Pediatric cancer risk in association with birth defects: A systematic review

Background Many epidemiological studies have examined associations between birth defects (BDs) and pediatric malignancy over the past several decades. Our objective was to conduct a systematic literature review of studies reporting on this association. Methods We used librarian-designed searches of the PubMed Medline and Embase databases to identify primary research articles on pediatric neoplasms and BDs. English language articles from PubMed and Embase up to 10/12/2015, and in PubMed up to 5/12/2017 following an updated search, were eligible for inclusion if they reported primary epidemiological research results on associations between BDs and pediatric malignancies. Two reviewers coded each article based on the title and abstract to identify eligible articles that were abstracted using a structured form. Additional articles were identified through reference lists and other sources. Results were synthesized for pediatric cancers overall and for nine major pediatric cancer subtypes. Results A total of 14,778 article citations were identified, of which 80 met inclusion criteria. Pediatric cancer risk was increased in most studies in association with BDs overall with some notable specific findings, including increased risks for CNS tumors in association with CNS abnormalities and positive associations between rib anomalies and several pediatric cancer types. Conclusions Some children born with BDs may be at increased risk for specific pediatric malignancy types. This work provides a foundation for future investigations that are needed to clarify specific BD types predisposing toward malignancy and possible underlying causes of both BDs and malignancy.


Methods
We used librarian-designed searches of the PubMed Medline and Embase databases to identify primary research articles on pediatric neoplasms and BDs. English language articles from PubMed and Embase up to 10/12/2015, and in PubMed up to 5/12/2017 following an updated search, were eligible for inclusion if they reported primary epidemiological research results on associations between BDs and pediatric malignancies. Two reviewers coded each article based on the title and abstract to identify eligible articles that were abstracted using a structured form. Additional articles were identified through reference lists and other sources. Results were synthesized for pediatric cancers overall and for nine major pediatric cancer subtypes.

Results
A total of 14,778 article citations were identified, of which 80 met inclusion criteria. Pediatric cancer risk was increased in most studies in association with BDs overall with some notable specific findings, including increased risks for CNS tumors in association with CNS abnormalities and positive associations between rib anomalies and several pediatric cancer types.

Conclusions
Some children born with BDs may be at increased risk for specific pediatric malignancy types. This work provides a foundation for future investigations that are needed to clarify PLOS

Introduction
Pediatric cancer is diagnosed in >14,000 U.S. children per year from the ages of 0-19 years and is the leading cause of disease-related death among children aged 1-14 years [1,2]. Although a few risk factors have been conclusively identified, including exposure to high dose radiation and certain genetic syndromes, the etiology underlying most cases remains unknown. Evidence accumulated over several decades suggests positive associations between birth defects (BDs) and pediatric malignancy. BDs affect~1 in 33 U.S. children and are defined as "structural changes present at birth that can affect almost any part or parts of the body" [3]. BDs are often categorized as major and minor with major anomalies generally considered those having "an adverse effect on the individual's health, functioning or social acceptability" and minor anomalies considered those having "limited social or medical significance". Major anomaly examples include spina bifida, cleft lip palate, and Down Syndrome [4]. Minor anomaly examples include low set ears, epicanthal folds, and simian crease [5]. Although certain genetic syndromes are known to increase pediatric cancer risk (e.g. Down Syndrome and leukemia), other BDs (including major and minor), independent of known cancer predisposition syndromes, may also be associated with an increased risk.
Recent evidence suggests that 8.5% of children with cancer harbor germline mutations in well-known cancer genes that predispose them to the development of early-onset malignancy [6]. However, when considering a more broad predisposition definition including family history, co-morbidities, and types of pediatric cancer, up to about a third of children with cancer may have a genetic predisposition [7]. Many pediatric cancer predisposition syndromes involve defects in normal development and many pediatric cancers arise from immature cell types (e.g. the "blastomas"). Thus, identifying and understanding connections between abnormal fetal or childhood development and the risk of developing cancer will have implications for surveillance, prognosis, risk stratification and potential personalized therapeutics.
To summarize evidence on associations between pediatric malignancy and BDs, we conducted a systematic literature review to identify articles reporting primary human epidemiological research. This work provides a foundation for future studies and may help to identify high risk populations for malignancy among individuals with BDs that will enable improved surveillance, mechanistic research, targeted treatment and outcomes.
author conducted a few PubMed literature searches based on her expert opinion to identify additional qualifying articles.

