We linked 3 data sources: the California Birth Defects Monitoring Program (CBDMP) registry, the California Cancer Registry (CCR), and the live birth and death files from the California State Office of Vital Records. This study included 3 221 849 live births recorded between 1988 and 2004 in California counties covered by the CBDMP registry.

Data on birth defects were drawn from cases ascertained through the CBDMP’s surveillance program, a population-based active surveillance system for collecting information on births with major congenital malformations in California counties. (Cases with only minor anomalies are not identified in the registry.) Diagnostic and demographic data (including information on chromosomal anomalies) were collected by program staff from multiple sources of medical records for all liveborn and stillborn (defined as a fetus ≥20 weeks’ gestational age) infants.⁹ Most structural birth defects diagnosed within 1 year of delivery are ascertained; overall ascertainment has been estimated as 97% complete.¹⁰

Since 1988, the CCR has maintained a legislatively mandated population-based surveillance system for all newly diagnosed cancers, excluding basal and squamous cell carcinomas of the skin, among all California residents, and quality control studies indicate 99% ascertainment.¹¹ Routinely collected CCR data include detailed case demographic, diagnostic, and treatment characteristics. The CCR follows a rigorous, active surveillance protocol modeled after the National Cancer Institute’s Surveillance, Epidemiology and End Results program. The present study includes information on cancer cases diagnosed between 1988 and 2006 in children aged <15 years.

Using probabilistic record linkage (LinkPlus), we linked children in the cancer file to California live birth certificates (birth years 1988-2004). The personal identifiers common to both databases and used for linkage were name, date of birth, sex, and race/ethnicity. Birth defects cases are routinely linked to vital statistics records by the CBDMP, using probabilistic record linkage and personal identifiers common to both databases (eg, name, date of birth). Reasons for lack of matching to vital records include adoption and, for the cancer cases, birth outside of California. We subsequently matched those cancer cases for which we identified a birth certificate to the birth defects registry by a unique numeric identifier available from the Office of Vital Records. We also linked all of the births to the state death files to identify any children who died, and censored these children at date of death (linked by name, date of birth, and sex).

Each potential case with a link to both the birth defects registry and to the cancer registry was reviewed by a clinical geneticist (S.R.) to determine whether each case had only a minor birth defect (eg, polydactyly, branchial cleft cyst) or a defect secondary to the tumor (eg, hydronephrosis secondary to a kidney tumor, hydrocephalus associated with a brain tumor). Infants with either of these conditions were excluded from the total cohort considered to have both a birth defect and cancer.

We compared the risk of cancer in the 2 cohorts (those born with major birth defects and those born without major birth defects between 1988 and 2004) using Cox proportional hazards models to generate hazard ratios (HRs) and 95% CIs, based on person-years at risk. We censored each individual at time of cancer diagnosis, death, end of study period, or age 15 years. We analyzed the risk for all cancers combined, as well as for major subtypes, such as leukemia, neuroblastoma, and central nervous system (CNS) tumors, and some specific cancers, including acute lymphoblastic leukemia, acute myelogenous leukemia, and non-CNS germ cell tumors. We classified the cancers in accordance with the International Classification of Childhood Cancer, Third Edition; however, we included intracranial and intraspinal germ cell tumors with the CNS tumors and excluded renal carcinomas from Wilms tumors (nephroblastomas).

We examined all birth defects combined and groupings of birth defects defined based on British Pediatric Association Classification of Diseases codes, as modified by the Centers for Disease Control and Prevention (http://www.cdc.gov/ncbddd/birthdefects/documents/MACDPcode0807.pdf), a classification scheme similar to the International Classification of Diseases. The 3-digit codes reflect broad birth defect groupings. Because the overall results were likely to be driven by the well-known association between Down syndrome and leukemia, we examined risks in children with and without chromosomal anomalies separately, and we excluded children diagnosed with leukemias from our analysis of the overall risk of cancer in children with birth defects but no chromosomal anomalies.

