A multicenter study of the prognosis and treatment of adult brain ependymal tumors

Authors

  • Michele Reni M.D.,

    Corresponding author
    1. Department of Radiochemotherapy, San Raffaele Hospital Scientific Institute, Milan, Italy
    • Department of Radiochemotherapy, San Raffaele Hospital Scientific Institute, Via Olgettina 60, 20132 Milan, Italy
    Search for more papers by this author
    • Michele Reni, Alba Ariela Brandes, Giovanna Cavallo, Francesca Vastola, Umberto Basso, and Eugenio Villa are affiliated with the Gruppo Italiano Cooperativo Neuro-Oncologia.

    • Fax: (011) 39 02 26437625

  • Alba Ariela Brandes M.D.,

    1. Department of Medical Oncology, Azienda Ospedale Università, Padua, Italy
    Search for more papers by this author
    • Michele Reni, Alba Ariela Brandes, Giovanna Cavallo, Francesca Vastola, Umberto Basso, and Eugenio Villa are affiliated with the Gruppo Italiano Cooperativo Neuro-Oncologia.

  • Vittorio Vavassori M.D.,

    1. Department of Radiotherapy, Ospedale di Circolo e Fondazione Macchi, Varese, Italy
    Search for more papers by this author
  • Giovanna Cavallo M.D.,

    1. Department of Oncology, Bellaria Hospital, Bologna, Italy
    Search for more papers by this author
    • Michele Reni, Alba Ariela Brandes, Giovanna Cavallo, Francesca Vastola, Umberto Basso, and Eugenio Villa are affiliated with the Gruppo Italiano Cooperativo Neuro-Oncologia.

  • Federica Casagrande M.D.,

    1. Department of Neurosurgery, Ospedale di Circolo e Fondazione Macchi, Varese, Italy
    Search for more papers by this author
  • Francesca Vastola M.D.,

    1. Department of Medical Oncology, Azienda Ospedale Università, Padua, Italy
    Search for more papers by this author
    • Michele Reni, Alba Ariela Brandes, Giovanna Cavallo, Francesca Vastola, Umberto Basso, and Eugenio Villa are affiliated with the Gruppo Italiano Cooperativo Neuro-Oncologia.

  • Alessandro Magli M.D.,

    1. Department of Radiotherapy, Ospedale di Circolo e Fondazione Macchi, Varese, Italy
    Search for more papers by this author
  • Alberto Franzin M.D.,

    1. Department of Neurosurgery, San Raffaele Hospital Scientific Institute, Milan, Italy
    Search for more papers by this author
  • Umberto Basso M.D.,

    1. Department of Medical Oncology, Azienda Ospedale Università, Padua, Italy
    Search for more papers by this author
    • Michele Reni, Alba Ariela Brandes, Giovanna Cavallo, Francesca Vastola, Umberto Basso, and Eugenio Villa are affiliated with the Gruppo Italiano Cooperativo Neuro-Oncologia.

  • Eugenio Villa M.D.

    1. Department of Radiochemotherapy, San Raffaele Hospital Scientific Institute, Milan, Italy
    Search for more papers by this author
    • Michele Reni, Alba Ariela Brandes, Giovanna Cavallo, Francesca Vastola, Umberto Basso, and Eugenio Villa are affiliated with the Gruppo Italiano Cooperativo Neuro-Oncologia.


Abstract

BACKGROUND

The current analysis of outcomes in a large series of adult patients with intracranial ependymal tumors contributes to the characterization of the primary prognostic factors and to the therapeutic management of this rare disease, for which limited information is available in the literature.

METHODS

The authors analyzed data on patient and tumor characteristics, treatment, and survival in a series of 70 patients age > 17 years with pathologic diagnoses of brain ependymal tumors from 4 institutions.