Eligibility criteria and review process
The citation lists obtained from each database search were evaluated for initial eligibility by at least two coauthors (KA, JL, and KJ). We excluded review articles, editorial commentaries, meeting abstracts, articles focused solely on syndromes with known genetic etiologies, on treatment or clinical outcomes, solely on adults (! 18 years), and articles not providing information on risks specific to children with the exception of the Gelberg et al. study that reported risks for osteosarcoma from 0-24 years [12]. In addition, lab-based studies and case reports/series without external comparison groups were excluded. Reviewers classified each article initially based on title review (and abstract if eligibility was unclear from the title) as: 1) eligible, 2) unclear eligibility, and 3) ineligible. Following reconciliation of articles with coding disagreement by the senior author and reviewers, articles assigned to category one were abstracted.

Data collection
Data were abstracted from each article using a pre-designed form that captured: study population, study design and description, sources of information on BDs and cancer, study inclusion and exclusion criteria, age groups studied, birth years, cancer diagnosis years, BD case definition, BD types examined, cancer types examined, subject numbers, analysis method, measure of association reported, and key findings. Any additional information relevant to interpretation of study results was captured in a comment field. To the extent possible, we abstracted the number of subjects for each comparison group and included this information where the numbers were reported directly or required minimum assumptions to calculate. Any articles that were identified as ineligible during the abstraction phase were subsequently excluded. During our update of the review, we also abstracted information where reported on classification of major and minor anomalies, maximum age at which BDs were ascertained, and cancer risks by age. We did not attempt to reclassify anomalies from the original reports for summary purposes.

Quality assessment
We employed a modified Newcastle-Ottawa Scale (NOS), designed for quality assessment of observational studies [13], that uses a star assignment system to determine overall quality for case-control, nested case-control, case-cohort, and cohort studies. Case series studies with external comparison groups were not evaluated. Evaluation criteria are described briefly below.
Case-control, nested case-control, and case-cohort studies. A) Selection. Up to four stars were awarded to studies based on: 1) adequate case definition, 2) representativeness of cases, 3) control selection, and 4) adequate control definition. A case definition was considered adequate if cases were defined based on either clinical and/or histopathological validation of their cancer diagnosis or were identified through cancer registry data. Cancer cases were considered representative if they were community/population-based versus hospital-based. For control selection, a star was given if controls were selected from the same population as cases. The control definition was considered adequate if the authors provided information to indicate that controls did not have a history of the outcome. B) Comparability between the groups. Up to two stars were awarded to studies matching on or adjusting for maternal age and child's sex, potential confounders of the association between pediatric malignancy and BDs. C) Ascertainment of the exposure. Up to three stars were awarded to studies if: 1) BDs were ascertained through medical records, physical exam, or birth certificate data, 2) they used the same method of BD ascertainment for cases and controls, and 3) they reported a similar non-response rate for both cases and controls (<10% difference) or the study used registry data for BD ascertainment.
Cohort studies. A) Selection. Up to four stars were awarded to studies based on: 1) representativeness of the exposed cohort, 2) selection of the non-exposed cohort, 3) ascertainment of exposure, and 4) demonstration that the outcome of interest was not present at the start of the study. Studies were given one star each if the exposed cohort registry/population included all children from a defined geographic region was representative of the community, if the nonexposed cohort was drawn from the same community as the exposed cohort, if medical records or BD registry information was used to identify exposed subjects or if the exposure was based on birth certificate data, and if it was indicated that cancer was not present at the start of the study.
B) Comparability. Comparability criteria were the same as for case-control studies. C) Outcome. Up to two stars were awarded based on: 1) assessment of the outcome and 2) follow-up length. One star was awarded if the outcome was assessed independently or through medical records or record linkage (i.e. through administrative data through ICD codes). One star was given if the follow-up period was long enough for outcomes to occur (we set this at ! 6 years with consideration for the age distribution of pediatric cancer). Table 2. Characteristics of included studies on the association between birth defects and childhood cancer.

