Adult ependymal tumors: prognosis and the M. D. Anderson Cancer Center experience

Abstract

Ependymomas in adults are rare and often misdiagnosed. This study reports on a series of adult patients with confirmed ependymoma treated at The University of Texas M. D. Anderson Cancer Center (MDACC). Patients aged >17 and with ependymoma were identified, and clinical data were collected by retrospective chart review. Descriptive statistics were used to describe the clinical data, Kaplan–Meier methods were used to generate survival curves, and Cox proportional hazards models were used to evaluate the association of clinical characteristics with survival. This series included 123 adult patients [51% male; median age 39 years (18–72)]. Forty had tumors in the brain, 80 in the spine, and 3 had both. The majority were Grade I/II lesions (108) vs Grade III (anaplastic; 15). Eighteen patients had tumors that were reclassified as ependymal tumors at MDACC. The most common presenting symptom was pain, with an average of 4 symptoms reported prior to diagnosis. Sixty-three percent of patients had a gross total resection, and 49% received radiation therapy. Average follow-up was 5.5 years, and 13% had died. Median time to recurrence was 21 months (Grade II) brain and 18 months (Grade III). Worse outcome measured by overall and progression-free survival were associated with brain location (P = .01, P = .04) and tumor anaplasia (P = .0025, P = .001). An MIB-1 > 10 was associated with worse outcome (P = .03). Tumor grade and brain location are associated with a worse prognosis. Reclassification of ependymoma by neuropathologists is common. Results of this study have lead to a multicenter study to further define important diagnostic and prognostic variables for adults with ependymoma.

Ependymomas in adults are rare, accounting for only 3% of all primary central nervous system tumors diagnosed each year in the United States.¹ Although ependymomas may arise in various regions throughout the central nervous system, the frequency of location varies by age.^2,3 Spinal cord ependymomas are most prevalent in adults, whereas this location is quite rare in the pediatric population.² Conversely, infratentorial tumors are common in children but less frequent in adults.^4,5 Both age groups have a similar incidence of supratentorial tumors.

The classification of ependymomas is based on the degree of variability of cellular pleomorphism, the number of mitotic figures, cellular density, and the degree of tumor infiltration (invasion) into surrounding brain tissue. The WHO classification system categorizes ependymomas into either Grade I, II, or III.⁶ Grade I tumors, such as subpependymoma, are considered benign and are often cured with complete surgical excision. Grade II ependymomas, referred to as ependymoma, tend to be slow growing and often amenable to surgical excision. Grade III ependymomas, also known as anaplastic ependymomas, are characterized by a higher proliferative rate and a greater tendency to infiltrate surrounding brain or disseminate into cerebrospinal fluid causing drop metastases. Although distinct grades have been designated, the criteria by which a patient's tumor is assigned to a given grade remain controversial, particularly in pediatric tumors.³

Because ependymomas are relatively rare in adults; early series either focused exclusively in children or included a mixture of children and adults.^7–9 In addition, these studies often reported on a small number of patients, using data that spanned several decades, thereby introducing differences in treatment approaches that may influence outcome.^10–12 To date, a large, multicenter analysis of the diagnosis and clinical course of adult patients treated in the United States has not been reported. Recent reviews have included data collected through SEER,^13,14 but the veracity of the results are questioned because of concern regarding the accuracy of the diagnosis. For example, in the report by Metellus et al.,¹⁵ 41% of the cases reviewed had been misclassified as ependymoma by outside pathologists. More recently, the review of the SEER date by McGuire et al.¹³ raised concern regarding the unexpected low incidence of Grade III ependymomas.

Currently, the standard therapy for all ependymomas includes surgical removal with an attempt at total surgical excision. The decision to proceed with radiation is based on several factors. These include tumor grade, extent of tumor resection, and the presence or absence of tumor dissemination, either as multifocal disease or spread into the cerebrospinal fluid.^11,16 Complete surgical resection is often not possible because of the location of the tumor and the concern for damage to surrounding eloquent brain or spinal cord during surgery or the presence of dissemination at the time of diagnosis. Complete surgical removal resulting in cure is unlikely for Grade III tumors, even if post-operative imaging demonstrates no residual tumor, as these tumors have a greater propensity to infiltrate surrounding brain or spinal cord parenchyma and have a higher proliferative rate. These patients routinely undergo radiotherapy to the tumor bed and surrounding brain or spinal cord. Craniospinal radiation is generally reserved for patients with evidence of CNS dissemination. The role of chemotherapy either before or following radiation remains uncertain, although this approach has proven to be effective with medulloblastoma. Recurrent ependymoma remains a challenge without a uniformly accepted approach. Some patients undergo an additional course of radiation therapy, and others are treated with a variety of chemotherapy regimens but with an overall modest response rate.

