Journal of Neuro-Oncology

, Volume 98, Issue 1, pp 83–92

Primary spinal cord glioma: a Surveillance, Epidemiology, and End Results database study

Authors

    • Department of Radiation OncologyUniversity of Rochester Medical Center
  • Mahlon D. Johnson
    • Department of PathologyUniversity of Rochester Medical Center
  • Joohee Sul
    • Department of NeurologyUniversity of Rochester Medical Center
  • Nimish A. Mohile
    • Department of NeurologyUniversity of Rochester Medical Center
  • David N. Korones
    • Department of PediatricsUniversity of Rochester Medical Center
    • Department of Medicine, Division of OncologyUniversity of Rochester Medical Center
  • Paul Okunieff
    • Department of Radiation OncologyUniversity of Rochester Medical Center
  • Kevin A. Walter
    • Department of Medicine, Division of OncologyUniversity of Rochester Medical Center
    • Department of NeurosurgeryUniversity of Rochester Medical Center
Clinical Study - Patient Study

DOI: 10.1007/s11060-009-0054-7

Cite this article as:
Milano, M.T., Johnson, M.D., Sul, J. et al. J Neurooncol (2010) 98: 83. doi:10.1007/s11060-009-0054-7

Abstract

To characterize the overall survival (OS) and cause specific survival (CSS), and variables affecting outcome, in patients with primary spinal cord astrocytoma (SCA) and ependymoma (SCE). About 664 patients with SCA and 1,057 patients with SCE were analyzed using the Surveillance, Epidemiology, and End Results database. For grade 1, 2, 3 and 4 SCA, the 5-year OS was 82, 70, 28 and 14%; the 5-year CSS was 89, 77, 36 and 20%. For SCA, lower grade, younger age, and undergoing resection significantly improved OS and CSS; treatment without radiotherapy was favorable for CSS. Smaller tumor size also improved survival. For grade 1, 2, and 3 SCE, the 5-year OS was 92, 97 and 58%; the 5-year CSS was 100, 98 and 64%. For SCE, lower grade, younger age, and undergoing resection significantly improved OS and CSS; treatment without radiotherapy was favorable for OS. Smaller tumor size did not confer a survival benefit. Patients with resected grade 2 spinal cord glioma who did not receive radiotherapy fared well with respect to OS and CSS. For patients with spinal cord glioma, the variables of histology, grade, age and undergoing resection are significant predictors of outcome. Though treatment with radiotherapy was associated with worse outcomes, this may reflect a bias in that patients who underwent radiotherapy were perhaps more likely to have had adverse risk factors. Given the retrospective nature of this study, specific recommendations about which situations warrant radiotherapy cannot be determined.

Keywords

Spinal cord ependymomaSpinal cord astrocytomaSpinal cord gliomaRadiotherapy

Introduction

Primary spinal cord tumors are uncommon. Spinal cord glioma (SCG) represents <30% of all spinal cord tumors, with an overall incidence of approximately 0.22 SCG per 100,000 person years [1]. Standard treatment options for SCG include resection, radiotherapy and combined modality resection and radiotherapy [239]. Some series show that patients with SCG fare better than those with intracranial gliomas [6, 29, 39, 40] though other studies do not [4, 7, 27]. Several studies demonstrate that spinal cord astrocytomas (SCA) fare worse than spinal cord ependymomas (SCE) [8, 13, 14, 21, 25, 30].

Previously reported adverse risk factors for SCE include: more advanced age [17, 26, 27], younger age [39], poor performance status/neurologic function [6, 17, 19, 27, 36], higher grade [6, 17, 21, 22, 27, 29, 39], less extensive resection [9, 12, 16, 17, 26, 29, 35, 38, 39, 41, 42], and sub-therapeutic radiation dose [2, 4, 6, 8, 10, 19, 27]. The variables of age [10, 19, 30], extent of resection [10, 21, 27, 30], and grade [41] were not significant in other studies. Previously reported adverse risk factors for SCA include: poor performance status [34], more advanced age [20, 33, 37, 39, 43], younger age [24], higher grade [3, 13, 18, 24, 25, 33, 34, 37, 39, 43], and less extensive resection [8, 21, 39, 44], though less extensive resection was not an adverse risk factor in most studies [12, 18, 20, 24, 33, 37], with higher grade tumors [12, 18, 37], as well as lower grade tumors [12, 20, 24, 34]. More extensive resection was an adverse risk factor in one recent study [43]. Generally grade 4 SCG has a dismal prognosis, though long-term survivors after aggressive resection and radiotherapy have been reported [25].