Reference (location)
BD ascertainment Pediatric cancer case ascertainment: cancer types, age group, diagnosis dates/years, case identification source

Comparison group ascertainment
Case-control studies Stewart et al., 1958 [18] (England and Wales) Maternal interview Overall, 0-9 years, 1953-1955 (died of cancer), Registrar General Controls were selected from birth registers matched to cases on age, sex, and locality. Ager et al., 1965 [39] (Minnesota, U.S.) Maternal interview/medical record verification. The authors note that "lesser conditions, such as nevi, were not recorded".
Leukemia, 0-4 years, 1953-1957 (died of leukemia), Minnesota death certificates Two control groups were included with a total of two controls per case. If available, sex-matched sibling controls were included with the birthdate closest to the index child and who had reached an age at the time of interview that was the same or older as the age of the index child when they died. Neighborhood controls were matched on sex and birth date within one year. Swerdlow et al., 1982 [91] (United Kingdom) The

Renal tumors nested case-control studies
Lindblad et al., 1992 [81] 110 (NR) 550 (NR) WT Any BD NR The authors note that reported BD associations were not confirmed.    (Continued) generally weaker associations for older children and when the follow-up period started at the time of BD diagnosis [38].

Leukemia (Table 3)
We identified 36 articles reporting findings for associations between leukemia and hematological malignancies and BDs.
Cohort studies (n = 11). In six studies, positive associations were observed between leukemias and any BDs/major BDs ranging from 1.1-21.97 [27][28][29][30][31]33]. Associations were generally weaker in studies/analyses with DS cases were excluded [27-29, 31, 32, 34-36]. Rankin et al. noted, however, that "the association between congenital anomalies and childhood leukemia remained after exclusion of Down Syndrome cases" [30]. Notably, Dawson et al. reported a positive association between other leukemias (types other than ALL or AML) and non-chromosomal BDs [34]. Finally, Sun et al. examined nervous and circulatory system BDs and risk of lymphatic and haematopoietic tissue malignancies in children without chromosomal anomalies and reported positive associations that were stronger for infants versus children 1-15 years. Associations were generally weaker when follow-up time was counted from the BD diagnosis rather than from birth [38].
Case series with an external comparison group studies (n = 1). Narod et al. examined a number of specific abnormalities and did not report any consistently increased risks for leukemia [53].

Lymphoma (Table 3)
We identified 17 articles reporting findings for associations between lymphoma and BDs.
Cohort studies (n = 7). Lymphoma was positively associated with any BD in three of four cohort studies [30,31,33,35]. Fisher et al. and Janitz et al. reported positive associations between non-chromosomal anomalies and lymphoma [35,36], while both Botto et al. and Dawson et al. observed risks for lymphoma/lymphoma subtypes in both directions in children with non-chromosomal BDs and those not known to be related to cancer [32,34]. Certain musculoskeletal anomalies were associated with lymphoma in one study [33].
Case series with an external comparison group studies (n = 1). Narod et al. reported findings for cardiac septal defects, genitourinary abnormalities, and spine and rib abnormalities with a significant positive association for spine and rib abnormalities for the British Columbia registry comparison group only [53]. (Table 3) We identified 31 articles reporting findings for associations between CNS tumors and BDs.