We identified a birth cohort of 3 221 849 live births for the period 1988-2004, coinciding with the study population in which surveillance for birth defects was conducted. In this cohort, the CBDMP identified a total of 65 585 infants with structural birth defects (2%). The phenotypes of these infants are displayed in Table I by birth defect category. The categories are not mutually exclusive, because many infants had multiple defects.

We identified 4869 children with cancer in the birth cohort, representing 81.5% of all cancers occurring in children living in the study counties during the study period (ie, 81.5% were linked to birth certificates). The distribution of cancer types in these children is shown in Table I. A total of 270 children were included in the CBDMP registry as well. Of these 270, 48 were excluded after detailed case review for having a defect considered minor or likely secondary to the cancer, leaving 222 infants with both major birth defects and cancer. In these 222 infants, the most common cancers were leukemias and CNS tumors.

Table II lists cancer risks in the 6327 children with chromosomal anomalies, including 3923 with Down syndrome. As expected, the risk for leukemia was highly elevated (HR, 28.99; 95% CI, 23.07-36.42). The HR for acute myelogenous leukemia was much higher than that for acute lymphoblastic leukemia. These children also had an increased risk of Wilms tumor (HR, 13.43; 95% CI, 5.54-32.55).

Table II also shows the risks of specific types of childhood cancer in infants born with birth defects but without chromosomal anomalies. In contrast to children with chromosomal anomalies, this group was not at increased risk for leukemia (HR, 0.96; 95% CI, 0.66-1.38). However, they were at increased risk for the other types of cancer examined, namely solid tumors including lymphomas, CNS tumors, neuroblastomas, and non-CNS germ cell tumors.

Demographic characteristics for the children born without chromosomal anomalies, comparing children with and without birth defects and with and without cancer, are shown in Table III. The 4 groups demonstrated no significant differences in terms of sex, birth year, or race/ethnicity. The mothers of infants with birth defects and cancer tended to be older and more educated than the mothers of the other groups of children.

The risk of childhood cancer in children with specific types of birth defects, excluding children with chromosomal anomalies or leukemias, is delineated in Table IV. Infants diagnosed with any of the birth defect phenotypes except cleft lip or palate had a ≥2-fold greater risk of developing cancer during childhood.

Further analyses to investigate the specific cancer phenotype in children who were previously diagnosed with birth defects were hindered by small sample sizes (data not shown), and thus we cannot firmly infer specific birth defect and specific cancer associations. We can make some tentative observations, however, including that in children born with “other congenital anomalies of the nervous system” (British Pediatric Association Classification of Diseases [BPA] 742), CNS tumors were most common (occurring in 47% of the cases with BPA 742 and cancer), and that in children born with “other congenital anomalies of the heart” (BPA 746), lymphomas were most common (38%).

To identify new potential etiologic clues underlying birth defects and childhood cancers, we undertook a large population-based North American study to examine whether the incidence of childhood cancer was elevated in children with structural birth defects. Our findings confirm previous observations that children born with chromosomal abnormalities are at increased risk for developing cancer in childhood, as demonstrated by, for instance, the known association between Down syndrome and leukemia.

Our findings extend the knowledge base beyond these previous observations, however. We found that children with nonchromosomal birth defects had an elevated risk of cancer (HR, 1.58; 95% CI, 1.33-1.87), particularly for CNS tumors, lymphomas, neuroblastomas, and germ cell tumors, but not for leukemia. Furthermore, we observed that risk of developing cancer in childhood was substantially increased (2- to 3- fold) in children with nearly every structural birth defect phenotype, with the exception of cleft lip or palate. More specifically, and more tentatively, we found that CNS tumors were more common in children with a history of CNS birth defects and that lymphomas were more common in children with a history of congenital heart defects.

