Childhood Brain Tumor Epidemiology: A Brain Tumor Epidemiology Consortium Review

Brain and central nervous system (CNS) tumors are the most common solid tumor and the second leading cause of cancer-related death in individuals 0 to 19 years of age in the United States and Canada (1, 2). The objective of this review is to summarize the descriptive and analytic epidemiology of childhood brain tumors (CBT) with a specific focus on studies from the past decade (since 2004) and to delineate future directions in CBT epidemiology research that are needed for progress in the field. We have included studies published primarily since 2004 pertaining to CBT descriptive and analytical epidemiology. We note that there is no precise definition of CBTs, and the tumor types included vary between studies, which can make them difficult to compare.

There are >100 different histologic subtypes of CNS tumors with the incidence of each varying by age and histologic subtype. Childhood CNS tumor incidence varies by country from 1.12 to 5.14 cases per 100,000 persons with the highest incidence in the United States (Table 1). CBTs are more common in males, though this varies by histologic type. In the United States, whites and Asians-Pacific Islanders have a higher CBT incidence than blacks and American Indians/Alaska Natives, whereas non-Hispanics have higher incidence than Hispanics. Subtype incidence and survival rates are reviewed below and in Tables 1 and 2.

Case ascertainment methodology, completeness, and standard populations used for age adjustment of rates vary between cancer registries, making it challenging to compare statistics across registries. In addition, registries vary on when they began to include the reporting of benign brain tumors. For example, in the United States, registration of nonmalignant tumors was not required by law and, therefore, limited before 2004. Final confirmation of CNS tumors can also vary by histologic type and by region; even in the United States, some tumors are not microscopically confirmed but are confirmed radiographically. However, across registries, the standard approach is to include both brain tumors and other CNS tumors in all statistics. Therefore, all comparison statistics must be interpreted with these caveats in mind.

Gliomas that arise from glial cells are the most common CBT (3). Incidence and survival vary significantly depending on location and histologic type (reviewed below and in Tables 1 and 2).

Pilocytic astrocytoma [World Health Organization (WHO) grade 1] is one of the most common CBTs, representing approximately 17% of all CNS tumors in 0- to 14-year-olds (4). Incidence rates in population-based analyses range from 0.74 to 0.9 cases per 100,000 persons (Table 1). These tumors are usually nonmalignant, although some progress to higher-grade tumors (5, 6). Pilocytic astrocytomas have a high overall 10-year survival rate at >96% (4).

Brain stem tumors represent approximately 10% of all pediatric CNS tumors with the most common being diffuse intrinsic pontine glioma (DIPG; ref. 7). DIPG prognosis is dismal with >90% of cases dying within 2 years of diagnosis (8). These tumors are rarely biopsied, and as a result, their true incidence from cancer registry datasets is difficult to assess (8).

Other glioma types are less common in children. Diffuse astrocytomas (WHO grade II) account for approximately 5% of all tumors in children ages 0 to 14 years, with a U.S. incidence rate of 0.28/100,000 (Table 1; ref. 4). High-grade astrocytomas (WHO grades III and IV) are less common, with incidence rates of 0.08 for anaplastic astrocytoma and 0.14 for glioblastoma (4).

Embryonal tumors are theorized to develop in embryonic cells remaining in the CNS after birth. There are three major embryonal tumor types with distinct differences in age at diagnosis and survival: primitive neuroectodermal tumor (PNET), medulloblastoma, and atypical teratoid/rhabdoid tumor (ATRT; ref. 9). Overall embryonal tumor incidence ranged from 0.28 to 0.80 cases per 100,000 children ages 0 to 14 years (Table 1) with a 10-year relative survival rate of 55.5% (Table 2; ref. 4).

Average annual age-adjusted incidence rates for PNET ranged from 0.08 to 0.21 cases per 100,000 children. PNET survival improves with increasing age with U.S. population data from 2001 to 2006 showing 1-year survival rates of 31%, 88%, and 95% for children ages 0 to 1, 1 to 9, and 10 to 19 years, respectively (10). On the basis of the 1993 WHO criteria, histologically similar tumors that are classified as PNETs and medulloblastomas when they occur supratentorially and infratentorially, respectively (11). Before this, tumors were considered PNETs regardless of tumor location.

Medulloblastomas are the most common embryonal tumors with average annual age-adjusted incidence rates ranging from 0.20 to 0.58 cases per 100,000 persons. An analysis of U.S. population data from 2001 to 2006 reported 1-year survival rates of 52%, 90%, and 92% for children ages 0 to 1, 1 to 9, and 10 to 19 years, respectively (10). Molecular analysis has identified four distinct medulloblastoma subtypes that correlate strongly with survival (12). No population-based studies of subtype-specific survival have been reported, but in an international meta-analysis, children with WNT tumors had a 95% 10-year overall survival. Children with SHH tumors, group 3, and large cell anaplastic tumors had 51%, 50%, and 32% 10-year survival, respectively (13).

ATRT is a rare embryonal tumor that most commonly occurs in children <3 years old. Average annual age-adjusted incidence rates range from 0.07 to 0.14 per 100,000 persons (14, 15). Prognosis is generally poor, though survival increases with age (14–18). Overall, median survival is usually between 6 and 18 months (16, 18–20). Most analyses show that ATRTs are more common in males (15, 16, 21) and among whites (15, 22). A systematic diagnostic approach for ATRT was not common until 2005; before that, these tumors were frequently misclassified, mostly as medulloblastomas or PNET (14).