CNS and miscellaneous intracranial and intraspinal neoplasms
Case-control and nested case-control studies (n = 19). Risk estimates for associations between CNS tumors and any BDs/major BDs ranged from 1.0 to 4.7 [15][16][17][55][56][57][58][59][60]. Two studies did not report risk estimates; however, one reported no association and the other reported a decreased frequency of BDs (excluding minor BDs) in cases versus controls [61,62]. Altmann et al. reported an increased odds of astrocytoma in children with any BD [17]. However, three studies [58,59,63] found weak to no evidence that children with any BD/ major BDs have an increased risk for specific CNS subtypes with the exception of other gliomas [58] where a 2-3 fold increased risk was reported. Importantly, these results were unchanged after excluding seven CNS tumor cases with Neurofibromatosis Type 1 [58]. Partap et al.'s results contrast with these findings by showing increased odds of a BD history for MBs and PNETs compared to controls and inverse associations for gliomas [57].
Gold et al. reported a strong positive association only between brain tumors and club foot [64]. Another study reported universally positive associations for hand, foot, eye, nose, mouth, and ear minor BDs [20]. Birthmarks/deformities were not associated with brain tumors in one study [65]. Altmann et al. reported highly significant and strong associations between CNS tumors and nervous system and eye/face/neck abnormalities [17]. Finally, several investigators reported that brain tumor cases with varying subtypes had higher odds of rib abnormalities than controls [23][24][25][26].
Cohort studies (n = 11). Six studies reported positive associations for any BD and CNS tumors ranging from 1.11 to 2.9 [27][28][29][30][31]33]. Fisher et al. reported increased HRs for CNS tumors in both children with and without chromosomal BDs of 1.87 and 1.80 [36]. Botto et al. reported an increased brain tumor incidence in children with structural BDs that was stronger for other brain tumors [32]. Dawson et al. reported a weak imprecise association of 1.26 between BDs not known to be related to cancer and CNS tumors [34]. Janitz et al. reported increased risks in children with non-chromosomal BDs that decreased with increasing age [35].
For specific BD types, nervous system abnormalities were strongly associated with CNS tumors in three studies [27,33,38]. Finally, circulatory system malformations were positively associated with CNS tumors in one study [38].
Case series with an external comparison group studies (n = 1). Narod et al. reported inconsistent or inverse associations for total BD and most specific abnormalities with the exception of hydrocephalus where both RRs were positive and significant [53].
Neuroblastoma and other peripheral nervous cell tumors (Table 3) We identified 26 articles reporting findings for associations between NB and BDs.
Cohort and case-cohort studies (n = 11). Children with any BD had a consistently higher risk of NB across studies with RRs from 1.1-20.3 [27-31, 33, 75]. Fisher et al. and Botto et al. [32,36] also reported that non-chromosomal BDs were associated with NB and other peripheral nervous system tumors with individuals in the BD group having a >2-fold higher risk, while Dawson et al. reported a lower risk estimate of 1.41 for individuals with BDs not known to be related to cancer [34]. Finally, a case-cohort study reported a similar percentage of cases and sub-cohort members had BDs recorded in their birth records [76]. For specific anomalies, Agha et al. reported more observed than expected cases of SNS tumors in those with other anomalies of the digestive system [33].
Case series with an external comparison group studies (n = 1). Narod et al. reported results that were imprecise and in opposite directions for total BD. Generally positive associations were observed for specific BDs, particularly for cardiac and gastrointestinal abnormalities [53].

Retinoblastoma and eye tumors (Table 3)
We identified 10 articles reporting findings for associations between RB/eye tumors and BDs.
Case-control studies (n = 2). Results were mixed with Mann et al. finding no BDs in RB cases and Altmann et al. reporting a strong positive OR of 15.0 [16,17]. For specific BDs, strong associations were reported for chromosomal anomalies in a single study.
Case series with an external comparison group studies (n = 1). Narod et al. reported inconsistent results for associations between RB and BDs and strong associations between RB and cataracts and ventricular septal defects [53]. (Table 3) We identified 21 articles reporting findings for associations between renal tumors and BDs.

