Congenital anomalies (i.e., birth defects) are one of the strongest and most consistent risk factors for childhood cancer. Birth defects are generally categorized as chromosomal or non-chromosomal anomalies.[1] The role of chromosomal anomalies on childhood cancer risk has been described. For example, children with Down syndrome (DS) have a 20-fold increased risk of acute lymphoblastic leukemia (ALL) compared to those without DS.[2, 3] Similarly, children with chromosome 13q14 deletion syndrome, characterized by dysmorphic facial features, have increased risk of retinoblastoma.[4] Four recent population-based registry linkage studies in the United States (U.S.)[2, 5–7] suggest that children with non-chromosomal anomalies may also be more likely to develop cancer compared to their unaffected contemporaries.

Evidence of shared biological pathways for congenital anomalies and cancer is limited, but possible mechanisms proposed include non-genetic exposures (e.g., environmental exposures) that lead to both conditions;[2] somatic mutations in developmental genes early in embryogenesis leading to tissue mosaicism;[8] or chromosomal microdeletions that include both developmental and cancer predisposition genes.[7] The biological underpinnings of these associations are likely to vary by specific birth defect and specific cancer type.

Few studies have evaluated possible associations of specific non-chromosomal anomalies with specific cancer types, largely due to the rarity of both childhood cancer and congenital anomalies. Relatively large study sizes can be conducted in different geographic regions using population-based linked health registry data allowing uniform measurement of both the congenital anomaly and cancer incidence. Such large linked databases provide rich opportunities to examine the associations of specific anomalies, particularly those that are not chromosomal in origin, with specific cancers. Using linked population-based birth-cancer-registry-hospital discharge data from Washington State in a case-control epidemiological study, we examined the relationships of congenital anomalies with childhood cancers, with a focus on major non-chromosomal anomalies.

Childhood cancer cases were more likely than controls to have mothers aged 35 years or older, to be white, or to have a birthweight >4000g (Table 1). The most common types of cancer were leukemia (28%), central nervous system (CNS) tumors (22%), and lymphoma (11%).

A greater proportion of cases (7%) than controls (5%) had at least one major congenital anomaly identified (OR: 1.46, 95% CI 1.28–1.65) (Table 2). The presence of any minor anomaly in the absence of a major anomaly (OR: 0.52, 95% CI 0.24–1.10) or an unspecified anomaly that could not be classified as major or minor (OR: 1.01, 95% CI 0.90–1.14) did not differ markedly in cases and controls. The ORs for childhood cancer increased with increasing numbers of major anomalies, from 1.35 (95% CI 1.17–1.55) for a single anomaly to 2.79 (95% CI 1.44–5.43) for 3 or more anomalies. When only non-chromosomal anomalies were considered in relation to any cancer, the OR remained increased (OR: 1.35, 95% CI 1.18–1.54). A similar pattern was observed after restriction of analyses to children with normal birthweight (2500 -<4000g) with term (37+ weeks) deliveries.

Increased ORs for childhood cancer were observed for all anomalies, with the exception of oral clefts (OR: 0.56, 95% CI 0.20–1.52), club foot (OR: 0.77, 95% CI: 0.18–3.22), dactyly (OR: 0.99, 95% CI: 0.23–4.20), spina bifida (OR: 0.84, 95% CI: 0.11–6.38), other CNS anomalies (OR: 0.37; 95% CI: 0.04–3.54), and other circulatory anomalies (OR: 0.89, 95% CI: 0.41–1.93). The effect sizes varied: the greatest OR was observed for chromosomal anomalies (OR: 7.52, 95% CI 5.21–10.84). Large and positive ORs were also associated with gastrointestinal anomalies (OR: 3.07, 95% CI 1.85–5.11) and CNS anomalies (OR 2.99, 95% CI: 1.71–5.19). Within general anomalies, selected specific conditions had increased ORs, including microcephalus (OR: 6.64, 95% CI 1.94–22.75), hydrocephalus (OR: 3.95, 95% CI 1.45–10.74), anal atresia (OR: 4.75, 95% CI 1.49–15.19), and Down syndrome (OR: 10.86, 95% CI 7.02–16.81).

ORs were increased for the association of anomalies in relation to childhood cancer diagnosed in all age groups, although the magnitudes of the associations were greatest for cancers diagnosed at <5 years of age (Table 3). Modestly increased ORs with CIs including one were noted for cancer diagnosed between 5–9 years of age, although statistically significant associations were noted for cancers diagnosed in the older (10–19 years) age group.