Previous investigations examined the relationship between childhood cancer and birth defects in population-based datasets,^2-8 but did not take our approach to exclude chromosomal defects. These studies varied in their definition of what constituted a birth defect, but all reported an increased risk of cancer in children with birth defects. A Canadian study comparing cohorts of children with and without birth defects found a 2-fold higher risk of all types of cancer, particularly leukemia, CNS tumors, and sympathetic nervous system tumors, in the children with birth defects.² They also found a 6-fold greater risk of cancer in the first year of life in the children with birth defects. The 2 population-based cohort studies conducted in the United States were based on relatively small cohorts of children with birth defects (approximately 19 000 in one study and 10 000 in the other).^4,5 The largest study to date addressing this issue was published by researchers in Norway and Sweden.³ These researchers linked data on more than 5 million births to cancer registries in the 2 countries to evaluate cancer risks in individuals with birth defects. They computed standardized incidence ratios and found that children with birth defects had an overall increased risk of 1.7, with especially high risks associated with Down syndrome and nervous system malformations. Notably, although most of the population-based studies to date conducted analyses or subanalyses excluding children with Down syndrome, none systematically evaluated cancer risks in children without chromosomal anomalies as was done in the present study.

Several previous case-control studies based on interview data also have examined cancer risk in children with birth defects. Four of these studies reported increased risks for cancer, generally in the 2-fold range, in children with birth defects.^12-15 Similar to our results, 2 studies of leukemia that excluded Down syndrome found no overall difference in the risk of birth defects between leukemia cases and controls.^16,17 Hospital-based clinical studies also have noted a high prevalence of morphological abnormalities in children with cancer.^18-20

All of the previous studies were conducted in primarily non-Hispanic white children in northern Europe, Canada, Australia, and the United States. The cohort of California births included in the present study was racially and ethnically diverse: 56% Hispanic, 25% non-Hispanic white, 10% Asian, and 7% non-Hispanic black. Interestingly, our results are consistent with those of the previous studies in predominantly non-Hispanic white populations.

An important strength of the present study is that data on birth defect diagnoses were obtained by active abstraction of diagnostic data from hospital reports and medical records through the first year of life. This approach ensures relatively complete ascertainment in the study population.⁹ Previous studies of birth defects and childhood cancer tended to be limited by passive ascertainment and/or ascertainment only during the newborn period.^2,3 A limitation of the present study is that birth defect diagnoses were limited to relatively broad groupings based on British Pediatric Association Classification of Diseases codes, thereby restricting our ability to interrogate specific birth defects.

Despite our opportunity to evaluate cancer and birth defects in a large and well-characterized population-based sample, the co-occurrence of these outcomes is rare, and our focus was on even rarer nonchromosomal defects. Small sample size is a study limitation that hindered our ability to make more narrow inquiries into specific birth defect–cancer phenotype associations, and thus these connections cannot be firmly inferred. In addition, our ability to examine potential explanatory variables in more depth was limited. However, we note that our results were not substantially different after adjustment for maternal age (data not shown). We chose not to adjust for birth weight, because it might be influenced by the presence of birth defects and/or act as a mediator between birth defect–cancer associations. Differential outmigration from California based on birth defect status could affect our risk estimates; data to address this potential bias were not available.

Etiologies for structural birth defects and childhood cancers remain elusive. Improved clue-finding strategies are needed to unravel these etiologies. The present study takes a first step toward identifying new potential etiologic clues underlying birth defects and childhood cancers by studying the co-occurrence of the 2 broad phenotypes. We found significant co-occurrence of various solid tumors, but not leukemias, with nonchromosomal birth defects. Because teratogenesis and some forms of carcinogenesis represent errors in early growth, the 2 mechanisms likely share biologic underpinnings.²¹ Indeed, we hypothesize that dysregulation in human development plays an important role in the etiology of childhood cancer, especially solid tumors. For example, defects in specific homeobox genes might be related to some solid tumors and birth defects, and some CNS tumors have been linked to the Shh pathway, a key in regulation of brain development.²² In contrast, leukemia might hinge on a single gene defect in clonal proliferation of leukocytes and be less closely related to aberrant developmental pathways.

Continued inquiries into the predisposition to cancer in young children with structural birth defects should help elucidate etiologic commonalities between these 2 outcomes. Indeed, molecular tools are available to examine multiple genes simultaneously in children, and the next logical step is to explore the genomic profiles of those children with cancer with and without birth defects. For now, the identification of structural birth defects in the first year of life that indicate possible unrecognized tumor predisposition should serve to alert the pediatrician to a child at increased risk for cancer and contribute significantly to our understanding of the causes of childhood cancer.