We undertook this study to carefully evaluate a series of adult patients with ependymoma who were seen at M. D. Anderson Cancer Center (MDACC). The pathologic diagnosis was confirmed by a single neuropathologist (Kenneth Aldape) and the clinical course of each patient was extracted from the electronic medical record database. This survey provides important information on the clinical course of patients, including prognostic factors for recurrence and survival and can serve as historical data for the evaluation of new therapeutic trials recognizing the difficulty in successfully performing randomized studies in this rare disease.

Materials and Methods

Study Population

The MDACC Neuro-Oncology Database and the Neuropathology Tissue Database were reviewed, and all patients greater than 17 years of age with a diagnosis of ependymoma or an ependymoma variant treated at MDACC between 1990 and 2007 were included in the study. Fifty-nine patients (48%) had a follow-up within 2 years prior to analysis (Jan. 2007–Dec. 2008).

All tumor tissue samples were reviewed to confirm the diagnosis. A final sample of 123 adult patients was included in the analysis. Clinical data, including age at diagnosis, gender, tumor site, type and length of presenting symptoms, treatment, recurrence data, and survival, were collected by retrospective chart review.

Statistical Considerations

All statistics were performed using SAS version 9.1 (SAS Institute, Inc.). Descriptive statistics were used to describe the clinical characteristics. Survival curves were generated using the Kaplan–Meier method. Overall survival (OS) was calculated from the date of initial surgery that established the diagnosis to the date of death or last follow-up. Progression-free survival was calculated from the date of initial surgery to the date of recurrence, follow-up, or death. Time to event outcomes were compared by means of the log-rank test. Clinical and biological characteristics were analyzed for their association with survival using Cox proportional hazards models. Predictors used in the Cox proportional hazards regression model were reviewed to assess the need for transformation based on martingale residual plots. Because of the small number of events, multivariate analysis could not be performed. Therefore, only univariate analysis with a P value cutoff of .05 was reported.

Results

Diagnosis

Table 1 presents the characteristics of the study population at the time of diagnosis, including the tumor grade at presentation and the original tumor location. A summary of treatment and survival is provided in Table 2. The analysis included 123 adults with a confirmed diagnosis of an ependymal tumor. Men comprised 51% of the study population. The overall median age was 39 years (range 18–72) for the entire cohort. The tumor was located exclusively in the brain in 40 patients, either supratentorially (n = 17, 42%) or in the posterior fossa (n = 23, 58%). The tumor was spinal in location in 80 patients, and the remaining 3 patients had involvement of the brain and spine at the time of diagnosis. Tumors involving the spine were primarily Grade II lesions (n = 79), although 2 patients were subsequently found to have Grade III lesions at recurrence. For tumors involving the brain, 75% were Grade II lesions (n = 30) and the remaining tumors were Grade III at diagnosis. Eighteen adult patients (15%) had a prior diagnosis other than ependymoma that was subsequently reclassified as ependymoma when seen at MDACC. Forty-four percent of these cases were located in the brain, and 2 of these were based on tissue from a gross total resection. The majority (56%) of the reclassified cases were located in the spine, and most had been classified by an outside pathologist as astroctyomas. Two cases initially classified as Grade II ependymomas, by outside pathologists, were classified as Grade III tumors on review at MDACC.

Table 1.

Patient characteristics

	All	Brain	Spine
Total	123	40	80
Gender
Male	63 (51%)	20 (50%)	42 (53%)
Female	60 (49%)	20 (50%)	38 (47%)
Age
Range	18–72	19–72	18–69
Mean/median	40/39	40/38	42/40
Location
Supratentorial		16 (40%)
Infratentorial		23 (58%)
Malignancy
Grade II	112 (91%)	30 (75%)	79 (99%)
Grade III	11 (9%)	10 (25%)	1 (0.01%)
Grade III at recurrence	15	13	2

Table 2.