Small patient numbers, and differences in patient populations between different studies complicate the interpretation of these findings. With rare exceptions [39], the published outcome on SCG represents single institution retrospective studies with small patient numbers, spanning several decades.

This study offers a descriptive analysis of patients with SCG included in the Surveillance, Epidemiology, and End Results (SEER) database, with the goal of better characterizing the overall survival (OS), cause specific survival (CSS) and risk factors affecting patient outcome.

Methods

Patient database

The SEER database is a longitudinal database that collects information from 17 cancer registries, representing 26% of the United States population (http://seer.cancer.gov/). Serial registry data are de-identified, submitted to the United States National Cancer Institute on a biannual basis, and are available for researchers. The SEER 17 limited-use registry was used [45]. Nine of the registries date back to 1973–1975, four registries date back to 1992, and four registries date back to 2000.

A total of 1,814 patients with primary SCG (excluding primitive neuroectodermal tumors), coded in the SEER database as “Primary site-labeled” = “C72.0-Spinal cord” were included in the present analysis. Patients coded as “Primary site-labeled” = “C72.1-Cauda equina” and “C70.1-Spinal meninges” were not included.

In the SEER database, “Grade” was scored as a measure differentiation, and not meant to reflect grade as scored by any recognized grading system. “Grade” was classified as: “Well differentiated; Grade I,” (n = 175) “Moderately differentiated; Grade II,” (n = 337) “Poorly differentiated; Grade III,” (n = 60) “Undifferentiated; anaplastic Grade IV,” (n = 138) or “Unknown” (n = 1,104), based upon explicit mention of differentiation or grade in the pathology report. From 2004, ependymomas were characterized as ‘malignant’, ‘borderline’ or ‘benign.’ Myxopapillary ependymoma and subependymoma were considered, ‘borderline’ malignancies. ‘Benign’ and ‘borderline’ tumors were not included prior to 2004; however, ‘malignant’ myxopapillary ependymoma was included in the ICD-O-2 and ICD-O-3 fields as “Myxopapillary ependymoma, malignant” or “Ependymoma, NOS” and ‘malignant’ subependymal tumors were included as “Subependymal glioma.”

The “Histology/behavior,” “Histology recode-Brain groupings” and “Grade” fields were used to define tumor grade for the purpose of this study, with the understanding that this “grade” does not reflect a WHO grade. For this study, the following assumptions were made: tumors classified as pilocytic astrocytoma, myxopapillary ependymoma or subependymoma in the SEER database histology fields were scored as grade 1, even if the grade in the SEER database was scored as “Moderately differentiated; Grade II” or “Unknown”; tumors classified as “Benign” or “Borderline malignancy” in the behavior code fields were scored grade 1; tumors classified as fibrillary astrocytoma in the SEER database histology fields were scored as grade 2, even if the grade in the SEER database was scored as “Unknown”; tumors classified as “Anaplastic astrocytoma” or “Anaplastic ependymoma” in the SEER histology fields were scored as grade 3, even if the grade in the SEER database was scored as “Undifferentiated; anaplastic; Grade IV” or “Unknown”; tumors classified as glioblastoma in the SEER database histology fields were scored as grade 4, even if the grade in the SEER database was scored as “Poorly differentiated; Grade III” or “Unknown.” Astrocytomas not specifically coded as “Anaplastic astrocytoma” in the SEER database histology fields were scored as grade 3 if categorized as “Poorly differentiated; Grade III” and grade 4 if categorized as “Undifferentiated; anaplastic; Grade IV.” Ependymomas not specifically coded as “Anaplastic ependymoma “in the SEER database histology fields were scored as grade 3 if categorized as “Poorly differentiated; Grade III.” Cases scored in the SEER database as “Unknown” grade as well as: “Astrocytoma, NOS,” “Glioma, malignant,” “Mixed glioma,” “Oligodendroglioma, NOS” or “Ependymoma, NOS” in the “Histology/behavior” field; and “Astrocytoma, NOS,” “Glioma, NOS,” “Mixed glioma,” “Oligodendroglioma” or “Ependymoma/anaplastic ependymoma” in the SEER database histology fields were considered to have unknown grade for this analysis.