Established familial cancer syndromes (gene) that increase brain tumor susceptibility include neurofibromatosis type 1 (NF1), neurofibromatosis type 2 (NF2), tuberous sclerosis (TSC1 or TSC2), Li–Fraumeni (TP53 or CHEK2), nevoid basal cell carcinoma (PTCH), Turcot (APC), Cowden (PTEN), hereditary retinoblastoma (RB1), and Rubinstein–Taybi (CREBBP; refs. 23–26).

Findings from studies of CBT risk among family members vary substantially. A 2008 review (27) including publications as early as 1959 reported that although most studies observed positive associations specific to brain tumors, there was borderline statistical evidence for an increased risk. Siblings of childhood CNS cancer cases consistently showed increased risks of developing a childhood CNS tumor, with a higher risk seen if both children had medulloblastoma or PNET diagnoses. Risk was also reported to be higher among relatives if the index child was diagnosed at ≤4 years old. Children also had an increased risk of developing a nervous system tumor if a parent also had this tumor type. The SEARCH international brain tumor case–control study, which included 1,200 CBT cases and 2,218 controls from Australia, Canada, France, Israel, Italy, Spain, and the United States, reported no significant associations between CBTs and brain tumor history in close relatives, with ORs of 0.8 [95% confidence interval (CI), 0.5–1.3], 1.3 (95% CI, 0.7–2.3), and 1.1 (95% CI, 0.6–1.9) for astroglial (n = 620), PNETs (n = 244), and other CBT (n = 324) subtypes, respectively (28).

Parental age at birth may serve as a marker for inherited somatic changes in aging parental germlines. Hemminki and colleagues (29) previously reported that offspring of older fathers (>40 years at the child's birth) were at increased CBT risk with no maternal age effect in a cohort study that included 1,617 CBT cases diagnosed at ages 0 to 14 years. A more recent Swedish analysis (30) of CNS tumors diagnosed in 0- to 4-year-olds (n = 977) indicated higher risks associated with paternal age >40 years after maternal age adjustment [incidence rate ratio (IRR) = 1.69; 95% CI, 1.21–2.35], particularly for astrocytoma. In contrast, Johnson and colleagues (31) reported an increased childhood CNS tumor risk in association with maternal age after paternal age adjustment in a U.S. multistate record linkage study including >3,500 cases (OR_{per 5-year age increase} = 1.08; 95% CI, 1.03–1.14). Only astrocytomas and ependymomas were associated with an increased risk. The authors also reported stronger IRRs for children diagnosed at younger ages.

Recent research has addressed the role of maternal genetic variation in genes that may influence the in utero environment. In spite of the potential importance of this mechanism in the etiology of CBTs, few assessments of maternal genetic effects have been performed. To our knowledge, there is only one small report that used a case parent triad study design (32) of the role of maternal variation in xenobiotic detoxification genes and the risk of childhood medulloblastoma (33), where it was reported that the maternal EPHX1 rs1051740 genotype was associated with medulloblastoma risk (RR, 3.26; 95% CI, 1.12–9.53). Larger studies are needed to explore the role of maternal genetic effects in CBT susceptibility.

Studies consistently suggest inverse associations between adult gliomas and allergic conditions (34). In children, a 2008 U.K. study including 575 cases diagnosed at <15 years of age and 6,292 controls indicated that maternally reported asthma decreased CNS tumor risk (OR, 0.75; 95% CI, 0.58–0.97), particularly for medulloblastoma/PNETs (OR, 0.43; 95% CI, 0.23–81). However, this result was not confirmed in a participant subset for whom medical records were available (OR, 1.20; 95% CI, 0.74–1.94), which could be due to the diagnosis not being present or not being recorded (35). CNS tumors were not associated with eczema (OR, 0.94; 95% CI, 0.74–1.18), but there was a significant inverse association for children with both asthma and eczema (OR, 0.48; 95% CI, 0.28–0.81; ref. 36). A study of 272 matched case–control pairs reported an inverse association between CBTs diagnosed between 0 and 15 years old and asthma (OR, 0.55; 95% CI, 0.33–0.93), which was stronger for ependymoma (OR, 0.15; 95% CI, 0.18–1.21). No association with eczema was found. Overall, CNS tumor risk was increased with use of asthma controllers (e.g., inhaled corticosteroids; OR, 2.55; 95% CI, 0.79–8.20) or asthma relievers (e.g., beta agonists; OR, 1.62; 95% CI, 0.57–4.63; ref. 37). Finally, CEFALO, a study conducted in Denmark, Norway, Sweden, and Switzerland that included 352 CBT cases diagnosed from 7 to 19 years and 646 controls, found no association with any atopic condition (asthma, wheezing, eczema, allergic rhinitis; OR, 1.03; 95% CI, 0.70–1.34) and some evidence for reverse causality; an inverse association between CBTs and having a current (OR, 0.76; 95% CI, 0.53–1.11) but not past (OR, 1.22; 95% CI, 0.86–1.74) atopic condition was found (38). Altogether, allergic conditions may be a protective factor for CBT development, but further research is needed.

Studies, before 2004, of markers of infection and CBT risk have yielded mixed results (39–41). More recently, higher risks of CBTs among first-born children versus those with higher birth order and lower risks among those who attended daycare as an infant have been reported. Altieri and colleagues (42) compared the incidence of brain tumors in the Swedish Cancer Registry based on the number of total siblings, older siblings, and younger siblings. When compared with cases diagnosed at <15 years old with no siblings, the relative risk (RR) for cases with ≥3 younger siblings was increased for astrocytoma (RR, 1.34), medulloblastoma (RR, 2.30), ependymoma (RR, 2.61), and meningioma (RR, 3.71). Shaw and colleagues (43) reported that CBT risk was elevated for having siblings (OR, 1.4; 95% CI, 0.9–2.3) and being at least second born (OR, 1.7; 95% CI, 1.2–2.4).

