Perinatal characteristics and risk of neuroblastoma

Abstract

Neuroblastoma (NB), a tumor of the sympathetic nervous system, is the most common infant malignancy. The etiology of NB is largely unknown. We explored the association between birth record variables and subsequent NB development in a population-based case-cohort study in Minnesota by linking the birth and cancer registries. NB cases included 155 children born during 1976–2004 who were diagnosed from 28 days through 14 years of age. The comparison group included 8,752 individuals randomly sampled from the birth cohort of cases. Cox proportional hazards regression was used to calculate hazard ratios (HRs) and 95% confidence intervals (CIs). Controlling for birth year and sex, maternal history of one fetal loss (HR = 1.7, 95% CI 1.2–2.5), maternal prenatal drug-use (recorded starting in 1992) (HR = 5.7, 95% CI 2.3–14) and child's small size for gestational age (HR = 2.1, 95% CI 1.1–4.0) were significantly associated with NB. Age group specific analyses indicated that maternal hypertension (HR = 3.0, 95% CI 1.3–7.2) and maternal age <20 years (HR = 2.6, 95% CI 1.1–6.1) increased risks for infant NB only. Our study provides evidence that a few perinatal exposures as recorded in birth records may play a role in NB etiology. © 2008 Wiley-Liss, Inc.

Neuroblastoma (NB) is a rare childhood tumor with an annual incidence of 9.1 cases per million children under age 15 years in the United States. NB is most commonly diagnosed in children <12 months, representing the most frequently diagnosed infant tumor. The incidence of NB is higher in whites than blacks and in males than females.1

NB is a sympathetic nervous system (SNS) tumor that originates from the neural crest.2 Although NB can arise anywhere in the SNS, it is most commonly found in the adrenal gland.3 NB is a heterogeneous tumor with prognosis dependent on the stage, the molecular and histological characteristics of the tumor and the age of diagnosis. The extent of the disease is generally inversely correlated with prognosis with the exception of stage 4S disease, found in infants, that has limited metastases and often spontaneously regresses. Amplification of the MYCN protooncogene is the most frequent genetic abnormality found in NB and is associated with later stage tumors and an unfavorable prognosis. MYCN amplification is more commonly found in children older than 1 year3 and consequently NB in older children is associated with a lower survival rate than in infants.1 Other commonly identified genetic abnormalities include: hyperdiploidy, 17q gain, and loss of heterozygosity at 11q, 14q, and 1p.2

Little is known about what causes NB or whether environmental factors play a role. The primitive nature of the tumor cells in addition to young diagnosis age suggest that pre- and perinatal exposures may be etiologically relevant. No definitive risk factors for NB have been identified with inconsistent associations reported for most previously examined exposures. Studies have examined prenatal exposure to maternal tobacco,4–8 alcohol,4, 5, 7, 9, 10 maternal hair dye5, 11 and drug-use.5, 7, 12–15 Other commonly examined exposures include those pertaining to sociodemographics, reproductive and pregnancy history, labor and delivery, and infant characteristics.6, 8, 9, 12, 16–21

To gain further understanding of whether perinatal exposures are involved in the etiology of NB, we evaluated associations between factors recorded in birth records and NB in a population-based case-cohort study in Minnesota using prospectively collected exposure information. In addition, because of the variation in the clinical and molecular characteristics of NB diagnosed during infancy versus later on, we also examined the risk factors in age-stratified analyses.

CI, confidence interval; HR, hazard ratio; IRB, institutional review board; MCSS, Minnesota Cancer Surveillance System; NB, neuroblastoma; ref., reference; SNS, sympathetic nervous system; SRI, serotonin reuptake inhibitors.

