Cerebellar location may predict an unfavourable prognosis in paediatric high-grade glioma.

BACKGROUND: High-grade glioma (HGG) of the cerebellum accounts for only 5% of paediatric HGG. Since little is known about these tumours, the present study aimed at their further characterisation.
METHODS: Twenty-nine paediatric patients with centrally reviewed cerebellar HGG were identified from the HIT-GBM/HIT-HGG database. Clinical and epidemiological data were compared with those of 180 paediatric patients with cortical HGG.
RESULTS: Patients with cerebellar tumours were younger (median age of 7.6 vs 11.7 years, P=0.028), but both groups did not differ significantly with regard to gender, tumour predisposing syndromes, secondary HGG, primary metastasis, tumour grading, extent of tumour resection, chemotherapy regimen, or radiotherapy. Except for an increased incidence of anaplastic pilocytic astrocytoma (APA) in the cerebellar subset (20.7% vs 3.3%; P<0.001), histological entities were similarly distributed in both groups. As expected, tumour grading had a prognostic relevance on survival. Compared with cortical HGG, overall survival in the cerebellar location was significantly worse (median overall survival: 0.92 ± 0.02 vs 2.03 ± 0.32 years; P=0.0064), and tumour location in the cerebellum had an independent poor prognostic significance as shown by Cox-regression analysis (P=0.019).
CONCLUSION: High-grade glioma represents a group of tumours with an obviously site-specific heterogeneity associated with a worse survival in cerebellar location.

Although brain tumours frequently arise in the posterior fossa beyond the first year of life (Pollack, 1999), only 5% of all HGGs arise in the cerebellum, accounting for <1% of all paediatric brain tumours (Kramm ). Since previously published series are small or mixed up with adult patients (Chamberlain ; Hayostek ; Djalilian and Hall, 1998; Grahovac ), merely little is known about paediatric HGG in this specific site.

The present study with 29 centrally reviewed paediatric HGG cases of the cerebellum aimed at a further characterisation of cerebellar HGG in children and adolescents with special emphasis on clinical characteristics and potential prognostic factors. Furthermore, we asked if these tumours may represent a clinically distinct subset of paediatric HGG. To investigate this hypothesis, paediatric patients with cerebellar HGG were compared with a control group of 180 paediatric patients with non-thalamic supratentorial HGG which had developed in cortical structures. This control group was chosen because both cerebellar and cerebral cortical tumours represent tumour populations with a supposedly facilitated surgical access, and extent of tumour resection still represents a strong prognostic parameter in paediatric HGG (Kramm ). Results were critically reviewed in the light of the present literature.

Patients and methods

Patients' characteristics and inclusion/exclusion criteria

Patient data were obtained from the HIT-GBM/HIT-HGG database of the Society of Paediatric Oncology and Haematology in Germany, Austria, and Switzerland (Gesellschaft für Paediatrische Onkologie und Haematologie, GPOH). The HIT-GBM/HIT-HGG database contains clinical data of all patients enrolled in the various HIT-GBM/HIT-HGG trials since 1994 (Wolff , 2006, 2010a, 2010b). For the present study, the following inclusion criteria were defined:

(A) Histopathological diagnosis of an HGG as defined by the WHO classification of central nervous system tumours (Louis ) confirmed by central neuropathological review (The German Brain Tumour Reference Center, Department of Neuropathology, Bonn, Germany). High-grade gliomas included glioblastoma multiforme (GBM) WHO IV (M9440/3), anaplastic astrocytoma WHO III (M9401/3), anaplastic oligodendroglioma WHO III (M9451/3), anaplastic oligoastrocytoma WHO III (M9382/3; synonym: anaplastic mixed glioma WHO III), APA WHO III (M9421/3), anaplastic ganglioglioma WHO III (M9505/3), pleomorphic xanthoastrocytoma with signs of anaplasia (M9424/3), giant cell glioblastoma WHO IV (M9441/3), and glioblastoma with sarcomatous components WHO IV (M9442/3; gliosarcoma WHO IV).

