Pediatric Brain Tumors: Descriptive Epidemiology, Risk Factors, and Future Directions

Primary brain tumors are the most common solid tumors in children and the leading cause of cancer mortality in this population, in high-income countries. Pediatric brain tumors (PBT) are heterogeneous in histopathology, molecular features, and prognosis, and they are classified into two major categories including glial and neuronal tumors (1). The most common forms of glioma in children are astrocytomas, oligodendrogliomas, ependymomas, brain stem gliomas, and optic nerve gliomas. Another rare, but often fatal glial tumor that occurs in children is diffuse intrinsic pontine glioma, or DIPG. The majority of neuronal tumors are embryonal tumors of which the most common types are: medulloblastoma, atypical teratoid/rhabdoid tumors, and central nervous system primitive neuroectodermal tumors (CNS PNET; refs. 2, 3). The term PNET was removed with the 2016 World Health Organization Classification of CNS tumors (3). The new classification is based on amplification of the C19MC region on chromosome 19 (19q13.42). Embryonal tumors with abundant neuropil and true rosettes, ependymoblastomas, and some medulloepithelioma were reclassified as embryonal tumor with multilayered rosettes C19MC-altere. Without C19MC amplification, they should be classified as embryonal tumor with multilayered rosettes, NOS (not otherwise specified) or medulloepithelioma, depending on their histologic features. The term CNS embryonal tumor, NOS is used for CNS PNETs without classifiable genetic mutations (3). Table 1 summarizes the most commonly occurring brain tumor histologies in children (3, 4).

The frequency of different histologic subtypes of PBTs varies by age. According to the Central Brain Tumor Registry of the United States (CBTRUS), in children 0–14 years old, glioma accounted for 53% of all primary brain and other CNS tumors. Among gliomas, the majority were pilocytic astrocytoma (33%) followed by other low-grade gliomas (27%). In addition, 15% of all primary CNS tumors were embryonal tumors of which medulloblastoma (62%) and atypical teratoid/rhabdoid tumors (15%) were the most common histologic subtypes (5).

Incidence and prognosis of PBTs varies greatly and depends on various factors including tumor histology, tumor location, age at diagnosis, race, ethnicity, and sex. Despite their prevalence and clinical importance, knowledge on the etiology and molecular characterizations of pediatric brain tumors is limited. In this review article, serving as an update to the review article by Johnson and colleagues (6), we summarize the descriptive epidemiology and the current knowledge on etiology of primary pediatric brain tumors.

The incidence of PBTs differs by age, sex, geography, race, and ethnicity. In the United States, from 2012 to 2016, the incidence of all primary brain and other CNS tumors in children and adolescents <20 years of age was 6.06 per 100,000 children. Approximately, 58% of cases were malignant and 42% were nonmalignant (4). The incidence was reported to be higher in non-Hispanics compared with Hispanics (6.35 vs. 5.14 per 100,000) as well as in Whites compared with Blacks (6.29 vs. 4.71 per 100,000). The largest differences were observed in incidence of neuroepithelial tissue tumors and cranial and spinal nerves tumors between non-Hispanics and Hispanics, while between Blacks and Whites, the largest differences were found in incidence of neuroepithelial tissue tumors, cranial, spinal nerves tumors, germ cell tumors, and tumors of sellar region (4). In addition, the incidence of all primary brain and other CNS tumors was higher in girls compared with boys (6.13 vs. 5.98 per 100,000; ref. 4); however, this is not consistent with previous reports indicating higher incidence for most histologies in boys compared with girls (5.44 vs. 5.07 per 100,000; for all brain and CNS tumors), during 2007–2011 in the United States (5). The observed elevated incidence for all primary brain and other CNS tumors in girls might be driven by meningioma in 15–19 years old females that were not included in the previous studies, reported on the basis of 0–14 years old children. Among other histologies, medulloblastoma was more common in males compared with girls (0.60 vs. 0.38 per 100,000; ref. 5). According to the CBTRUS report, among children 0–4 years of age diagnosed with brain and other CNS tumors, the highest incidence was attributable to pilocytic astrocytomas (1.15 per 100,000); however, the incidence of this histologic subtype decreased with advancing age. Among children ages 5–9, pilocytic astrocytoma (1.04 per 100,000) followed by malignant glioma (0.88 per 100,000) showed the highest incidence. In addition, the highest incidence of medulloblastoma was observed among children 5–9 years of age (0.59 per 100,000). Among children ages 10–14 and 15–19, the highest incidence was attributable to tumors of the sellar region (0.86 per 100,000) and tumors of the pituitary (2.30 per 100,000), respectively (4).