Material and methods

Study population

Childhood cancer cases diagnosed from 1988 to 2004 at ages 28 days through 14 years were ascertained by the Minnesota Cancer Surveillance System (MCSS) and matched to Minnesota birth records through probabilistic record linkage.22 The comparison group (referred to as the subcohort) was selected from the Minnesota birth registry by randomly sampling 4 individuals per childhood cancer case frequency matched on birth year. Subcohort members who did not survive past the neonatal period (>28 days) were excluded for birth years occurring in 1980 or later (the earliest period for which neonatal deaths were recorded). Because sampling was performed to enable analyses of multiple childhood cancer diagnoses, the ratio of NB cases to subcohort members is greater than 4:1. All other childhood cancer cases were excluded from analyses [except for those randomly sampled into the sub-cohort (n = 15)]. NB case diagnoses included ICD-0-3 codes 9490 (neuroblastoma) and 9500 (ganglioneuroblastoma).23 The University of Minnesota Institutional Review Board (IRB), the Minnesota Department of Health IRB and the MCSS approved all protocols for data use.

Variables

Variables collected in birth records are noted in detail in the Tables and included parental sociodemographic characteristics, maternal reproductive history, pregnancy conditions, labor and delivery procedures, and infant characteristics. Two gestational age variables were available in birth records: (i) imputed from last menstrual period and (ii) the physician's estimate, with the physician's estimate being used where the imputed estimate was missing [n = 620 (7%)]. Size for gestational age was calculated according to the method of Brenner et al.24 The birth record database collected information on maternal drug-use beginning in 1992 and included 9 categories: none, cocaine, heroin, phencyclidine, methamphetamine, amphetamine, other, blank and not classifiable. The birth record allowed physicians to name “other” drugs which were entered into the electronic database verbatim and abstracted by birth registry personnel for our study.

Statistical analyses

SAS version 9.1 (Cary, NC) was used for all analyses. A modified version of the stratified Cox proportional hazards model, appropriate for the sampling strategy used in case-cohort study designs,25 was employed to calculate hazard ratios (HRs) and 95% confidence intervals (CIs). Person-years of follow-up was calculated as the time from birth to cancer diagnosis date, age 15 years, or Dec. 31, 2004, whichever came first with modification of follow-up time for NB cases who were also selected in the subcohort (n = 1).25 Birth year and sex adjusted analyses were first conducted, followed by selection of variables with a p-value of ≤0.2 for further examination in multivariate models. Fisher's exact test was used to determine the statistical significance where the number of exposed cases for binary variables was <5. Modification of the risk of NB associated with birth factors by age group (<12 months vs. ≥12 months) was examined by comparing risk estimates for variables between age group models. Variables that had notably different age-group risk estimates are reported.

Results

One hundred sixty-six NB cases diagnosed from 0 to 14 years were recorded by MCSS from 1988 to 2004. Of these, a total of 155 cases (93%) diagnosed >28 days were born in Minnesota and were eligible for our study. The frequency of NB diagnoses declined with age with 37% (n = 57) of cases being diagnosed during infancy (28 days to 1 year), followed by 23% (n = 35), 28% (n = 44) and 12% (n = 19) being diagnosed from 1 to 2 years, 2 to 4 years and 4 to 14 years, respectively.

No significant associations were found between NB and sociodemographic characteristics: maternal age, education or marital status (Table I). Subjects whose mothers were of white compared to non-white race were at a marginally significant increased risk of NB (HR = 2.0, 95% CI 1.0–4.0). Risks associated with paternal characteristics were similar to those for maternal characteristics (data not shown).