Several studies suggest infectious exposures during older childhood increase brain tumor risk, whereas earlier infections reduce brain tumor risk. Harding and colleagues (44) reported that children who had no social contact with other infants in the first year of life displayed an increased CNS tumor risk versus those who had such early exposures (OR, 1.37; 95% CI, 1.08–1.75), particularly among medulloblastoma cases (OR, 1.78; 95% CI, 1.12–2.83). In addition, children who attended informal (OR, 0.86; 95% CI, 0.68–1.09) or formal (OR, 0.93; 95% CI, 0.68–1.26) daycare showed slightly reduced risks versus those reporting social contact only. Shaw and colleagues (43) reported that CBT risk was reduced for subjects who attended daycare for >1 year or were breastfed, whereas Harding and colleagues (45) found no association between breastfeeding and CBTs (OR, 1.01; 95% CI, 0.85–1.21). Most recently, Andersen and colleagues (46) reported that glioma (OR, 2.93; 95% CI, 1.57–5.50) and embryonal tumor (OR, 4.21; 95% CI, 1.24–14.30) cases had more frequent sick days with infections in the first 6 years of life versus controls. However, the timing of infections in relation to the first year of life versus later in childhood was not evaluated. One common observation from these studies is that level of risk often varies by age at diagnosis and tumor type.

Congenital anomalies (CA) and birth characteristics have been examined as putative risk factors for pediatric CNS tumors (47–49). Among large studies, 45,200 children with CAs were identified in the Canadian Congenital Anomalies Surveillance System and matched to 45,200 children without CAs identified through the Ontario Birth Certificate File. The Ontario Cancer Registry was then used to identify 212 newly diagnosed cancers in the matched cohorts. The authors observed a 2.5-fold increased CNS cancer risk in association with CAs that was stronger for children <1 year old (5.5-fold greater risk). Those with nervous system anomalies had an approximate 6-fold increased rate of primary CNS tumors (50).

Using two population-based national birth registries in Sweden and Norway, Bjorge and colleagues (51) linked birth and cancer registry data to examine risk of multiple pediatric cancer types in association with birth defects. Specifically, children with nervous system malformations were at elevated risk of CNS cancers in both countries, particularly Norway.

Fisher and colleagues (52) linked data from the California Cancer Registry (CCR) to the Birth Defects Monitoring System for the period of 1988 to 2004 among children ages 0 to 14 years. There were 4,869 children identified with cancer, among whom 222 had a major birth defect. The authors reported a 1.87-fold (95% CI, 0.6–5.79) and 1.80-fold (95% CI, 1.28–2.53) elevated risks of CNS tumors among children with and without nonchromosomal and chromosomal anomalies, respectively.

A second study linking the CCR to California birth certificates examined birth anomalies and CNS tumor risk among children ages 0 to 14 years old between 1988 and 2006 (53). In this study, 4,560 newly diagnosed CNS tumors were identified of which 3,733 cases (82%) could be linked to the birth registry. Cases were then individually matched to four controls (n = 14,932). Medulloblastomas and PNETs were more elevated in children with birth defects, with age-stratified analyses revealing stronger risks for younger children (OR, 1.7; 95% CI, 1.12–2.57 and OR, 2.9; 95% CI, 1.68–5.05 for children <2 and <1 year(s) old, respectively). This study was limited by the inability to capture birth defect information after hospital discharge.

In one of the largest studies to date, Bjorge and colleagues (54) conducted a nested case–control study to examine fetal growth in relation to cancer development in Nordic children born between 1967 and 2010 using population-based birth registries. Each case (n = 17,698) was matched to 10 controls (n = 172,422). Both higher birth weight (RR_{≥4,500 g vs. 3,000–3,499 g} = 1.3; 95% CI, 1.1–1.3) and increasing head circumference (RR_{39–45 cm vs. 33–36 cm} = 1.7; 95% CI, 1.2–2.3; P trend < 0.001) were associated with childhood CNS cancer risk. In a similar but smaller study including the same four Nordic nations, Schmidt and colleagues (55) conducted a nested case–control study to examine the impact of fetal growth (including birth weight) on CNS tumor risk among children ages 0 to 14 years who were diagnosed with a CNS tumor between 1985 and 2006. This study matched 3,443 CNS cases identified from national cancer registries to 16,169 birth registry controls and found a significant gestational age-adjusted association between birth weight >4,500 g and risk of all CNS (RR, 1.27; 95% CI, 1.03–1.55), and embryonal (RR, 1.8; 95% CI, 1.2–2.8) tumors but not other histologic subtypes.

Milne and colleagues (56) examined the relationship between fetal growth measured as proportion of optimal birth weight or length and CNS tumor development diagnosed between 1980 and 2004 in children ages 0 to 14 years. Among >600,000 live births, 183 pediatric CNS tumors were identified. There were no statistically significant associations between fetal growth factors and CNS tumor development.

