Treatment Outcomes of Patients with Ependymoma Receiving Radiotherapy: A Single Institution Experience Open Access

Ependymomas are rare neuroectodermal tumors originating in the central nervous system (CNS). They are reported to have an annual incidence of 0.43 patients per 100,000 population in the USA [1, 2]. In total, ependymomas represent 1.8% of all CNS tumors and 6.8% of all gliomas [2], with a male to female ratio of 1.3:1. These tumors are especially prevalent in the pediatric population, accounting for 5–8% of all primary CNS tumors, and over 90% of these cases occur as intracranial (IC) lesions [3].

Earlier research has indicated that World Health Organization (WHO) grade III anaplastic ependymoma is linked to poorer outcomes [4], while achieving surgical gross total resection (GTR), even through multiple operations, tends to improve outcomes [5‒8]. The prognosis of ependymoma is also related to the tumor location and site-specific molecular genetics. For instance, histological subclassification significantly influences prognosis in tumors located in the supratentorial (ST) or posterior fossa (PF) areas but not in the spinal (SP) cord area. Additionally, certain gene signatures (MMP9, TOP2A, TKTL1, COL3A1, LAMB1, COL4A2, TGFBI, GRIA1, F5, and SLC14A1) have been significantly associated with a worse recurrence-free survival [9‒11]. In the updated 2021 WHO classification of CNS tumors, ependymoma is classified into ten distinct entities based on the site, histology, and molecular features of the tumors [12].

The treatment of ependymoma involves a combination of surgery, radiotherapy (RT), and, in some cases, post-operative chemotherapy. While surgery is the first critical step in treating ependymoma, the role of postoperative RT has been established as a standard of care for WHO grade III ependymoma, incompletely resected ependymomas of lower grades for IC ependymomas in adults, as well as ependymomas of the SP cord [8]. Previously, young children were often exempt from RT due to potential somatic, endocrine, and cognitive impacts; however, postoperative chemotherapy alone has yielded poor results [13‒15]. With the advancements in conformal RT techniques, postoperative RT up to 59.4 Gy in 1.8 Gy fractions is now recommended for pediatric patients with IC disease, regardless of the extent of resection, even in children as young as 18 months of age [4, 8, 16]. In this retrospective study, we aimed to share our institution’s 17-year experience of incorporating RT as part of a multimodal treatment approach for patients with ependymoma, and we assess the clinical outcomes and failure patterns of these patients.

This retrospective study was conducted with the approval of the Research Ethical Committee (REC) of National Taiwan University Hospital (REC number 202103029RINB). This was a retrospective analysis of 32 patients with ependymomas who received RT as part of multimodal treatment at our institution between January 2004 and June 2022. Pediatric patients were defined as those aged between 0 and 21 years at the time of diagnosis. All histological grades of ependymoma were classified according to the 2016 WHO classification [17]. Primary tumor sites of both IC and SP cord were included in the analysis. While molecular analysis was not conducted for most of the patients, the results of any performed molecular analysis were recorded. All patients underwent initial magnetic resonance imaging (MRI), despite imaging of both brain and spine not being an institutional requirement. Patients who underwent palliative-intent treatment at the onset were excluded.

Surgical resection was performed in the majority of patients, and the extent of resection was classified according to previously established criteria [4, 7, 16]: (1) GTR: no residual tumor was observed under the microscope and on postoperative imaging. (2) Near-total resection (NTR): minimal residuum, defined as a tumor measuring 0.5 cm or smaller in the greatest dimension or <1.5 cm², was present on postoperative imaging. (3) Subtotal resection (STR): less than minimal residuum was observed on postoperative imaging.

RT was frequently performed as an initial adjuvant treatment to the tumor bed and residual tumors. However, this study also includes cases of definitive RT and RT after local recurrence. Additional RT beyond the initial treatment course was also analyzed. All RT techniques were included. Except for 1 patient who underwent CyberKnife radiosurgery, the remaining patients were treated with a total dose ranging from 45 to 66 Gy in patients with IC ependymoma and from 40 to 63 Gy in those with SP ependymoma, mostly delivered in fraction sizes of 1.8–2 Gy per fraction. Data were extracted from electronic medical records, imaging records, as well as printed or handwritten records.