RESULTS

The 5- and 10-year overall survival (OS) rates (± standard errors) were 67% ± 6% and 50% ± 8%, respectively. The 5- and 10-year failure-free survival (FFS) rates were 43% ± 7% and 24% ± 6%, respectively. Younger age and infratentorial tumor location were associated with longer survival. Among patients with Grade 2 ependymoma (n = 51), 21 (41%) received no postsurgical treatment. These 21 patients had a 5-year OS rate of 78% ± 10% and a 10-year OS rate of 68% ± 13%; the 5- and 10-year FFS rates for these patients were 47% ± 12% and 12% ± 11%, respectively. Twenty-six patients with Grade 2 ependymoma (51%) received postoperative radiotherapy (RT). These 26 patients had a 5-year OS rate of 71% ± 9% and a 10-year OS rate of 59% ± 11%; the 5- and 10-year FFS rates for these patients were 54% ± 10% and 34% ± 10%, respectively. Among patients with Grade 2 ependymoma, neither OS nor FFS differed significantly between those who did not receive postoperative RT and those who did; however, these two groups were heterogeneous with respect to prognostic factors. On multivariate analysis, RT use exhibited a trend toward improved OS and was significantly predictive of improved FFS.

CONCLUSIONS

The current analysis does not rule out the possibility that deferral of RT at the time of recurrence could have a detrimental effect on FFS or OS in patients with Grade 2 ependymoma, regardless of the degree of ablation. The role of postoperative RT for patients who undergo imaging-based macroscopic total resection remains to be addressed. Cancer 2004. © 2004 American Cancer Society.

Ependymal tumors are rare, accounting for 3% of all intracranial neoplasms.1 These tumors are of neuroectodermal origin and can arise from ependymal cells in the obliterated central canal of the spinal cord, the filum terminale, the choroid plexus, or the white matter adjacent to the highly angulated ventricular surface1; in addition, these neoplasms can be caused by the migration of fetal ependymal cell residue from periventricular areas in the brain parenchyma.2

Brain ependymal tumors can be classified as subependymomas (World Health Organization [WHO] Grade I), ependymomas (WHO Grade II), or anaplastic ependymomas (WHO Grade III).3 Ependymal tumors commonly are infratentorial, particularly in the fourth ventricle.4, 5 Supratentorial ependymomas arise more frequently in the lateral ventricles than in the third ventricle. Subependymomas typically are located in the fourth ventricle or in the lateral ventricles.

Due to the extremely low incidence of ependymal tumors in adults, most reported series of these neoplasms involve childhood tumors, are retrospective, include limited numbers of patients, span several decades (thus hampering the interpretation of results, because of changes in grading systems and in diagnostic and therapeutic policies), and have limited statistical power. As a consequence, universally accepted therapeutic management strategies and prognostic factors for the adult population are lacking. From this perspective, the analysis of outcome data from a large series that reflects different therapeutic trends in ordinary clinical practice could contribute to improved characterization of the main prognostic factors and the optimal therapeutic management strategies for ependymal tumors in adults. In the current article, we report the results of our analysis of the impact of tumor-, patient-, and treatment-related variables on survival in a retrospective series of 70 adult patients from a total of 4 institutions (San Raffaele Hospital Scientific Institute, Milan, Italy; Azienda Ospedale Università, Padua, Italy; Ospedale di Circolo e Fondazione Macchi, Varese, Italy; and Bellaria Hospital, Bologna, Italy).

MATERIALS AND METHODS

Study Population

Each center provided data on patient age, gender, performance status, tumor site and grade, date of diagnosis, treatment performed (including extent of surgery), irradiated volume and radiation doses, date of treatment failure, second-line treatment, and survival. Inclusion criteria were 1) pathologic diagnosis of Grade 2 ependymoma or anaplastic ependymoma; 2) disease localized in the brain; and 3) age > 17 years. Performance status and treatment type were not used to exclude patients. Patients with ependymoblastomas, subependymomas, or spinal cord ependymal tumors were excluded. In total, the 4 centers contacted submitted data on 70 patients diagnosed between May 1980 and December 2001.

Therapeutic Management

Therapeutic data were available for 66 patients. Of the remaining four patients, two died perioperatively and two were lost to follow-up immediately after surgery; these patients were excluded from further analysis. The planned treatment was surgery alone for 24 patients (36%), surgery followed by conventional external-beam radiotherapy (RT) for 36 patients (55%), surgery followed by stereotactic RT for 2 patients, surgery followed by RT and chemotherapy for 3 patients, and surgery followed by chemotherapy for 1 patient. As a general rule and also within each institution, the choice of postsurgical therapeutic strategy (which ranged from no treatment to craniospinal irradiation) appeared not to be influenced by either the extent of resection or the tumor site; instead, the postsurgical strategy depended on the attending physician. The irradiation volume included the tumor bed plus a margin in 18 patients, the whole brain in 7 patients, and the craniospinal axis in 14 patients.