Using CCR data to examine birth characteristics and CNS tumor risk in children ages 0 to 14 years old between 1988 and 2006, MacLean and colleagues (57) matched each child with a CNS tumor (n = 3,733) to four controls identified through the California birth certificate database, resulting in 14,932 controls. There was an increased CNS cancer risk in the highest weight category (>4,000 g) among high-grade gliomas, whereas among low-grade gliomas, those in the lowest weight category (<2,500 g) appeared to be protected against CNS tumors. This study indicates that separation of CNS subtype is warranted in studies of birth characteristics and CNS tumor risk.

Finally, in a 2008 meta-analysis, eight studies were identified that examined CBT risk in association with birth weight. Data from over 1.7 million children/young adults (0–18 years old) were analyzed, with 4,162 primary diagnoses of astrocytoma, medulloblastoma, or ependymomas combined. Most cases were identified through cancer registries, and the predominant study design was case–control. The authors found that high birth weight (>4,000 g) was predictive of both astrocytoma (OR, 1.38; 95% CI, 1.07–1.79) and medulloblastoma (OR, 1.27; 95% CI, 1.02–1.60), but not ependymomas, which was only examined in few studies (58).

High-dose radiation to the head and neck for treatment of cancer or other conditions is an established CBT risk factor (59). Radiotherapy for acute lymphoblastic leukemia is associated with a particularly high risk with several studies published in the 1990s (reviewed in ref. 60) showing increased brain tumor risks (gliomas, PNETs) in children who received prophylactic CNS irradiation (usually a cumulative dose of ∼25 Gy). The latency between radiotherapy and subsequent brain tumor development has been estimated at 7 to 9 years with a higher risk for younger children (60). It has also been broadly accepted for several decades that in utero diagnostic radiation exposure is associated with a small-to-moderate dose-dependent increase in childhood cancer risk, including brain tumors (61). Recent studies examining ionizing and nonionizing radiation exposure are reviewed briefly below with study details provided in Tables 3 and 4, respectively.

The Childhood Cancer Survivor Study reported that radiation therapy for a first primary cancer (most were leukemia) was associated with a significant 7.1-fold increased risk of a subsequent CNS tumor (62).

A Danish study examined CBT risk associated with neonatal diagnostic X-ray exposure (vs. no exposure) and observed a 2-fold positive nonsignificant association (63). A Swedish study of individuals born between 1975 and 1984 examined the association between medical record–abstracted prenatal X-ray abdominal exposures and CBTs and found no increased risk overall but an almost 2-fold increased risk for PNETs (64). A U.S. study of medulloblastoma/PNETs examined risks associated with maternally reported postnatal diagnostic X-rays and reported no significant associations for head, dental, or any X-ray exposure versus no exposure (65). A U.K. study reported associations between cancers and medical record–abstracted prenatal diagnostic radiation exposure. On the basis of 25 and 41 exposed CBT cases and controls, respectively, no significant association for prenatal or early infancy radiation exposure was observed (66).

Two studies examined childhood/adolescence CT scans and subsequent brain tumor development. A U.K study employing a retrospective cohort study design that included 176,587 CT scan–exposed individuals reported increased risks of subsequent brain tumor development for the exposed group (67). An Australian study, that included approximately 11 million individuals, also reported significant positive associations between CT scan exposure and brain tumor development with risk generally decreasing with increasing age at first exposure, years since first exposure, and increasing calendar year of first CT scan (68).

An ecological study conducted in Florida in response to an observed excess of childhood brain and other nervous tissue cancers in the 1990s found no evidence to indicate that the observed excess was related to nuclear plant installation in St. Lucie County in 1976 (69).

Sources of nonionizing radiation that have been studied for their role in CBT risk predominantly include radio frequency/microwave (e.g., cell phones, AM and FM radio, televisions, and microwaves) and extremely low frequency magnetic fields (ELF-MF; e.g., power lines and electrical wiring) that are classified as possibly carcinogenic by the International Agency for Cancer Research (70). Studies have shown no significant associations between nonionizing radiation exposure and CBTs as summarized briefly below.

A South Korean case–control study examined associations between residential AM-radio transmission exposures and CBTs. No significant associations were observed by residential distance to the AM-radio transmitter or for estimated radio-frequency radiation exposure (mV/m) for the fourth versus first quartile (71).

A 2008 meta-analysis that examined associations between residential magnetic field exposure and CBTs reported no significant associations in a number of different analyses (72).

Kheifets and colleagues (73) conducted a pooled analysis of 10 U.S. and European studies and found no evidence for an association between ELF-MFs and CBTs.

A U.K. registry-based case–control study that examined maternal radio-frequency exposure from macrocell cellular phone base stations (masts) and mast proximity during pregnancy and offspring CBT risk reported no association for mother's exposure to masts during early pregnancy or for modeled power density birth address (74).

CBTs and cellular phone use in 7- to 19-year-olds were examined in a European multicenter case–control study (CEFALO); the authors reported no significant association between CBTs and regular cellular phone use versus nonuse (75).

Medications containing amides or amines (e.g., barbiturates, antiepileptics, and antihistamines) may be converted to carcinogenic N-nitroso compounds (NOCs including N-nitrosamines and N-nitrosoamides) upon ingestion through reaction with dietary nitrate in the stomach. Two studies before 2004 did not find statistical evidence for an association between maternal exposure to nitrosable drugs and offspring CBTs (76, 77). Likewise, a large study published in 2006 that included 1,218 CBT cases and 1,218 controls found little support for an association between CBTs and medications containing amines or amides (OR, 1.01; 95% CI, 0.82–1.24) overall or for astroglial (OR, 1.01; 95% CI, 0.78–1.31), PNET (OR, 1.09; 95% CI, 0.75–1.60), or other glial (OR, 1.01; 95% CI, 0.71–1.44) subtypes. No significant associations were found when data were analyzed by age group (≤5 vs. >5 years) or class of drugs (barbituates, antiepileptics, antihistamines, neurally active drugs, diuretics, sex hormones, or antiemetics; ref. 78).