Patients at our institution underwent regular MRI follow-up, usually focusing on the primary site (brain or SP cord) unless neuro-axis spread was suspected. Recurrence was defined as the occurrence of new lesions on follow-up imaging. Progressive disease was evaluated based on the 1.1 criteria of Response Evaluation Criteria in Solid Tumors (RECIST) [18].

The primary objective of this study was to identify prognostic factors affecting survival and failure patterns following multimodal therapy with RT. Secondary endpoints included survival outcomes, the rate of secondary malignancies, and treatment-related toxicity. Local progression-free survival (LPFS) was defined as the time from diagnosis to the occurrence of local recurrence, progression, death, or the last follow-up. Distant metastasis-free survival (DMFS) was defined as the time from diagnosis to neuro-axis spread beyond the original site (IC vs. SP), death, or last follow-up. Progression-free survival (PFS) was defined as the time from diagnosis to local recurrence, progression, distant metastases, death, or the last follow-up. Overall survival (OS) was defined as the time from diagnosis to any cause of death or the last follow-up [4, 19, 20].

Variables collected for analysis included age at diagnosis, sex, primary site, tumor size, extent of surgical resection, total dose of RT, and RT techniques. LPFS, DMFS, PFS, and OS were estimated using Kaplan-Meier analysis. Univariate Cox regression analysis was conducted to identify potential prognostic factors. Treatment-related toxicity of RT was assessed and graded based on the Common Toxicity Criteria for Adverse Events (CTCAE), version 5.0. Neurological deficits present at diagnosis and noted during follow-up were not considered as treatment-related toxicity in this analysis. All statistical analyses were conducted with IBM SPSS Statistics, version 26.

A total of 32 patients from our database were included in this retrospective analysis. The patient characteristics and treatment details are summarized in Table 1. Among the patients, 9 (28.1%) were pediatric. The majority of patients were diagnosed with IC ependymoma (25, 78.1%), with 15 (60%) originating in the PF. WHO grade II ependymomas were the most prevalent (15, 46.9%), followed by grade III anaplastic ependymoma (10, 31.3%). All WHO grade III tumors were IC, while all but one WHO grade I tumor (histological type unspecified) were SP. The SP WHO grade I tumors were all myxopapillary ependymomas, though the WHO grade has since been upgraded to grade II (12). The median diameter of IC tumors was 5 cm, while SP tumors had a median diameter of 8 cm. Only 1 (3.3%) patient with IC ependymoma of the PF had evidence of neuro-axis dissemination at initial presentation.

Upfront surgical resection was performed in 31 patients (96.9%). Among them, 27 patients (84.4%) received adjuvant RT, and 4 patients (12.5%) underwent aggressive RT upon local recurrence (Fig. 1). Three patients (9.4%) who received adjuvant treatment also underwent adjuvant chemotherapy. The chemotherapy regimens included bevacizumab and etoposide for one cycle followed by bevacizumab and cisplatin for two cycles; vincristine, carboplatin, and cyclophosphamide for one cycle followed by vincristine, carboplatin, and etoposide for one cycle (American Clinical Neurophysiology Society [ACNS] 0121 regimen) (16); and vincristine for four cycles. Additionally, 1 patient (3.1%) underwent definitive CyberKnife radiosurgery as the initial treatment (Fig. 1).

Extent of initial resection was as follows: GTR (13 of 31, 42.0%), NTR (5, 16.1%), and STR (13, 42.0%). All patients with STR underwent adjuvant RT. One patient with initial STR had a second, NTR following adjuvant chemotherapy. In univariate analysis, this patient was grouped as having GTR/NTR.

