. Subependymomas and the myxopapillary variant were excluded from this analysis. Furthermore, ependymomas located in the spinal cord also were excluded, because the clinical situation of patients with spinal tumors is not comparable to that of patients with intracranial tumors. The tumor specimens were obtained during the first surgery performed. In addition to scoring the histologic features, immunostaining for the cell proliferation-related antigen Ki-67 (MIB-1 antibody) was performed. The results were correlated with clinical parameters such as gender, age, tumor site, and treatment and were evaluated in univariate and multivariate analyses.
MATERIALS AND METHODS
The clinical charts and the initial surgical pathologic specimens from 108 patients with ependymoma were retrieved from the files of the Department of Neurosurgery and Pathology of the Erasmus Medical Center (Rotterdam, The Netherlands). Patients had been admitted between 1970 and 2000. Collection of clinical data was closed on December 31, 2000. In two patients, histologic material was no longer available. For all other patients, the histologic diagnosis was revised, and the tumors were divided into subependymomas (n = 8 tumors), myxopapillary ependymomas (n = 9 tumors), and nonmyxopapillary, nonsubependymomas (n = 89 tumors). Because the histology of myxopapillary ependymomas and subependymomas varies little, and proliferation indices invariably appear to be low, these 2 ependymoma variants (n = 17 tumors) were left out of the evaluation of the impact of individual histologic parameters. Because the clinical situation of patients with intracranial ependymomas is essentially different from that of patients with tumors located in the spinal cord, the latter (n = 12 tumors, including 9 myxopapillary ependymomas) also were excluded from the current analyses, leaving 69 intracranial ependymomas (nonmyxopapillary and nonsubependymomas) for further study.
Tumor Histology and Scoring of Individual Histologic Features
The individual histologic parameters of the nonmyxopapillary, nonsubependymomas were recorded by two independent reviewers (P-P.Z. and J.M.K.). The group of 69 ependymomas was subdivided further into cellular ependymomas (n = 34 tumors), clear cell ependymomas (n = 8 tumors), papillary ependymomas (n = 20 tumors), and tanycytic ependymomas (n = 7 tumors) (Fig. 1). The following histologic parameters were scored: rosettes, microcysts, giant cell formation, nuclear atypia, cell density, blood vessel density, endothelial proliferation, hyalinization of vessel walls, thrombosis, perivascular cuffing, calcifications, necrosis, mitoses, and the MIB-1 labeling index (LI) (Table 1). An outline of these parameters and scoring criteria are shown in Table 1. The rosettes identified consisted of ependymal rosettes and perivascular pseudorosettes. Rosettes, microcysts, giant cells, endothelial proliferation, hyalinization of vessel walls, thrombosis, perivascular lymphocytic cuffs, calcifications, and necrosis were scored in a simple present versus absent dichotomy (Table 1). The presence of nuclear atypia and the density of blood vessels were scored as low or high relative to the scores of these features in the entire group of tumors. The tumor cell density was scored as low, intermediate, and high according to the absolute numbers counted in an area of magnification in a × 20 objective (Table 1). The mitotic index was based on the number of mitoses in 10 fields of a × 40 objective and was calculated as the percentage of nuclei (Table 1). The MIB-1 LI was based on the tumor areas with the highest mitotic index and calculated in 10 high-power fields (HPF) as the percentage of immunopositive nuclei (Table 1).