A German case–control study of 399 CNS cases diagnosed between 1992 and 1994 and 2,057 controls evaluated associations between maternally reported medications and CBTs and found no significant associations between CNS tumors and diuretics/antihypertensives (OR, 1.65; 95% CI, 0.73–3.74), “pain relievers” (OR, 1.0; 95% CI, 0.6–1.67), antinauseants or antiemetics (OR, 1.15; 95% CI, 0.68–1.96), or cold medications (OR, 0.81; 95% CI, 0.55–1.21). The authors also reported ORs of 1.23 (95% CI, 0.71–2.12) and 0.92 (95% CI, 0.68–1.24) for associations between offspring CNS tumors and high blood pressure/edema during pregnancy treated with and without drugs, respectively (79).

A Taiwanese pregnancy cohort study examined maternal use of herbal medicines (Coptidis Rhizoma, An-Tai-Yin, and other herbs) and reported an increased HR for brain tumors in association with Coptidis Rhizoma (HR, 4.79; 95% CI, 1.28–17.91; ref. 80).

A Swedish registry–based linkage study used medical record data to examine associations between maternal medication ascertained from medical records and offspring CBTs from 0 to 14 years. No significant associations were found for alimentary tract medicines (mainly antacids and laxatives), vitamins and iron, folic acid (FA), diuretics, antiinfectives (antifungals, penicillin, antibiotics), analgesics [Aspirin/NSAID, Opioids, Paracetamol (acetaminophen), antiemetics, antihistamines, neuroleptics], antiasthmatics (oral and inhalation therapy). In contrast, maternal antihypertensives (OR, 2.7; 95% CI, 1.1–6.5) were positively associated with offspring CBTs, especially for β-blockers (OR, 5.3; 95% CI, 1.2–24.8; ref. 81).

A German case–control study reported significantly increased risks for CNS tumors overall (OR, 1.56; 95% CI, 1.01–2.40), medulloblastoma (OR, 2.07; 95% CI, 1.03–4.17), and astrocytoma (OR, 2.26; 95% CI, 1.09–4.69), but not ependymoma (OR, 1.23; 95% CI, 0.37–4.13) in association with maternal prenatal antibiotic use. For maternal antibiotic exposure, including the three months before pregnancy through pregnancy, the associations were less strong and not significant for CNS tumors overall (OR, 1.37; 95% CI, 0.92–2.05), medulloblastoma (OR, 1.79; 95% CI, 0.92–3.48), astrocytoma (OR, 1.79; 95% CI, 0.87–3.70), or ependymomas (OR, 0.95; 95% CI, 0.28–3.17; ref. 82).

Maternal antibiotic use during pregnancy was also examined in a Canadian case–control study that included 272 case–control pairs. Cases were diagnosed at <15 years of age from 1980 to 1999. A nonsignificant positive association between CBTs and prenatal antibiotic exposure (OR, 1.7; 95% CI, 0.8–3.6) was reported (43).

Relatively consistent evidence from earlier studies for a protective effect of prenatal vitamins on offspring CBT risk (reviewed in ref. 83) has been reported. A 2007 German case–control study (79) reported no significant association between CBTs and maternally reported vitamin, folate, and/or iron supplements (OR, 1.07; 95% CI, 0.85–1.34). A U.S. case–control study of 315 medulloblastomas/PNETs diagnosed from 0 to 5 years old from 1991 to 1997 reported no association for periconception (OR, 1.2; 95% CI, 0.8–2.1) or mid-pregnancy (OR, 1.1; 95% CI, 0.7–1.6) dietary folate intake when comparing the highest with lowest quartile of intake (84). In a later report, the authors reported an OR of 0.7 (95% CI, 0.4–1.0) for preconception multivitamin use. For dietary folate with supplements, the periconception and mid-pregnancy ORs for the highest versus lowest intake quartile category (≥380 μg vs. <267 μg) were 0.5 (95% CI, 0.3–0.9) and 0.3 (95% CI, 0.5–1.3) with a significant trend for increasing periconceptional intake (P = 0.007; ref. 85). A Swedish study reported a nonsignificant inverse association for FA supplementation (OR, 0.6; 95% CI, 0.3–1.1; ref. 81).

A 2010 case–case study compared maternal FA supplement intake in nervous system tumors (n = 44) versus mesodermal tumor (n = 155) cases diagnosed in children ages 0 to 14 years old during the 2004 to 2006 period. The ORs for ≥400 versus <400 μg/day were 0.34 (95% CI, 0.10–1.06), 0.19 (95% CI, 0.06–0.6), 0.57 (95% CI, 0.33–0.99), and 0.94 (95% CI, 0.79–1.14) for preconceptional, <21 days gestation, <36 days gestations, and any period, respectively. Multivitamin supplementation was also inversely associated with CNS tumors for first (OR, 0.29; 95% CI, 0.09–0.92), second (OR, 0.18; 95% CI, 0.02–1.35), and any (OR, 0.22; 95% CI, 0.07–0.68) trimester intake (86). A 2012 Australian study of 327 CBT cases diagnosed from 0 to 14 years between 2005 and 2010 and 867 controls reported inverse associations during pre-pregnancy for maternal FA supplement intake (OR, 0.68; 95% CI, 0.46–1.01) and FA supplement without iron, vitamins B6, B12, C, or A intake (OR, 0.55; 95% CI, 0.32–0.93). No significant associations were found for FA supplement intake during trimester 1 or 2/3. Associations were also inverse for pre-pregnancy use for low-grade gliomas (LGGs) (n = 109) and medulloblastomas/PNETs (n = 47) for any FA versus no FA supplement use (87).