(B) Patient age 0–17 years at the time of initial diagnosis.

(C) Primary location within the cerebellum or cerebral cortex. Tumours that could not be accurately defined as primarily cerebellar or cortical were excluded.

By applying these criteria, 209 paediatric patients were identified: 29 with cerebellar HGG and 180 with HGG affecting the cortex (in the following, the term ‘cortical HGG' will be used for this HGG subgroup). Informed consent for data storage and statistical analyses had been given by all patients and/or their parents at the time of enrolment in the various HIT-GBM/HIT-HGG trials in accordance to the Declaration of Helsinki.

HIT-GBM/HIT-HGG treatment protocols

Eligible paediatric patients older than 3 years but younger than 18 years with HGG and/or diffuse intrinsic pontine glioma had been enrolled in the various HIT-GBM/HIT-HGG trials since 1994. In all these trials, best feasible tumour resection was recommended before starting chemo- and/or radiotherapy. Standard fractionated radiotherapy (59.4 Gy total dose; daily fractions of 1.8 Gy over 6–7 weeks) was also common for all trials. In HIT-GBM-A, oral etoposide and trofosfamide were given during radiotherapy and continued for 1 more year after radiotherapy (Wolff ). In HIT-GBM-B, cisplatinum, etoposide, and ifosfamide were given in two cycles concomitantly to radiotherapy followed by intravenous low-dose cyclophosphamide and subcutaneous interferon-gamma (Wolff ). In HIT-GBM-C, weekly vincristine injections were added to the concomitant HIT-GBM-B radiochemotherapy. Chemotherapy with cisplatinum, etoposide, ifosfamide, and vincristine was continued during maintenance treatment (Wolff ). The concomitant radiochemotherapy of HIT-GBM-C was also adopted for HIT-GBM-D followed by a maintenance therapy with prednisolone, vincristine, and lomustine (CCNU). The therapeutic efficiency of two courses of high-dose methotrexate before radiotherapy was studied for feasibility in a pilot study (HIT-GBM pilot D) (Wolff ) and then as a randomised question in HIT-GBM-D. Children younger than 3 years of age were primarily treated with surgery and chemotherapy alone following the HIT-SKK chemotherapy recommendation for infant patients with brain tumours (Rutkowski ). Central neuropathological and central neuroradiological reviews were highly recommended in HIT-GBM-A and HIT-GBM-B and became mandatory for the HIT-GBM-C and HIT-GBM-D trials. The extent of tumour resection was determined on the basis of early post surgical imaging and/or the neurosurgical report. Gross-total tumour resection was defined as 100% macroscopic removal of the tumour mass. Subtotal tumour resection was defined as removal of <100% but ⩾90% of the tumour mass, partial tumour resection as <90% resection. Biopsies were either performed as an open or stereotactic biopsy without tumour debulking.

Molecular biologic characterisation

TP53 accumulation was studied in 49 patients by immunohistochemistry using standard techniques, and H3.3 mutation analysis was performed in 6 cerebellar HGG by pyrosequencing as previously described (Gessi ).

Statistical analysis

Statistical analysis was retrospectively performed using the statistical package for social studies (SPSS Inc, San Francisco, CA, USA). Overall (OS) and event-free survival (EFS) were determined by Kaplan–Meier analysis and log-rank testing. An event was defined as tumour relapse or progression, occurrence of a secondary malignancy, or death from any cause.

