Fax: (011) 33 (0) 4 91 49 25 18
Supratentorial ependymomas: Prognostic factors and outcome analysis in a retrospective series of 46 adult patients
Article first published online: 9 MAY 2008
Copyright © 2008 American Cancer Society
Volume 113, Issue 1, pages 175–185, 1 July 2008
How to Cite
Metellus, P., Figarella-Branger, D., Guyotat, J., Barrie, M., Giorgi, R., Jouvet, A., Chinot, O. and for the Club de Neuro-Oncologie de la Société Française de Neurochirurgie and the Association des Neuro-Oncologues d'Expression Française (2008), Supratentorial ependymomas: Prognostic factors and outcome analysis in a retrospective series of 46 adult patients. Cancer, 113: 175–185. doi: 10.1002/cncr.23530
- Issue published online: 20 JUN 2008
- Article first published online: 9 MAY 2008
- Manuscript Accepted: 13 FEB 2008
- Manuscript Revised: 9 FEB 2008
- Manuscript Received: 27 NOV 2007
- French Society of Neurosurgery and the Association of Neurologists
- Neuro-oncologists of French Expression
- prognostic factors;
Ependymomas account for 2% of all intracranial tumors in adults. Supratentorial ependymomas are less common than their infratentorial counterparts. To the authors' knowledge to date, the predictive values of surgery, histology, and patient-related prognostic factors for these tumors remain unresolved. The authors report a series of adult patients with supratentorial ependymomas to characterize the roles of surgery and histology in tumor control.
The authors retrospectively studied a homogenous population of 46 adult patients who had supratentorial ependymomas from 24 French neurosurgical centers between 1990 and 2004. All clinicoradiologic and follow-up data were analyzed, and a central pathologic review was performed by 2 certified neuropathologists.
The mean (±standard error) 5- and 10-year overall survival rates for the entire population were 57.1% ± 8.7% and 41.8% ± 9.9%, respectively. The 5- and 10-year progression-free survival rates for the entire cohort were of 33.8% ± 8.1% and 25.4 ± 8%, respectively. On both univariate and multivariate analysis, age <55 years, greater extent of surgery, and lower histologic grade were associated with longer overall and progression-free survival. However, longer progression-free survival but was not considered a candidate variable for the multivariate model, because data were available for only 34 of 46 patients.
In association with age and extent of surgery, histologic grade was identified as a major prognostic factor in adult supratentorial ependymomas. The application of a simple and reproducible grading scheme using objective anaplastic criteria appeared to be both useful practically and clinically applicable. The role of adjuvant radiotherapy for patients with incompletely resected, low-grade ependymomas needs to be investigated further. Cancer 2008. © 2008 American Cancer Society.
Adult intracranial ependymoma is a relatively rare brain tumor entity that accounts for 2% to 5% of all intracranial neoplasms.1–5 Adult supratentorial ependymomas are less common than their infratentorial counterparts, and the prognostic significance of various parameters, including patient age, tumor location, volume of resection, and adjuvant therapy protocol, for patients with these tumors remains a considerable source of debate.5–14 Because of the extremely low incidence of supratentorial ependymal tumors in adults, most reported series of these neoplasms involve childhood tumors, are retrospective, include limited numbers of patients, and have limited statistical power. Furthermore, most series span several decades, thus hampering the interpretation of results because of changes in histologic grading, in diagnosis, and in therapeutic tools and policies. Consequently, these neoplasms continue to generate considerable controversy with regard to their rational clinical management.2–4
However, several studies have pointed out some histologic and behavioral differences between supratentorial ependymomas and their infratentorial counterparts. A recent report on intracranial ependymoma demonstrated site-related differences in the molecular biology of these neoplasms, raising the question of whether infratentorial and supratentorial ependymomas represent molecularly distinct entities.6
We report herein the results of a multi-institutional retrospective analysis of 46 intracranial supratentorial ependymomas in adult patients who were diagnosed between 1990 and 2004. This cohort study was conducted by the French Neurosurgical Society (SFNC), the French Speaking Association of Neurologists and Neuro-Oncolgists (ANOCEF), and the French Neuropathological Society (SFNP) to determine whether age, preoperative clinical status, tumor location, extent of surgery, histologic features, and postoperative radiotherapy (RT) affect overall survival (OS) and progression-free survival (PFS).