Renal tumors
Case-control and nested case-control studies (n = 10). Wilkins et al. reported similar frequencies of any BD in WT cases and controls [77], while two other groups reported positive associations [16,17]. Stronger associations were reported for non-WT associated BDs than WT-associated BDs in one study [78]. Two studies reported significant positive associations between WT and spina bifida [79,80]. WT was also reported as positively associated with eye/ face/neck BDs and chromosomal BDs [17]. WT/renal carcinomas were positively associated with rib abnormalities in three studies [23,24,26]. Finally, Lindblad et al. reported that associations between BDs and WT were not confirmed in a nested case-control study [81].
Cohort and case-cohort studies (n = 9). Risk estimates for cohort and case-cohort studies examining associations between any BD and WT or renal tumors ranged from 1.0-3.2 [27,28,30,31,33,82]. Fisher et al. reported significant risks for WT only in children with chromosomal BDs [36]. In two studies excluding children with chromosomal anomalies and syndromes known to predispose toward cancer, results were mixed [32,34].
Case series with an external comparison group studies (n = 2). Breslow et al. compared the rate of different BD types in a case series to two external comparison groups and observed consistent strongly increased rates of aniridia, double-collecting system, and hemihypertrophy [83]. Narod et al. reported consistent findings across the two different comparison groups for genitourinary and reproductive organ associated BDs. Renal carcinomas were strongly associated with spine and rib malformations in one study [53].
Hepatic tumors (Table 3) We identified 10 articles reporting associations between hepatic tumors and BDs.
Case-control studies (n = 3). There was no significant association between HB and any BDs in one study [16], while another reported a non-significant 9.3-fold increased odds of BDs in hepatic tumor cases versus controls [17]. For specific abnormalities, both spina bifida and genitourinary conditions were positively associated with HB in one study [84]. Other abnormalities were examined only in single studies.
Cohort and case-cohort studies (n = 6). One study reported three hepatic tumor cases in individuals with BDs and none in the comparison group [33]. Carozza et al. reported no association between liver tumors and any BD [31], while Spector et al. reported an almost 6-fold increased risk of HB in association with any BD [85]. Both Botto et al. and Dawson et al. reported strongly increased IRRs for liver tumors in individuals with BDs after exclusion of children with chromosomal BDs and syndromes known to be associated with cancer [32,34]. In the Janitz et al. study, there were too few hepatic tumor cases to examine statistical associations [35]. Digestive and chromosomal anomalies were positively associated with hepatic tumors in one study [33].
Case series with an external comparison group studies (n = 1). Narod et al. reported positive associations between spina bifida and genitourinary abnormalities and HB [53]. (Table 3) We identified 12 articles reporting associations between BDs and bone tumors.