Treatment and outcome characteristics

	All	Brain	Spine
Treatment characteristics
Gross total resection	78 (63%)	23 (58%)	53 (66%)
Radiation therapy	60 (49%)	29 (73%)	30 (38%)
With Grade II	49 (82%)	19 (66%)	29 (97%)
With Grade III	11 (18%)	10 (34%)	1 (3%)
At diagnosis	44 (73%)	21 (72%)	22 (73%)
At recurrence	9 (15%)	5 (17%)	4 (13%)
With Grade II and at diagnosis	36 (60%)	13 (50%)	22 (84%)
With Grade II and at recurrence	6 (10%)	3 (12%)	3 (12%)
With Grade III and at diagnosis	8 (13%)	8 (31%)	0 (0%)
With Grade III and at recurrence	3 (5%)	2 (8%)	1 (4%)
Chemotherapy	16 (13%)	10 (25%)	6 (7.5%)
With Grade II	9 (56%)	4 (40%)	5 (83%)
With Grade III	7 (44%)	6 (60%)	1 (17%)
At diagnosis	6 (38%)	5 (50%)	1 (17%)
At recurrence	10 (62%)	5 (50%)	5 (83%)
With Grade II and at diagnosis	2 (13%)	1 (10%)	1 (17%)
With Grade II and at recurrence	7 (44%)	3 (30%)	4 (66%)
With Grade III and at diagnosis	4 (25%)	4 (40%)	0 (0%)
With Grade III and at recurrence	3 (18%)	2 (20%)	1 (17%)
Outcome characteristics
Time to follow-up (y)
Range	0–20.75	0.25–20.75	0–18.5
Mean/median	5.5/5	5.5/4	5.5/5
Time to first/second recurrence
Median
First recurrence		21 months (Grade III: 18 months)	25 months
Second recurrence		16.5 months (Grade III: 12 months)	20 months
Overall survival		221 months (Grade III: 67 months)
Died	16 (13%)	11 (28%)	5 (6%)

Interestingly, 15% (n = 18) of patients also had a second systemic cancer diagnosis. Other cancer diagnoses included 2 patients with breast cancer, 2 with prostate cancer, and 1 diagnosis of thyroid cancer, adenocarcinoma of the lung (as well as transitional cell of the bladder), sarcoma, pituitary adenoma, cartilaginous tumor, and a basal cell cancer.

Clinical Presentation

The most common presenting symptom was pain for both brain (70% of patients) and spinal (85% of patients) ependymomas. As expected, patients with brain lesions often presented with mental status changes (50%) and problems with coordination (45%), whereas patients with spine tumors commonly experienced weakness (30%) and sensory changes (70%). On average, patients experienced 4 distinct symptoms prior to diagnosis. Patients with spine tumors had neurologic symptoms for an average of 8 months prior to diagnosis (range 2–38 months), whereas brain tumor patients had symptoms for an average of 3.5 months (1–18 months). Table 3 outlines these results.

Table 3.

Presenting symptoms

Presenting symptoms	All	Brain	Spine
Range	0–13	1–7	1–13
Mean/median	4/4	4/4	4/3
Most common		Pain (70%)	Pain (85%)
		Mental status (50%)	Sensory (70%)
		Coordination (45%)	Weakness (30%)
		Nausea/vomiting (30%)	Bladder (25%)
		Vision (30%)	Bowel (11%)
		Seizures (15%)	Sexual (4%)
		Weakness (13%)
		Sensory (12%)
		Emotional (13%)
Mean time to diagnosis		3.5 months	8 months
Range		1–18 months	2–38 months

Initial Treatment

Treatment summaries are provided in Table 2. All patients underwent an initial surgical procedure. By surgical report, 63% (n = 78) had a gross total resection. Radiation treatment was used as the initial post-surgery treatment in 49% of patients, and 13% (n = 16) were also treated with chemotherapy. Documented regimens included single agent carmustine (BCNU); cyclophosphamide, cisplatin, and etoposide; or procarbazine, vincristine, and CCNU (PCV).

Prognosis

Patients in this study were followed for a median time of 5 years (range 0 to 21 years). Recurrence was more common in patients with brain lesions (48% vs 23% in the spine). For patients with brain tumors, the median time to first recurrence was 21 months, and from first to second recurrence it is 10.5 months. For patients with spine tumors, time to first recurrence was 25 months and to second recurrence was 20 months. For those with Grade III lesions, median time to first recurrence was 18 months and to second recurrence was 12 months. Sixteen patients (13%) were deceased at the time of the analysis. Overall median survival for those with Grade III lesions was 67 months. For patients with brain lesions, median survival was 221 months. Overall survival for patients with spine lesions could not be calculated due to the small number of events.

