Maternal and perinatal characteristics in relation to neuroblastoma
Neuroblastoma is the most common malignancy among infants, but risk factors remain poorly understood. Because most patients present in the first few years of life, it has been hypothesized that prenatal and perinatal exposures may contribute to the pathogenesis of neuroblastoma.
A population-based case-control study was conducted by using linked birth and cancer registry records from 1980 to 2004 in Washington State. Maternal and infant characteristics from birth and hospital discharge records for 240 cases of neuroblastoma and 2400 controls were compared.
Neuroblastoma was associated with the presence of major congenital abnormalities (odds ratio [OR], 6.86; [95% CI], 2.92–16.08), particularly with cardiac abnormalities (OR, 5.84; 95% CI, 1.93–17.66), even after excluding abnormalities near the primary tumor. A borderline association was observed with maternal gestational diabetes (OR, 1.84; 95% CI, 0.98–3.47). The magnitude of both associations was greater when the analysis was limited to children who were diagnosed at age <1 year.
The findings from this population-based study supported prior case-control studies that identified an etiologic link between neuroblastoma and congenital abnormalities. However, to the authors' knowledge, the association between neuroblastoma and maternal diabetes has not been reported previously and requires further study. Cancer 2007 © 2007 American Cancer Society.
Neuroblastoma is the most common malignancy of infancy and the second most common solid tumor in children overall. Approximately 650 cases are diagnosed in the United States annually.1 It is derived from the neural crest cells that ordinarily mature to form the sympathetic nervous system. With a median age of diagnosis around 19 months,2 it has been hypothesized that prenatal and perinatal exposures may be important in the pathogenesis of neuroblastoma. Although epidemiological studies have been performed over the past 20 years, consistent risk factors have yet to be identified.1 In particular, there are conflicting findings associated with infant characteristics, such as birth weight, gestational age,3–7 and congenital abnormalities,4, 6, 8–12 and parental characteristics, such as prior fetal loss, medications, and environmental and occupational exposures.13–26 Given the uncertainty regarding these earlier results, we conducted a population-based case-control study to evaluate maternal reproductive history, birth history, and infant characteristics and their association with subsequent neuroblastoma by using linked cancer registry and birth hospital discharge data for Washington State.
MATERIALS AND METHODS
Case and Control Ascertainment
For the case group, patients who were diagnosed with neuroblastoma between 1980 through 1992 were identified from the Cancer Surveillance System (CSS) of Western Washington, which has been affiliated with the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program since 1974. The CSS identifies incident cases of cancer in 13 western Washington counties (representing approximately 70% of the state's population). Cases diagnosed between 1993 through 2004 were identified within the Washington State Cancer Registry (WSCR), which was established as part of the Centers for Disease Control and Prevention's National Program of Cancer Registries and, in collaboration with the CSS, has been coordinating surveillance of incident cases state-wide since 1993. The WSCR also exchanges data with 30 other states to ascertain residents who are diagnosed/treated for cancer out of state.
Cases for this study were identified within CSS and WSCR using the International Classification of Diseases for Oncology-version 3 (ICD-O-3) codes 9490 and 9500 by screening the registry records of children aged <20 years. Codes within the registries updated are retrospectively to reflect current ICD-O-3 coding. All records of children who met these criteria (N = 282) were linked to Washington State birth certificate records from 1980 to 2004 to identify neuroblastoma cases born in Washington (N = 240 cases). Ten controls were selected randomly for each case from the remaining birth certificates and were frequency matched on year of delivery (N = 2400 controls).
Parental, perinatal, and infant characteristics recorded on the birth certificate were available for analysis. Given changes in the birth certificate format, some exposures were recorded differently over the study period. For individuals who were born in 1987 or later, data from the Comprehensive Hospital Abstract Reporting System (CHARS) data base also were linked. CHARS was created in 1987 by the Washington State Department of Health and consists of all hospital discharge records of individuals who are hospitalized at nonfederal facilities. Linkage of the birth certificate to CHARS for the birth hospitalization of the mother and infant provided additional discharge diagnoses codes (up to 9 fields per hospitalization of International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes), procedure codes (up to 9 fields, also based on ICD-9-CM), and insurance payer billed at discharge.