|Rosettes||Ependymal rosettes (small tubules with lumina of variable sizes) and pseudorosettes (perivascular rosettes)||Present/absent|
|Microcysts||Microscopic cysts of variable sizes within the tumor tissue lined by tumor cells, variably filled with mucinous material||Present/absent|
|Giant cells||Cells of significantly larger sizes than surrounding tumor cells, at least 4 times the square cytoplasm, usually pleomorphic, with large pleomorphic nuclei or multinucleated||Present/absent|
|Nuclear atypia||Hyperchromasia of tumor cell nuclei relative to nuclei of normal cells present in the slide (usually vascular cells), with pleomorphism (aberrant nuclear shape of individual nuclei) and polymorphism (different sizes of nuclei between different cells)||Low/high|
|Cell density||The number of cells per field in a × 10 objective||Low if < 100, intermediate if between 100 and 500, high if > 500|
|Blood vessel density||The number of blood vessels (irrespective of endothelial/pericytic proliferation) per 10 fields of a ×20 objective||Low if < 5, high if > 5|
|Endothelial proliferation||Proliferation of endothelial cells in the blood vessel walls||Present/absent|
|Hyalinization of vessel walls||Thickening of the blood vessel walls not by endothelial proliferation, showing amorphic, homogenous, pink-staining (H & E) material; present irrespective of the size of the blood vessel||Present/absent|
|Thrombosis of vessels||Present/absent|
|Perivascular lymphocytic cuffs||Present/absent|
|Mitotic index||The number of mitoses scored as the percentage of nuclei present in 10 fields of a ×40 objective||< 1; 1-5, and > 5|
|MIB-1 labeling index||Calculated as the percentage of immunopositive nuclei in 10 fields of a ×40 objective||< 1%; 1-5%, and > 5%|
Immunohistochemistry for MIB-1
Sections that measured 4 μm in thickness were deparaffinized in xylene and hydrated through descending ethanol gradients. Endogenous peroxidase activity was blocked with peroxide (3% in methanol for 20 min). After rinsing with water and phosphate buffered saline (PBS), antigen retrieval was accomplished by placing the sections in a microwave oven and heating to a temperature of 100 °C in citrate buffer, pH 6.0, for 15 minutes.
After preincubation with normal goat serum diluted 1:10 in PBS, MIB-1 antibody (mouse monoclonal; diluted 1:100 in PBS\bovine serum albumen; Immunotech, Marseille, France) was incubated for 1 hour at room temperature. The Ki-67 nuclear antigen is expressed in the nucleus of proliferating cells during the cell cycle (i.e., G1 phase, S-phase, G2 phase, and M-phase) but is absent from resting (G0) cells. The monoclonal antibody recognizes native Ki-67 antigen and recombinant fragments of the Ki-67 molecule. After they were rinsed in PBS, the sections were incubated with Ultra Sense biotinylated goat antipolyvalent (IL Immunologic, Duiven, The Netherlands) for 10 minutes at room temperature, rinsed with PBS, labeled with Ultra Sense streptavidin peroxidase (IL Immunologic) for 10 minutes at room temperature, rinsed with PBS again, and subsequently visualized by placing the slides in a 3,3′-diaminobenzidine-tetrahydrochloride solution (Fluka; Buchs, Switzerland) for 7 minutes in the dark at room temperature. After rinsing, a slight counterstaining with Mayer hematoxylin was performed.
Postoperative death, which was defined as death within the first 4 postoperative weeks, was not considered tumor-related death. Survival curves were calculated and compared by using the Kaplan–Meier method and the log-rank test. Factors that were found to be significant on univariate analysis were entered into a multivariate analysis (Cox regression). P values (2-sided) of 0.05 were considered significant.
Of the 69 patients with intracranial, nonmyxopapillary, nonsubependymomas who were included in the current analysis, 38 patients were male and 31 patients were female. Patient ages ranged from birth to 65 years, with a mean age of 26.7 years. Twenty-two patients (32%) were younger than 15 years. By the end of the clinical data collection, 29 patients were alive, 23 patients had died because of their tumor, 5 patients had died of causes unrelated to their tumor, and 12 patients had died in the postoperative period. Neither tumor localization (i.e., infratentorial or supratentorial localization; P = 0.47), patient age (P = 0.21), nor gender (P = 0.20) were found to have a significant effect on survival.
Of 69 patients, 21 patients underwent complete resection, 16 patients underwent partial resection, and 21 patients underwent partial resection and received radiotherapy. Furthermore, three patients underwent a biopsy, including only one patient who received additional radiotherapy; three patients underwent complete tumor removal with additional radiotherapy; three patients underwent partial resection followed by chemotherapy or chemotherapy and radiotherapy; and two patients underwent partial tumor resection and chemotherapy. However, the numbers of patients in these categories were too small for statistical analysis. The survival of patients who underwent complete tumor resection differed significantly from the survival of patients who underwent partial resection with or without radiation therapy (P < 0.001) (Fig. 2). A comparison of survival between patients who underwent their first surgery before 1985 and after 1985 did not reveal a significant difference.