The SEER “Site specific surgery” and “Surgery of primary site” fields describe the extent of surgical resection. Cases in the SEER database categorized as “Unknown if surgery performed” were considered to not have surgery if the “Diagnostic confirmation” field was “Clinical diagnosis only” or “Radiography without microscopic confirmation.” Cases in the SEER database categorized as “Unknown if surgery performed” were considered to have had resection if diagnostic confirmation was “Direct visualization without microscopic confirmation.” Cases in the SEER database categorized as having undergone surgery, but diagnosed by “Clinical diagnosis only” or “Radiography without microscopic confirmation” were assumed to have undergone the procedure, but without pathologic confirmation. The SEER database has a field describing if radiotherapy was administered, though neither the radiation dose nor extent of field are available. There is no field to describe if chemotherapy was administered. The SEER database does not provide information about the location of the tumor or number of sites within the spine, or whether the cerebral spinal fluid was seeded with metastatic cells.

Statistical analysis

Stata version 9.2 (StataCorp, College Station, TX) was used for data analysis. Actuarial OS and CSS were calculated using the Kaplan–Meier method [46]. The log-rank test was used for univariate analysis (UVA) comparison of survival between subgroups. Cox proportional hazards model was used for multivariate analyses (MVA) [47], performed using variables significant on UVA. Patients with unknown surgery and/or unknown radiation delivery were excluded from the UVAs and MVAs.

The MVAs in which grade is included omit a large number of patients whose grade is unknown. The majority of patients with unknown grade were identified as having ependymoma, and the treatment and outcome of those patients with unknown grade are similar to patients with low grade SCG (see “Results”). It is a reasonable assumption that most of these patients had low grade SCG. In an effort to further explore the variables significant on UVA for OS and CSS, additional MVA models were performed in which unknown grade was re-categorized as grade 2 (“grade unknown → 2” MVA models).

Results

Patient and tumor characteristics

Table 1 summarizes the patient and tumor characteristics. Patient age ranged from <1 to 91 years old. Most patients were white. Close to half were diagnosed in the 2000s (due in part to the addition of four registries in 2000). Most patients with unknown grade had ependymoma. Tumor size was unavailable in 224 patients (diagnosed before 1983) and unknown in 1,104; size ranged from microscopic to 160 mm (median 27 mm).
Table 1

Patient and tumor characteristics

 

Alla

Astrocytoma

Ependymoma

Total number

1,814

664

1,057

Age

 <1–9

148 (8%)

119 (18%)

20 (2%)

 10–19

187 (10%)

117 (18%)

61 (6%)

 20–39

535 (29%)

180 (27%)

329 (31%)

 40–59

674 (37%)

169 (25%)

482 (46%)

 ≥60

270 (14%)

79 (12%)

165 (16%)

Race

 White

1,539 (85%)

531 (80%)

937 (88%)

 Black

158 (9%)

88 (13%)

55 (5%)

 Otherb

105 (6%)

43 (6%)

55 (5%)

Male

1,004 (55%)

375 (56%)

580 (55%)

Female

810 (45%)

289 (44%)

478 (45%)

Year diagnosed

 1970s

145 (8%)

84 (13%)

53 (5%)

 1980s

269 (15%)

140 (21%)

110 (10%)

 1990s

512 (28%)

195 (29%)

285 (27%)

 2000s

888 (49%)

245 (37%)

609 (58%)

Grade

 1

407 (22%)

180 (27%)

221 (33%)

 2

323 (18%)

155 (23%)

146 (14%)

 3

150 (8%)

117 (18%)

28 (3%)

 4

82 (5%)

82 (12%)

 Unknown

852 (47%)

130 (20%)

662 (63%)

aIncludes 11 patients with oligodendroglioma and 81 patients with unspecified glioma of whom 54 have unknown grade

bExcluding 12 unknown

For the entire cohort, on UVA, SCA fared significantly worse than SCE with respect to OS and CSS (P < 0.00001).