Survival for patients with PBTs also varies by histology, tumor location, age at diagnosis, race, and ethnicity. The 10-year survival for children ages 0–19 diagnosed with malignant brain and other CNS tumors was estimated at 72% with lowest (17%) and highest (96%) survival rates being attributable to glioblastoma and pilocytic astrocytoma, respectively. In addition, in the United States, 96% of children 0–19 years old with nonmalignant tumors survived 10 years after diagnosis (4). Overall, tumors located in the brain stem showed the poorest survival compared with tumors located at any other site, while tumors of the cranial nerves showed the highest survival (5). Also, survival is better for children diagnosed at an older age, because younger children cannot be treated as intensively as older children (1, 5). Therefore, the difference in survival by age is more pronounced for medulloblastoma and PNETs because their treatment depends more on radiotherapy (5). Survival was reported to be poorer among non-Hispanic Blacks and Hispanic patients compared with non-Hispanic Whites (7–10); in the United States for the period of 2001–2008, 5-year relative survival was reported as 77.6% for non-Hispanic White patients, 69.8% for non-Hispanic Black patients, and 72.9% for Hispanic patients (7). In contrast, among adults ages 18 years or older, in the United States, during 2000–2014, the survival for many tumor types was reported to be poorer in non-Hispanic Whites, while it was relatively comparable in Hispanics and Blacks (11).

There is an established increased risk of PBTs associated with rare single-gene disorders or genetic syndromes, which may occur de novo or may be inherited. However, only a small proportion (∼4%) of PBTs are attributable to these rare autosomal dominant or autosomal recessive disorders. The most common genetic syndromes (and their related genes) predisposing to nervous system tumors include: neurofibromatosis type 1 (NF1), neurofibromatosis type 2 (NF2), tuberous sclerosis complex (TSC1 or TSC2), Li-Fraumeni (TP53), Gorlin syndrome (PTCH1), familial adenomatous polyposis (APC, MMR), glioma susceptibility 3 (BRCA2), and biallelic mismatch repair deficiency (MSH2, MLH1, MSH6, PMS2; refs. 12–14).

A modest risk of developing CNS tumors among the siblings of PBT cases has been reported. In particular, a higher risk was observed if both children have been diagnosed with medulloblastoma and PNET. Children with a parent diagnosed with a CNS tumor showed an elevated risk of developing brain tumors; however, these observed associations were based on small numbers of affected families. In general, there is limited evidence for an association between family history of cancer and nonsyndromic PBTs (6, 15).

Because of rarity of PBTs and further rarity of familial PBTs, little knowledge is available on genetic variants contributing to the genetic architecture of familial PBTs. Backes and colleagues (16) performed a study in a family with two unaffected parents and two siblings diagnosed with glioblastoma. By using whole-exome sequencing, they identified three significant pathways containing at least three affected genes, including: focal adhesion, extracellular matrix–receptor interaction, and complement and coagulation cascades. Of all the identified genes, 32 genes were located on chromosomes 1, 11, and 22. More specifically, the affected genes were accumulated on 22q12.2 and 1p36.33 (16).

Germline mutations associated with sporadic PBTs vary by histologic subtype, and about 10% of sporadic PBT cases harbor a predisposition mutation. To date, the conducted studies have been mainly focused on high-penetrant germline mutations in known cancer predisposition genes; therefore, the contribution of rare high-penetrant mutations in the risk of PBTs is largely unknown (17). Recently, a large study, performed on childhood high-grade glioma by using whole-exome sequencing, identified that the rare germline variants associated with risk of PBTs are mainly located in 24 genes largely involved in DNA repair and cell-cycle pathways, predominantly in the TP53 and NF1 genes (18). In addition, Waszak and colleagues, by employing rare variant burden analysis, estimated that 6% of medulloblastoma diagnoses are attributable to germline mutations and identified APC, BRCA2, PALB2, PTCH1, SUFU, and TP53 as consensus medulloblastoma predisposition genes. They reported that the prevalence of genetic predispositions differs among molecular subgroups of medulloblastoma, with the highest prevalence being attributable to patients in the Sonic HedgeHog (SHH) subgroup. Also, correlations between specific germline mutations and development of specific molecular subgroups of medulloblastoma were detected (19). Furthermore, they identified ELP1 as the most common medulloblastoma predisposition gene and found that ELP1 rare variants occurred in 14% of medulloblastoma SHH subgroup and elevated the prevalence of genetic predisposition to 40% among patients in this molecular subgroup (20). Begemann and colleagues, by investigating 1,044 medulloblastoma cases, identified that heterozygous germline mutations in the G protein–coupled receptor 161 (GPR161) gene was exclusively associated with SHH subgroup and accounted for 5% of infants with SHH subgroup in their medulloblastoma cohort (21).