MATERIALS AND METHODS
A multi-institutional database search on adult intracranial ependymomas that included 24 French university hospital neurosurgical centers was conducted by the SFNC, the ANOCEF, and the SFNP. Inclusion criteria were confirmation by 2 independent neuropathologists (D.F.-B. and A.J.) of intracranial ependymoma diagnosis in patients of both sexes, aged ≥18 years, underwent surgery between 1990 and 2004 for a supratentorial tumor, and no previous brain irradiation for any intracranial pathology. After a central pathologic review of 103 patients who had supratentorial tumors that initially were diagnosed as ependymomas, 46 patients who had a confirmed diagnosis of ependymoma were eligible for this multicentric retrospective study. The clinical and radiologic treatments and follow-up data were collected by a senior neurosurgeon (P.M.) and a senior neuro-oncologist (M.B.).
The following clinicoradiologic data were collected: patient age at surgery, sex, presenting symptoms, preoperative and postoperative Karnofsky performance status (KPS) score, quality of surgical resection based on postoperative magnetic resonance imaging (MRI) scans, and complementary treatment protocol (RT and chemotherapy) if received. Tumor location was divided into 3 groups: parenchymal, third ventricle (V3), and lateral ventricle (LV) ependymomas. Contrast enhancement and associated preoperative and postoperative hydrocephalus was analyzed. Patients were considered metastatic if MRI scans demonstrated intracranial or spinal tumor dissemination and/or if cerebrospinal fluid (CSF) analysis revealed abnormal cells. Outcome measures included OS and PFS.
Except for patients who died during the perioperative period, the extent of surgery was evaluated by postoperative MRI scans. Perioperative mortality was defined as death that occurred within 3 months postsurgery and included neurosurgical complications and systemic complications, such as pulmonary embolism, sepsis, or cardiac failure. For these patients, the extent of surgery was evaluated by a single neurosurgeon (P.M.) according to an extensive review of all operative protocols. Surgical resection was classified as either complete (patients who had no residual tumor on MRI scans or who achieved total tumor removal based on operative protocol examination) or incomplete for all patients.
Regarding irradiation, the following data were collected: doses, time to surgery (adjuvant, at recurrence, or at progression) and location—cranial (focal or panencephalic) or craniospinal. For patients who received chemotherapy or underwent radiosurgery, no data were collected concerning protocol or doses, but the precise time to surgery was noted.
For all patients, the slides that were used for diagnosis, which either were stained with hematoxylin and eosin and/or were embedded in paraffin blocks, were sent to the same neuropathologist. Pathologic examination was conducted centrally by 2 senior neuropathologists (D.F.-B. and A.J.). Subependymomas and ependymoblastomas were excluded according to the World Health Organization (WHO) classification.7 Ependymomas were classified first as WHO grade 2 or WHO grade 3 (anaplastic).8 Second, ependymomas were graded according to the Marseille neograding system based on objective anaplastic features. The following criteria were assessed and quantified: necrosis (present vs absent), microvascular proliferation (present vs absent), and mitotic count in 10 consecutive high-power fields (<5 vs ≥5). An ependymoma was considered low grade if 0 or 1 of the criterion was present and high grade if 2 or 3 criteria were present. The Ki-67 labeling index was determined in 34 tumors (paraffin-embedded blocks and specimen fixed in formalin).
In doubtful cases, immunohistochemistry (immunoperoxidase with avidin-biotin complex after antigen retrieval on Ventana devices) was performed to reach a diagnosis. Expression of the following antigens was determined with the appropriate antibodies: glial fibrillary acidic protein (GFAP) (polyclonal Dakopatts), epithelial membrane antigen (clone E29), synaptophysin (polyclonal Dakopatts), keratin (clone AE1/AE3/PCK26 and KL1), and vimentin (clone V9).
Values for categorical variables were expressed as percentages. Survival was estimated by using the Kaplan-Meier method, and survival curves were compared by using the log-rank test. The effect of potential risk factors that were associated with disease-free survival and OS were evaluated with Cox proportional-hazards models. All statistical tests were 2-sided, and the threshold for statistical significance was P = .05. Analyses were performed with SPSS for Windows (version 15.0; SPSS Inc., Chicago, Ill).