Maternal conditions that were identified among ≥20 women from birth certificates were screened for in the diagnosis codes from the mother's CHARS hospital discharge records and included first-trimester bleeding (ICD-9-CM code 640), anemia (codes 280–285 and 648.2), established diabetes mellitus (codes 250, 357.2, 362.0, 366.41, and 648.0), gestational diabetes mellitus (code 648.8), genital herpes (code 054.1), established hypertension (codes 401–404, 642.0–642.2, 642.7, and 642.9), gestational hypertension (code 642.3), preeclampsia (codes 642.4–642.5), and tobacco use (codes 305.1 and V15.82). Other selected characteristics that were identified by both birth certificates and CHARS records of the mother or infant included eclampsia (code 642.6), infant of a diabetic mother (code 775.0), and infant congenital abnormalities (codes 740–759). Congenital abnormalities that were indicated by checked boxes on birth certificates included anencephaly; hydrocephalus; microcephalus; spina bifida; other central nervous system abnormalities; heart defects; other circulatory or respiratory abnormalities; cleft lip and palate; tracheoesophageal fistula; gastroschisis; diaphragmatic hernia; omphalocele; anal atresia; other gastrointestinal abnormalities; renal agenesis; urogenital abnormalities, including malformed genitalia; musculoskeletal defects, including limb reductions, club feet, and digit abnormalities (eg, polydactyly); and other unspecified abnormalities. All procedures were approved by the Institutional Review Boards at the Washington State Department of Health and the University of Washington prior to the conduct of this study.
Unconditional multivariable logistic regression was used to calculate odds ratio (OR) estimates and 95% confidence intervals (95% CI) associated with selected exposures. Analyses also were stratified by age of neuroblastoma diagnosis (age <1 year or ≥1 year), tumor type (ganglioneuroma/ganglioneuroblastoma [ICD-O-3 code 9490] or neuroblastoma [ICD-O-3 code 9500]), and extent of tumor spread (local and regional spread or distant metastases). Variables that were considered for their possible effects on the associations of interest included infant sex, birth year (5-year interval linear splines), gestational age (20.0–36.9 weeks, 37.0–42.0 weeks, or 42.1–45.0 weeks), birth weight (<2500 g, 2500–3999 g, or ≥4000 g), parental age (5-year interval linear splines) and ethnicity (white, black, Hispanic, Asian/Pacific Islander/Native American, and other/unknown), and maternal residence (western Washington vs the rest of Washington state). In general, adjustment for any of these variables did not alter risk estimates meaningfully (>10%); therefore, estimates adjusted only for birth year are presented in the analyses discussed below unless noted otherwise. Tests of trend were calculated by modeling selected exposures as a grouped linear term using logistic regression. All analyses were carried out by using the STATA software package (version 9; Stata Corporation, College Station, TX).
Because the identification of cases prior to 1993 was conducted only in western Washington, the distribution of case characteristics before and after 1993 was examined (Table 1). The distributions of diagnosis age and ICD-O-3 codes did not differ significantly between the 2 periods. The differences in distribution of histologic grade and tumor stage reflect the smaller percentages of unspecified recorded for cases diagnosed between 1993 and 2004 compared with cases diagnosed in earlier years.
Table 1. Characteristics of Neuroblastoma Cases (N = 240)
|Age at diagnosis, y|
| 9490: Ganglioneuroma, ganglioneuroblastoma||34||14.2||14.5||14|
| 9500: Neuroblastoma||206||85.8||85.5||86|
| Well differentiated||5||2.1||2.4||1.9|
| Moderately differentiated||3||1.3||1.2||1.3|
| Poorly differentiated||16||6.7||2.4||8.9|
|SEER tumor staging|
| Localized disease||31||12.9||6||16.6|
| Regional disease (direct extension or local lymph node involved)||48||20||10.8||24.8|
| Distant metastatic disease (including distant lymph node involvement)||98||40.8||16.9||53.5|
There was no consistent association of neuroblastoma with parental age, education level, marital status, or insurance status (Table 2). The association of white ethnicity approached statistical significance compared with nonwhite ethnicities. However, after adjusting for both maternal and paternal ethnicity simultaneously, only the association of maternal white ethnicity maintained borderline significance (OR, 1.48; 95% CI, 0.95–2.32). Among parental age groups, the extremes of paternal age were associated more strongly with neuroblastoma, although most estimates were not significant. Overall, 72% of cases (n = 172) and 66% of controls (n = 1593) had birth certificates that indicated maternal residence within the 13 western counties surveyed by the CSS (P = .10). Adjustment for maternal county of residence (13 western counties covered by CSS vs the rest of the state) did not affect our risk estimates.