Finally, two ecological studies reported CBT incidence trends in association with mandatory population FA fortification of grain and cereal products. When CBT incidence before and after fortification (1985–1997 to 1998–2006) was compared in a Canadian study, IRRs for children ages 0 to 4 and 5 to 9 years old of 0.95 (95% CI, 0.75–1.19) and 0.91 (95% CI, 0.73–1.13) were found, respectively (88). A similarly designed U.S. study examined CBT incidence patterns from 1986 to 2008 for children diagnosed between 0 and 4 years old by comparing rates for those who were estimated to be in utero before versus after fortification in 1998. On the basis of 573 prefortification and 454 postfortification CBT cases, the authors reported significantly lower incidence rates after fortification versus before for PNETs (IRR, 0.56; 95% CI, 0.37–0.84) and ependymomas (IRR, 0.7; 95% CI, 0.51–0.97) but not other brain tumor types or overall. Trend analyses indicated that the data were consistent with a fortification effect for PNETs but not ependymomas (89).

Studies of rodents and non-human primates have provided evidence that maternal intake of dietary NOCs, particularly N-nitrosamides, induces brain tumors in offspring. However, their contribution to human CBTs is less clear. Direct NOC sources include nitrite-cured and smoked meat, fish, cheese, and beer, whereas vegetables containing nitrites that can undergo conversion to NOCs are an indirect NOC source (reviewed in ref. 90). A 2004 meta-analysis that included 1,226 CBT cases and 1,768 controls from seven studies reported a summary RR of 1.68 (95% CI, 1.30–2.17) for the association between CBTs and maternal cured meat intake during pregnancy versus no intake (91). Since 2004, a study of 315 medulloblastoma/PNET cases and 315 controls reported no overall association between maternal prenatal cured meat intake and offspring CBTs; however, maternal high cured meat intake in combination with low vitamin C intake increased risk (OR, 1.5; 95% CI, 1.0–2.3; P = 0.08; ref. 85). An international case–control study of 1,218 CBT cases and 2,223 controls diagnosed from ages 0 to 19 years old from 1982 to 1992 reported positive associations with ORs ranging from 1.8 to 2.5 across astrocytoma subtypes (92). Finally, one study of 202 cases and 286 controls examined the association between maternal cured meat consumption and CBTs, which was modified by glutathione S-transferase (GST) genotypes involved in NOC inactivation. Increasing risk with increasing frequency of maternal cured meat consumption in children without GSTT1 (OR, 1.29; 95% CI, 1.07–1.57) was reported (93). Altogether, a causal connection between maternal intake of NOCs and CBTs is possible; however, other nutrients as well as genetic factors may modify risk.

Alcohol exposure in utero is a known toxin to the developing CNS. However, in agreement with earlier studies (94), a recent large case–control study did not support maternal consumption as a risk factor for offspring CBTs (95).

The relation between maternal tobacco exposure and CBTs has been studied previously in case–control studies with no significant associations (96–99). Only one recent prospective Swedish linkage study found an association with maternal smoking for CBTs (benign and malignant tumors, HR, 1.24; 95% CI, 1.01–1.53; ref. 100). A recent study assessed neonatal blood spots from 202 cases for genetic polymorphisms that metabolize tobacco-smoke chemicals and reported that the EPHX1 H139R polymorphism, one of nine polymorphisms that metabolizes polycyclic aromatic hydrocarbons, had a positive interaction OR for both maternal and paternal smokers with CBTs in offspring (P interaction = 0.02 and 0.10; ref. 101).

Although several studies suggest a causal relation between residential pesticide exposure and CBTs (reviewed in 2007, see ref. 102; see also residential pesticides section below), results from studies of parental occupational exposure are less consistent. These inconsistencies could be due to heterogeneous definition of “child” (ranging from 0 to 30 years old), difficulties in separating parental occupational exposure from residential use generic definitions of pesticides (i.e., “pesticides” instead of a specific compound), and inconsistent definitions of exposure time windows.

In a Northern England cancer registry study (n = 843 CNS tumors; ages 0–24 years old), no association between likely paternal occupational pesticide exposure (at the time of the child's birth) and childhood CNS tumor risk was found when comparing cases with other cancer controls in contrast to population controls where inverse associations were observed for all CNS tumors (OR, 0.44, 95% CI, 0.31–0.63) and astrocytomas (OR, 0.48; 95% CI, 0.27–0.87). However, the inverse associations disappeared when the results were stratified by urban versus rural residence (103).

In the U.S. Atlantic Coast Childhood Brain Cancer Study, 421 case–control pairs <10 years were analyzed (104). A positive association was observed between paternal exposure to herbicides from both residential and occupational sources in the two years before the child's birth and astrocytoma (OR, 1.8; 95% CI, 1.1–3.1) with no evidence of an increased risk for fungicides or insecticides.

The Australian Study of Childhood Brain Tumors (Aus-CBT) included 256 cases and 819 controls ages 0 to 14 years old. Although the authors concluded that exposure to pesticides in preconception as well as during pregnancy is associated with an increased CBT risk, the evidence was less clear for parental occupational exposure specifically. Only 13 fathers were classified as “exposed” to occupational pesticides in the year before pregnancy (OR, 1.36; 95% CI, 0.66–2.80; ref. 105).