Demographic data are summarized in Table 1. The mean patient age was 44 years (range, 18-76 years). Increased intracranial pressure was the most frequent symptom in V3 and LV tumors (78% of intraventricular location), whereas epilepsia, focal deficit, and behavioral changes (37%, 59%, and 42%, respectively, of parenchymal tumors) were observed mostly in parenchymal tumors. Hydrocephalus requiring shunting or endoscopic ventriculostomy was observed in intraventricular tumors only. No correlation was observed between age, sex, clinical status, and grade or extent of surgery (results not shown). However, a significant correlation was observed between age (<55 years) and tumor location (Pearson chi-square test; P = .034). Patients who had tumors located in the V3 were significantly younger than patients who had other tumor locations.
|Characteristic||No. of patients|
|Mean±SD||44 ± 17|
|Preoperative KPS score|
|Lateral ventricle||8 (17.4)|
|Extent of surgery, MRI-based|
|Surgical mortality||3 (6.5)|
|Grade 2||21 (45.7)|
|Grade 3||25 (54.3)|
|Marseille grading system|
|Low grade||21 (45.7)|
|High grade||25 (54.3)|
|Cranial alone||25 (54.3)|
|Recurrence/disease progression||28 (60.9)|
|Local isolated||22 (47.8)|
|Death during follow-up||19 (41.3)|
Of the 103 patients who were included in the central pathologic review, only 46 patients fulfilled diagnostic criteria of ependymomas. Based on both pathologic and immunohistochemical features, the following diagnoses were excluded: glioblastomas, oligodendrogliomas and mixed oligoastrocytomas, pilocytic astrocytomas, subependymomas, papillary tumors of the pineal gland, central neurocytomas, papillary glioneuronal tumors, metastases, and papillary meningiomas (Table 2).
|Result||No. of patients (%)||Repartition among initial misdiagnosis, %|
|Confirmed diagnosis of ependymoma||46/103 (44.7)||—|
|Misdiagnosis/differential diagnosis||57/103 (55.3)||100 [n=57]|
|Oligodendrogliomas and mixed oligoastrocytomas||7 (6.8)||12.3|
|Pilocytic astrocytoma||1 (0.9)||1.8|
|Papillary tumor of the pineal gland||12 (11.6)||21.1|
|Central neurocytomas||4 (3.9)||7|
|Papillary glioneuronal tumors||2 (1.9)||3.6|
|Papillary meningiomas||1 (0.9)||1.8|
Tumors were graded in 2 different manners according to the WHO classification and the Marseille grading system. All anaplastic ependymomas that were included in this study retained features of ependymomas in some areas, including perivascular pseudo rosettes in all tumors and true ependymal rosettes in a few tumors. Moreover, cells that formed perivascular pseudo rosettes had long cell processes that were immunostained with anti-GFAP antibody. According to the WHO classification, 21 patients had grade 2 ependymomas (45.7%), and 25 patients had grade 3 ependymomas (54.3%). Similarly, according to the Marseille grading system, 21 patients had grade 2 ependymomas (45.7%), and 25 patients had grade 3 ependymomas (54.3%). However, 2 grade 2 tumors in the WHO classification were reclassified as high-grade ependymomas (2 or 3 anaplastic criteria) according to the Marseille grading system, and 2 grade 3 WHO tumors were reclassified as low-grade ependymomas (0 or 1 anaplastic criterion) according to the Marseille grading system.