Table I. Demographics and Risk of Neuroblastoma

Variable	No. in subcohort (%)	No. of cases (%)	HR1	95% CI
1 Adjusted for sex and birth year. 2 Collected in birth years 1989–2004.
Maternal race
White	7,909 (91.3)	145 (94.8)	1.0	ref.
Non-white	754 (8.7)	8 (5.2)	0.48	0.23–0.99
Missing	89	2
Maternal ethnicity2
Hispanic	160 (3.5)	7 (6.0)	1.62	0.73–3.60
Non-Hispanic	4,475 (96.5)	109 (94.0)	1.0	ref.
Missing	349	7
Maternal age
<20	753 (8.6)	15 (9.7)	1.16	0.65–2.08
20–24	2,052 (23.4)	34 (21.9)	1.01	0.66–1.57
25–29	3,040 (34.7)	53 (34.2)	1.0	ref.
30–34	2,070 (23.7)	32 (20.6)	0.78	0.50–1.22
≥35	837 (9.6)	21 (13.5)	1.13	0.67–1.90
Missing	0	0
Maternal education
Less than 12 years	902 (10.9)	16 (10.5)	1.05	0.59–1.85
12 years	3,173 (38.3)	51 (33.6)	1.0	ref.
More than 12 years	4,211 (50.8)	85 (55.9)	1.08	0.76–1.54
Missing	466	3
Birth place of mother
U.S. or U.S. territory	8,227 (94.1)	145 (94.2)	1.0	ref.
Foreign country	517 (5.9)	9 (5.8)	0.80	0.40–1.60
Missing	8	1
Parental marital status
Married	6,990 (79.9)	118 (76.1)	1.0	ref.
Not married	1,754 (20.1)	37 (23.9)	1.06	0.72–1.54
Missing	8	0

No significant associations were found between NB and maternal reproductive, pregnancy, and labor and delivery characteristics with the exception of significantly elevated risks for NB in association with a maternal history of one past fetal loss compared to none (HR = 1.7, 95% CI 1.2–2.4), and maternal drug-use during pregnancy (HR = 5.9, 95% CI 2.4–14.4) (Table II). Specific drugs reported for mothers of NB cases were: marijuana (n = 1), cocaine (n = 1), levothyroxine (n = 1), a nonspecified other drug (n = 1) and the serotonin reuptake inhibitor (SRI) antidepressants: fluoxetine and venlavaxine (n = 2). The frequency of reported drug-use in case mothers was significantly different from subcohort mothers for cocaine (p = 0.02), levothyroxine (p = 0.02) and SRI antidepressants (p = 0.004) (data not shown).

Table II. Maternal Reproductive History, Pregnancy Conditions, Labor and Delivery Procedures, and Risk of Neuroblastoma

Variable	No. in subcohort (%)	No. of cases (%)	HR1	95% CI
1 Adjusted for sex and birth year. 2 Collected in birth years 1989–2004. 3 Too few cases to fit model. Fisher's exact test was used to calculate p-values. 4 Pre-pregnancy or gestational. 5 Collected in birth years 1992–2004. 6 Collected in birth years 1976–1979 and 1989–2004.
Number of previous live births
None	3,449 (40.1)	69 (45.4)	1.0	ref.
1–2	4,253 (49.4)	68 (44.7)	0.81	0.58–1.14
3 or more	903 (10.5)	15 (9.9)	0.85	0.48–1.50
Missing	147	3
Prior fetal loss
None	6,713 (78.2)	108 (70.1)	1.0	ref.
One	1,348 (15.7)	40 (26.0)	1.72	1.18–2.49
Two or more	518 (6.0)	6 (3.9)	0.62	0.27–1.43
Missing	173	1
Maternal hypertension
No	7,680 (95.8)	136 (93.2)	1.0	ref.
Yes	337 (4.2)	10 (6.8)	1.66	0.86–3.20
Missing	735	9
Maternal anemia2
No	4,613 (98.3)	117 (99.2)	3	p = 0.72
Yes	80 (1.7)	1 (0.8)
Missing	291	5
Maternal diabetes4^,5
No	3,195 (96.7)	90 (97.8)	3	p = 0.99
Yes	110 (3.3)	2 (2.2)
Missing	215	2
Tobacco use during pregnancy2
No	4,011 (85.1)	93 (80.2)	1.0	ref.
Yes	702 (14.9)	23 (19.8)	1.42	0.89–2.27
Missing	271	7
Alcohol use during pregnancy2
No	4,587 (97.7)	115 (97.5)	3	p = 0.75
Yes	106 (2.3)	3 (2.5)
Missing	291	5
Drug use during pregnancy5
No	3,177 (98.6)	76 (92.7)	1.0	ref.
Yes	45 (1.4)	6 (7.3)	5.72	2.32–14.1
Missing	298	12
Intrauterine procedures
None	8,183 (97.6)	146 (97.3)	2	p = 0.79
Any	203 (2.4)	4 (2.7)
Missing	366	5
Induction of labor6
No	3,879 (73.0)	89 (74.2)	1.0	ref.
Yes	1,432 (27.0)	31 (25.8)	0.87	0.57–1.32
Missing	225	4
Caesarean section
No	6,954 (82.4)	119 (79.3)	1.0	ref.
Yes	1,485 (17.6)	31 (20.7)	1.13	0.75–1.69
Missing	313	5