Statistical analysis of cerebellar and cortical HGGs was defined by the following parameters: histological tumour grading (WHO grade III/WHO grade IV), gender (male/female), median age (⩽7 years/>7 years, only cerebellar subset) and extent of tumour resection (gross-total resection and no gross-total resection). The prognostic relevance regarding EFS and OS in cerebellar HGG was compared for the subgroups ‘age', ‘grading', ‘gender', and ‘extent of tumour resection' as defined above by Kaplan–Meier analysis and for both cortical and cerebellar HGGs by multivariate Cox-regression analysis for the same subgroups. The relevance of tumour location (cerebellar vs cortical) on OS and EFS was elucidated by both univariate Kaplan–Meier and multivariate Cox-regression analysis for both the total group and a subset of patients with GBM. To compare the epidemiological distribution of disease/treatment characteristics between the two HGG sites, the two-sided χ2-test was performed to determine significant differences regarding gender, distribution of grade III and IV tumours, histological diagnosis, secondary HGG, tumour predisposing syndromes, primary metastases, secondary metastasis, extent of tumour resection (total resection/subtotal resection/partial resection/biopsy and no resection), radiotherapy, chemotherapy regimen, and early response assessment (usually after radiotherapy or first SKK cycle around week 8 with the exception of the methotrexate group of HIT-GBM-D patients who were analysed at week 10), and accumulation of TP53. The difference of age distribution was evaluated by an unpaired, two-sided t-test. For all statistical analyses, the significance level was set at P<0.05.

Results

Clinical and histological parameters as well as treatment characteristics of paediatric HGG of the cerebellum and the cerebral cortex are listed in Table 1. It appeared that cerebellar and cortical HGGs display many similar clinical characteristics. Sex ratio as well as incidence rates of secondary HGG, primary metastases, secondary metastases, and tumour predisposing syndromes did not differ significantly between both tumour sites. However, patients with cerebellar HGG were significantly younger than patients with cortical HGG (median age 7.6 vs 11.7 years, P=0.028). The histological entities were found in both sites with a similar distribution except for APAs WHO grade III that were significantly associated with a cerebellar location. The frequency of WHO grade III and IV gliomas in cortex and cerebellum was also similar. In addition, few tumours were characterised for H3.3 mutation and TP53 accumulation. TP53 status did not differ significantly between both sides. Noteworthily, no H3.3 mutation was found in six cerebellar tumours.

Table 1

Clinical and treatment characteristics of paediatric HGG of the HIT-GBM/HIT-HGG database located in the cerebral cortex or the cerebellum