Location, Grade, and Extent of Surgery
Data for tumor location, tumor grade, and extent of surgery are summarized in Table 3. Twenty-two patients (47.8%) had parenchymal tumors, 16 patients (34.8%) had V3 tumors, and 8 patients (17.4%) had LV tumors. A significant correlation was observed between tumor grade and tumor location. The frequency of low-grade and grade 2 tumors was significantly greater in patients with V3 tumors (Pearson chi-square test; P = .022 for both WHO and Marseille grade), whereas the frequency of grade 3 tumors was significantly greater in patients with parenchymal tumors (Pearson chi-square test; P = .017 for WHO grade). Pearson chi-square tests failed to demonstrate a significant correlation between parenchymal location and high grade in the Marseille grading system; however, there was a trend toward a correlation (P = .07). Patients with LV tumors were distributed equally according to low or high histologic grade. Postoperative MRI scans were available for all patients who remained alive at 3 months postoperatively and constituted the reference group for evaluating the extent of surgery. Only 9 patients (19.6%) benefited from an early (<3 days) postoperative MRI scan. Thirty-seven patients had a perioperative spinal MRI scan, but CSF cytology was available for only 16 patients (34.8%). There was no significant correlation between location and extent of surgery or between grade (WHO and Marseille grade) and extent of surgery.
|Characteristic||No. of patients (%)||No. of patients with specified characteristic (%)|
|Tumor location||Extent of surgery||WHO grade||Marseille grading system|
|Parenchymal||V3||VL||GTR+||GTR−||2||3||Low grade||High grade|
|<55||31 (67.4)||11 (35.5)||14 (45.2)*||6 (19.4)||17 (54.8)||14 (45.2)||16 (51.6)||15 (48.4)||16 (51.6)||15 (48.4)|
|≥55||15 (32.6)||11 (73.3)||2 (13.3)*||2 (13.3)||6 (40)||9 (60)||5 (33.3)||10 (66.7)||5 (33.3)||10 (66.7)|
|Men||18 (39.1)||8 (44.4)||7 (38.9)||3 (16.7)||7 (38.9)||11 (61.1)||5 (27.8)||13 (72.2)||7 (38.9)||11 (61.1)|
|Women||28 (60.9)||14 (50)||9 (32.1)||5 (17.9)||16 (57.1)||12 (42.9)||16 (57.1)||12 (42.9)||14 (50)||14 (50)|
|Preoperative KPS score|
|≤80||27 (58.7)||13 (48.1)||11 (40.7)||3 (11.1)||10 (37)||17 (63)||12 (44.4)||15 (55.6)||12 (44.4)||15 (55.6)|
|>80||19 (41.3)||9 (47.4)||5 (26.3)||5 (26.3)||13 (68.4)||6 (31.6)||9 (47.4)||10 (52.6)||9 (47.4)||10 (52.6)|
|Parenchymal||22 (47.8)||NA||NA||NA||15 (68.2)||7 (31.8)||6 (27.3)*||16 (72.7)*||7 (31.8)||15 (68.2)|
|V3||16 (34.8)||NA||NA||NA||4 (25)||12 (75)||11 (68.8)*||5 (31.2)*||11 (68.8)*||5 (31.2)*|
|VL||8 (17.4)||NA||NA||NA||4 (50)||4 (50)||4 (50)||4 (50)||3 (37.5)||5 (62.5)|
|Extent of surgery|
|GTR+||23 (50)||15 (63.2)||4 (17.4)||4 (17.4)||NA||NA||10 (43.5)||13 (56.5)||10 (43.5)||13 (56.5)|
|GTR−||23 (50)||7 (30.4)||12 (52.2)||4 (17.4)||NA||NA||11 (47.8)||12 (52.2)||11 (47.8)||12 (52.2)|
|2||21 (45.7)||6 (28.6)*||11 (52.4)*||4 (19)||10 (47.6)||11 (52.4)||NA||NA||19 (90.5)||2 (9.5)|
|3||25 (54.3)||16 (64)*||5 (20)*||4 (16)||13 (52)||12 (48)||NA||NA||2 (8)||23 (92)|
|Marseille grading system|
|Low grade||21 (45.7)||7 (33.3)||11 (52.4)*||3 (14.3)||10 (47.6)||11 (52.4)||19 (90.5)||2 (9.5)||NA||NA|
|High grade||25 (54.3)||15 (60)||5 (20)*||5 (20)||13 (52)||12 (48)||2 (8)||23 (92)||NA||NA|
|No||24 (52.2)||10 (41.7)||9 (37.5)||5 (20.8)||12 (50)||12 (50)||15 (62.5)*||9 (37.5)*||14 (58.3)||10 (41.7)|
|Yes||22 (47.8)||12 (54.5)||7 (31.8)||3 (13.6)||11 (50)||11 (50)||6 (27.3)*||16 (72.7)*||7 (31.8)||15 (68.2)|
Follow-up Data and Patterns of Failure