The association between birth characteristics of the child and NB are reported in Table III. No significant associations were found between NB and child sex, birth weight, season of birth, plurality, gestational age, congenital abnormalities, Apgar scores or assisted ventilation. Small size for gestational age was associated with a significantly increased risk of NB of ∼2-fold (95% CI 1.1–4.2).

Table III. Child's Birth Characteristics and Risk of Neuroblastoma

Variable	No. in subcohort (%)	No. of cases (%)	HR1	95% CI
1 Adjusted for sex and birth year. 2 Too few cases to fit model; Fisher's exact test was used to calculate p-values. 3 Imputed gestational age based on date of last menstrual period, if missing used physician's estimate when available. 4 Collected in birth years 1980–2004. 5 Collected in birth years 1981–2004. 6 Collected for birth years 1989–2004.
Child's Gender
Male	4,502 (51.5)	87 (56.1)	1.21	0.88–1.67
Female	4,243 (48.5)	68 (43.9)	1.0	ref.
Missing	7	0
Birth weight
<2,500 g	452 (5.2)	10 (6.5)	1.17	0.60–2.28
2,500–4,000 g	7,076 (81.2)	122 (78.7)	1.0	ref.
>4,000 g	1,184 (13.6)	23 (14.8)	1.10	0.70–1.73
Missing	40	0
Season of birth
Winter	2,009 (23.0)	41 (26.5)	1.12	0.71–1.74
Spring	2,273 (26.0)	32 (20.6)	0.80	0.50–1.28
Summer	2,285 (26.1)	40 (25.8)	1.0	ref.
Fall	2,185 (25.0)	42 (27.1)	1.11	0.72–1.72
Missing	0	0
Plurality
Singleton birth	8,528 (97.4)	152 (98.1)	2	p = 0.99
Multiple birth	224 (2.6)	3 (1.9)
Missing	0	0
Gestational age3^,4
<37 weeks	657 (8.3)	14 (9.3)	1.04	0.59–1.82
≥37 weeks	7,251 (91.7)	136 (90.7)	1.0	ref.
Missing	292	4
Size for gestational age3^,4
Small	271 (3.4)	10 (6.7)	2.06	1.06–4.03
Average	5,710 (72.4)	103 (69.1)	1.0	ref.
Large	1,905 (24.2)	36 (24.2)	1.01	0.68–1.49
Missing	314	5
Any congenital abnormality
No	8,095 (98.5)	143 (98.6)	2	p = 0.92
Yes	122 (1.5)	2 (1.4)
Missing	538	10
One minute apgar score5
≤7	1,798 (23.4)	39 (26.2)	1.18	0.81–1.71
>7	5,886 (76.6)	110 (73.8)	1.0	ref.
Missing	336	5
Five minute apgar score5
≤7	275 (3.6)	8 (5.4)	1.59	0.76–3.32
>7	7,396 (96.4)	141 (94.6)	1.0	ref.
Missing	349	5
Assisted ventilation6
None	4,549 (98.3)	110 (97.3)	2	p = 0.45
Any	79 (1.7)	3 (2.7)
Missing	356	10

In multivariate models that included subjects born from 1992 to 2004 (the years drug-use was recorded) controlling for maternal race, sex, prior fetal loss, maternal hypertension, tobacco-use, size for gestational age and birth year, the association between NB and maternal drug-use remained significant (HR = 6.4, 95% CI 2.4–17.3) (data not shown).