Survival

For the survival analysis, only those patients with a diagnosis of ependymoma or anaplastic ependymoma were included in the analysis. Factors included in the analysis were gender, ethnicity, tumor grade, tumor location (brain vs spine and infratentorial vs supratentorial), and extent of resection. On univariable analysis, decreased OS was associated with location in the brain (vs spinal cord) (P = .01) and Grade III (vs Grade II) diagnosis (P = .0025; Table 4 and Figs 1 and 2). A worse recurrence-free survival was also associated with diagnosis of anaplastic (Grade III) ependymoma (P < .001) and location in the brain (P = .04; Table 5 and Figs 3 and 4).

Table 4.

Analysis of survival

Predictor	Variable	Hazard ratio (95% CI)	P value
Tumor grade	Grade III	1.000	.0025
	Grade II	0.197 (0.069, 0.565)
Ethnicity	Non-white	1.000	.6600
	White	1.401 (0.312, 6.298)
Gender	Female	1.000	.7527
	Male	0.843 (0.291, 2.443)
MIBa	MIB-1 < 10%	1.000	.0278
	MIB-1 ≥ 10%	3.497 (1.145, 10.644)
Location of tumor	Spine	1.000	.0107
	Brain	5.301 (1.473, 19.07)
Location in brain	Supratentorial	1.000	.4452
	Infratentorial	0.627 (0.190, 2.076)
Resection extent	Incomplete resection	1.000	.6624
	Complete resection	0.780 (0.255, 2.385)

^aMIB-1 data available on only 63 patients.

P values in bold are significant.

Fig. 1.

Overall survival by tumor location (brain vs spine).

Fig. 2.

Overall survival by tumor grade (Grade II vs III).

Table 5.

Analysis of progression-free survival

Predictor	Variable	Hazard ratio (95% CI)	P value
Tumor grade	Grade III	1.000	<.0001
	Grade II	0.151 (0.064, 0.358)
Ethnicity	Non-white	1.000	.8512
	White	1.110 (0.372, 3.310)
Gender	Female	1.000	.4590
	Male	1.383 (0.586, 3.266)
MIBa	MIB-1 < 10%	1.000	.0061
	MIB-1 ≥ 10%	4.694 (1.556, 14.163)
Location of tumor	Spine	1.000	.0404
	Brain	2.522 (1.041, 6.110)
Location in brain	Supratentorial	1.000	.3424
	Infratentorial	0.588 (0.197, 1.760)
Resection extent	Incomplete resection	1.000	.2583
	Complete resection	0.593 (0.240, 1.467)

^aMIB-1 data available on only 63 patients.

P values in bold are significant.

Fig. 3.

Recurrence-free survival by tumor grade (Grade II vs III).

Fig. 4.

Recurrence-free survival by tumor location (brain vs spine).

In a subset of these patients (n = 63) with adequate tissue for additional analysis, the impact of MIB-1–labeling index on risk of recurrence (Table 4 and Fig. 5) and progression-free survival (Table 5 and Fig. 6) was evaluated. We found it inadequate to look at MIB-1 as a continuous variable, because of the significant impact of an MIB-1 of >10% resulting in a nonlinear relationship between survival and MIB-1. We therefore dichotomized MIB-1, looking at differences in very high MIB-1 (>10%) and those with MIB-1 ≤ 10%. The risk of death for those with an MIB-1 of >10% is 3.5 times greater than for those with an MIB of ≤10% (OR 3.497; CI 1.145, 10.644; P = .0278), and risk of recurrence for those with an MIB-1 of >10% was 4.7 (1.556, 14.163, P = .006). When we evaluated the impact of MIB-1 on survival based on lesion location, a trend was identified which indicated that those patients with brain lesions with a high MIB-1 may have a higher risk of death, but this was not statistically significant (P = .066). We also analyzed the impact of MIB-1 on prognosis in the group with exclusively anaplastic tumors. The results suggest that MIB-1 does not enhance outcome prediction, including survival in this subpopulation of Grade 3 ependymomas. A high MIB-1 was associated with shorter progression-free survival for the group overall and may reflect the importance of proliferative rate in Grade II lesions, and with a larger sample, importance may be found in those with anaplastic tumors.