At the end of the follow-up analysis, 27 patients (58.7%), including 18 disease-free patients, were still alive after a median duration follow-up of 62 months (range, 14-217 months). Nineteen patients (41.3%) died during follow-up. Of those patients, the causes of death included ependymoma progression in 16 patients (65.8%), complications within 3 months after surgery (operative mortality) in 3 patients (15.8% of patients who died; 6.5% of the entire population). Twenty-eight patients had recurrent or progressive disease (60.9% of patients); all 28 of those patients benefited from cerebral and spinal MRI, but CSF histology was available for only 25 patients. Disease progressed as an isolated local recurrence in 22 patients (47.8% of patients; 78.6% of failures). Disseminated disease within the central nervous system (CNS) was identified in 6 patients (13% of patients; 21.4% of failures), including, 4 patients who developed distant intracranial metastasis only (8.7% of patients; 14.3% of failure), 2 patients who presented with only spinal metastasis (4.3% of patients; 14.3% of failure), and no patient who presented with combined distant intracranial and spinal metastasis. All of these patients presented with associated focal progressive or recurrent disease. Of the 6 patients who presented with disseminated disease within the CNS, 5 patients (83.3%) had histologic grade 3 tumors according to the WHO classification, and 6 patients (100%) had high-grade tumors according to the Marseille grading system. Histology (both WHO and Marseille grade) was correlated significantly with the occurrence of metastasis (P < .005) but not with the extent of surgery (P > .05).
The 5- and 10-year OS rates for the entire cohort were 57.1% ± 8.7% and 41.8% ± 9.9%, respectively. On univariate analysis (Table 4), age <55 years (P < .0001), preoperative KPS score >80 (P = .034), macroscopic (gross) total removal (GTR) (P = .007), and low Marseille grade (P = .037) were associated with longer survival; whereas WHO grade (P = .064) had borderline significance (Fig. 1). On multivariate analysis (Table 4), age <55 years (P < .0001), GTR (P = .010), and low (0 or 1 in 3 criterion) Marseille grade (P = .021) were confirmed as prognostic indicators. KPS, sex, and adjuvant treatment exhibited no independent association with OS.
|Variable||No. of deaths/No. of patients (%)||Univariate analysis|
|Mean ± SE, %||Log rank||Multivariate analysis|
|5-year OS rate||10-year OS rate||P||RR||95% CI|
|Overall||19/46 (41.3)||57.1 ± 8.7||41.8 ± 9.9||—||—||—||—|
|<55||9/31 (20)||72.3 ± 9.9||51.1 ± 12.5||1.00||—|
|≥55||10/15 (66.6)||22.3 ± 12.7||22.3 ± 12.7||9.25||2.7-32.248|
|Men||9/18 (50)||42.2 ± 14.5||31.7 ± 14.2|
|Women||10/28 (35.7)||64.7 ± 10.7||47.2 ± 13.2|
|Preoperative KPS score||.034||.543|
|>80||5/19 (26.3)||75 ± 11||64.3 ± 13.7|
|≤80||14/27 (51.9)||41.6 ± 12.5||25 ± 11.8|
|Extent of surgery||.007||.010|
|GTR+||5/23 (21.7)||69.1 ± 12.1||69.1 ± 12.1||1.00||—|
|GTR−||14/23 (60.9)||45.2 ± 11.8||18.1 ± 11||4.17||1.41-12.37|
|2||6/21 (28.6)||75.1 ± 11.5||54.7 ± 15|
|3||13/25 (52)||42.1 ± 11.6||31.6 ± 12.6|
|Marseille grading system||.037||.021|
|Low grade (0 or 1 crit/3)||5/21 (23.8)||73.5 ± 12.4||61.3 ± 15.2||1.00||—|
|High grade (2 or 3 crit/3)||14/25 (66)||44.7 ± 11.3||26.8 ± 11.9||4.33||1.25–14.94|
|<10%||3/12 (25)||75 ± 15.8||60 ± 18.4|
|≥10%||9/22 (40.9)||49.1 ± 13.8||32.7 ± 16.2|
|No||7/24 (29.2)||73.5 ± 9.4||63 ± 12.6|
|Yes||12/22 (41.3)||41.5 ± 12.4||24.9 ± 11.8|
The 5- and 10-year PFS rates for the entire cohort were of 33.8% ± 8.1% and 25.4 ± 8%, respectively. On univariate analysis (Table 5), age <55 years (P < .0001), low Marseille grade (P = .025), WHO grade 2 (P < .046), and a Ki-67 labeling index <10% (P = .001) were associated with longer PFS (Fig. 2). On multivariate analysis (Table 5), only age <55 years (P = .005), GTR (P = .013), and low Marseille grade (P = .005) were confirmed as significant independent prognostic indicators. The Ki-67 labeling index was not considered a candidate variable for the multivariate model, because data were available for only 34 of 46 patients.