To determine whether the increased risk associated with maternal drug-use during pregnancy was specific to NB, we also examined whether the prevalence of maternal drug-use was increased in subjects with other frequently diagnosed childhood cancers present in the complete dataset. The proportion of mothers of cases of Wilms tumor (3.8%, n = 3), soft tissue sarcoma (4%, n = 2), CNS tumors (0%, n = 0), lymphoma (0%, n = 0) and leukemia (0.4%, n = 1) with reported drug-use during pregnancy was not significantly different from that in the subcohort (1.4%, n = 45) (data not shown).

Some differences in risk factors for NB notably varied by age group (<12 months vs. 12 months or older) in birth year and sex adjusted models. Factors associated with an increased risk in the younger but not older age group were: maternal hypertension (HR_{<12 months} = 3.0, 95% CI 1.3–7.2; HR_{≥12 months} = 1.0, 95% CI 0.4–2.7) and maternal age < 20 years (HR 2.6_{<12 months}, 95% CI 1.1–6.1; HR_{≥12 months} = 0.6, 95% CI 0.3–1.5). Maternal drug-use during pregnancy was a strong risk factor for both younger and older children (HR_{<12 months} = 7.6, 95% CI 2.1–27; HR_{≥12 months} = 4.6, 95% CI 1.4–15.9).

Discussion

The etiology of NB is obscure. In this population-based case-cohort study, significantly increased risks for NB were associated with a maternal history of one fetal loss, small size for gestational age and maternal drug-use during pregnancy. In addition, some risk factors varied by age with significant positive associations between NB and maternal age <20 years and maternal hypertension limited to infants.

The association between maternal history of fetal loss and NB has been examined in a number of studies with inconsistent results (Table IV). Three studies have reported increased risks for a prior history of induced abortions,6, 18, 27 while one study has reported a decreased risk for 2 or more prior therapeutic abortions.20 Maternal history of any fetal loss or miscarriage specifically has been positively associated with NB in some studies,7, 9, 10, 18, 20 but not in others.6, 17, 19, 27 Results from our study are also inconsistent with a significant increased risk for a history of one, but not 2 fetal losses; the latter result could be due to low power to detect a significant association or to random variation.

Table IV. Relevant Findings of Studies Examining Neuroblastoma and Birth Chararacteristics