Molecular analyses were performed on 7 patients with IC ependymomas. Three patients diagnosed at ages 4, 6, and 7 years were found to have PF group A ependymoma with loss of H3K27me3, and one 39-year-old patient had PF group B ependymoma with preserved H3K27me3. One patient diagnosed at 16 years of age had ZFTA gene fusion-positive ST ependymoma, and another diagnosed at 24 years of age had RELA fusion-positive ST ependymoma. A 74-year-old man presented with a YAP1 fusion-positive ST ependymoma. The detailed characteristics of the patients are listed in Table 2.

The patients included in this study were treated with RT techniques including intensity-modulated radiation therapy (20, 62.5%), 3D conformal radiation therapy (7, 21.9%), tomotherapy (3, 9.4%), CyberKnife radiosurgery (1, 3.1%), and a combination of 3D conformal radiation therapy for cranio-spinal irradiation (CSI) and intensity-modulated radiation therapy for the IC tumor bed in 1 patient. In patients with IC ependymomas, one received 21 Gy in 3 fractions with CyberKnife to a total biologically effective dose (BED) of 35.7 (as calculated with an alpha beta ratio of 10). For patients treated with conventional RT, the total doses of RT ranged from 45 Gy to 66 Gy, with a median of 54 Gy and a mean of 54.7 Gy. The total BED ranged from 53.10 to 79.20, with a median of 64.26 and a mean of 65.46. Of note, 1 patient with neuro-axis dissemination at diagnosis received 18 Gy in 10 fractions to the PF following CSI with 36 Gy in 20 fractions and a subsequent boost of 5.4 Gy in 3 fractions to the residual disease. In patients with SP ependymomas, the total doses of RT ranged from 40 to 63 Gy, with a median of 46.8 Gy and a mean of 48.8 Gy. The total BED ranged from 48 to 72.9, with a median of 55.22 and a mean of 57.78. Of note, 1 patient received 45 Gy in 25 fractions to the tumor bed extending from T12 to sacrum, followed by a boost of 18 Gy in single fraction to the residual tumor.

Salvage treatment for recurrent ependymoma included surgical resection, RT, chemotherapy, and immunotherapy. Five patients received a second round of RT as salvage management, among whom one received a total of four rounds of RT (three IC and one SP) and one received a total of three rounds (all IC).

At a median follow-up of 6.25 years from diagnosis, ranging from 0.26 to 19.58 years, 11 (34.4%) patients remained disease free, 1 (3.1%) had stationary disease, 19 (59.4%) had progression, and 1 was lost to follow-up shortly after completion of RT, and the outcome is unknown. Of the patients with progression, 15 (78.9%) only had local progression, 1 (5.3%) only had neuro-axis progression, and 3 (15.8%) had both local and neuro-axis progression. At the final follow-up, 5 (15.6%) patients died, all of whom had locally progressive disease.

At 3 years of follow-up, 14 patients had disease progression, with 13 cases of local progression and 3 cases of neuro-axis progression, with one occurring in isolation, one occurring after previous local progression, and one occurring simultaneously with local progression. More than half of all local recurrences occurred within 3 years (13 of 18, 72.2%). Among them, 3 patients with local recurrence of IC ependymoma died.

The 3-year LPFS, DMFS, PFS, and OS rates were 56.3%, 89.6%, 53.0%, and 88.4%, respectively, whereas the 5-year LPFS, DMFS, PFS, and OS rates were 48.5%, 89.6%, 45.1%, and 88.4%, respectively (Fig. 2a–d). The median DMFS and OS were not reached, while the median LPFS and PFS were 3.94 and 3.05 years, respectively. Overall, five deaths occurred, with 2 patients having multiple local and neuro-axis recurrences (five and six recurrences, respectively).