|No. with disease progression/No. of patients (%)||Mean ± SE, %||Log rank||Multivariate analysis|
|5-year PFS rate||10-year PFS rate||P||RR||95% CI|
|Overall||28/46 (60.9)||33.8 ± 8.1||25.4 ± 8||—||—||—||—|
|<55||16/31 (51.6)||44.2 ± 10.5||31.6 ± 10.7||1.00||—|
|≥55||12/15 (80)||8.5 ± 8.1||8.5 ± 8.1||3.9||1.51-10.28|
|Men||11/18 (61.1)||33.8 ± 13||22.6 ± 12.7|
|Women||17/28 (60.7)||34.5 ± 10.3||27.6 ± 10.3|
|Preoperative KPS score||.085||.819|
|>80||10/19 (52.6)||44.2 ± 12.7||35.3 ± 12.9|
|≤80||18/27 (66.7)||24.9 ± 10.5||16.6 ± 9.7|
|Extent of surgery||.017||.013|
|GTR+||11/23 (47.8)||44.7 ± 12.4||35.8 ± 12.8||0.345||0.15–0.80|
|GTR−||17/23 (73.9)||23 ± 9.7||15.3 ± 9||1.00||—|
|2||10/21 (47.6)||53 ± 13||35.4 ± 13.4|
|3||18/25 (72)||18.7 ± 8.9||18.7 ± 8.9|
|Marseille grading system||.025||.005|
|0 or 1 Crit/3||9/21 (42.9)||58.1 ± 13.2||38.7 ± 14.2||1.00||—|
|2 or 3 Crit/3||19/25 (76)||17.3 ± 8.4||17.3 ± 8.4||4.1||1.53–11.02|
|<10%||4/12 (33.3)||79.5 ± 13.1||47.7 ± 19.1|
|≥10%||16/22 (72.7)||8.4 ± 7.7||8.4 ± 7.7|
|No||14/24 (34.4)||32.8 ± 11.4||32.8 ± 11.4|
|Yes||14/22 (40.7)||26 ± 11.5||17.3 ± 10.4|
Adjuvant RT impact on OS and PFS
Among patients with low-grade and high-grade ependymomas (WHO or Marseille grade), no significant impact of complementary treatment (RT or chemotherapy) was observed in terms of PFS or OS, even when the analysis was stratified according to the quality of surgical resection.
In the entire population, 16 patients received chemotherapy (34.8%). All patients who received chemotherapy received concomitant or previous RT. According to the Marseille grading system, 12 patients had high-grade tumors, and 4 patients had low-grade tumors. Six of those 16 patients underwent GTR, and 12 of the 16 patients had died at the last follow-up.
Adult supratentorial ependymomas are rare CNS tumors that continue to generate considerable controversy with regard to their clinical management. The lack of widely accepted and recognized prognostic factors leads to the absence of standardized therapeutic guidelines. In the current study, we analyzed potential clinical and pathologic prognostic factors in the most important and homogenous population of adult patients with supratentorial ependymomas treated in the microsurgical era.