	Cerebellar HGG	Cortical HGG
Total number	29	180
Age at diagnosis			P=0.028a
Median (years)	7.6	11.7
Mean (years)	8.3	10.5
Range (years)	0.8–16.9	0–17.9
Gender			NSb
Female	10 (34.5%)	76 (42.2%)
Male	19 (65.5%)	104 (57.8%)
Histological tumour grading			NSb
WHO III	10 (34.5%)	66 (36.7%)
WHO IV	19 (65.5%)	114 (63.3%)
Histological diagnosis
Anaplastic pilocytic astrocytoma	6 (20.7%)	6 (3.3%)	P<0.001b
Anaplastic astrocytoma	4 (13.8%)	43 (23.9%)	NSb
Anaplastic ganglioglioma	0	5 (2.8%)	—
Glioblastoma multiforme	17 (53.3%)	96 (58.6%)	NSb
Gliosarcoma	0	4 (2.2%)	–
Giant cell glioblastoma	2 (6.9%)	14 (7.8%)	NSb
Anaplastic oligoastrocytoma	0	6 (3.3%)	—
Anaplastic oligodendroglioma	0	2 (1.1%)	—
Pleomorphic xanthoastrocytoma with anaplasia	0	4 (2.2%)	—
H3.3 mutation			—
Yes	0	—
No	6 (100%)	—
Unknownc	23 (79%)	180 (100%)
TP53 mutation			NSb
Yes	4 (57%)	22 (52%)
No	3 (43%)	20 (48%)
Unknownc	22 (76%)	138 (77%)
Secondary high-grade glioma			NSb
Yes	1 (3.6%)	13 (4.1%)
No	27 (96.4%)	151 (95.9%)
Unknownc	1 (3.4%)	16 (8.9%)
Tumour predisposition syndromes			NSb
Yes	2 (6.9%)	7 (4.1%)
No	27 (93.1%)	164 (95.2%)
Unknownc		9 (4.0%)
Primary metastasis			NSb
Yes	4 (13.8%)	11 (7.0%)
No	25 (86.2%)	158 (93.0%)
Unknownc		11 (6.1%)
Secondary metastasis			NSb
Yes	8 (38.1%)	37 (30.3%)
No	13 (61.9%)	85 (69.7%)
Unknownc	8 (27.6%)	58 (31%)
Extent of surgical resection			NSb
Gross-total tumour resection	9 (31.0%)	70 (39.1%)
Subtotal tumour resection	12 (41.4%)	37 (20.7%)	P=0.014b
Partial tumour resection	5 (17.2%)	45 (25.1%)	NS
Biopsy/no tumour surgery	3 (10.3%)	27 (15.1%)	NS
Unknownc		1 (0.6%)
Radiotherapy			NSb
Yes	24 (85.7%)	149 (88.7%)
No	4 (14.3%)	19 (11.3%)
Unknownc	1 (3.4%)	12 (6.7%)
Chemotherapy			—
No	0	11 (6.3%)
HIT-GBM-A	1 (3.4%)	11 (6.3%)
HIT-GBM-B	3 (10.3%)	13 (7.5%)
HIT-GBM-C	6 (20.7%)	36 (20.7%)
HIT-GBM-D	12 (41.4%)	70 (40.2%)
HIT-SKK	5 (17.2%)	17 (9.8%)
Other chemotherapy	2 (6.9%)	19 (10.9%)
Unknownc		6 (3.3%)
Early response			NS
Overall complete response	10 (41.6%)	57 (38.5%)
Complete response/Continuing complete response	2/8 (8.3%/33.3%)	18/39 (12.2%/26.4%)
Partial response	2 (8.3%)	18 (12.2%)
Stable disease	4 (16.7%)	40 (27.0%)
Progressive disease	8 (33.3%)	33 (22.3%)
Unknownc	5 (17.2%)	32 (17.8%)
Relapse therapy			—
Chemotherapy	10 (63%)	59 (78%)
Surgery	1 (6%)	3 (4%)
Radiotherapy	2 (13%)	1 (1%)
None	3 (19%)	13 (17%)
Unknownc	13 (45%)	104 (58%)

Abbreviations: —=not determined; HGG=high-grade tumour; NS=not significant; WHO=world health organisation.

t-test.

χ2-test.

Each percentage of unknown data refers to the total patient number of each group.

No significant difference was found regarding therapy: while frequency of gross-total tumour resection seemed to be increased in cortical HGG (39.1% vs 31.0%), in cerebellar HGG subtotal resection was more frequent (41.4% vs 20.7%). Hence, in 72.4% of cerebellar HGG and 59.8% of cortical HGG at least 90% tumour resection could be achieved (not significant). Most patients received radiotherapy, while 11 patients with cortical HGG did not receive any chemotherapy, all patients with cerebellar HGG did. Distribution of chemotherapy regimen between both sites was comparable (Table 1). Regarding early response assessment after radiotherapy (weeks 8–10), both groups did not differ significantly, and complete remission (CR) was found in 41.6% of cerebellar HGG vs 38.5% of cortical HGG.

In our cohort of 29 patients with cerebellar HGG, tumour grading (WHO grade III versus IV) was strongly associated with survival. Patients with WHO grade III tumours had a superior OS and EFS in comparison to patients with WHO grade IV tumours as shown by Kaplan–Meier estimate (Table 2) and multivariate Cox-regression analysis (Table 3). There was no significant prognostic relevance of the extent of tumour resection on survival: gross-total tumour resection did not translate into a significantly improved OS or EFS in both univariate and multivariate analysis. In addition, in cerebellar HGG age (⩽7 vs >7 years) and gender were not significantly associated with survival as shown by Kaplan–Meier analysis and Cox-regression analysis (Tables 2 and 3).