Table 2. Parental Characteristics of Neuroblastoma Cases and Controls*
|Maternal age at birth, y|
| Black||7||2.9||108||4.5|| || |
| Hispanic||17||7.1||194||8.1|| || |
| Asian/Pacific Islander/Native American||10||4.2||200||8.3||1.00||Ref|
| Other/unknown||2||0.8||55||2.3|| || |
|Maternal education, y§|
|Insurance billed at hospital discharge||
|Paternal age at birth, y|
| Black||9||3.8||109||4.5|| || |
| Hispanic||13||5.4||166||6.9|| || |
| Asian/Pacific Islander/Native American||13||5.4||158||6.6||1.00||Ref|
| Other/unknown||19||7.9||311||13.0|| || |
|Paternal education, y§|
There were no significant associations of neuroblastoma with the number of prior pregnancies or births, with multiple gestation births compared with singletons, with the method of delivery, or with any level of miscarriages or therapeutic abortions (Table 3). Among women who had ≥3 prior pregnancies, the OR for neuroblastoma associated with ≥3 miscarriages versus none was 2.28 (95% CI, 0.94–5.55). Risk estimates associated with miscarriages at ≥20 weeks' gestation versus none consistently were increased, although the increases were not significant. These associations were greater for cases who were diagnosed at age <1 year (≥3 prior miscarriages: OR, 3.65; 95% CI, 1.24–10.72; prior miscarriages at ≥20 weeks' gestation: OR, 2.25; 95% CI, 0.87–5.84).
Table 3. Maternal Reproductive History and Risk of Neuroblastoma*
|Multiple gestation birth|
|Prior miscarriage <20 wks gestation‡|
|Prior miscarriage ≥20 wks gestation‡|
|Prior therapeutic abortion¶|
Neuroblastoma was not associated significantly with first-trimester bleeding, anemia, genital herpes, established or gestational hypertension, oligohydramnios, preeclampsia, or prenatal smoking (Table 4). The following maternal conditions were reported rarely (<20 records) in the birth certificates but were not distributed differentially among cases and controls: eclampsia, epilepsy, heart disease, polyhydramnios, and renal disease (data not shown). There was a suggestion of an increased risk of neuroblastoma associated with maternal diabetes, particularly gestational diabetes (OR, 1.84; 95% CI, 0.98–3.47).
Table 4. Selected Maternal Health Conditions and Exposures During Pregnancy and the Risk of Neuroblastoma*
| All types||12||5||72||3||1.71||0.91–3.22|
| Gestational only§||12||5.8||67||3.3||1.84||0.98–3.47|
|Prenatal smoking (cigarettes/day)¶|
When the association with gestational diabetes was examined for neuroblastoma diagnosed at age <1 year, the OR was 2.82 (95% CI, 1.24–6.39). For neuroblastoma diagnosed at any age with distant metastases, the OR associated with gestational diabetes was 2.53 (95% CI, 1.17–5.44). In contrast, the OR for gestational diabetes in relation to neuroblastoma diagnosed at age ≥1 year was 1.23 (95% CI, 0.49–3.11), and the OR associated with local or regional tumor spread at any age was 1.20 (95% CI, 0.43–3.35). These estimates were similar whether the birth certificate or CHARS data were used to identify women with gestational diabetes, and the combined data resulted only in improved statistical precision.
Cases and controls generally were similar with respect to sex, gestational age, birth weight, and gestational size adjusted for age (Table 5). Although the risk estimates increased with greater gestational age, birth weight, and size for gestational age, the trends for all 3 characteristics were not significant. Premature infants <34 weeks' gestation (OR, 0.77; 95% CI, 0.28–2.16), infants with very low birth weight (<1500 g: OR, 0.51; 95% CI, 0.07–3.79), and infants with very high birth weight (≥4500 g: OR, 1.29; 95% CI, 0.54–3.06) also were not associated with neuroblastoma compared with term gestation and normal weight infants, respectively.
Table 5. Infant Characteristics and Risk of Neuroblastoma*
|Gestational age, wks|
|Birth weight, g|
|Gestational size for age|
| Small for age||20||8.3||217||9||0.91||0.56–1.48|
| Appropriate for age||181||75.4||1805||75.2||1.00||Ref|
| Large for age||29||12.1||229||9.5||1.26||0.83–1.91|
| Major only‡||9||3.8||14||0.6||6.86||2.92–16.08|
| Cardiac only||5||2.1||9||0.4||5.84||1.93–17.66|
| Urogenital only||2||0.8||25||1||0.85||0.20–3.62|
Cases were more than twice as likely as controls to have any congenital abnormality reported by either the birth certificate or the CHARS record (Table 5). When only major abnormalities, as defined by Cordier et al,27 were considered, the risk estimate associated with neuroblastoma was higher (OR, 6.86; 95% CI, 2.92–16.08). Three of 9 cases who had major abnormalities had primary tumors on the same side of the diaphragm as the abnormality. If those 3 cases were excluded, then risk estimates were diminished but remained significant (OR, 4.54; 95% CI, 1.72–11.98). When the risk of neuroblastoma was evaluated by abnormality subtype, there was an association with cardiac abnormalities (OR, 5.84; 95% CI, 1.93–17.66) that remained after the exclusion of cases and controls with patent ductus arteriosus (OR, 7.04; 95% CI, 1.97–25.21).