The group of intracranial nonmyxopapillary, nonsubependymomas consisted of 34 cellular ependymomas, 20 papillary ependymomas, 8 clear cell ependymomas, and 7 tanycytic ependymomas. No significant differences in survival were found between patients who had tumors of these subtypes. In univariate analysis, both the mitotic index (P = 0.04) and the MIB-1 LI (P = 0.014) had a significant influence on survival (Table 2; Figs. 3, 4). A posteriori testing revealed that patients who had ependymomas with the lowest number of mitoses differed significantly from the other two groups. Specifically, patients who had tumors with < 1 mitosis per 10 HPF (× 40 objective) had significantly better survival compared with patients who had tumors with > 1 mitosis per 10 HPF. A posteriori testing of the subgroups of patients according to the MIB-1 LI indicated that the group with MIB-1 LI > 5% differed significantly from the other 2 groups and was associated with worse survival. Furthermore, in univariate analysis, the scores for nuclear atypia influenced survival significantly (P = 0.03): A high nuclear atypia score was associated with worse survival (Fig. 5). The absence of microcysts, high blood vessel density, and the absence of the features of vascular hyalinization demonstrated trends toward a worse survival, but none of those variables achieved significance (Table 2).
|Histologic feature||Frequency (n = 69)||P value|
|Univariate analysis||Multivariate analysis|
|Ependymal and pseudorosettes||31|
|Blood vessel density|
|Hyalinization of vessel walls|
|Perivascular lymphocytic cuffs|
|MIB-1 labeling index (n = 67)|
In the multivariate analysis (Cox regression), features that were found to have significant (and near significant) influence on survival in the univariate analysis were tested for their independent influence on survival. Among the variables presence of microcysts, nuclear atypia, blood vessel density, hyalinization of vessel walls, mitotic index, and MIB-1 LI, only the mitotic index (P = 0.027) and the MIB-1 LI (P = 0.023) were identified as factors with independent prognostic power (Table 2). The mitotic index and the MIB-1 LI were correlated highly (correlation coefficient, 0.582; P < 0.001). The effect of blood vessel density as an independent factor remained near significant (P = 0.067) (Table 2).
The literature on factors that influence the outcome of patients with ependymomas contains many contradictory results. Discrepancies extend even to parameters such as patient age that are not subject to inconsistencies in definition or scoring. In some reports, age was not mentioned as an independent significant variable that influenced survival.1, 4, 10, 11 Conversely, in other reports, age was the only statistically significant prognosticator.5 In the current study, 22 patients were younger than age 15 years, and they did not fare worse or better than the older patients, and age did not appear to have a significant influence on survival. In the literature, data regarding the role of another objective parameter, namely, tumor site, also are inconclusive: Schild et al. and Merchant et al. recorded better survival for patients with ependymomas in the posterior fossa.2, 12 Conversely, in a study by Rawlings et al., worse survival for patients who had tumors with infratentorial localization reached a near significant value.13 However, in the analysis of 62 children with ependymomas by Akyuz et al., no influence of tumor site on survival was elucidated.3 Korshunov et al. and Merchant et al. also found that ependymomas in supratentorial regions were more anaplastic, which may offer an explanation for the worse outcomes those authors recorded for patients who had tumors in those sites.5, 6 In the current study, we found that the overall survival of patients with supratentorial ependymomas did not differ from that of patients with infratentorial ependymomas. Because the above-described studies, as the current one, were not performed in a prospective way, to our knowledge there are no reliable data for a sound analysis to date. One exception to the discrepancies in the literature regarding prognostic factors is the role of patient gender: No studies in the literature have identified this factor as important, in line with the current results.