Spinal cord astrocytoma

Table 2 summarizes the treatment with respect to surgery and radiotherapy for SCA. More patients with low grade SCA, as compared to high grade SCG, underwent resection alone.
Table 2

Treatment summaries

 

Total

Grade 1

Grade 2

Grade 3

Grade 4

Grade unknown

Astrocytoma

 Surgery alone

230

117 (68%)

48 (32%)

17 (15%)

12 (15%)

36 (29%)

 Surgery + radiotherapy

246

31 (18%)

61 (40%)

67 (58%)

49 (62%)

38 (31%)

 Radiotherapy alone

105

14 (8%)

31 (21%)

25 (22%)

11 (14%)

24 (19%)

 No surgery or radiotherapy

59

10 (6%)

11 (7%)

5 (4%)

7 (9%)

26 (21%)

Ependymoma

 Surgery alone

710

159 (74%)

97 (67%)

10 (36%)

444 (69%)

 Surgery + radiotherapy

253

46 (21%)

41 (28%)

18 (64%)

148 (22%)

 Radiotherapy alone

24

5 (2%)

3 (2%)

0

16 (2%)

 No surgery or radiotherapy

49

6 (3%)

3 (2%)

0

40 (6%)

Excluding patients with unknown surgery and/or radiotherapy

For grade 1, 2, 3 and 4 SCA, the 5-year OS was 82, 70, 28 and 14%. The 5-year CSS was 89, 77, 36 and 20%. Table 3 shows the OS and CSS for the entire cohort of patients with SCA as well as subgroups of interest. On UVA, younger age, lower grade, resection and no radiotherapy delivery were significantly favorable predictors of OS and CSS. There is a non-significant trend towards worse early survival (<4 years) of the 10–19 year group versus older adults. Gender (not shown), race (not shown), and decade of diagnosis did not significantly impact survival. Figure 1 depicts the Kaplan–Meier OS by grade.
Table 3

Patient outcome: astrocytoma

 

Overall survival

Cause specific survival

1 years (%)

5 years

(%)

10 years (%)

MS

(months)

P value

UVA

P value

MVA

P value

MVAa

1 years

(%)

10 years

(%)

P value

UVA

P value

MVA

P value

MVAa

All patients

79

58

52

133

   

83

62

   

Grade (as outlined in “Methods”)

 1

93

82

74

NR @ 360

<0.00001

<0.0001

<0.0001

96

88

<0.00001

<0.0001

<0.0001

 2

87

70

62

248

92

73

 3

65

28

23

19

70

30

 4

46

15

11

10

49

17

 ?

84

68

64

237

89

73

Grade (as scored in SEER database)

 Well differentiated; grade I

92

80

68

277

<0.00001

Not run

Not run

95

85

<0.00001

Not run

Not run

 Moderately differentiated; grade II

89

71

65

250

93

76

 Poorly differentiated; grade III

72

36

31

22

74

37

 Undifferentiated; anaplastic grade IV

58

23

17

15

64

25

 Unknown

80

64

60

186

84

67

Age

 <1–9

94

84

80

360

<0.00001

<0.0001

<0.0001

95

88

<0.00001

0.028

0.008

 10–19

71

59

54

181

74

60

 20–39

81

56

54

198

84

61

 40–59

81

53

43

66

85

52

 ≥60

62

40

24

23

74

53

Decade diagnosed

 1970s

79

52

44

72

0.38

Not run

Not run

81

52

0.33

Not run

Not run

 1980s

74

60

54

135

80

65

 1990s

80

58

53

NR @ 189

86

63

 2000s

82

59

NR @ 71

85

Resection

 No resection

70

47

38

52

<0.00001

0.002

0.004

75

47

<0.00001

<0.0001

0.001

 Resection

83

63

57

228

86

68

Radiotherapy

 No radiotherapy

86

75

69

NR @ 366

<0.00001

0.43

0.11

91

82

<0.00001

0.062

0.014

 Radiation delivered

75

47

40

53

77

49

aMVA model in which grade unknown is reassigned as grade 2

MS Median survival, NOS not otherwise specified, NR not reached, UVA univariate analyses, MVA multivariate analyses