Malignant bone tumors
Case-control studies (n = 7). One study reported no association between OS and any BDs [12]. Two studies reported positive but imprecise associations (ORs = 1.56 and 23.2) between bone cancer and any BD [16,17]. Minor BDs were associated with OS in one study (OR = 12.4) [20]. One study reported a positive association between bone tumors and chromosomal anomalies [17]. Finally, three studies reported inconsistent associations between rib anomalies and bone cancer (OS or ES) [23,24,26].
Cohort studies (n = 4). Mixed results were reported for any BD/BDs not known to be related to cancer and bone cancer in three studies [31,33,34]. Botto et al. reported a two-fold non-significant increased incidence of ES in individuals with non-chromosomal structural BDs [32]. An~4-fold excess of bone tumors was reported in children with other musculoskeletal BDs in one study [33].
Case series with an external comparison group studies (n = 1). Narod et al. reported consistently strong associations using two different comparison groups between bone tumors overall and spina bifida; however, the association was based on only one case. Strong positive associations were also reported between osteodystrophy and cataracts and ES using both comparison groups [53].
Soft tissue and other extraosseous sarcomas (Table 3) We identified 19 articles reporting associations between BDs and soft tissue and other extraosseous sarcomas.
Case-control studies (n = 8). Four studies reported positive associations between any BDs/major BDs and STS/RMS ranging from 1.3 to 7.9 [15][16][17]86]. Minor BDs were observed in 100% of RMS cases and only 35% of controls in one study [20]. Further evidence for a positive association between genitourinary malformations and STS was reported for RMS [17].
Cohort studies (n = 9). Five cohort studies examined associations between any BDs and STS/RMS, with all finding positive associations ranging from 1.9 to 4.1 [29-31, 33, 35]. Three studies reported positive associations between RMS and non-chromosomal (structural) BDs, while one study reported a non-significant inverse for STS for BDs not known to be related to cancer [32,[34][35][36]. Finally Sun et al., reported strong positive associations between nervous system BDs and mesothothelial and soft tissue cancers across age groups and weaker associations for circulatory system BDs [38].
Case series with an external comparison group studies (n = 2). One study reported markedly higher rates of CNS anomalies, upper alimentary tract/digestive systems, cardiopulmonary anomalies, and accessory spleens in the RMS cohort than in the two comparison groups [87]. Narod et al. reported that genitourinary and spine and rib malformations were positively associated with STS but this was inconsistent across comparison groups used for RR calculations. Cardiac septal defects were not significantly associated with RMS [53].
Germ cell tumors (GCT), trophoblastic tumors, and neoplasms of gonads (Table 3) We identified 16 articles reporting associations between BDs and GCTs, trophoblastic tumors, and neoplasms of gonads.
Case-control and nested case-control studies (n = 9). Several studies reported positive associations between any BDs and gonadal tumors/GCTs ranging from 1.1 to 9.12 [16,17,[88][89][90]. Genitourinary defects and inguinal hernias were positively associated with testicular tumors [91]. Johnson et al. reported a strong association with cryptorchidism in males [88]. Merks et al. reported an~3-fold increased odds of cervical anomalies in GCT cases compared to controls [26]. Hall et al. reported an increased odds of ear/face/neck anomalies in GCT cases and those with teratomas specifically [90]. In a nested case-control study, no cases had BDs in the 0-4 year old age group [92].
Cohort studies (n = 6). Among two cohort studies, Agha et al. reported no association between germ cell, trophoblastic and other gonadal carcinoma and any BD, while Carozza et al. reported a significant 5-fold increased risk [31,33]. Fisher et al., Botto et al., and Janitz et al. reported significant positive associations between GCTs/Gonadal and GCT and nonchromosomal BDs, while Dawson et al. reported a weak positive association for BDs not known to be related to cancer [32,[34][35][36]. GCTs were strongly associated with other musculoskeletal anomalies in one study [33].
Case series with an external comparison group studies (n = 1). Narod et al. reported consistent increased risks between gonadal and GCT tumors and total BDs and weak non-significant associations between these tumors and genitourinary defects. Musculoskeletal, spinal, and spine and rib malformations were consistently positively associated with [53].
Other tumors (Table 3) Several studies examined other childhood cancers and various subtype groupings for their association with BDs. For completeness, we include these results; however, given the heterogeneous nature of the BDs/cancer types and outcomes examined, we do not summarize the findings.

Quality assessment (S2 and S3 Tables)
The mean percent total quality point score was higher for cohort (88% ± 13%) than case-control (62% ± 19%) studies. The three case-cohort study reports that were based on the same parent study were of high quality each receiving 89% of the total quality points, while the three nested case-control studies received a mean of~85% ± 6% of the total quality points (data not shown).