Table 2

OS and EFS of 29 paediatric patients with HGG from the German HIT-GBM/HIT-HGG database with cerebellar tumour

		Event-free survival				Overall survival
	Number	One-year EFS	Two-year EFS	Five-year EFS	Log rank	One-year OS	Two-year OS	Five-year OS	Log rank
Gender					NS				NS
Female	n=10	20±12%	20±12%	0		30±14%	30±14%	15±13%
Male	n=19	42±12%	28±12%	14±12%		48±12%	29±11%	14±12%
Age at diagnosis					NS				NS
⩽7 years	n=16	39±13%	23±12%	12±10%		41±13%	41±13%	12±10%
>7 years	n=13	26±13%	26±13%	0		42±14%	21±13%	21±13%
Grading					P=0.0202				P=0.0079
WHO III	n=10	68±16%	56±17%	19±16%		79±13%	66±16%	44±21%
WHO IV	n=19	14±9%	7±6%	0		22±10%	11±7%	6±5%
Extent of tumour resection					NS				NS
Gross total resection	n=9	52±18%	39±17%	0		51±18%	38±17%	0
No gross-total resection	n=20	25±10%	19±10%	9±8%		37±11%	25±10%	17±10%

Abbreviations: EFS=event-free survival; HGG=high grade glioma; NS=not significant; OS=overall survival; WHO=World Health Organisation.

Table 3

Cox-regression analysis of overall survival and event-free survival in our cohort of HGG affecting the cortex (180 patients) or the cerebellum (29 patients) for the subgroups ‘gender', ‘age', ‘histological grading', and ‘extent of tumour resection'

	HGG of the cortex		HGG of the cerebellum		All HGG
	EFS	OS	EFS	OS	EFS	OS
Gender	NS	NS	NS	NS	NS	NS
Age	—	—	NS	NS	—	—
Grading	P<0.001	P<0.001	P=0.016	P=0.012	P<0.001	P<0.001
Tumour resection	P<0.001	P<0.001	NS	NS	P<0.001	P<0.001
Tumour site	—	—	—	—	NS	P=0.019

Abbreviations: —=not determined; EFS=event-free survival; HGG=high-grade glioma; NS=not significant; OS=overall survival.

In addition, a Cox-regression analysis of all HGG included the parameter ‘tumour site'.

Patients with HGG of the cerebellum exhibited a worse overall survival compared with cortical HGG. The median OS differed significantly by log-rank testing (P=0.0064) and was 2.03±0.32 vs 0.92±0.02 years in cortical and cerebellar HGGs, respectively (Figure 1). One-year OS was 77±3% vs 41±9%, 2-year OS 52±4% vs 33±9%, and 5-year OS 32±5% vs 16±9% for cortical and cerebellar HGGs, respectively. In contrary, median EFS was worse in cerebellar HGG but did not differ significantly: 0.91±0.09 vs 0.62±0.17 years in cortical and cerebellar HGGs, respectively. One-year EFS was 47±4% vs 34±9%, 2-year EFS 30±4% vs 25±9%, and 5-year EFS 24±4% vs 8±7% for cortical and cerebellar HGGs, respectively. In addition, multivariate Cox-regression analysis corroborated these findings and revealed an independent prognostic significance of tumour location for OS, but not for EFS (Table 3). These findings were also confirmed for the subset of patients with GBM including 96 and 17 patients with cortical and cerebellar diseases, respectively. Median OS was 1.53±0.23 vs 0.9±0.03 years (P=0.0079) and median EFS was 0.66±0.13 vs 0.53±0.17 years (not significant) in cortical vs cerebellar GBM. In multivariate Cox-regression analysis, location in the cerebellum was of independent prognostic significance for OS (P=0.037, Figure 2) but not for EFS. Both subgroups did not differ significantly with respect to age, gender, incidence of primary and secondary metastasis, tumour predisposing syndromes as well as extent of tumour resection, application of radiotherapy, chemotherapy regimens and early treatment response after 8–10 weeks (data not shown). In this subgroup, some tumours were screened for biological characteristics: none of the four cerebellar GBM exhibited a H3.3 mutation. Concerning TP53 status, half of the tumours screened in each location exhibited an accumulation: 14 out of 28 cortical and 2 out of 4 cerebellar GBM, respectively.