? = unknown

https://static-content.springer.com/image/art%3A10.1007%2Fs11060-009-0054-7/MediaObjects/11060_2009_54_Fig1_HTML.gif
Fig. 1

Kaplan–Meier overall survival of spinal cord astrocytoma by grade. Beyond 25 years, the total number of patients at risk is small (31), so survival is shown only up to 25 years

The survival outcomes by grade as classified in the SEER database were comparable to survival outcomes by grade as described in the “Methods”, with the exception of an appreciably better OS and CSS in patients classified as “Undifferentiated; anaplastic Grade IV” in the SEER database as compared to grade 4 in the present study. Also, patients classified as “Well differentiated; Grade I” in the SEER database experienced a slightly worse long-term survival as compared to patients classified as grade 1 in the present study.

Multivariate analyses of OS and CSS were run, in which age and grade were treated as continuous variables, while radiation delivery and resection were treated as discrete variables (yes/no). Younger age, lower grade and resection were significantly favorable variables. Additional MVA models were performed for OS and CSS, in which unknown grade was reclassified as grade 2. With the “grade unknown → 2” MVA model, age, grade and resection remained significant, while the use of radiotherapy (adverse factor) was significant for CSS.

To explore the potential impact of SCA tumor size, additional MVA models were run in which tumor size was included as a continuous variable. For OS, tumor size was significant (P = 0.048) in the MVA models in which unknown grade was excluded (98 patients analyzed) and borderline significant (P = 0.077) in the “grade unknown → 2” MVA models (111 patients analyzed); grade (P < 0.0001 in both models) and age (P = 0.005 and 0.007) were significant, while resection (P = 0.99 and 0.89) and use of radiotherapy (P = 0.41 and 0.83) were not significant. For CSS, tumor size was not significant in either MVA model (P > 0.2); grade (P < 0.0001 in both models) and age (P = 0.015 and 0.026) were significant, while resection (P = 0.65 and 0.92) and use of radiotherapy (P = 0.50 and 0.99) were not significant.

Spinal cord ependymoma

Table 2 summarizes the treatment with respect to surgery and radiotherapy for SCE. More patients with low grade SCE, as compared to high grade SCE, underwent resection alone.

For grade 1, 2, and 3 SCE, the 5-year OS was 92, 97 and 58%. The 5-year CSS was 100, 98 and 64%. Table 4 shows the OS and CSS for the entire cohort of patients with SCE as well as subgroups of interest. On UVA, younger age, lower grade, earlier decade of diagnosis, resection and no radiotherapy delivery were significantly favorable predictors of OS and CSS. The 10–19 year old age group experienced a non-significant trend towards worse OS compared to younger children and adults <60 years (5-year OS of 80 vs. 93%, P = 0.16). After 15–20 years, the OS of the 10–19 year old subgroup begins to exceed the OS of older adults. Gender and race (not shown) did not significantly impact survival. Figure 2 depicts the Kaplan–Meier OS by grade.
Table 4

Patient outcome: ependymoma

 

Overall survival

Cause specific survival

1 years

(%)

5 years

(%)

10 years

(%)

MS

(months)

P value

UVA

P value

MVA

P value

MVAa

1 years

(%)

10 years

(%)

P value

UVA

P value

MVA

P value

MVAa

All patients

97

90

81

300

   

99

93

   

Grade (as outlined in “Methods”)

 1

98

92

73

NR @ 315

0.002

*

*

100

93

<0.00001

*

*

 2

100

97

91

NR @ 375

100

98

 3

96

58

51

NR @ 264

96

56

 ?