Statistical Considerations

The clinical characteristics of the therapeutic subgroups were compared using the χ2 test or the Fisher exact test for categoric variables, depending on the sample size. Survival curves were generated using the Kaplan–Meier method.6 Overall survival (OS) was calculated from the date of surgery to the date of death or last follow-up, whereas failure-free survival (FFS) was calculated from the date of surgery to the date of recurrence, progression, death, or last follow-up. Survival data were reported as 5- or 10-year OS or FFS rates ± standard errors. The impact of clinical and therapeutic variables on survival was evaluated by comparing entire OS curves using the log-rank test.7 The independent prognostic value of each variable was assessed using the Cox model.8 Year of diagnosis was included as a continuous variable in the multivariate analyses to confirm that modern medical management did not influence outcomes more than treatment itself did. Backward stepwise regression was performed to identify the most powerful predictors of survival. All probability values were two sided. Analyses were performed using the Statistica 4.0 software package (Statsoft, Tulsa, OK).

RESULTS

Data on patient characteristics at diagnosis, tumor site and grade, and therapeutic management are summarized in Table 1. The median patient age was 34 years (range, 17–68 years), and the male-to-female ratio was 1:1.06. Lumbar puncture was performed at diagnosis for 15 of 45 patients (33%; data were not available for 21 patients) and yielded positive results for 2 patients (13%) undergoing macroscopic total resection of a Grade 2 infratentorial or supratentorial ependymoma.

Table 1. Patient and Tumor Characteristics for the Entire Study Population (n = 66)
CharacteristicNo. of patients (%)No. of patients (% of patients with specified characteristic)
WHO gradeTumor siteExtent of surgery
23InfratentorialSupratentorialR0R1
  1. WHO: World Health Organization; R0: total resection; R1: subtotal resection.

Age (yrs)      
 < 4041 (62)31 (76)10 (24)19 (46)22 (54)26 (63)15 (37)
 ≥ 4025 (38)20 (80)5 (20)10 (40)15 (60)13 (52)12 (48)
WHO grade      
 251 (77)26 (51)25 (49)32 (63)19 (37)
 315 (23)3 (20)12 (80)7 (47)8 (53)
Tumor site       
 Infratentorial29 (44)26 (90)3 (10)13 (45)16 (55)
 Supratentorial37 (56)25 (68)12 (32)26 (70)11 (30)
Extent of surgery      
 R039 (59)32 (82)7 (18)13 (33)26 (66)
 R127 (41)19 (70)8 (30)16 (59)11 (41)

Thirty-six patients were alive (with 17 free of disease) after a median follow-up period of 58.5 months (range, 12–265 months). Thirty patients died; causes of death included the ependymal tumor (n = 22), complications at second surgery (n = 2), unrelated causes during a disease-free period (n = 4), and a second neoplasm (n = 2). The 5- and 10-year OS rates for the entire series (n = 66) were 67% ± 6% and 50% ± 8%, respectively. The 5- and 10-year FFS rates were 43% ± 7% and 24% ± 6%, respectively.

On univariate analysis (Table 2), younger age (P = 0.02) and infratentorial site (P = 0.03) were found to be associated with longer survival, whereas gender, histologic grade, and extent of surgery had no effect on outcome. These results were confirmed when the analysis was repeated taking tumor-specific survival into account. The 5- and 10-year survival rates were 74% ± 8% and 60% ± 10%, respectively, for patients age < 40 years and 56% ± 11% and 36% ± 12%, respectively, for older patients. The 5- and 10-year survival rates were 70% ± 8% and 31% ± 12%, respectively, for patients with supratentorial tumors and 65% ± 10% (both at 5 years and 10 years) for patients with infratentorial tumors.

Table 2. Univariate Analysis of Overall Survival among Patients with Grade 2 or Grade 3 Ependymal Tumors (n = 66)
VariableOS ± SE (%)P
5 yr10 yr
  1. OS: overall survival; SE: standard error; WHO: World Health Organization.