Very few and generally small genetic association studies have been conducted on brain tumors in children and adolescents. To date, there is one published genome-wide association study (GWAS) of medulloblastoma. This study identified 13 genetic variants associated with medulloblastoma risk located in CD83 (6p23), MAGI2 (7q21.11), CSMD1 (8p23.2), DOCK1 (10q26.2), PTPRM (18p11.23), and 8q24.12 (22). The genetic variants associated with risk of PBTs have been mainly identified by candidate-gene association studies conducted on pooled histologic subtypes of PBTs (12). The identified genetic variants mainly belong to genes involved in xenobiotic detoxification (CYP1A1, GSTT1, GSTM1; refs. 23, 24), inflammation (NOS1; ref. 24), DNA repair (ERCC1, ERCC2, CHAF1A, XRCC1, EME1, ATM, GLTSCR1, XRCC4, PALB2; refs. 22, 24–26), and cell-cycle regulation (AICDA, CASP1, IRS2, EGFR, PTCH1; refs. 22, 25–27). It has been shown that the validated genetic variants identified by GWAS on adult glioma are also associated with risk of PBTs. These variants are predominantly located in TERT (5p15.33), RTEL1 (20q13.33), CCDC26 (8q24.21), and CDKN2BAS (9p21.3; refs. 28–30). A recent study was performed in a U.S. population to assess whether genome-wide ancestry differences are associated with risk of ependymoma. In addition, admixture mapping was conducted to detect associations with local ancestry. The results revealed significant associations between eastern European ancestral substructure and ependymoma risk among Hispanics and non-Hispanic Whites. Furthermore, a significant peak located at 20p13 was detected to be associated with increased local European ancestry (31). Given the limited knowledge available on the germline variants associated with PBTs and the rarity of these malignancies, utilizing various approaches, including Mendelian randomization, are needed for a better understanding of the etiology of PBTs and assessing their risk factors (32). Table 2 summarizes the identified genetic variants associated with PBT risk.

Despite the potentially important role of maternal genetics on the risk of PBTs by affecting the in utero environment of the developing embryo, limited knowledge is available on role of maternal genetic variations in the etiology of these tumors. Lupo and colleagues (33), in the only available study of its kind, investigated the role of maternal variation in xenobiotic detoxification genes and the risk of pediatric medulloblastoma using a case-parent triad study design. The results indicated that maternal variation in EPHX1 (rs1051740) was associated with elevated risk of pediatric medulloblastoma (relative risk = 3.26; 95% confidence interval, 1.12–9.53; ref. 33).

Exposure to moderate-to-high doses of ionizing radiation is the only established environmental risk factor for PBTs (34). Compared with adults, children are more radiosensitive and have a longer life expectancy to experience the carcinogenic effects of ionizing radiation (35). There is evidence that radiotherapy for early-onset childhood cancers, particularly children who received radiotherapy for acute lymphoblastic leukemia that included exposure to the brain, is correlated with an increased risk of brain tumor development later in life (12, 34, 36). In addition, some studies have reported that maternal diagnostic radiation during pregnancy is associated with an increased risk of brain tumors in offspring (34, 37). The effect of diagnostic radiation during early childhood on subsequent brain tumor risk was evaluated, and a 29% excess risk was reported for children exposed to one or more head CT scans (35, 37–39). This finding should be interpreted with caution because pre-existing cancer in children with high susceptibility may lead to undergoing more head CT scans (39).

The effects of non-ionizing radiation, including radiofrequency, microwaves, and extremely low-frequency (ELF) magnetic fields, on the risk of PBTs have been investigated by some studies. Despite the classification of radiofrequency fields as a possible carcinogen by the International Agency for Research on Cancer (IARC) in 2011, no significant associations were observed for cellular phone use or other radiofrequency radiation exposure by recent high-quality studies. In addition, in 2002, based on the available findings, IARC concluded that there are not sufficient data to classify ELFs as a risk factor for brain tumors (6, 12).