Discussion
Associations between BDs and pediatric cancer have been extensively studied. Overall conclusions on pediatric cancer risk in children with BDs are limited by heterogeneity in study design, subject selection (including inclusion/exclusion criteria), measurement, definitions, and length of follow-up for ascertaining BDs, as well as covariate adjustment in models. For example, in studies ascertaining individuals with BDs from BD surveillance systems, standardized definitions using published classification schemes were used to group individuals with BDs according to BD type (e.g. major, minor, and specific BD types), while for studies ascertaining BDs through parental questionnaire, coding schemes were often elusive. We strongly emphasize that future studies on this topic should employ and clearly report in their methods a standardized classification system such as that reported by Rasmussen et al. for the National Birth Defects Prevention Study [93], which will facilitate pooling and meta-analysis studies that are needed to more precisely quantify pediatric cancer risk in children with BDs. In spite of the differences in methodology used by studies on this topic that limit overall conclusions, several noteworthy findings emerged from this review.
An increased risk for pediatric cancer overall in association with BDs clearly exists with most case-control and cohort studies reporting positive associations. A seminal Nordic study of 5.2 million individuals, not included in our review because it did not present pediatric cancer specific risk estimates, linked medical birth and cancer registries to examine cancer risk in children with BDs [94]. Cancer risk was significantly increased in individuals with nonchromosomal BDs by~1.4-1.5 fold, with stronger risks at younger ages, which was also reported in several studies included in this review [34,35,38]. Moreover, a strong association was observed in both Norway and Sweden (>8 fold) between nervous system abnormalities and CNS tumors [94]. Collectively, our review and the Nordic data provide strong evidence for an overall increased cancer risk in children with BDs.
For leukemia, most studies excluding DS cases reported relatively weak or no evidence for an increased risk of leukemia in children with any BD/major BDs. For example, when considering the results from several cohort studies with overall higher quality than case-controls studies, generally weak associations were reported between BDs and leukemia when DS cases were excluded [28,29,32,[34][35][36] with the exception of one study [30]. It is noteworthy that consistent associations with rib anomalies and minor malformations have been reported in several studies [19, 20, 23-26, 50, 52]. These data suggest that minor BDs may be linked to leukemia development but most of the observed positive associations between major BDs and leukemia are likely explained by inclusion of DS cases.
For CNS tumors, strong consistent associations with CNS abnormalities were reported in several studies [17,27,33,38,53]. When interpreting these results, it is important to consider the timing of the BD diagnosis. BDs detected only as a result of tests or procedures associated with the tumor diagnosis may lead to over-ascertainment of BDs in cases compared to controls, a potential issue raised by Altmann et al. [17]. In addition, it is also important to consider whether the abnormality occurred secondary to the tumor, a concern noted by Narod et al. for the association between brain/spinal tumors and hydrocephalus [53]. Evidence against these possibilities was provided by Sun et al. who reported a strong positive association between nervous system abnormalities and CNS tumors when follow-up was initiated at the point of the BD diagnosis (i.e. prior to the cancer diagnosis) [38].
For NB, many studies reported consistent positive associations with BDs overall, although there was inconsistency for specific BD types [17, 26, 28-34, 36, 53, 66-75] with the possible exception of gastrointestinal BDs [17,53,69,71]. It is important to note that some of the observed anomalies were noted as secondary to the tumor [67,71]. Further research is needed to clarify specific abnormalities associated with NB and to confirm the timing of the abnormality as congenital.
For RB/eye tumors, most cohort studies reported consistent positive associations with any BD and structural defects [27,28,[30][31][32][33], some of which may be explained if cases with partial monosomy 13q were included that is associated with a number of anomalies and an increased RB risk [95]. Since~40% of RB tumors arise from a germline mutation [96], there is biological plausibility for developmental abnormalities associated with RB haploinsufficiency. However, despite mouse studies demonstrating lethality for embryos with Rb1-/-genotypes as well as neural and hematopoietic abnormalities; heterozygotes did not have any observable defects [97].