In the current study, we were unable to demonstrate an effect of the (predominant) histologic tumor subtype on survival, and this confirms the results from a number of articles published over several decades.14, 15 Some authors simply concluded that histology (either tumor type or tumor grade) does not have any measurable effect on patient outcome.5, 16, 17 Others recorded an effect of intraparenchymal growth of tumor tissue as a negative factor for prognosis compared with tumors that grew in an exophytic fashion.1, 18 These observations were made regardless of specific histologic features like the mitotic count or the MIB-1 LI. Conversely, in most studies in which it was concluded that anaplastic features had a negative influence on the clinical outcome, the growth patterns of the tumors were not taken into account.1, 4, 10, 12 It is well known that features such as high cell density and necrosis, which invariably are related to high-grade malignancy and rapid tumor progression in glial neoplasms (e.g., astrocytomas and oligodendrogliomas), do not have such connotations in patients with ependymomas.19 Furthermore, the high cellularity of ependymal neoplasms is not necessarily correlated with high indices of cell proliferation. Cellular ependymomas may vary considerably in terms of mitotic count or MIB-1 LI, variations that also were observed in the current study, and the effect of cell density on survival did not reach a significant level.
In the current investigation of 69 ependymomas, the MIB-1 LI influenced survival significantly, confirming the findings of various studies.10, 20, 21 In the literature, the percentages of the MIB-1 LI vary from 1%10 to 20%.22 We found a significant difference in survival between patients who had ependymomas with an MIB-LI < 5% or an MIB-LI > 5%. Rushing et al. concluded that the MIB-1 LI would make an objective indicator of tumor grade and that a high MIB-1 LI should be considered the most important factor in defining anaplasia in ependymomas.20 In addition, in the current study, the MIB-1 LI was identified as a factor with significant influence. Because the mitotic index also was identified as a significant factor, and because both parameters were correlated strongly, either the MIB-1 LI or the mitotic count should be included in a grading scheme for ependymomas.
In a recent multivariate analysis of 81 patients with ependymoma by Ho et al., 13 histologic parameters were tested for their prognostic significance.4 In a univariate analysis, the features cell density, mitotic count, the presence of endothelial proliferation, necrosis, thrombi, clear cells, dystrophic calcifications, bone, cartilage, and the MIB-1 LI reportedly had an impact on progression-free survival. The features endothelial proliferation, necrosis, and mitotic index also were associated successfully with patient outcome in earlier studies.4, 10, 22, 23 We were able to confirm the significant effects of the parameters of cell proliferation (e.g., mitotic count and MIB-1 LI) but failed to confirm the prognostic significance of any of the other parameters mentioned. Instead, we found that the feature nuclear atypia had a significant influence on survival. It is possible that the inclusion of myxopapillary ependymomas in the series by Ho et al. had an influence on their results, but there may be more factors responsible for the discrepancies. For instance, strict guidelines about how to score histologic features are mandatory for reliable comparisons between the results from different studies.
The number of blood vessels and the presence of vascular hyalinization showed statistical trends in their influence on survival, although no significant differences between the respective groups were reached. High vascular density and hyalinization of blood vessel walls both are related to advanced glial tumor development. Generally, high vascular density in gliomas is indicative of advanced tumorigenesis and generally is triggered under conditions of hypoxia.8 Hyalinization of vessel walls may be regarded as a degenerative phenomenon that often is observed in vessels of tumors with advanced-grade malignancy. In fact, blood vessels with hyalinized walls may be considered end-stage vessels that reflect advanced tumorigenesis, which would explain the association of this feature with high proliferative indices and a poor prognosis. The question why vascular hyalinization of tumor vessels would be correlated to some extent with clinical outcome in ependymomas and has never been identified in other glial neoplasms remains unanswered.
In conclusion, the current results confirm the significant effects of parameters of cell proliferation on the survival of patients with ependymomas. Because the factors mitotic index and MIB-1 LI are correlated strongly, a choice should be made between them for inclusion into a grading scheme. Furthermore, the current results also legitimate the inclusion of the parameter nuclear atypia in such a scheme. Only prospective studies with randomization of patients will reveal how therapeutic interventions influence the prognosis of patients with intracranial ependymomas.
The authors thank Mr. F. van der Panne for assistance with the photography.