Although there is consistent evidence for an association between personal medical history of allergies and decreased risk of adult glioma, inconsistent evidence of reduced risk of PBTs associated with allergic and atopic conditions (such as asthma, wheezing, and eczema), as well as early life exposure to infections, has been reported previously (6, 40–42). It is unclear whether history of allergies and atopic diseases decrease the risk of PBTs or PBTs prevents allergic and atopic conditions, further investigations are warranted to clarify the action of these effects (42).

Advanced parental age, as a marker for accumulated genetic aberrations in the parents' DNA, has been reported to be associated with an elevated risk of brain tumors in offspring (43). There is more consistent evidence of increased risk of PBTs associated with advanced paternal age at birth than advanced maternal age at birth (6, 44). Despite the extensive research on the association between parental occupational exposures and risk of brain tumors in offspring, inconsistent findings have been reported. However, the results from the studies of parental occupational/residential exposure to pesticides are more consistent and the meta-analysis studies indicate a positive association between risk of PBTs and exposure to pesticides (6, 12, 45, 46). In addition, positive associations between parental high socioeconomic status (47, 48) as well as maternal intake of dietary N-nitroso compounds (NOC) and risk of PBTs in offspring have been reported by meta-analysis (6, 12, 49).

Of the investigated birth characteristics, large studies and meta-analyses provide evidence that high birth weight (>4,000 g) is associated with an increased risk of pediatric CNS tumors, particularly astrocytoma and embryonal tumors (50, 51).

Approximately 7% of PBTs can be attributable to nonchromosomal structural birth defects, which is one of the most consistent risk factors for childhood cancer overall (12, 52). Large, population-based studies reported that birth defects are associated with approximately 2-fold elevated risk of brain tumors in children. (12, 53, 54). Children with CNS birth defects or with a neurologic anomalies showed an even higher susceptibility to developing PBTs (12, 55). Table 3 summarizes the suspected nongenetic risk factors associated with PBT risk.

PBTs represent a complex heterogeneous group of neoplasms with different histopathology, molecular features, and etiology. Various factors including tumor histology, tumor location, age at diagnosis, sex, race, and ethnicity are correlated with the incidence and prognosis of PBTs. Exposure to ionizing radiation and some rare genetic syndromes are the only established risk factors for PBTs; although relatively consistent evidence of positive associations for birth defects, markers of fetal growth, advanced parental age, maternal dietary NOCs, and exposure to pesticide has been reported (summarized in Fig. 1); however, the findings of these studies should be interpreted with caution as some of the studies are based on small sample sizes and exhibit some methodologic challenges.

Performing large, collaborative, and multi-institutional genetic studies based on SNP-array and next-generation sequencing data to identify common and rare germline variants associated with risk of PBTs, in different ethnic groups, is an important research priority. Considering the heterogeneity of PBTs, the studies that aim to evaluate histology-specific risk variants are also needed. Utilizing high-quality publicly available genetic and environmental data as well as data from cancer and birth defect registries are beneficial for studies of these rare tumors. Gene–environment interaction studies will play an important role to increase our understanding of the etiology of PBTs. Incorporating genetic and environmental data may lead to the development of comprehensive risk prediction models that could be leveraged for the prevention of these tumors. To conclude, the literature on the risk factors for PBTs is currently an amalgam of small, underpowered studies, many of which also suffer design flaws that limit their generalizability. As such, the PBT etiologic literature suffers from extensive publication bias. Thus, large-scale, well-powered systematic collaborative studies conducted by researchers from multiple institutions are warranted in the pediatric brain tumor research field to improve our knowledge of PBT risk factors; which would lead to the development of prevention measures and better management of pediatric brain tumors.

M. Adel Fahmideh reports grants from Cancer Prevention and Research Institute of Texas during the conduct of the study. No disclosures were reported by the other authors.

The sponsor had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the article; or decision to submit the article for publication.

This study was supported, in part, by the Research Training Award for Cancer Prevention Post-Graduate Training Program in Integrative Epidemiology from the Cancer Prevention and Research Institute of Texas (grant number RP160097, PI: M. Spitz).