Figure 1

Kaplan–Meier analysis of overall survival in paediatric patients with high-grade glioma of the cerebral cortex (

Figure 2

Kaplan–Meier analysis of overall survival in paediatric patients with GBM of the cerebral cortex (      

Discussion

High-grade glioma in children represents a non-homogenous group of glial tumours characterised by WHO grade III or IV histology. They account for 10% of all paediatric brain tumours, but only 5% of HGG arise in the cerebellum (Campbell ; Kramm ; Fangusaro, 2009). Since reports are rare and mainly studied adult patients (Chamberlain ; Hayostek ; Djalilian and Hall, 1998), only little is known about paediatric HGG in this specific site. Thus, the present study aimed at a further characterisation of paediatric cerebellar HGG and also investigated if HGG in this location may represent a distinct clinical subset in comparison with cortical HGG in paediatric patients.

Most epidemiologic characteristics were comparable between both sides and did not differ from previous reports mainly in adult patients including gender (Chamberlain ; Hayostek ), tumour predisposing syndroms (Fangusaro, 2009), incidence of primary metastasis (Benesch ), and tumour grading. In line with previous reports, gender did not influence prognosis (Hayostek ; Djalilian and Hall, 1998), while patients with WHO grade III tumours exhibited a superior OS and EFS compared with WHO grade IV tumours in cerebellar and cortical HGGs (Djalilian and Hall, 1998; Pollack, 1999; Fangusaro, 2009).

High-grade glioma usually affects children of all age groups with a median age of ∼9 years (Fangusaro, 2009). This is younger than the median age of our cohort with cortical HGG and older than the median age of our patients with cerebellar HGG. Obviously, there seems to be some differences in median age at diagnosis between the different affected sites of paediatric HGG: 11.5 years for thalamic HGG (Kramm ), 11.3 years for spinal HGG (Wolff ), 11.7 years for cortical HGG (this study), 8.0 years for pontine gliomas (Kramm ), and 7.6 years for cerebellar HGG (this study). Although not all of these differences are significant these observations may still reflect different subentities of paediatric HGG. Nevertheless, age (older and younger than median age) did not have a significant relevance on OS and EFS in our series of cerebellar HGG.

In our cohort, the histological entities were equally distributed over both sites except for an unexpected and unprecedented significantly increased incidence of APAs in the cerebellum. However, since APA was not introduced in the WHO classification of brain tumours until 2000, it cannot play the same role in previous paediatric series as published by Bertolone (patients enrolled between 1985 and 1990) (Bertolone ). Therefore, further studies are necessary to confirm the high incidence of APA in children with cerebellar HGG, which might be explained by the fact that WHO grade I pilocytic astrocytomas are frequently found within the posterior fossa (Fernandez ) and that WHO grade III APAs may follow the same pattern as their low-grade counterparts. Some of the APAs might also have arisen from previous subclinical WHO grade I pilocytic astrocytoma within the cerebellum since such malignant transformation had been reported before (Fangusaro, 2009; Otero-Rodriguez ; Peters ). However, malignant transformation of a subclinical WHO grade I pilocytic astrocytoma still seems an extraordinary clinical rarity and is often associated with prior irradiation or chemotherapy.

In paediatric HGG, the extent of tumour resection is a strong prognostic factor, and gross-total resection is associated with the best survival (Finlay ; Campbell ; Pollack, 1999; Kramm ). Although gross-total tumour resection was significantly associated in cortical HGG with a superior OS and EFS, this finding could not be confirmed in the subset of patients with cerebellar HGG. This is in contrast to adult patients with cerebellar HGG who had benefitted from complete tumour resection (Djalilian and Hall, 1998) and might be due to the small patient numbers in our study. Therefore, we still recommend aiming for a gross-total tumour resection whenever possible until further data will support (or disprove) our present findings of the impact of gross-total tumour resection in paediatric cerebellar HGG.