96

90

81

298

98

93

Grade 1–2 versus 3

    

0.0001

<0.0001

<0.0001

  

<0.00001

<0.0001

<0.0001

Grade (as scored in SEER database)

 Well differentiated; grade I

96

91

73

NR @ 315

<0.00001

Not run

Not run

100

92

<0.00001

Not run

Not run

 Moderately differentiated; grade II

100

97

91

222

100

98

 Poorly differentiated; grade III

100

38

32

100

<50

 Undifferentiated; anaplastic grade IV

94

54

45

72

94

48

 Unknown

97

90

81

298

98

93

Age

 <1–9

100

100

100

NR @ 294

<0.00001

<0.0001

<0.0001

100

100

0.002

0.070

0.002

 10–19

95

80

73

NR @ 390

96

84

 20–39

100

95

90

NR @ 386

100

94

 40–59

98

92

84

278

100

94

 ≥60

89

78

49

119

94

85

Decade diagnosed

 1970s

98

81

69

215

0.002

  

98

78

<0.00001

  

 1980s

95

87

77

280

  

96

89

  

 1990s

96

90

82

NR @ 190

  

98

94

  

 2000s

98

92

NR @ 71

  

99

  

Year of diagnosis

     

0.11

0.013

   

0.11

0.003

Resection

 No resection

86

71

65

292

0.002

0.005

0.001

94

77

0.003

0.011

0.003

 Resection

98

92

82

296

99

94

Radiotherapy

 No radiotherapy

97

92

84

NR @ 390

0.002

0.056

0.045

99

96

0.001

0.66

0.17

 Radiation delivered

98

89

75

249

99

87

aMVA model in which grade unknown is reassigned as grade 2

*When grade was included as a continuous variable, grade was not significant for OS (P > 0.1), while it was significant for CSS (P ≤ 0.001). The MVA models shown include grade as a discrete variable (1–2 vs. 3)

MS Median survival, NOS not otherwise specified, NR not reached, UVA univariate analyses, MVA multivariate analyses

? = unknown

https://static-content.springer.com/image/art%3A10.1007%2Fs11060-009-0054-7/MediaObjects/11060_2009_54_Fig2_HTML.gif
Fig. 2

Kaplan–Meier overall survival of spinal cord ependymoma by grade. Beyond 25 years, the total number of patients at risk is small (25), so survival is shown only up to 25 years

The survival outcomes by grade as classified in the SEER database were comparable to survival outcomes by grade as described in the “Methods”. Interestingly, patients classified as having grade 1 SCE experienced a non-significant trend towards worse OS versus those with grade 2 SCE (P = 0.063), though CSS was more similar.

Multivariate analyses of OS and CSS were run, in which age and year of diagnosis were treated as continuous variables, while grade (1–2 vs. 3), radiation (yes/no) and any resection (yes/no) were treated as discrete variables. Additional MVA models were performed for OS and CSS, in which unknown grade was reclassified as grade 2. Grade 1–2 versus 3 was significant in all MVA models. Younger age was favorable for OS and CSS. No radiation delivery was favorable for OS; resection was significantly favorable for OS and CSS. A more recent year of diagnosis was significantly favorable for OS and CSS in the “grade unknown → 2” MVA models.

To explore the potential impact of SCE tumor size, additional MVA models were run in which tumor size was included as a continuous variable. For OS, tumor size was not significant (P = 0.56) in the MVA model in which unknown grade was excluded (154 patients analyzed) or in the “grade unknown → 2” MVA model (P = 0.24; 352 patients analyzed). In both MVA models of OS, in which size was included, grade was not significant (P = 0.20 and 0.13), while younger age (P = 0.038 and <0.0001) and resection (P = <0.0001 and 0.001) were significant; radiotherapy delivery was significant (P = 0.050) in the “grade unknown → 2” model. For CSS, no variable, including tumor size (P = 0.91) was significant (P > 0.2 for all variables) in the MVA model in which unknown grade was excluded. For the “grade unknown → 2” CSS MVA model, grade (P = 0.025), younger age (P = 0.017), more recent year of diagnosis (P = 0.002), and resection (P = 0.010) were significantly favorable, while smaller tumor size (P = 0.12) and radiotherapy (P = 0.12) were not.