Study	Study population/exposure assessment method	Diagnosis period	Notable findings
a Information was available for 538 cases. b Information was available for 160 cases.
Johnson and Spitz16	Texas; 157 cases ≤14 years; 314 controls/birth certificates	1964–1978	Significant decreased risk for preterm birth (<37 weeks) (OR = 0.3, 95% CI 0.1–0.9) that was independent of ethnicity and birth weight. Significant increased risk for low birth weight term babies (OR = 3.2, 95% CI 1.1–9.2) compared to high birth-weight term babies.
Kramer et al.5	Pennsylvania; 104 cases (median diagnosis age 1 year), 101 random digit dialed and 86 sibling controls/telephone interview	1970–1979	Significant increased risk for maternal use of neurally active drugs (amphetamines, antidepressants, antipsychotics, muscle relaxants, prescription pain medications and tranquilizers) (OR = 2.8, 90% CI 1.3–6.0), antinauseants (OR = 1.7, 90% CI 1.1–1.8), diuretics (OR = 5.8, 90% CI 2.6–12.6) and hair dye (OR = 3.0, 90% CI 1.6–5.5) during pregnancy. Increased risk for sex hormone use during the 3 months before or during pregnancy (OR = 2.3, 90% CI 1.1–4.4) and for fluid retention during pregnancy (OR = 6.0, 90% CI 2.4–14.8). Similar percentage of miscarriages and stillbirths between mothers of cases and controls.
Neglia et al.17	Minnesota; 97 cases <9 years, 388 controls/birth certificates	1968–unspecified	Significant inverse association for maternal history of prior fetal death (OR = 0.5, 95% CI 0.2–0.8). More cases (6.6%) than controls (0.8%) had noted congenital abnormalities before newborn hospital discharge.
Schwartzbaum7	Tennessee (St. Jude Children's Research Hospital); 101 cases <9 years, 690 other childhood cancer controls/telephone interview	1979–1986	Significant increased risk for maternal history of any miscarriages or stillbirths (OR = 1.8, 95% CI 1.1–3.0), tranquilizers (OR = 2.1, 95% CI 1.1–2.3), nonprescription pain relievers (OR = 1.9, 95% CI 1.1–3.1) and cigarettes (OR = 1.9, 95% CI 1.1–3.2). Borderline significant increased risk for diuretics for hypertension (OR = 4.1, 95% CI 1.0–16.9).
Michalek et al.2	New York; 183 cases ≤14 years; 372 controls/telephone interview	1976–1987	Nonsignificant inverse association for maternal chronic high blood pressure (OR = 0.7, 95% CI 0.2–2.3). Significant increased risk for sex hormones taken for infertility (OR = 10.4, 95% CI 1.2–90).
Buck et al.6	New York; 155 cases 0–5 years, 310 controls/birth certificates and telephone interview	1976–1987	Significant reduced risks for preterm (<37 weeks) (OR = 0.4, 95% CI 0.1–0.9) and postterm (>42 weeks) (OR = 0.3, 95% CI 0.1–0.7) births; Nonsignificant increased risk for maternal history of induced abortion (OR = 1.8, 95% CI 0.8–4.1). No association for prior miscarriage (OR = 0.9, 95% CI 0.5–1.8).
Schuz 20019	Germany; 283 cases <8 years, 1,785 controls/parent self-administered questionnaire and telephone interview	1988–1994	Significant increased risks for gestational age < 37 weeks (OR = 3.4, 95% CI 1.7–6.7), birth weight < 2,500 g (OR = 3.1, 95% CI 1.5–6.4), and maternal alcohol consumption during pregnancy of >7 glasses/week (OR = 5.2, 95% CI 1.3–20.6) for stage III/IV NB. Significant increased risk for oral contraceptives or sex hormones during pregnancy (OR = 4.5, 95% CI 1.2–16.5) for stages I/II NB. No significant increased risk for maternal fetal loss (OR = 1.3, 95% CI 0.9–1.9) or hormonal treatment for infertility (OR = 1.1, 95% CI 0.5–2.3) for all stages.
Hamrick et al.,18 Cook et al.,13 Bluhm et al.14a	United States and Canada; 504 cases <19 years; 504 controls/telephone interview	1992–1994	Nonsignificant increased risk for 2 or more prior miscarriages (OR = 1.3, 95% CI 0.8–2.3). Marginally significant increased risk for 2 or more induced abortions (OR = 1.9, 95% CI 1.0–3.7). Weak positive association for maternal hypertension (OR = 1.4, 95% CI 0.9–2.1). Significant association for opioid agonist use (OR = 2.4, 95% CI 1.3–4.3) but not other drugs. Increased risk for maternal use of any illicit or recreational drug anytime between 1 month prior to pregnancy and childbirth (OR = 1.8, 95% CI 1.1–3.0) and for marijuana use during the first trimester (OR = 4.8, 95% CI 1.6–16.5).
Mchlaughlin10	New York; 360 NB cases 1 month to 19 years; 12,664 controls/birth certificates	1985–2001	Increased risk for birth weight > 3,500 g (vs. <3,500 g) (OR = 1.3, 95% CI 1.0–1.5). No significant association for small size for gestational age. Maternal history of prior fetal loss significantly increased NB risk (OR = 1.4, 95% CI 1.1–1.8) in unadjusted analyses but was not significant in adjusted analyses. Maternal hypertension increased NB risk overall (OR = 1.7, 95% CI 1.1–2.7) but was stronger in infants (OR = 2.5, 95% CI 1.2–5.2). Maternal drug-use notably increased NB risk in infants only (OR = 2.2, 95% CI 0.9–5.1).
Schuz et al.,15 Schuz and Forman26b	Germany; 157 cases ≤14 years; 2,057 controls/parent self-administered questionnaire and telephone interview	1992–1994	Nonsignificant association for maternal hypertension treated with drugs (OR = 2.0, 95% CI 0.9–4.4). Significant association for diuretics/hypertensive drugs (OR = 3.2, 95% CI 1.0–9.7). No association for pain relievers, anitnauseants or antiemetics, or cold medications. Nonsignificant increased risks for small size for gestational age (OR = 1.8, 95% CI 0.8–4.1) and large size for gestational age (OR = 1.4, 95% CI 0.9–2.3).
Chow et al.20	Washington state; 240 cases <20 years, 2,400 controls/birth certificates	1980–2004	Significant increased risk for major congenital abnormalities (OR = 6.9, 95% CI 2.9–16.1). Increased risk for previous pregnancy losses (OR = 1.6, 95% CI 0.7–1.7). Marginally significant increased risk for maternal gestational diabetes (OR = 1.8, 95% CI 1.0–3.5). Nonsignificant increased risk for >3 prior miscarriages (OR = 2.3, 95% CI 0.9–5.6) but not ≥2 prior therapeutic abortions (OR = 0.7, 95% CI 0.3–1.8). Nonsignificant decreased risks for maternal hypertension (established and gestational).
Urayama et al.19	California; 508 cases <5 years, 1,016 controls/birth certificates	1988–1997	Significant decreased risk for Hispanic (OR = 0.6, 95% CI 0.4–0.8) and other (OR = 0.6, 95% CI 0.4–0.9) race/ethnicity. Increased risk for NB diagnosed ≤12 months for postterm/high birth weight (OR = 7.0, 95% CI 1.1–45.6), preterm/low birth weight (OR = 1.6, 95% CI 0.6–4.0) and term/low birth weight (OR = 1.8, 95% CI 0.5–6.5) infants compared to term/normal birth weight infants. Significant increased risk for cesarean delivery (OR = 1.7, 95% CI 1.2–2.5) for NB diagnosed between 1 and 4 years and decreased risk for ≥3 previous pregnancies of the mother (OR = 0.4, 95% CI 0.2–0.7) for NB diagnosed during infancy. Significant increased risk for any congenital abnormality in the univariate analysis (OR = 1.8, 95% CI 1.1–2.9).
Munzer et al.27	France; 191 cases <15 years, 1,681 controls/telephone interview	2003–2004	Significant increased risk for congenital malformation (OR = 2.2, 95% CI 1.1–4.5) that was stronger in children < 1 year (OR = 16.8, 95% CI 3.1–90). Significant decreased risk for maternal history of spontaneous abortion (OR = 0.6, 95% CI 0.4–0.9). Nonsignificant positive association for induced abortions (OR = 1.4, 95% CI 0.9–2.2).