In the last few years, there is an increasing evidence of a site-specific genetic heterogeneity in paediatric HGG, for example, SETD2 mutations were restricted to the cerebral hemispheres in a series of 60 paediatric HGG (Fontebasso ), p53 immunoreactivity was frequently found with supratentorial tumour but was absent in cerebellar tumours in a series of 54 paediatric GBM (Ganigi ), and an amplification of 7q21-22 (including the CDK6 gene) was detected exclusively in supratentorial glioblastoma in a series of 18 children (Qu ). To elucidate, whether this genetic heterogeneity of cerebellar and cortical HGGs may translate into a site-specific varying survival, the HIT-GBM/HIT-HGG database was screened for HGG of the cortex. Cerebellar and cortical HGGs shared most of the epidemiological characteristics except for an increased incidence of APA in the cerebellum (Table 1). The extent of tumour resection, frequency of radiotherapy, and chemotherapy regimen did not differ among both groups, and furthermore, early response assessment after radiotherapy (i.e., weeks 8–10) was comparable between both sites.

Despite all these parameters so similar between both sites, prognosis in cerebellar HGG was worse both for the entire cohort and for a subset of GBM. Overall survival was significantly impaired in cerebellar HGG as shown by Kaplan–Meier analysis, and in contrary to a series of adult patients (Djalilian and Hall, 1998), tumour location in the cerebellum was found to be an independent prognostic parameter of OS by multivariate Cox-regression analysis. This finding was confusing regarding the increased incidence of APA in the cerebellum because survival in APA was favourable in previous reports (Tomlinson ; Krieger ). The present finding might be explained by a more effective salvage therapy in cortical tumours, because EFS did not differ significantly. However, there were numerous strategies that were similar in both groups (Table 1) and hence cannot explain the different survival.

Since the tumour site itself was identified as a prognostic parameter and any explanation for the observed clinical differences has probably to go beyond pure differences in neurosurgical accessibility in the different tumour localisations, a varying tumour biology may well explain clinical differences between the investigated tumour sites. In this regard, our previous findings, that clinical characteristics of paediatric HGG of the Thalamus do share more similarities with pontine HGG than with supratentorial HGG (Kramm ), correlate well with the high frequency of H3F3A K27 mutations in thalamic and pontine gliomas published just recently (Khuong-Quang ; Schwartzentruber ; Sturm ). In the present cohort, we could not show a different incidence of TP53 accumulation between cortical and cerebral HGGs. Noteworthily, no H3.3 mutations were documented in cerebellar HGG, so it is unlikely that a high incidence of the K27M H3.3 mutation might explain the poor prognosis of the cerebellar subset. However, these results must be assessed carefully and further studies are necessary, since the HIT-GBM/HIT-HGG database is clinically based, and unfortunately contains only fragmentary tumour biology data, so that in the present cohort, H3.3 and TP53 status was documented in less than one-fourth of patients.

In conclusion, the present study encompasses the largest series of paediatric cerebellar HGG to date. Only patients with central neuropathological review were analysed to ensure a high quality control level for the data presented here. Most surprisingly, gross-total tumour resection did not significantly improve survival in our cohort of cerebellar HGG which may be due to the low patient number with total tumour resection. In our series, cerebellar HGG shared most of the investigated epidemiological and treatment characteristics with HGG affecting the cortex. Despite these similarities, cerebellar HGG showed a significantly worse survival in comparison with cortical HGG, and tumour site appeared to represent an independent prognostic parameter. Future molecular analyses shall help to corroborate our observations of the obviously site-specific heterogeneity of paediatric HGG on a molecular genetic basis and further to elucidate the biology of cerebellar HGG in children.