Patient outcome: use of radiotherapy

Table 5 summarize the OS and CSS of patients with grade 2–4 SCG, treated with resection with or without adjuvant radiotherapy. Patients with resected grade 2 SCE treated with radiotherapy experienced a worse OS compared to those not treated with radiotherapy, though CSS was similar. Patients with resected grade 2 SCA treated with radiotherapy experienced a worse OS and CSS compared to those not treated with radiotherapy. In patients with resected high grade SCG, OS and CSS were not significantly impacted by whether or not radiotherapy was delivered.
Table 5

Radiotherapy

 

Number

10-year OS (%)

P value

10-year CSS (%)

P value

Grade 2 ependymoma s/p resection

 No radiotherapy

97

99

0.041

99

0.88

 Radiotherapy

41

80

100

 

Number

5-year OS (%)

P value

5-year CSS (%)

P value

Grade 2 astrocytoma s/p resection

 No radiotherapy

48

86

0.042

92

0.083

 Radiotherapy

61

74

79

Grade 3–4 astrocytoma s/p resection

 No radiotherapy

29

29

0.94

47

0.50

 Radiotherapy

116

21

28

Discussion

Spinal cord glioma is a rare malignancy, for which the management is challenging due to potentially devastating consequences of the disease as well as paucity of published data. Given the relative rarity, SCG is well-suited for retrospective studies, and the SEER database provides a large study cohort with long follow-up.

Several retrospective studies suggest that grade and histology are significant predictors of survival and disease control, with lower grade tumors and ependymomas having the best prognosis. The present study confirms these findings. Nevertheless, comparing outcomes of like-graded tumors between different studies, and even within the same study, can be a challenge due to the difficulty in accurately and reproducibly assigning a grade. Several factors complicate glioma grading, including: (1) tissue sampling errors; (2) variable levels of pathology expertise; (3) differing criteria with which grade is assigned; (4) availability and use different pathologic staging systems over the course of decades. Database discrepancies, inconsistencies and/or coding errors can also be problematic. Major limitations of this study are (1) a large percentage of patients with unknown grade; and (2) among those who were assigned a grade, a uniform grading system was not used. While after 2004, WHO grading was to be recorded in the optional “CS Site-Specific Factor 1” field, this field was scored as unknown for all patients, even if a grade was assigned in the SEER grade field, suggesting that registries did not opt to use the “CS Site-Specific Factor 1” field and/or used other grading systems.

In our study, greater age was an adverse predictor of OS and CSS for SCA and SCE, a finding which contradicts a recent multi-institutional pooled analysis [39]. Interestingly, the subgroup of patients aged 10–19 experienced a non-significantly worse OS and CSS as compared to younger children and non-elderly adults. These differences do not appear to be attributable to differences in the distribution of grade or histology (data not shown). Possible reasons for these differences are unknown.

Grade was highly significant for OS and CSS of SCG, consistent with innumerable other studies (see “Introduction”). Interestingly, the OS for grade 1 SCE was non-significantly worse than grade 2 SCE, which resulted in grade being non-significant when treated as a continuous variable in the MVA for SCE (Table 4). Low grade (grade 1–2) versus grade 3 SCE was highly significant in MVA models for OS and CSS.

Grade 1 myxopapillary ependymomas can be characterized by early progression, and relatively poor OS [48], perhaps contributing to the relatively worse OS of grade 1 SCE. Most (>95%) of the patients identified as having myxopapillary ependymoma in the SEER database were diagnosed after 2000, and thus a meaningful comparison could not be made in our analysis. Grade as a continuous variable was significant in MVA models for CSS, suggesting that the worse OS of grade 1 SCE in this study may also reflect confounding variables. This discrepancy may also reflect the difficulty in grading gliomas (as described above), as well as the manner in which grade was assigned in this study. As described in the “Methods”, from 2004, ependymomas were characterized as ‘malignant’, ‘borderline’ or ‘benign’, while prior to 2004, only ‘malignant’ ependymomas were included. While tumors which would have been scored as ‘borderline’ or ‘benign’ after 2004, may have been included as grade 1 ‘malignant’ tumors prior to 2004 (since the classification of grade 1 ependymomas as ‘malignant’, ‘borderline’ or ‘benign’ is a matter of semantics), perhaps many of these tumors were not included in the SEER database prior to 2004. This would potentially create a different patient population of grade 1 tumors after 2004. The long-term survival of patients in this study reflects patients registered when registries were instructed to include only ‘malignant’ gliomas. In our study, <35% of grade 1 tumors were diagnosed prior to 2004 (data not shown).