WT was associated with a number of abnormalities, some of which are due to known syndromic causes of WT including Beckwith-Wiedemann and WAGR (Wilms tumor, Aniridia, Genitourinary anomalies and intellectual disability) syndromes. WAGR syndrome is caused by a chromosome 11p deletion and is associated with genitourinary abnormalities and aniridia (absence of the colored part of the iris) [98]. In two recent cohort studies excluding children with chromosomal anomalies, associations between BDs and WT were weak and imprecise [32,36]. It is interesting to note, however, that a greater frequency of non-WT associated BDs were observed in children with WT versus controls in one study [78], a finding that could suggest an underlying genetic predisposition.
For liver cancer, where most cases are HBs, most studies provided evidence of a positive association with any BD [17,32,34,85] that may stem from genitourinary abnormalities that were positively associated with HB in three different study populations [53,84]. It is unclear if inclusion of Beckwith-Wiedemann Syndrome or Familial Adenomatous Polyposis cases in some of studies explains the observed associations.
Although a three studies detected positive associations between BDs (especially for bone BDs) and bone tumors or bone tumor subtypes [16,17,53], cohort study results are mixed and based on small numbers [31][32][33][34]. For STSs, increased risks in children with BDs are evident from most case-control and cohort studies [15-17, 20, 29-33, 35, 36, 86]. However, as with other cancer types, conclusions about particular abnormalities associated with risk are limited by the heterogeneous nature of BDs observed/assessed in different studies.
For germ cell and other gonadal tumors, there is a relatively consistent pattern of an increased risk in association with BDs across studies [16, 17, 31, 32, 34-36, 53, 88, 89] that could be partially be due to cryptorchidism but only one study provided risk estimates in males [88].
Although not completely consistent, one of the most intriguing findings is the observed association between rib anomalies and many different childhood cancer types [23][24][25][26]. Although some of these associations could be due to detection bias as a result of diagnostic tests for malignancy, the Zierhut et al. study was not subject to this type of bias [24].
The biology underlying associations between pediatric cancer and BDs is poorly understood. If the observed statistical associations are causal, a leading theory is that a common genetic abnormality impairing normal development may subsequently predispose toward both BDs and malignancy [99][100][101][102][103]. Since family history is absent in most individuals with non-syndromic BDs, genetic aberrations could arise through de novo mutations in the parental germline or through somatic mutations or epimutations arising early in development. For example, Beckwith-Wiedemann syndrome, can result from genetic and epigenetic mosaicism accompanied by a variably increased risk of malignancy (typically HB or WT) in the affected tissues depending on the causative genetic lesion [104]. The striking number of studies showing an association between rib abnormalities and pediatric/adolescent cancer risk may provide an important clue to underlying genetic commonalities. The finding has frequently been attributed to possible mutations in homeobox genes [25,53] but no genetic cause has been identified so far explaining this statistical observation.
Our study is the first systematic and most comprehensive review on this topic to date. However, there are limitations to consider. At the expense of specificity, we used broad search terms to capture articles examining associations between BDs and pediatric cancers, identifying >14,000 articles. To be as comprehensive as possible, we identified additional relevant articles through review of article references lists, IARC's review [11], and PubMed searches. Despite these efforts, it is still possible that we missed some eligible articles; however, it seems unlikely that they are systematically missing and therefore their exclusion should not bias our overall conclusions. In addition, we did not include gray literature [105] (e.g. meeting abstracts and PhD theses) and therefore publication bias may influence general conclusions.

Conclusions and future directions
Clear positive associations exist between BDs and pediatric cancer with evidence for increased risks for specific cancer/BD type combinations such as CNS abnormalities and CNS cancer, rib anomalies and a number of cancer types, and genitourinary abnormalities and HB. With advances in mutation detection through next generation sequencing technologies it may be possible to identify genetic causes underlying some of these cases, which will provide insight into overlaps between genes impacting both development and malignancy and provide a basis for identification of high risk populations among children with congenital abnormalities.