The association of neuroblastoma with abnormalities appeared to differ according to age at diagnosis and tumor stage. The OR for neuroblastoma diagnosed at age <1 year associated with the presence of major abnormalities was 11.10 (95% CI, 3.84–32.02); for neuroblastoma diagnosed at age ≥1 year, the OR was 4.61 (95% CI, 1.49–14.27). For the association of major abnormalities with local/regional neuroblastoma, the OR was 11.42 (95% CI, 4.46–29.20); for metastatic disease, the OR was 2.94 (95% CI, 0.66–13.18). Mothers who had infants with abnormalities were not more likely to have had prenatal ultrasounds compared with mothers who had infants without reported abnormalities (64% vs 63%, respectively). Finally, the association of neuroblastoma with abnormalities was unchanged after adjustment for maternal diabetes status. Aside from the instances mentioned above, for most characteristics and exposures, there were no consistent differences in risk estimates between different neuroblastoma categories: age of diagnosis (<1 year vs ≥1 year), ICD-O-3 code (9490 vs 9500), or tumor stage (local/regional vs distant metastases).
In this population-based case-control study, by using linked birth certificate-hospital discharge data, neuroblastoma was associated with gestational diabetes and congenital abnormalities, particularly among children who were diagnosed with neuroblastoma at age <1 year. We did not observe any association with other previously reported factors, such as high birth weight, low birth weight, or prematurity. The pathogenesis of neuroblastoma remains poorly understood; however, given the clinical heterogeneity, there well may be different pathways involved for patients who are diagnosed at younger ages versus older ages and for patients with local disease versus metastatic disease. Younger age and local disease at diagnosis are attributes typically associated with an improved prognosis that require less aggressive treatment compared with presentation at an older age with metastatic disease.2
Consistent with national SEER data,1 in our study, neuroblastoma was more common among males and among those of white race/ethnicity, and approximately 40% of cases were diagnosed at age <1 year. Most epidemiologic studies, including the current effort, did not report an association between most elements of maternal reproductive history and neuroblastoma. Consistent with 2 prior studies,7, 20 we observed a borderline increased risk associated with increased prior miscarriages (≥3) and late gestation miscarriages. However, in contrast to Hamrick et al,7 we did not observe an increased risk after therapeutic abortions.
There is no consensus among epidemiologic studies on the association of neuroblastoma with gestational age or low birth weight,3–7, 20 although few estimates reach statistical significance. Both values tend to be reported reliably regardless of the data collection method (eg, parental report, birth record).28, 29 Although ORs increased with increasing gestational age and birth weight in our study, the trends were not significant. Overall, given the modest range of risk estimates reported in the literature, it is unlikely that gestational age and birth weight are major risk factors for neuroblastoma.