In the univariate analysis (Table 3), the extent of resection was prognostic for LPFS and PFS in IC tumors, with STR displaying poorer outcomes than GTR or NTR (hazard ratio [HR] = 3.69, p = 0.018, for LPFS; HR = 3.20, p = 0.029, for PFS) (Fig. 3a and b). There was no significant difference in either LPFS or PFS between the GTR and NTR groups. Additionally, STR showed a trend toward poorer OS in patients with IC tumors (HR = 7.60, p = 0.070) (Fig. 3c) and in the LPFS, PFS, and OS of the overall population (HR = 2.49, p = 0.063, for LPFS; HR = 2.22, p = 0.097, for PFS; HR = 7.18, p = 0.079, for OS) (Fig. 3d–f). Additionally, WHO grade III tumors trended toward poorer DMFS when compared with that of WHO grade II tumors (HR = 7.27, p = 0.098), and a larger tumor diameter showed a non-significant correlation with improved PFS (HR = 0.81, p = 0.096), which could be attributed to larger tumors in SP ependymomas (mean diameter 6.44 cm vs. 4.91 cm for IC tumors), as IC tumors had non-significantly inferior LPFS and PFS (HR = 1.76, p = 0.375, for LPFS; HR = 1.92, p = 0.303, for PFS). Tumor location did not significantly affect the outcomes (Fig. 4a–d). In addition, age, sex, total RT dose (Gy), or total RT BED were not associated with the clinical outcomes (Table 3).

Nineteen patients (59.4%) experienced recurrence. Of the first recurrences, 17 (89.4%) were isolated local recurrences, 1 was an isolated neuro-axis recurrence, and the last patient had both local and neuro-axis recurrences. Most patients had a single recurrence (11 of 19, 57.9%), six (31.6%) had two recurrences, and two (10.5%) had three or more recurrences, both of whom had IC ependymomas. All 5 patients with neuro-axis dissemination (1 patient had SP seeding at diagnosis and thus was not categorized as having neuro-axis progression) had primary IC disease. Overall, 11 patients (57.9%) achieved long-term control (minimum follow-up 2.7 years), with either stable disease (5 patients) or disease-free status (6 patients). In patients with evaluable records, the majority of first local recurrences after RT occurred in the treatment field (13 of 16, 81.3%), mostly in IC ependymomas (12 of 13, 92.3%). One patient with an IC ependymoma had both infield and outfield IC recurrences. Two patients (12.5%) with pure outfield recurrence had SP ependymomas.

Out of 26 evaluable patients, 20 (76.9%) experienced at least one acute toxicity, and 6 (23.1%) had ≥ grade 2 toxicity. The most common acute toxicity was radiation dermatitis (n = 13, 50%), followed by fatigue (n = 7, 26.9%), nausea and vomiting (n = 5, 19.2%), and alopecia (n = 4, 15.4%). Notably, 1 patient developed cellulitis of the scalp secondary to grade 3 radiation dermatitis, whereas another experienced grade 3 fatigue and decreased appetite, likely due to brain edema.

Regarding late toxicities, regular assessments were not conducted, and the absence of toxicity was not consistently recorded in our medical records. One patient reported symptoms of dizziness, headache, insomnia, and nausea during follow-up, but as they had repeated IC recurrences, it is challenging to determine the extent to which these symptoms were attributable to RT. Similarly, another patient exhibited decreased memory function and total blindness that presented during the initial symptom onset, but their history of childhood seizure complicates the assessment of RT’s contribution to these symptoms. Notably, 1 pediatric patient was diagnosed with autism spectrum disorder 3 years into follow-up, an adult patient developed Parkinson’s disease 15 years after treatment, and another patient is currently under surveillance for pituitary tumor due to Cushing’s syndrome. All 3 patients received IC RT.

Furthermore, 1 patient developed 2 secondary malignancies, namely, endometrial endometrioid adenocarcinoma and retroperitoneal leiomyosarcoma, both diagnosed 8 years after ependymoma treatment.