In a meta-analysis of this topic including studies published between 1985 and 2008, there was a significant positive association between CBTs and paternal (OR, 1.40; 95% CI, 1.20–1.62) but not maternal occupational pesticide exposure (106). In a second meta-analysis of studies published between 1974 and 2010, a positive association between parents who had potential prenatal occupational pesticide exposure (including farm/agricultural workers, pesticide applicators, pesticide manufacturers, and others such as gardeners, greenhouse workers, etc.) and offspring brain tumors was reported after combining all case–control (summary OR, 1.30; 95% CI, 1.11–1.53) or cohort (summary rate ratio, 1.53; 95% CI, 1.20–1.95) studies (107).

Previous studies examining the possible association between parental occupational ELF exposure in different exposure time windows and CBTs are inconsistent (108). A recent U.K. registry-based case–control study of CNS tumors examined associations with likely paternal occupational exposure to the broad category of radiation or electromagnetic fields, applying an occupational exposure matrix to jobs reported on birth certificates (109). No association was observed with all CNS tumors combined or for ependymomas, astrocytomas, PNETs, other gliomas, or other specified intracranial and intraspinal neoplasms.

A 2009 Canadian study of 548 incident CBTs (ages 0–14 years old) and 760 control subjects assessed potential associations with indicators of maternal occupational ELF exposure based on individualized exposure estimates or a job exposure matrix (JEM) applied to job history information collected during interview (110). Positive associations between average ELF exposure ≥90th percentile (0.30 μT) but not cumulative or peak exposure in the two-year period before conception were observed for CBTs overall (OR, 1.4; 95% CI, 1.0–2.1) and astroglial tumors specifically (OR, 1.5; 95% CI, 1.0–2.4). Positive associations between average ELF exposure ≥90th percentile (0.28 μT) during the pregnancy period were also observed for all CBTs (OR, 1.5; 95% CI, 1.1–2.2) and astroglial tumors (OR, 1.6; 95% CI, 1.1–2.5).

A German case–control study of 444 child CNS tumors (0–14 years) from the German Childhood Cancer Registry and 444 controls recruited through resident registration office files examined associations with preconceptional parental occupational exposure to ELF estimated using a JEM applied to lifetime occupational histories collected during interview (111). No clear association was observed with preconceptional paternal occupational ELF exposure >0.2 μT (OR, 1.06; 95% CI, 0.84–1.34) or >1 μT (OR, 1.19; 95% CI, 0.81–1.75). Similarly, no association was observed with preconceptional maternal occupational ELF exposure >0.2 μT (OR, 0.88; 95% CI, 0.58–1.33).

In recent analyses, a Taiwanese case–control study including 74 incident brain tumor cases <30 years old and 170 controls reported preliminary results that CBTs were associated with maternal preconceptional occupations for electronic parts and component manufacturing (OR, 11.81; 95% CI, 1.20–116.3) and the textile and garment industry (OR, 7.25, 95% CI, 1.18–31.0; ref. 112). In contrast, analyses of the complete study of 202 CBT cases and 501 controls revealed no associations with parental (paternal or maternal) or personal occupation or industry for brain tumor risk overall, or with glioma specifically, nor with parental exposure to petrochemicals. A complete list of the industries/occupations studies is not included in the article. Selected industries/occupations studied included agriculture, forestry, fishing; electricity, gas, and water; clerks; plant and machine operators and assemblers; construction; and craft and related trades workers (113).

An Australian population-based case–control study examined associations between parental occupational exposure to engine exhausts and brain tumors in children ages 0 to 14 years (114). A total of 306 CBT cases were examined with 950 controls. Estimates of paternal and maternal exposure to diesel, petrol, and other exposures one and two years before birth and the year following birth were inferred based on a decision-rule approach. Significant positive associations for maternal (OR, 2.03; 95% CI, 1.09–3.81) and paternal (OR, 1.38; 95% CI, 1.02–1.86) diesel exhaust exposure any time before birth were observed. Paternal occupational exposures to petrol and other exhausts were also studied; however, no significant association at any time before birth was found (OR, 1.31; 95% CI, 0.92–1.87). No associations between CBTs and occupational exposures to other exhausts were observed. Associations between CBTs and paternal occupational paint exposure were also examined with small positive associations reported for any paternal exposure before the child's birth (OR, 1.32; 95% CI, 0.90–1.92) or in the conception year (OR, 1.25; 95% CI, 0.69–2.28; ref. 115).

In a case–control study of 1,218 CBT cases (0–19 years) and 2,223 controls in seven countries, there was a significant positive association between paternal preconceptional occupational exposure to polycyclic aromatic hydrocarbons (PAH) and risk of all childhood brain (OR, 1.3; 95% CI, 1.1–1.6) and astroglial tumors (OR, 1.4; 95% CI, 1.1–1.7; ref. 116). These results are generally consistent with previous studies on parental occupational PAH exposure (117).

A South Korean study that linked birth to death records from 1995 to 2004 classified parental occupation (listed on birth certificate) as nonmanual (e.g., legislators, professionals, office workers), manual (e.g., skilled agricultural, forestry and fishery workers, craft and affiliated craft workers, device and machine operators), and economically inactive (e.g., students, homemakers, soldiers). For CNS tumors, the HRs for paternal and maternal occupation were not significant (comparing manual and inactive work with nonmanual work). Likewise, when stratified by paternal education level, paternal occupation was not associated with childhood CNS tumors (118).