Age (yrs)   
 < 4074 ± 860 ± 10 
 ≥ 4056 ± 1136 ± 120.02
Tumor site   
 Infratentorial65 ± 1065 ± 10 
 Supratentorial70 ± 831 ± 120.03
Gender   
 Male64 ± 947 ± 11 
 Female69 ± 953 ± 110.18
Extent of surgery   
 Radical80 ± 759 ± 10 
 Less than radical50 ± 1040 ± 120.21
WHO grade   
 271 ± 754 ± 8 
 354 ± 1436 ± 170.17

On multivariate analysis (Table 3), only age was confirmed to be a prognostic indicator, while disease site had borderline significance (P = 0.08). Gender, extent of surgery, histologic grade, and year of diagnosis exhibited no significant independent association with OS.

Table 3. Multivariate Analysis of Overall Survival among Patients with Grade 2 or Grade 3 Ependymal Tumors (n = 66)
VariableSubgroupsOdds ratio (95% CI)P
  • CI: confidence interval.

  • a

    Backward stepwise regression confirmed that age (P = 0.01) was the most powerful predictor of survival.

AgeaContinuous variable1.04 (1.01–1.07)0.01
Tumor siteSupratentorial vs. infratentorial0.49 (0.19–1.08)0.08
GenderFemale vs. male1.63 (0.76–3.40)0.22
Extent of surgeryMacroscopic total vs. subtotal1.62 (0.77–5.74)0.21
Tumor gradeGrade 2 vs. Grade 31.85 (0.69–4.90)0.16
Year of diagnosisContinuous variable0.97 (0.45–1.22)0.25

Due to the limited number of cases (n = 15), no meaningful comparisons could be drawn for patients with anaplastic ependymoma. Postoperative RT was prescribed to 11 of these 15 patients. Five patients were treated with craniospinal irradiation; 3 of these 5 experienced disease recurrence (at 7, 18, and 21 months, respectively) and died of disease (at 7, 52, and 69 months, respectively), whereas 2 patients were alive and free of disease (at 88 and 165 months, respectively). Two patients received whole-brain irradiation; both experienced recurrence (at 9 and 34 months, respectively) and died of disease (at 10 and 41 months, respectively). Four patients received RT to limited fields; 3 experienced recurrence within the tumor bed (at 4, 13, and 34 months, respectively), whereas 1 was free of disease at 57 months. All 4 patients who received RT to limited fields were alive (at 13, 38, 59, and 66 months, respectively). Four patients with anaplastic ependymoma did not receive postoperative RT. One of these 4 patients received 5 cycles of lomustine and experienced recurrence at 20 months; this patient received salvage RT and was alive and free of disease at 30 months. For the remaining 3 patients who did not receive postoperative RT, no postoperative treatment was administered, either due to concomitant non-Hodgkin lymphoma (n = 1; patient died at 37 months) or for unknown reasons (n = 2; patients experienced recurrence at 7 and 11 months, respectively). One of the two patients whose reasons for not receiving postoperative treatment were unknown underwent salvage therapy with subsequent surgery and RT and was alive and free of disease at 57 months, whereas the other patient died of tumor recurrence.

Among patients with Grade 2 ependymoma (n = 51), postoperative management was found to be variable; this finding reflected the absence of standards of care and also the variability in the personal opinions of the attending physicians. Two patients with Grade 2 ependymoma underwent postoperative radiosurgery; both experienced recurrence (at 12 and 42 months, respectively) and died (at 20 and 103 months, respectively). Two other patients received RT preceded or followed by chemotherapy; these patients experienced recurrence (at 33 and 68 months, respectively) and subsequently died (at 52 and 110 months, respectively). Twenty-one patients (41%) did not receive postsurgical treatment (Table 4).

Table 4. Patient and Tumor Characteristics for Individuals with Grade 2 Ependymoma
CharacteristicNo. of patients (%)P
No RTRT
  1. RT: radiotherapy; R0: total resection; R1: subtotal resection.