The presence of a chromosomal anomaly was generally associated with greater ORs for most types of cancer than was the presence of non-chromosomal anomalies (Fig 1). Non-chromosomal anomalies were associated with greater than two-fold increased risk of hepatoblastoma (OR: 2.50, 95% CI 1.13–5.53) and germ cell tumors (OR: 2.38, 95% CI 1.41–4.03), but also with increased risk for neuroblastoma (OR: 1.93, 95% CI 1.32–2.83) and soft-tissue sarcomas (OR: 1.71, 95% CI 1.10–2.65). The presence of a chromosomal anomaly was associated with large increased risk for leukemia (OR: 21.65, 95% CI: 14.57–32.15), retinoblastoma (OR: 14.30, 95% CI: 4.38–46.72), and renal tumors (OR: 4.70, 95% CI 1.14–19.45). Increased ORs for all other cancer types examined except CNS tumors in relation to chromosomal anomalies were also observed, although the estimates were imprecise and confidence intervals included one.

We explored the associations of specific anomaly types in relation to specific types of childhood cancer (Fig 2). The largest ORs were observed for the presence of chromosomal anomalies in relation to leukemia (OR: 21.65, 95% CI 14.57–32.15) and retinoblastoma (OR: 14.30, 95% CI 4.38–46.72), and for the presence of gastrointestinal anomalies in relation to soft-tissue sarcoma (OR: 12.17, 95% CI 4.86–30.46). CNS tumors were associated with CNS anomalies (OR: 6.05, 95% CI 2.75–13.27) but not with other anomalies. Most other non-chromosomal anomalies were associated with several types of cancer. The presence of a gastrointestinal anomaly was associated with increased ORs for germ cell, leukemia, neuroblastoma, and soft-tissue sarcoma. Heart anomalies were associated with hepatoblastoma, neuroblastoma, and other unspecified malignancies.

Congenital anomalies have been associated with childhood cancer in several prior studies. Our observed overall increased risk for cancer in relation to congenital anomalies is consistent with results of other U.S. population-based data linkage studies based on data from California,[6] Texas,[2] Oklahoma,[5] and a pooled analysis of data from three other states (Utah, Arizona, and Iowa, i.e., UTAZIA).[7] Similar results have been reported in population-based health registry studies in Australia,[15] Canada,[16] the United Kingdom,[17] and Norway and Sweden.[18] While the association of chromosomal anomalies with childhood cancer occurrence has been fairly well established in previous population-based data linkage studies, with population-based studies reporting estimates of >10-fold increased risk.[2, 6, 7], the majority of anomalies are non-chromosomal in origin. Our study lends to the growing body of evidence that non-chromosomal anomalies are also associated with childhood cancer risk.[2, 5–7] Additionally, our results indicate that an increasing number of non-chromosomal anomalies was more strongly associated with increased cancer risk compared to those children with only one non-chromosomal anomaly (OR for three or more anomalies: 3.11, 95% CI: 1.54–6.11 vs. OR for one anomaly: 1.29, 95% CI: 1.12–1.49). This may suggest that children with previously unidentified multiple malformation syndromes (and no obvious chromosomal anomaly) may be at a significant risk of cancer.

Our data confirm the association between chromosomal anomalies and childhood cancer, including the well documented association of Down syndrome and acute leukemia. Although there have been efforts to identify factors associated with acute lymphoblastic leukemia in children with Down syndrome (e.g., maternal health conditions and irradiation), most results have been null.[19, 20] Our results also support an association between having a chromosomal anomaly with risk of retinoblastoma, which is consistent with other studies.[2] It is likely the primary driver behind this association is an autosomal deletion of 13q14, which includes the RB1 gene, a germline predisposition gene for retinoblastoma.[21] We also observed an association of chromosomal anomalies with renal tumors (e.g., Wilms tumor). Notably, Wilms tumor, aniridia, genitourinary anomalies, and mental retardation (i.e., WAGR syndrome) are a set of conditions associated with a deletion on 11p13, which includes the WT1 gene.[22]

Our results indicate an increased risk of childhood cancer in relation to presence of anomalies for cancers diagnosed in all age groups. When only non-chromosomal anomalies were considered, the risk estimates generally remained increased, supporting results of other studies indicating an association with childhood cancer diagnosed at different ages, but with a slight decrease in risk with attained age.[5] We also observed increased cancer risk in relation to increasing number of anomalies present, which is consistent with one earlier report.[18] Also consistent with other population-based data linkage studies that examined non-chromosomal anomalies, we observed increased risks of selected cancer types. We found associations between having major non-chromosomal anomalies and increased risks of neuroblastoma, hepatoblastoma, soft-tissue sarcoma, and germ cell tumors. With the exception of leukemia, retinoblastoma, and bone tumors, we observed increased risks among all other cancer types however these associations were not statistically significant.

Notably, our study supports other recent population-based registry linkage studies in demonstrating the relationship between non-chromosomal anomalies and childhood cancer.[5–7] Our overall effect estimate for the risk of cancer among children with non-chromosomal anomalies (OR = 1.35) was only slightly attenuated compared to those reported by Janitz et al. (HR = 2.50)[5], Botto et al. (incidence rate ratio [IRR] = 2.00),[7] and Fisher et al. (HR = 1.58).