In our cohort, which consisted of patients treated for ependymoma with curative intent RT, we observed promising long-term OS outcomes, with a 5-year OS rate of 88.4%. In a study of the International Society for Pediatric Oncology (SIOP) Ependymoma I, which enrolled pediatric patients (age 3–21) with IC ependymoma who underwent GTR followed by RT or STR followed by chemotherapy and subsequent RT, Ritzmann et al. [7] demonstrated a 5-year OS rate of 69.3%. Another prospective study, ACNS 0121, conducted by Merchant et al. [16] focused on pediatric patients who received GTR/NTR followed by RT or STR, chemotherapy, and secondary surgery and subsequent concurrent chemoradiotherapy and reported a 5-year OS rate of 83.8%. However, it is important to note that these data are focused on IC ependymomas in the pediatric populations [7, 16]. In comparison, our study population consisted mostly of adult patients. According to Ostrom et al. [2], the 5-year OS rate for patients with brain and other CNS tumors from population-based data in the Central Brain Tumor Registry of the United States (CBTRUS) was 84.2%.

Despite nearly sixty percent of our cohort having experienced local or neuro-axis recurrence, we achieved favorable survival outcomes. Our 5-year PFS rate of 45.1% is lower compared to ACNS 0121 (5-year EFS 62.7%) and similar to the findings of the SIOP Ependymoma I study (5-year EFS 49.5%) [7, 16]. The high 5-year OS rate of 88.4% in our cohort may be attributed to successful salvage strategies, including RT, chemotherapy, and surgical resection, employed for recurrent or progressive diseases. Notably, among the 4 patients who did not initially receive adjuvant RT but underwent aggressive RT after recurrence, two achieved a disease-free status without further progression. Although the other 2 patients experienced recurrence after RT, one had stable disease after surgical resection at a follow-up of 7.5 years, and the other had stationary disease under observation at a follow-up of 6.3 years.

The primary aim of our study was to identify prognostic factors and failure patterns that could guide the treatment of ependymoma. A study conducted by Paulino et al. [6] at the Children’s Hospital of Iowa involving 49 patients with IC ependymoma identified GTR and low-grade histology as favorable prognostic factors for local control. These findings align with the reports of other investigators [4, 5, 7, 8]. Despite the small number of patients in our analysis, STR was significantly correlated with worse LPFS and PFS in IC ependymomas and showed a trend toward worse OS in IC ependymomas and in LPFS, PFS, and OS in the overall population. However, there was no difference in the outcomes between patients receiving GTR and those receiving NTR. Thus, our results further support previous findings highlighting the negative impact of the STR on the clinical outcomes of patients with ependymomas [4‒8].

In our study, the incidence of neuro-axis dissemination was rare, both on initial diagnosis (less than 5%) and in first recurrences (2 of 19, 10.5%, one of which was simultaneous with local recurrence). Notably, while IC location did not significantly increase the risk of neuro-axis dissemination, only patients with IC tumors had neuro-axis spread, and none of the patients initially presenting with SP ependymoma had IC metastasis. Our rate of SP metastasis (10.5%) was consistent with previous studies, which reported rates ranging from 9.1% to 14% [5, 21‒23]. However, it is important to acknowledge that this percentage may be underestimated as the majority of patients did not receive regular MRI imaging of the spine. Paulino et al. [6] also found a significant correlation between neuro-axis dissemination, local failure, and high-grade histology. Although our patient sample size was too limited to adequately analyze the effect of WHO grade on DMFS, WHO grade III showed a trend toward worse DMFS outcomes than that of WHO grade II. We observed only one instance of isolated neuro-axis metastasis as the first recurrence, which reinforces the notion that local control remains the main challenge in ependymoma treatment [24].