Age (yrs)   
 > 404 (19)13 (50) 
 ≥ 4017 (81)13 (50)0.04
Tumor site   
 Infratentorial11 (52)13 (50) 
 Supratentorial10 (48)13 (50)0.87
Extent of surgery   
 R016 (76)15 (58) 
 R15 (24)11 (42)0.04

At the time of the current report, 15 patients (71%) were alive after a median follow-up period of 64 months (range, 15–136 months). The median survival duration was 50+ months. The 5- and 10-year OS rates were 78% ± 10% and 68% ± 13%, respectively, and the median time to treatment failure (TTF) was 25+ months. The 5- and 10-year FFS rates were 47% ± 12% and 12% ± 11%, respectively. Twenty-six patients (51%; Table 4) received only standard postoperative external-beam RT. External-beam RT consisted of 32–66 grays (Gy; median, 54 Gy) administered to local fields only in 13 cases; 45–50 Gy (median, 45 Gy) administered to the whole brain, with or without a boost dose administered to a local field (local field dose, 45–55 Gy; median, 54 Gy), in 5 cases; and 36–54 Gy (median, 42 Gy) administered to the whole brain, with or without a boost dose administered to a local field (local field dose, 50–66 Gy; median, 54 Gy), plus 30–36 Gy (median, 30 Gy) administered to the spinal axis in 8 cases. Among patients with Grade 2 disease who received standard external-beam RT only, 13 (50%) were alive after a median follow-up period of 58 months (range, 6–265 months). The median survival duration among patients with Grade 2 disease who received external-beam RT only was 56+ months. The 5- and 10-year OS rates were 71% ± 9% and 59% ± 11%, respectively, and the median TTF was 48+ months. The 5- and 10-year FFS rates were 54% ± 10% and 34% ± 10%. Neither OS nor FFS differed significantly between patients who received external-beam RT and those who underwent surgery only (P = 0.39 and P = 0.17, respectively; Figs. 1, 2); however, patient and tumor characteristics differed between the two groups, as patients who did not receive postoperative RT were significantly younger and more frequently underwent radical surgery compared with patients who did receive RT (Table 4).

Figure 1.

Overall survival among patients with Grade 2 disease who were treated with surgery alone (dashed line) or with surgery plus external-beam radiotherapy (solid line).

Figure 2.

Time to tumor progression among patients with Grade 2 disease who were treated with surgery alone (dashed line) or with surgery plus external-beam radiotherapy (solid line).

On multivariate analysis, younger age was the only variable found to be predictive of improved OS, while RT use and infratentorial disease site possessed borderline statistical significance (P = 0.08; Table 5). On multivariate analysis of FFS, infratentorial disease site (P = 0.006), RT use (P = 0.04), and negative resection margins (P = 0.006) were found to be significantly predictive of more favorable outcome (Table 6).

Table 5. Multivariate Analysis of Overall Survival among Patients with Grade 2 Ependymal Tumors Treated with Surgery Alone or with Surgery plus External-Beam Radiotherapy (n = 47)
VariableSubgroupsOdds ratio (95% CI)P
  • CI: confidence interval.

  • a

    Backward stepwise regression confirmed that age (P = 0.02) was the most powerful predictor of overall survival.

AgeaContinuous variable1.07 (1.02–1.13)0.01
Tumor siteSupratentorial vs. infratentorial0.40 (0.15–1.09)0.08
GenderFemale vs. male1.15 (0.43–3.06)0.78
Extent of surgeryMacroscopic total vs. subtotal1.10 (0.38–3.14)0.87
Radiotherapy useNo vs. yes0.25 (0.06–1.16)0.08
Year of diagnosisContinuous variable0.54 (0.26–1.15)0.12
Table 6. Multivariate Analysis of Failure-Free Survival among Patients with Grade 2 Ependymal Tumors Treated with Surgery Alone or with Surgery plus External-Beam Radiotherapy (n = 47)
VariableSubgroupsOdds ratio (95% CI)P
  • CI: confidence interval.

  • a

    Backward stepwise regression confirmed that extent of surgery (P = 0.003), tumor site (P = 0.005), and radiotherapy use (P = 0.04) were the most powerful predictors of failure-free survival.