The extent of surgery has emerged as one of the most significant predictors of outcome in patients with intracranial ependymomas.1, 4, 5, 9-13 Note, however, that some authors observed no correlation between the extent of surgery and prognosis.3, 14-16 In a previous adult supratentorial ependymomas series, surgical resection reportedly was significantly prognostic only in univariate analysis.17 In our series, the extent of resection (based on postoperative MRI studies) was a major prognostic factor in terms of survival and recurrence. In fact, GTR was associated significantly with better OS and PFS in our population in both univariate and multivariate analysis. The extent of surgery also was correlated with CSF dissemination and the metastatic rate in some reports.5, 9, 18 This was not the observed in our study; however, others have reported that histologic grade was correlated strongly with CSF dissemination and metastasis.5, 9, 18
The prognostic value of histologic findings remains a controversial issue that most likely is attributable to sample size, variability in the definition of anaplasia, discrepancies in histologic diagnoses, and the inclusion in some series of ependymoblastomas and subependymomas, which exhibit different biologic behavior and should be analyzed separately.1, 4, 5, 9, 19-22 It is important to emphasize that, in the current study, 57 of 103 patients (55.3%) initially had an incorrect diagnosis of ependymoma. Four classes of pathologic misdiagnoses can be distinguished. The first includes central neurocytomas, metastatic carcinomas, and papillary meningiomas. The second class of misdiagnoses concerns confusion between subependymomas and ependymomas. The third class results from the emergence of new pathologic entities, such as papillary tumors of the pineal region, glioneuronal tumors, and oligodendrogliomas with neurocytic differentiation. The fourth and last class of misdiagnoses corresponds to gliomas, which frequently are misdiagnosed as ependymomas. Therefore, we believe that, in patients who have tumors with a supratentorial parenchymal location, the pathologist carefully should exclude oligodendrogliomas, mixed oligoastrocytomas, or glioblastomas, before assessing the diagnosis of ependymoma. Because olig2 staining usually is lacking in ependymomas, it may be useful to differentiate true ependymomas from others gliomas based on this molecular marker expression.8 In a recent series, Reni et al did not observe a correlation between histologic grade and survival.1 However, in their study, no central pathologic review was planned, and the role of histology in predicting outcome may have been masked by the various definitions of anaplasia used by the pathologists involved in the diagnostic process. The lack of impact of tumor grade on survival in their report also may be attributable to the relatively small number of anaplastic ependymomas. Some other reports have denied the prognostic value of histology.16, 19, 23 However, most recent series have demonstrated a significant increase in OS and PFS for patients with low-grade ependymomas.4, 5, 7, 9, 14, 24, 25 Korshunov et al, in a recent single-institution study of 258 intracranial ependymomas that included 143 adults who were treated in the microsurgical era, reported that the grade of tumor malignancy was a cornerstone for prognosis.5 Metellus et al also reported that histology was a major prognostic criterion in a retrospective study of 152 intracranial ependymomas in adults.26 The lack of consensus regarding ependymomas anaplastic criteria may explain the conflicting data concerning histologic grade and prognosis. For this reason, we have elucidated the Marseille grading system, a simple and more reproducible grading scheme for classifying ependymomas. Only 3 criteria were taken into account: necrosis, microvascular proliferation (present/absent), and mitotic count (threshold, 5). We observed that patients who had ependymomas that met 0 or 1 criterion had a significantly better prognosis than patients who had ependymomas that met 2 or 3 criteria. Although WHO grade had a trend toward a correlation with OS and PFS in univariate analysis; in our study, only the Marseille grading system was correlated significantly with OS and PFS in multivariate analysis. These criteria were used previously to determine prognostic significance in pediatric ependymomas; however, in that particular population, only ependymomas that met all 3 criteria differed statistically from others.7 Our results are consistent with those of Korshunov et al.5 In the current study, histologic grade was correlated significantly and strongly with OS and PFS in both univariate and multivariate analysis. These results are strengthened by the finding that a central pathologic review was planned and conducted by 2 confirmed neuropathologists (D.F.