Two prior studies examined the association of neuroblastoma with maternal diabetes and did not report significant associations.7, 18 However, 1 study assessed only chronic diabetes,18 and the other study did not specify diabetes type.7 In our study, the overall prevalence of gestational diabetes (3%) was within the range cited by most studies.30 In a prior validation study, it was noted that linkages between Washington state birth certificates and CHARS improved substantially the reporting accuracy of any form of maternal diabetes along with other maternal conditions, such as genital herpes and chronic and gestational hypertension.31 In that study, with medical records as the gold standard, the true-positive rate of gestational diabetes improved from 64% to 93% after such linkages, whereas the false-positive rate was only 1%. Therefore, the ascertainment of maternal gestational diabetes in the current study may be reasonably complete, at least beginning in 1987 with the availability of CHARS data. Any under ascertainment of maternal conditions would be expected to be nondifferential, biasing the results toward the null.32
Although no studies linking neuroblastoma to diabetes have been reported previously to our knowledge, some biologically plausible pathways may exist. Children with Beckwith-Wiedemann syndrome, a congenital overgrowth syndrome in which abnormal insulin-like growth factor 2 (IGF-2) imprinting may occur, have a greatly increased risk of hepatoblastoma, Wilms tumor, and neuroblastoma.33, 34 IGF-2 is a mitogen that is expressed widely during fetal development and acts through multiple receptors, including the insulin receptor, during the fetal period. Abnormal IGF-2 expression may play a role in the development of some neuroblastoma subtypes,35 although loss of IGF-2 imprinting has not been observed in neuroblastoma cells, in contrast to reports in other embryonal cancers, such as rhabdomyosarcoma and Wilms tumor.36, 37 However, other epigenetic alterations in the same gene cluster (11p15) have been reported in neuroblastoma, although their role in oncogenesis remains unclear.38
A significant association between neuroblastoma and congenital abnormalities has been reported by some10–12 but not all4, 6, 8, 9 prior investigators who examined this correlation. Differences in study results may be related to the methods by which children with congenital abnormalities were ascertained. Studies based on medical chart review,11 birth defect registries,9, 10 parental self-report,12 and birth certificate/hospital discharge codes4, 6, 8 feature different biases, but ascertainment of abnormalities is imperfect for all methodologies.39–41 In our study, an association was established by using either birth certificates or hospital discharge codes. Overall, abnormalities probably were under reported, because children were observed only during their birth hospitalization. In addition, abnormalities that were recorded as other on a birth certificate could not be classified as major abnormalities in the absence of clarifying hospital discharge data. The number of discharge diagnosis fields also was limited; therefore, infants with multiple diagnoses may not have had all abnormalities recorded. However, any misclassification of abnormality status in our study also should be nondifferential, biasing results toward the null.32 In fact, when the analyses were restricted to major abnormalities that were less likely to be missed immediately after birth and more likely to be recorded in discharge records,39 the association with neuroblastoma became stronger.
When abnormality subtypes were examined, we observed a significantly elevated risk associated with cardiovascular abnormalities, in agreement with 2 previous studies.11, 12 Neural crest cells play a role in proper cardiac and great vessel development during fetal development.42 In a study that examined the echocardiograms of pediatric cancer patients, an elevated incidence of cardiac abnormalities was reported in patients who had neuroblastoma compared with patients who had acute lymphoblastic leukemia (20% vs 4%; P < .001).43 Three of 14 patients with neuroblastoma who had cardiac abnormalities in that study were diagnosed with neuroblastoma incidentally during the work-up for their cardiac issues. Such surveillance bias may exaggerate the relation between congenital abnormalities and neuroblastoma, at least in infants. Infants are more likely to have neuroblastoma that can regress spontaneously without treatment, a phenomenon that is not observed in older children. However, in the current study, the association of neuroblastoma with both major abnormalities and cardiac lesions remained significantly elevated in children who were diagnosed at age ≥1 year.
The current study does have some limitations. Prior to 1987, data on mothers' medical characteristics were available only from birth certificates. Birth certificates generally are considered reliable for characteristics like delivery method, number of prior births, birth weight, and gestational age but may be less sensitive (although very specific) for other elements of maternal reproductive history.29, 31 Compared with medical records, most data on Washington state birth certificates and hospital discharge records generally have been associated with an error rate of ≤1%.31, 44 In the current study, the estimates that were produced by using birth certificates generally were consistent with those from hospital discharge records, and statistical precision was enhanced after combining the 2 data sources. Bias also may have been introduced by using a control group from the state-wide population throughout the study period but using a case group that originated only from western Washington prior to 1993, which primarily would raise an issue with geographically referenced exposures. However, the majority of the state's population resides in western Washington, and adjustment for western Washington versus the rest of the state in multivariable regression did not change the risk estimates. When the analysis was stratified by year of diagnosis before 1993 and from 1993 onward, the risk estimates associated with diabetes and abnormalities were higher from 1993 onward. Finally, risk estimates for older children also may have been biased by migration to other states. However, there is no apparent reason why cases would be predisposed to move more frequently than controls. WSCR does attempt to ascertain state residents who leave the state to seek treatment elsewhere.
In summary, we observed suggestive associations of neuroblastoma with gestational diabetes and congenital abnormalities, especially cardiac abnormalities. These associations were stronger for patients who were diagnosed at a younger age. These findings support prior studies that linked neuroblastoma to congenital abnormalities. However, to our knowledge, the association between neuroblastoma and maternal diabetes has not been reported previously and needs to be confirmed.
We thank the Washington State Department of Health for allowing data access and Mr. Bill O'Brien for programming assistance