AgeContinuous variable1.02 (0.99–1.06)0.18
Tumor siteaSupratentorial vs. infratentorial0.24 (0.09–0.63)0.006
GenderFemale vs. male0.76 (0.34–1.70)0.51
Extent of surgeryaMacroscopic total vs. subtotal4.16 (1.57–10.9)0.006
Radiotherapy useaNo vs. yes0.30 (0.10–0.91)0.04
Year of diagnosisContinuous variable1.00 (0.59–1.69)0.99

The subset of 31 patients with Grade 2 ependymoma who underwent macroscopic total resection warranted further characterization. Of the 16 patients who received no further treatment after surgery, 7 (43%) were alive and free of treatment failure at 44 months (range, 15–114 months), 2 died (1 due to suicide and 1 due to lung emphysema, at 1.5 and 15 months, respectively) without evidence of recurrence, and 7 (43%) experienced recurrence after a median TTF of 35 months (range, 20–84 months). The 7 patients who experienced recurrence had a median age of 28 years (range, 17–41 years), and 3 of these 7 had infratentorial disease. All but 1 patient who experienced recurrence underwent salvage surgery plus RT (n = 5) or radiosurgery (n = 1); two patients also received chemotherapy. Five of these 7 patients were alive (at 8, 9, 31, 40, and 116 months after recurrence, respectively), and the remaining two died (at 50 and 78 months after recurrence, respectively). Of the 15 patients who received external-beam RT after surgery, 3 were alive and free of treatment failure (at 39, 121, and 124 months, respectively); 2 patients died of other causes while free of treatment failure (at 20 and 126 months, respectively), and 10 patients (66%) experienced recurrence after a median of 64 months (range, 5–124 months). The 10 patients who experienced recurrence had a median age of 40.5 years (range, 24–60 years), and 2 of these 10 had infratentorial disease. All but 2 patients who experienced recurrence received salvage surgery (n = 7) or chemotherapy alone (n = 1); chemotherapy was administered after surgery in 5 cases. Three patients were alive (at 1, 2, and 22 months after recurrence, respectively), and the remaining 7 died a median of 13 months (range, 0–106 months) after recurrence.

On univariate and multivariate analysis, no difference in OS was observed between patients who received whole-brain irradiation and those who received only local-field irradiation (P = 0.41) or between patients who received craniospinal irradiation and those who did not receive spinal irradiation (P = 0.30). On univariate analysis, a trend toward increased FFS was observed among patients who received whole-brain RT (P = 0.08) compared with those for whom RT was administered to more limited volumes; craniospinal irradiation did not affect FFS (P = 0.13). These findings were not confirmed with multivariate analysis.

Administration of radiation doses to the tumor bed was relatively uniform. Nineteen of 26 patients received between 50 and 56 Gy. Consequently, analysis of the impact of radiation dose on survival was not performed.

DISCUSSION

Sound evidence regarding the main prognostic factors and the optimal therapeutic management strategy for intracranial ependymal tumors in adults is lacking. These tumors are quite rare in adults, and until recently, a series that focused on a strictly adult population was not available in the literature. Although a clear correlation between age and outcome has not been demonstrated, adult patients appear to exhibit at least a trend toward more favorable prognosis relative to pediatric patients; 5-year survival rates consistently are greater than 50% for adult patients, whereas they range from 14% to 60% for pediatric patients.9–13 It also has been suggested that pediatric ependymomas may behave more aggressively because of the immaturity of neural tissue in children9, 13 and that age-related differences in outcome may be based on differences in cytogenetic aberration patterns between younger and older patients.14 In addition, considerations regarding side effects and treatment tolerability are quite different for pediatric patients and adult patients. Thus, we believe that it is prudent to report results from adult and pediatric populations separately. Recently, a small number of original articles and a review article have contributed to the body of knowledge regarding adult intracranial ependymal tumors,15–18 and the current article further addresses this topic.

Prognostic Factors

The reported 5- and 10-year survival rates for adults with ependymal tumors range from 57% to 77% and from 43% to 77%, respectively.9, 10, 15, 16, 18, 19 The findings in the current study (5-year OS, 67%; 10-year OS, 50%) were similar. Results also were comparable with respect to 5- and 10-year FFS rates, which were 43% and 24%, respectively, in the current series and 42–55% and 27–43%, respectively, in the literature.10, 15, 16, 18