The majority of recurrences in our patients were amenable to local salvage, with approximately half of the patients achieving long-term control. Interestingly, in our series, a patient with initial SP metastasis received subtotal IC resection followed by adjuvant CSI, PF boost, and residual tumor boost and remained disease free at a final follow-up of 11.2 years. Despite treating patients with a total dose similar to the 2018 European Association of Neuro-Oncology (EANO) guidelines of 54–59.4 Gy for IC ependymomas and 45–54 Gy for SP ependymomas [7], most recurrences, including those occurring after salvage RT, were infield. This raises the question of whether dose escalation could improve local control. In an Italian Association of Pediatric Hematology Oncology (AEIOP) study, a prescription of 70.4 Gy in 1.1 Gy fractions administered twice daily resulted in a 5-year PFS rate of 56% [25]. In a subsequent AEIOP protocol, patients with IC ependymoma underwent RT with 59.4 Gy in 1.8 Gy fractions, followed by either second-look surgery when possible or an 8 Gy boost in 4 Gy fractions to residual tumors [4]. This approach yielded a 5-year EFS of 65.4%, suggesting potential benefit in an additional boost beyond 59.4 Gy [4]. However, a retrospective analysis of the National Cancer Database published in 2019 showed no survival benefit when dose-escalated radiation of ≥59.4 Gy was administered [26]. It is important to note the retrospective nature of this analysis. Additionally, dose escalation of RT carries the risk of increased radiation toxicities, including a decrease in intelligence quotient in pediatric patients [27], as well as damage to endocrine functions [28, 29]. Nevertheless, identifying patients with ependymoma who may benefit from dose escalation of RT could improve local control in this population.

There are limitations to our study, including its retrospective nature and inherent selection bias, as well as the simultaneous inclusion of SP, infratentorial, and ST ependymomas. Although ependymal tumors are generally more prevalent in the pediatric population, our cohort primarily consisted of adult patients. This may be partially attributed to reluctance on the part of patient families, surgeons, or pediatric specialists to refer pediatric patients for RT treatment. This limitation is unfortunate as most prospective trials on ependymoma focus on the pediatric population. The small number of patients in our study makes it challenging to achieve statistical significance in determining prognostic factors. However, it is encouraging to observe that our findings align with the existing literature. The heterogeneity in our patient population, particularly in terms of tumor location initial radiation intent (adjuvant vs. aggressive salvage RT upon recurrence), lacking a high enough number of patients in the salvage group, further complicates drawing conclusive statements regarding the necessity of immediate adjuvant treatment. Another limitation is the lack of molecular and genetic characteristics in many cases of our patients, as this molecular and genetic information is characterized by important prognostic factors for ependymoma [9, 30, 31]. However, in the present study, 7 patients had information on molecular and genetic characteristics, including the presence of YAP1 or RELA gene fusions, and the aforementioned two genes represented distinct subgroups and prognoses of ependymoma [9, 30, 31]. Further analysis should be conducted in the future with larger cohorts and separate analyses based on the treatment regimen, tumor location, histology, and molecular and genetic characteristics of ependymoma.

Our results indicate that the extent of resection is a significant prognostic factor for local control in patients with IC ependymomas. Multimodal treatment, including RT for both IC and SP ependymomas, yields exemplary OS rates for patients with ependymoma. Considering most recurrences occur infield of RT, dose escalation of RT in suitable patients with ependymoma could be explored in future studies.

This research was conducted with the approval of the Research Ethical Committee (REC) of National Taiwan University Hospital (REC number 202103029RINB). The patients’ medical data were anonymized prior to access and analysis. The Institutional Review Board has waived the need for written informed consent from study subjects because all potentially patient-identifying information was removed prior to data analysis.

The authors declare no competing financial interests.

This research was funded by National Science and Technology Council, Taiwan, No. 111-2314-B-002-010-, No. 111-2811-B-002-095-, and No. 110-2314-B-002-219-MY3 and National Taiwan University Hospital, Taiwan, No. NTUH 113-S0087.

T.T.-F.L. and S.-H.K. contributed to the study design; J.C.-H.C., Y.-H.C., F.-M.H., K.-H.L., C.Y.-H., C.-W.W., and S.-H.K. treated patients; T.T.-F.L. and S.-H.K. were involved in data analysis and interpretation; and T.T.-F.L. and S.-H.K. wrote the manuscript, which was revised and approved by all coauthors.

The data generated or analyzed during this study are not publicly available for protection of the privacy of research participants but may be available from the corresponding author (S.-H.K.) upon reasonable request.