-B. and A.J.). The finding that tumor grade influenced the outcome of patients with supratentorial ependymomas, independent of others factors, should be considered in the design and analysis of future prospective trials that involve adult patients. In the current analysis, the Ki-67 immunolabeling index was available for 34 patients. On univariate analysis, a Ki-67 index <10 was correlated significantly with a better prognosis only for PFS. Because of the lack of data for this prognostic factor, we did not use it in the Cox proportional-hazards model constructs. A review of the literature on the tumor cell proliferation index in ependymomas indicated a considerable variation in Ki-67 immunolabeling fractions7, 18, 25, 27–29 That variation most likely is attributable to tissue fixation, staining protocols, and mode of quantification. For these reasons, in the current multicentric study, the Ki-67 immunolabeling index was measured in paraffin-embedded tumor blocks from the same laboratory by the same neuropathologist (D.F.-B.). Our results were consistent with those reported by Wolfsberger et al and others and underline the potential interest of assessing the Ki-67 index in adult intracranial ependymomas for outcome prediction in a routine diagnostic setting.7, 18, 24, 25, 27 Our results also indicate that, among patients with supratentorial ependymomas, anaplastic histology is more common in tumors associated with parenchymal location than in those located in V3. This significant correlation also was reported by Schwartz et al.17
Among the patient-related prognostic factors that we identified, age <55 years had positive impact on survival. Age <55 years was associated with a better prognosis in multivariate analysis in terms of OS but not for PFS. These results are consistent with those reported by Reni et al but differ from those reported by Guyotat et al.1, 3 These are the only 2 reports in the literature that analyze survival in adult intracranial ependymomas with regard to different age groups. According to the median age of our population 2 cutoff ages were chosen: age 45 years and age 55 years. Only the latter group exhibited a difference in survival analysis.
In Tables 4 and 5, PFS and OS estimates are shown at 5 years and 10 years. Even with a median follow-up of 62 months, only 5 of 46 patients in the current series had a follow-up >10 years. Thus, the estimates made at 10 years should be interpreted with caution, and these results reinforce the need for larger cohort of patients with longer follow-up.
There is a widespread opinion that postoperative irradiation should be included in the standard care of patients with high-grade ependymomas.11, 13, 14, 30 This was the most common attitude observed in our retrospective study. However, for patients with low-grade ependymomas, especially when complete tumor excision can be achieved, the role of RT remains controversial, as observed in the current series. Recently, Rogers et al reported the impact of RT in a series of 45 patients (including 25 patients who received RT) who essentially had grade 2 ependymomas (96%) of the posterior fossa.13 Those authors concluded that adjuvant RT significantly improves tumor control but not OS and, thus, recommended the use of postoperative RT regardless of the extent of surgical resection. In our study, the issue was different, because 54.3% of our patients had high-grade ependymomas. However, we were unable to determine whether the rationale for postoperative RT in patients with incompletely resected, low-grade supratentorial ependymomas was justified because of the sample sizes in our population. In our opinion, there still is not enough strong evidence that supports the benefit of RT in patients with incompletely resected, supratentorial, low-grade ependymomas to recommend adjuvant RT in this situation. A “wait-and-see” policy also may be discussed for these patients, with RT reserved for recurrent disease. Only a prospective randomized study will be able to resolve this issue.
To our knowledge, the current series represents the largest report on intracranial supratentorial ependymomas in adults in the microneurosurgical era. This study and analysis of the literature further highlights that complete tumor removal is a main prognostic factor and is the treatment of choice in adult supratentorial ependymomas. Application of reproducible diagnostic criteria for ependymoma grading has highlighted the key role of histology in clinical outcome. These 2 issues are strengthened by the finding that, in this multicentric study, a central pathologic review was conducted by 2 confirmed neuropathologists and that postsurgical residual disease was evaluated on MRI studies in almost all patients. No meaningful conclusion can be reached on the magnitude of impact of RT on clinical outcome in incompletely resected, low-grade supratentorial ependymomas and in completely resected, high-grade ependymomas. Thus, prospective clinical trials are warranted to ascertain the actual place of postoperative RT in this population.
We thank all of the neurosurgeons and pathologists who participated in this multi-institutional study. We particularly thank Dr. A. Maues de Paula, Dr. C. Fernandez, Dr. C. Bouvier, and C. Cazeaux for their help in the management of central pathological review and technical assistance. We also thank Maryna Blankenstein-Gabert for help with the English-language editing.