In our study, small compared to average size for gestational age increased the risk of NB. Five studies have reported NB risks in association with small size for gestational age or comparable measures with 3 studies finding increased risks ranging from 1.6 to 3.215, 16, 19 and 2 studies finding no association.10, 20 Small size for gestational age can be due to a variety of factors including: poor transplacental nutrient transfer, intrauterine infection, prenatal alcohol exposure and maternal hypertension that may be related to NB.

This association between maternal drug-use and NB has been examined in several studies5, 7, 9, 12–14, 18 (Table IV) with most creating broad groupings of drug classes based on the types of signs or symptoms they are intended to treat. Although, positive associations between maternal drug-use and NB have frequently been reported, the data are insufficient to determine whether any specific drug or drug class is involved in NB etiology. Increased risks have been reported for neurally active drugs,5 maternal use of diuretics (OR = 4.1, 95% CI 1.0–16.9), tranquilizers (OR 2.1, 95% CI 1.1–4.3) and nonprescription pain relievers (OR = 1.9, 95% CI 1.1–3.1),7 sex-hormones for infertility,12 oral contraceptives or sex hormone use,9 any drug-use for infant NB10 and prenatal intake of opioid agonists, codeine and marijuana,13, 14 cocaine or crack.14 In total, it is unclear which drugs may be mediating the effect or if the observed associations are due to recall and selection biases that can occur in case-control studies collecting information through survey-based methods.

