The degree of tumour removal, particularly a gross total resection (GTR), appears the most favourable and consistent clinical prognostic factor reported from retrospective case series, with 5-year survival rates being reported as high as 80 %, compared to 22–47 % for cases of incomplete resection [11–13]. However, literature suggests that for approximately one-third of childhood intracranial ependymomas, GTR is not achieved. This is particularly the case for infratentorial tumours extending to involve the lower cranial nerves and associated vasculature, where aggressive resection can be associated with substantial morbidity.

There is insufficient evidence that intracranial location impacts on patient prognosis, independent of that explained by surgical accessibility and the ease of achieving a GTR. There is data, albeit limited, to suggest complete resection for certain supratentorial ependymomas can be curative [14]. Indeed, the North American Children’s Oncology Group (COG) currently recommends only a post-operative observation strategy for grade II supratentorial tumours where GTR has been achieved.

Historically, a younger age at diagnosis has frequently been associated with shorter patient survival. Possible explanations for this include an age-related difference in underlying tumour biology, or the practice of delaying or avoiding adjuvant radiotherapy to the developing central nervous system. However, whilst the majority of studies reveal an association with age, this is not consistent as some studies report no prognostic difference between contrasting paediatric age groups [15].

A distinct difference in survival outcomes between anaplastic and classical ependymomas, according to the WHO grading system, remains unproven. This probably reflects the difficulty of applying such a classification scheme in clinical practice due to intratumoural heterogeneity, inter-observer variability and the inability to uniformly define anaplasia. Whilst some studies have observed an adverse prognosis for anaplastic ependymomas, other studies have refuted this. Unfortunately, alternative classification schemes have also failed to prove either reproducible or prognostic [16].

As stated, advances in tumour molecular analysis (high-throughput gene expression and genomic arrays, tissue microarray analysis by immunohistochemistry and fluorescence in situ hybridisation, etc.) have enabled new biological subgroups of ependymoma to be identified. For instance, gene expression profiling of posterior fossa ependymomas has delineated at least two distinct prognostic groups [9, 17]. One such set of tumours (termed ‘Group A’ or ‘1’) were often located laterally in the cerebellopontine angle, frequently demonstrated overexpression of genes including laminin-alpha 2 (LAMA2), or genes involved in mesenchymal/extracellular matrix development such as Tenascin-C (TNC), and were associated with a younger patient age, recurrence and metastasis. The contrasting group of ependymomas (‘Group B’ or ‘2’) were primarily located centrally, found predominantly in adolescents or adults and were less invasive. Consequently, the progression-free survival (PFS) and overall survival (OS) rates for Group A/1 ependymomas were significantly poorer than for Group 2/B tumours. Likewise, biological prognostic subgroups of supratentorial ependymomas have been identified according to the expression of genes implicated in neuronal differentiation such as NFL70, where overexpression independently conferred an improved PFS [10].

Gain of chromosome 1q is the most common chromosomal imbalance observed in paediatric ependymomas, when compared to adult tumours [11]. Nevertheless, it is only observed in approximately 20 % of cases [18]. Several retrospective genomic studies performed in both North America and Europe, including analyses of uniformly treated clinical trial cohorts, have confirmed an adverse prognostic role for 1q gain, being associated with a worse patient PFS and OS. Studies have also proposed incorporating 1q gain into future patient risk group stratifications in order to tailor therapy more appropriately [19]. The precise mechanism of 1q gain and why, in turn, this seems to impart a worse outcome for patients remains unclear.

Gain of another chromosomal region, 9q33-34, has been shown to occur relatively frequently at recurrence, in addition to being associated with an adverse prognosis in paediatric ependymoma. This region encompasses the loci of genes thought to be implicated in ependymoma pathogenesis such as NOTCH1 and TNC [18].

Several biological markers have been proposed as correlates of adverse outcome in paediatric ependymoma from retrospective case series of paediatric ependymoma. These include overexpression of human telomerase reverse transcriptase (hTERT) [20] and its nuclear chaperone Nucleolin [21], the extracellular matrix proteins TNC and LAMA2 [9, 22], p53 [23], EGFR [24, 25], the transcription factor EVI-1 [26] and the stem cell marker Nestin [27].

However, all of these putative prognostic candidates have not been completely validated, either by generating reproducible findings in different retrospective cohorts or indeed in a prospective clinical trial setting. Until this is undertaken, definitive conclusions regarding their effectiveness as prognostic markers in paediatric ependymoma should be considered with caution.

Evidence has shown, however, that ependymoma can no longer be considered a single biological entity. Indeed, as suggested by Grill and colleagues, due to the biological complexity and diversity of paediatric ependymoma, future prognostic predictions may require a set of several biomarkers which have relevance for a particular molecular subset of the tumour, as is currently the case for medulloblastomas [18].