Fig. 5.

Recurrence-free survival by MIB.

Fig. 6.

Progression-free survival by MIB.

Discussion

The management of adult patients with ependymoma remains controversial, primarily from the paucity of clinical investigations and conflicting results regarding factors associated with prognosis and survival. This is the first paper to describe the diagnosis and clinical course of adult patients with ependymal tumors treated in the United States.

The study has several limitations, including the retrospective nature and relatively limited amount of tumor tissue. However, despite these limitations, several significant prognostic and clinical factors were found to be significant and warrant further discussion, and the confirmation of histologic diagnosis in all evaluated cases increases the veracity of these findings.

In this review, 20% of the cases had been diagnosed as another histologic type of neoplasm prior to expert review. Seven cases were stereotactic biopsy samples, and the difficulty in diagnosis may have been impacted by the small sample size. The difficulty with making the diagnosis has previously been reported, and has lead to questions regarding the accuracy of studies based on large databases such as the Surveillance, Epidemiology and End Results (SEER) Program. The significance of proper classification for treatment planning and prognosis cannot be understated.

Controversy exists regarding the exact cellular origin of ependymoma and the prognostic significance of grade of tumor. In our series as well as others, Grade III tumors primarily occur within the brain, with nearly all spine tumors being a Grade II tumor or a variant. It has been postulated that this may reflect a different disease in the brain than what occurs in the spine. Molecular profile studies support this contention and may explain the differences in tumor biology and prognosis. It is imperative that a more refined classification system is developed to provide clinically meaningful classification for treatment purposes.

Another interesting finding is that despite the relatively young median age of this group, 15% of patients had a second cancer diagnosis. Ependymomas have been reported in patients with neurofibromatosis type II (NFII), but in this series, only 2 of the 18 patients with a second malignancy and an ependymoma had NFII. These diagnoses do not fit a known cancer-associated genetic mutation, but suggest that further investigation into genetic predisposition may be warranted.

Symptoms at presentation are increasingly recognized for their importance in not only quality of life, but also the effect on functional status and tolerance of treatment. In this study, patients with spine tumors had symptoms for an average of 8 months prior to diagnosis. This may represent the vague nature of a symptom such as back pain, but nearly 70% also had alterations in sensory function, 30% had weakness, and 25% had bowel and bladder dysfunction. Education of those health care providers involved in the initial assessment of these symptoms and the importance of imaging in any patient with back pain in addition to neurologic symptoms seems relevant. Patients with brain tumors had symptoms for a significantly shorter period of time. These presenting symptoms included mental status changes (50%), problems with coordination (45%), vision (30%), and seizures (15%). Strategies to manage these symptoms to improve quality of life and function are needed.

Maximal safe resection and use of radiation therapy are standard treatment approaches in patients with ependymoma. Most studies to date have demonstrated that gross total resection is an important prognostic variable for patients with ependymoma. In our report, 63% of patients were reported to have undergone a gross total resection. However, we were unable to demonstrate a relationship between extent of resection and survival. This may be a result of the retrospective nature of this review, without confirmation on postoperative magnetic resonance imaging. Tumor grade and location, as well as an MIB-1 value of >10% were predictive of worse outcome in this sample. Although the criteria for distinguishing low grade from anaplastic ependymomas in children are undergoing revision, the current classification in adult tumors appears to be predictive of outcome.

Further research is needed to fully define prognostic factors and evaluate treatment and clinical course of adult patients with ependymoma. The data provided here can serve as a clinical reference for the presentation and correlates of prognosis on which additional studies can be formulated. These endeavors will be accelerated by an international effort called The Collaborative Ependymoma Research Network (CERN, www.CERN-foundation.org) that has been established to promote multicenter clinical trials to improve the diagnosis and treatment of these rare tumors. Within the CERN, a follow-up study is planned compiling clinical and tumor-based data from an international cohort of samples, extending these results of the MDACC experience. This work will hopefully assist in further delineating specific prognostic and predictive molecular characteristics of tumors to further refine treatment approaches.

Conflict of interest statement. None declared.

Funding

The authors would like to thank the Collaborative Ependymoma Research Network (CERN) for their support of this work.