We did note specific cancer types associated with having non-chromosomal anomalies. For example, the OR for neuroblastoma in relation to non-chromosomal anomalies (OR = 1.90) was largely consistent with two previous U.S. registry linkage studies evaluating the risk of this malignancy in children with non-chromosomal anomalies (HR = 2.85[6] and IRR = 2.21[7]). Also, we observed a positive association of non-chromosomal anomalies with hepatoblastoma, consistent with the only other U.S. registry linkage study (UTAZIA study) to evaluate this particular relationship, although the effect size was larger in that assessment (IRR = 14.47 vs. our OR = 2.45).[7] Having a non-chromosomal anomaly was associated with soft-tissue sarcomas, consistent with the one U.S. registry linkage study evaluating this specific relationship.[5] Our observed association with renal tumors (OR = 1.71) was stronger than reported in the California (HR = 1.45) and UTAZIA (IRR = 1.03) studies.[6, 23] Finally, our observation of a positive association between germ cell tumors and non-chromosomal anomalies supports results of other studies that were able to assess this association.[5, 6, 23]

In our assessment, non-chromosomal anomalies overall were not strongly associated with leukemia, lymphoma or CNS tumors, consistent with results of the other U.S. registry linkage studies.[5, 6, 23], despite differences in birth defect surveillance across studies. (Washington does not have an active birth defects surveillance program as in California, Utah, Iowa, and Oklahoma.) Notably, the prevalence of congenital anomalies was slightly higher in our assessment (5.5%) when compared to these states (e.g., ~4% [2, 5]), however, the ascertainment of anomalies was independent of case status in our assessment, and therefore uniform for those children who did and did not develop cancer, which reduces the likelihood of differential misclassification.

Aside from an association of CNS anomalies with CNS tumors (an association that may be due to reverse causation given the majority of these anomalies were hydrocephalus-related), there was no strong evidence that non-chromosomal anomalies were likely to be specifically associated with childhood cancer occurring in the same organ system or anatomic location, although our ability to investigate this was limited by small numbers. Although neuroblastoma was associated with heart and gastrointestinal anomalies, it was also associated with musculoskeletal and skin anomalies. Few studies have been able to examine associations of specific non-chromosomal anomalies with specific cancer types, but of these, a generally consistent finding is an association of CNS defects with CNS tumors,[8, 18, 24] as we observed.

An important strength of our study was use of linked population-based health registry data, allowing us to avoid some biases that may be present in clinic-based or interview studies. We also increased the sensitivity of birth defect ascertainment by utilizing specific diagnostic codes in addition to birth record information. We were able to examine specific congenital anomalies. Our study must also be considered in the light of certain limitations. In order to identify major and minor anomalies, we used the classification system developed by Rasmussen and colleagues,[13] which utilizes CDC-BPA codes that are more specific than the ICD-9 codes available to us for this study. Despite our ability to utilize linked hospital discharge records, our ascertainment of anomalies is likely less complete than for studies using data from active birth defects surveillance programs.[5, 15–17] However, several birth defects surveillance programs only monitor specific anomalies, whereas we evaluated all congenital anomalies. It is also possible that some children in the control group may have moved out-of-state and been diagnosed with cancer elsewhere, however as childhood cancer is quite rare, any effect of this is likely minimal and would bias towards the null. Because of the availability of birth and cancer registry data during different time periods, some children diagnosed in earlier study years at older ages would not have been included, however sensitivity analyses restricting subjects to only those with similar opportunity (e.g., at least 5 years; at least 10 years) to have been identified in the cancer registry did not substantially alter results (S2 Table). We were also limited in our ability to evaluate possible associations with minor anomalies which may not be detected until later in a child’s life, and would not appear on birth certificates or hospital discharge records for the birth hospitalization. Finally, children with some types of congenital anomalies may die prematurely and therefore lack the opportunity to develop childhood cancer, which could possibly attenuate our associations.[5]

The etiologies of most non-chromosomal anomalies are largely unknown,[1] despite evidence that factors such as maternal obesity, prenatal smoking, and some chemical or environmental exposures may increase the occurrence of certain defect types.[1, 25] Our knowledge of childhood cancer causes is similarly limited, with few recognized external etiologies (e.g., ionizing radiation) although common variation and intrinsic factors such as birthweight and parental age are consistently associated with childhood cancers. Identification of factors associated with progression from defect presence to cancer occurrence, or of shared pathways (genetic and environmental) for both conditions may elucidate potential mechanisms to modify cancer risk. Future assessments should include pooling efforts across multiple regions. This will optimize our ability to identify associations between specific congenital anomalies and specific cancers. The ultimate goal of this work would be to inform screening strategies for children at high risk of developing cancer.