The possible correlation between tumor grade and survival is controversial,5, 9, 11, 12, 18, 20–22 with differences in findings probably being attributable to sample size, anatomic tumor location, variability in the definition of anaplasia,23–25 discrepancies in histologic diagnoses,26 and the inclusion in some series of ependymoblastoma and subependymoma, which exhibit different biologic behavior and should be analyzed separately. Anaplastic ependymoma is even less common in the adult population, accounting for approximately 20% of all ependymal tumors in the current series; this figure is consistent with previously reported rates of 17–28%.4, 9, 16, 27, 28 Due to the limited sample size, solid conclusions regarding survival differences between patients with Grade 2 ependymal tumors and those with Grade 3 ependymal tumors could not be made. In other series, histology was not assessed16 or was found to be a significant independent predictor of survival only on univariate analysis,15 which prevented any firm conclusions from being made, due to differences between compared groups in terms of the distribution of prognostic factors. When the current study was designed, no central pathologic review was planned. Thus, the role of histology in predicting outcome may be masked by the varying definitions of anaplasia used by pathologists who were involved in the diagnostic process at different centers over the extended period considered in the current study; nonetheless, the purpose of the current review was to analyze results and to devise management guidelines, essentially based on local pathology reports, for ordinary clinical practice. The lack of impact of tumor grade on survival also may be attributable to the relatively small number of patients in the anaplastic ependymoma group; this limitation would only allow the detection of a large effect on survival.

Another controversial tumor-related prognostic factor is location, whose predictive role is confounded by patient age, tumor resectability, the presence or absence of a capsule, and follow-up duration.4, 5, 9–11, 18, 19, 20, 29, 30 Infratentorial location was found to be associated with a trend toward more favorable prognosis, independent of age and extent of surgical resection, both in the overall population and in the subset of patients with Grade 2 disease. Furthermore, patients with infratentorial Grade 2 ependymoma had a significantly lower risk of treatment failure than did patients with supratentorial ependymoma. Supratentorial ependymomas appear to be associated with less favorable prognosis because they are more likely to exhibit peripheral infiltrative growth into the brain parenchyma and because they are less likely to be entirely encapsulated, rendering surgical resection troublesome.4, 5, 9, 11, 15, 30 Furthermore, supratentorial tumors exhibit higher mitotic activity than do infratentorial tumors.31

Among patient-related factors, age > 40 years had a negative impact on survival. This finding was confirmed on multivariate analysis. The cutoff value of 40 years was chosen based on the median age of the study population (34 years) and on recently reported results from a series of patients with low-grade glioma.32, 33 Age was not identified as a prognostic factor by Guyotat et al.,15 who chose a different cutoff point; Guyotat et al. actually observed a difference in survival that was similar to the one reported in the current series (16%), but the difference was not statistically significant, probably due to the limited number of patients or the lack of stratification. In the series investigated by Guyotat et al.,15 older patients had a higher rate of infratentorial tumors, Grade 2 histology, and macroscopic total resection compared with younger patients; these characteristics confounded the impact of age on survival. In another series, female gender was found to be associated with increased survival19; in the current series, survival curves for men and women overlapped.

The prognostic value of extent of resection also was assessed, and this assessment did not yield clear evidence of an independent effect on survival.5, 9, 12, 18, 19, 34 For this analysis, confounding factors included the unreliability of surgical assessment of the degree of ablation35 and differences in the composition of compared groups with respect to age, performance status, and tumor grade, site, and size. In the current series, as in previous studies,15, 16 no statistically significant difference in overall survival was observed on univariate or multivariate analysis. Extent of resection was predictive of TTF. Again, the absence of a significant effect on OS may be attributable to the limited number of patients or to the limited number of events, particularly in the group of patients who underwent macroscopic total resection or received salvage treatment; however, we believe that the limited accuracy of the assessment of extent of resection is the primary reason why no effect was observed. This hypothesis is supported by the high rate (63%) of recurrence among patients who were classified as having ‘totally resected’ disease. Postoperative magnetic resonance imaging, which allows more accurate evaluation of the degree of resection, was not performed routinely in the current series; thus, a possible association between the extent of imaging-based surgical resection and outcome cannot be excluded.