Lastly, a case–control study of 11,119 CNS tumors diagnosed at ages <15 years from the Great Britain National Registry of Childhood Tumors and 11,039 matched controls from birth records examined potential associations with 33 paternal occupational exposures as well as social class based on job title recorded at birth (119). Significant positive associations were observed with “definite” paternal occupational exposure to animals (OR, 1.40; 95% CI, 1.01–1.96) and lead (OR, 1.18; 95% CI, 1.01–1.39) and a significant inverse association with metal working (oil mists; OR, 0.87; 95% CI, 0.75–0.99). There was also a significant inverse association with paternal social class. Associations were also observed with paternal occupational exposures and specific CNS subtypes; however, they were somewhat sensitive to adjustment for paternal social class.

Pesticides are designed to act on the nervous system with some being carcinogenic in animal models (120). Numerous studies have investigated the potential impact of use of household pesticides, pest extermination services on CBT development (117, 121). Overall, these studies suggest positive associations with CBTs. In a meta-analysis, increased CBT risks in association with paternal pesticide use (herbicides, insecticides, and fungicides) in the home or garden during the prenatal (OR, 1.48; 95% CI, 1.22–1.80) or postnatal (OR, 1.66; 95% CI, 1.11–2.49) periods were reported (106). A U.S. case–control study found that parental use of garden or lawn herbicides was significantly associated with childhood astrocytoma (RR, 1.9; 95% CI, 1.2–3.0) for children ages ≤10 years old but not PNETs and medulloblastomas (104). However, another case–control study of PNET/medulloblastoma reported an association with use of pesticides on the lawn during pregnancy (OR, 1.6; 95% CI, 1.0–2.5) or childhood (OR, 1.8; 95% CI, 1.2–2.8) for children ages ≤6 years old at diagnosis (122).

Finally, a case–control study investigated genetic variation in the paraoxonase (PON1) gene that is involved in the organophosphorus insecticide metabolism and CBT risk. CBT risk increased with each PON1-108T allele the child carried in children exposed to residential insecticides (OR, 2.6; 95% CI, 1.2–5.5) but not among unexposed children (OR, 0.9; 95% CI, 0.5–1.6; ref. 123). Research in an expanded study population also found the association with PON1 (OR, 1.8; 95% CI, 1.1–3.0) and FMO1-9536A (*6) allele (OR, 2.7; 95% CI, 1.2–5.9) among pesticide-exposed children exclusively (124).

Established CBT risk factors remain limited to ionizing radiation exposure and certain cancer syndromes. However, accumulating evidence suggests relatively consistent support from larger studies and meta-analyses for positive associations between advanced parental age, birth defects, markers of fetal growth, CT scans, maternal dietary NOCs, and residential pesticide exposure (summarized in Table 5 and Fig. 1).

A priority area for future CBT epidemiologic research is the elucidation of both common and rare genetic risk variants that modify risk. Although it is well established that certain genetic syndromes strongly increase CBT risk, no genome-wide association studies that identify common risk variants were published at the time of this review. Identification of common genetic loci as well as potential parental genetic loci that modify CBT risk overall and by subtype will inform CBT biology. Identification of rare germline CBT risk variants through genome sequencing studies will also be an important future research priority.

With the exception of higher doses of ionizing radiation, no definitive environmental risk factors for CBT development exist. However, our review suggests that maternal dietary intake of NOCs, prenatal vitamin supplementation, and residential pesticide exposure may increase risk. Inherent limitations of case–control studies, including small sample sizes, survey measurement error, and selection bias, make it difficult to reach definitive conclusions. Finally, emerging evidence from two administrative data analyses suggests that exposure to CT scans increases brain tumor risk, emphasizing the importance of minimizing radiation exposure from diagnostic tests to the extent possible in children to mitigate cancer risk.

In summary, it is likely that the greatest gains in understanding of CBT etiology in the near future will come from genomic studies that identify genetic factors that modify CBT risk overall and by subtype. It will also be important to identify interactions between genetic and environmental factors and to conduct studies that integrate germline and somatic tumor sequence data to determine how germline variation influences tumor mutation profiles and prognosis. For progress in these areas to occur, a coordinated investment in systematic collection of clinically annotated biospecimens (both tumor and normal) from a large number of CBT cases should be an international priority because cancer is a leading cause of death in children and CBTs have the highest cancer mortality rate among childhood cancers (125). In the United States alone, there are >2,000 children diagnosed each year with brain tumors, representing a substantial population that could be approached for research participation during clinic visits where recruitment success has been shown to be higher (126). A clear need also exists for increased international coordination to make samples available through standardized processes to all researchers with meritorious proposals. A step forward in this direction was recently achieved through funding of a biospecimen bank that will store samples from children diagnosed with cancer at a Children's Oncology Group Institution that includes 220 centers in North America, Saudia Arabia, Australia, New Zealand, and Europe (127).

Jennifer Cullen is Director, Epidemiologic Research, Center for Prostate Disease Research. No potential conflicts of interest were disclosed by the other authors.

All of the co-authors began the collaboration for this paper at the 2013 meeting of the Brain Tumor Epidemiology Consortium (BTEC) in Chicago.

M.C. Turner was funded by a Government of Canada Banting Postdoctoral Fellowship. Q.T. Ostrom and J.S. Barnholtz-Sloan were supported in part by the Case Comprehensive Cancer Center Support Grant (NIH/NCI P30 CA043703).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.