Tumor size was explored as a variable potentially affecting outcome. For SCA, smaller tumor size appears to favorably impact OS, but not CSS. Interestingly, whether patients underwent resection or radiotherapy were not significant when SCA tumor size was included in the MVA. For SCE, tumor size did not appear to impact survival outcomes; whether patients underwent resection remained significant when SCE tumor size was included in the MVA models. Because many patients did not have tumor size recorded (111 patients with SCA and 352 patients with SCE were analyzed for tumor size), and because tumor size, in general, is not a reliable variable, caution should be used in interpreting these results.

The role of adjuvant radiotherapy after resection of SCG is not well understood. For high grade SCG, given the relative rarity and generally poor prognosis, the incremental benefit radiotherapy offers in terms of tumor control and survival is not well characterized. In our retrospective analysis, the OS and CSS of patients with grade 3–4 SCG were not significantly impacted by radiation delivery, albeit perhaps partially attributable to biases inherent in retrospective analyses. Nevertheless, radiotherapy is known to be effective against gliomas and remains a standard adjuvant treatment for high grade SCG.

For grade 2 SCG, the benefit of adjuvant radiotherapy is more controversial. Several retrospective studies have shown that after gross total resection of grade 2 SCG, post-operative radiotherapy is not needed [16, 18, 19, 25, 32]. Certainly, with modern surgical and imaging techniques, total resection of grade 2 SCG is more readily attainable, and close observation with serial imaging is reasonable [15, 16, 19, 25, 28, 29, 3133, 39].

After subtotal resection of low grade SCG, several studies advocate post-operative radiotherapy [2, 12, 16, 17, 21, 23, 25, 34, 35] and/or treated all (or nearly all) patients in their series with radiotherapy (and therefore cannot assess the efficacy of withholding radiation) [4, 68, 10, 11, 13, 14, 22, 29, 30, 33, 38]. However, it is not known which patients might benefit from close observation, reserving radiotherapy or additional surgery for salvage, as opposed to upfront radiotherapy [26]. At least one study demonstrates no benefit from radiotherapy after subtotal resection of low grade SCE [26]. A modern pooled analysis has advocated deferring radiotherapy in patients with subtotally resected low grade SCE, while offering upfront radiotherapy for low grade SCA (which yielded a significantly improved progression-free survival) [39]. In a recent retrospective study, post-operative radiotherapy for all infiltrative (grade 2–4) SCA yielded a significantly improved OS [43].

In our analysis, patients with resected grade 2 SCG, specifically SCA, who underwent radiotherapy fared worse than patients who did not undergo radiotherapy. This most likely reflects a bias in that patients who were offered therapeutic radiotherapy perhaps had more adverse risk factors (i.e. larger tumor, close or positive margins, multifocal disease), a notion confirmed by a large multi-institutional pooled analysis [39]. Therefore retrospectively comparing the outcome of those treated with upfront radiotherapy versus those not undergoing upfront radiotherapy is likely impacted by this bias.

Conclusions

For patients with SCG, age, histology and grade are significant predictors of outcome. Though treatment with radiotherapy was associated with worse outcomes, this likely reflects a bias in that patients who received radiotherapy were perhaps more likely to have had adverse risk factors (i.e. larger tumor, close or positive margins). Patients with grade 2 SCG who did not undergo radiotherapy after resection fared quite well with respect to OS and CSS. Given the retrospective nature of this study, specific recommendations about which situations warrant radiotherapy cannot be derived from our analyses.

Copyright information

© Springer Science+Business Media, LLC. 2009