Treatment

There is a general consensus that postoperative irradiation should be a part of standard care for patients with anaplastic ependymoma, and this was, in fact, the prevalent attitude noted in the current series. Because low-grade ependymoma is the prevalent histologic type in the adult population, and because treatment after surgery for patients with high-grade tumors is relatively homogeneous, no meaningful conclusions can be reached regarding the magnitude of the impact of RT on outcome or the optimal RT dose and treatment volume. General consensus recommends the delivery of 60 Gy to local fields.18 Nonetheless, the role of RT in treating Grade 2 ependymoma is controversial. In fact, the impact of RT on survival is not clearly supported by consistent statistical data, as no randomized trial has been conducted and retrospective studies have been uncontrolled or involved comparisons with historical controls,20, 29, 35 while data on survival after surgery alone are severely limited in modern series.

Recently, a number of smaller series have reported positive outcomes in children20, 27, 28, 35–40 or adults15 with totally resected Grade 2 intracranial ependymoma who did not receive RT after surgery. Reserving RT for recurrent disease appears to be an attractive strategy for patients with Grade 2 ependymoma, provided that the behavior of this malignancy is similar to that of low-grade glioma, in which no detrimental impact on OS was observed when a ‘wait and see’ policy was followed after surgery.41 Nonetheless, the current data regarding the impact of RT withdrawal on tumor control and survival must be interpreted with caution. Although univariate analyses detected no significant difference in OS and FFS rates between patients treated with surgery alone and those receiving postoperative irradiation, and although salvage therapy appeared to be more effective in the former group, the differences between the compared populations in terms of prognostic variables complicated the interpretation of these findings. In fact, when the analysis was balanced based on the primary known prognostic factors, RT was found to be associated with a clear trend, albeit one of borderline statistical significance, toward increased OS and with a significant trend toward increased FFS. The decreased risk of death or treatment failure among patients who received RT was independent of age, disease site, and extent of surgical resection and therefore was also applicable to patients who underwent macroscopic total resection.

It should be kept in mind that negative selection bias with respect to patients chosen by neurosurgeons to receive RT cannot be ruled out and that the impact of RT on outcome could be greater than what is reported in the current study. In fact, among the 27 patients with totally resected disease who did not die of other causes, the treatment failure rate was 77% (10 of 13 patients) in the group that received RT and 50% (7 of 14 patients) in the group that underwent surgery only; thus, the hypothesis regarding selection bias appears likely to be correct. In addition, for patients with totally resected disease who experienced recurrence, the use of RT delayed recurrence by approximately 30 months relative to patients who did not receive RT (median TTF, 35 months vs. 64 months).

Limited information is available on optimal treatment volume. Because the inability to eradicate the primary tumor in both low-grade and high-grade ependymomas remains the single most important factor related to treatment failure,19 and because no major difference in survival was detectable between patients who received whole-brain irradiation and those who had RT administered to more limited volumes, it appears to be reasonable to use modern conformal techniques to deliver RT to the presurgical tumor bed along with an added margin of 1–2 cm for both low-grade5, 18, 19, 21 and high-grade lesions.4, 18, 19, 29 The role of therapeutic or prophylactic craniospinal irradiation cannot be addressed by the current series, as spinal puncture at diagnosis and at recurrence was not routinely performed, and only a limited number of patients received craniospinal irradiation.

Little is also known about the optimal radiation dose to be administered, and patients in the current series were treated in a relatively uniform manner; thus, meaningful conclusions regarding dosing could not be drawn. It is probable that 54–60 Gy should be delivered to the tumor bed, with the dose to the optic chiasm limited to 55 Gy, the dose to the upper cervical spinal cord limited to 54 Gy, and the dose to the optic nerves limited to 50.4 Gy whenever possible.42

To our knowledge, the current study represents the largest report to date on adult patients with ependymal tumors. Multivariate analyses were performed to reduce selection biases and differences in terms of prognostic variables between compared groups. There have been no prospective trials involving adults with ependymal tumors. Such considerations notwithstanding, the current study suffers from all of the drawbacks associated with retrospective studies. In addition, due to the limited number of events, confidence intervals are large, and some results might not be completely reliable. Thus, our findings should be interpreted with caution and confirmed after an extended follow-up period. Nonetheless, in our opinion, the current analysis does not allow us to rule out the possibility that deferral of RT until recurrence may have a detrimental effect on FFS and/or OS for patients with Grade 2 ependymoma, regardless of the degree of tumor ablation. It is likely that a subset of patients can be treated with surgery alone; an instrument that is more accurate and reliable than subjective assessment in detecting postsurgical residual disease could be useful for identifying this subpopulation.

Ancillary