Interestingly, 4 of the reported drugs used by case mothers during pregnancy in our study target neural processes (cocaine, marijuana, fluoxetine and venlafaxine). However, because only a small percentage of cases had reported maternal drug-use and the specific drugs that were used were unique to each case, we cannot conclude with any certainty whether this finding has a biological basis or is due to chance. It is nevertheless interesting to note that the drugs reported in our study by case mothers are known to cross the placenta,28–30 that NB cells express the serotonin receptor, 5-HT₃31 that can bind fluoxetine and cocaine,32, 33 and that maternally derived serotonin has been shown to regulate gene expression, proliferation, migration and morphogenesis of neural crest cells, from which NB derives, in mouse embryos.34 Whether drugs, particularly those that act in the serotonin pathway, can influence malignant transformation of neural cells has not been examined. An alternative biological explanation for our results is the possibility that some of the drugs recorded in NB cases mark maternal depression, which has been estimated to affect 7–15% of pregnancies in developed countries.35 The relationship between maternal depression and childhood cancer to our knowledge has not been previously investigated.

The relation between NB and maternal hypertension or hypertension drugs has been reported in several studies (Table IV). Increased risks for NB associated with diuretics or diuretics/antihypertensive agents have been reported in 3 studies.5, 7, 15 An increased NB risk associated with maternal hypertensive conditions has been reported in 2 studies,10, 18 while 2 others have reported inverse associations.12, 20 A study that examined the independent effects of hypertensive drug-use and maternal hypertension showed a weak association between maternal hypertension without drug treatment (OR 1.3, 95% CI 0.9–2.0) and a stronger association for treated mothers (OR 2.0, 0.9–4.8),15 suggesting that diuretic/antihypertensive-use specifically may be related to NB. In our study, maternal hypertension increased NB risk in infants only. One previous study reported the association between NB and maternal hypertension by age group,10 finding a stronger risk in infants (OR 2.5, 95% CI 1.2–5.2) than older children (OR 1.4, 95% CI 0.8–2.5). Conclusions about antihypertensive drug-use or maternal hypertension in the etiology of NB must be made cautiously, as it has been shown that NB tumors can produce catchecholamines36, 37 that could in turn cause maternal hypertension.37

The increased risk of infant NB associated with young maternal age has not been reported previously by others.7–9, 13, 14, 18–20 However, most studies have not reported age group specific estimates, and so it is unclear whether this could represent a real association or a chance finding.

Positive associations between NB and congenital abnormalities have been frequently reported.17, 19–21, 27, 38–40 The frequency of congenital abnormalities reported in the birth records of NB cases was not increased relative to that of controls in our study. However, the prevalence of congenital abnormalities recorded in birth records was only half that expected from national data that reports that 3% of infants are born with a major birth defect,41 most likely due incomplete ascertainment and recording of birth defects on the birth certificate. Because of this limitation and the small number of NB cases in our study, our data are insufficient to confirm or refute previous findings of a positive association between NB and birth defects.

Strengths of our study include prospective data collection and random sampling of individuals for the comparison group. Previous retrospective case-control studies that used survey-based methods to measure exposures are limited by the potential for recall and selection biases. Since exposure information was assessed prior to disease knowledge in our study, this type of information bias cannot explain our findings. Selection bias occurs when the exposure distribution of the comparison population is not representative of the source population.42 Because our comparison group was assembled from a random sample of the birth cohort of cases, selection bias is also unlikely to explain our results.

The use of birth records to measure perinatal characteristics has limitations. Variables that have been reported to be poorly measured in birth records include tobacco and alcohol exposure, obstetric procedures and delivery events, while others such as parental sociodemographic information, birth weight, Apgar scores and delivery method are considered reliable.43 Drug-use, which is passively collected on birth certificates, was underreported in our study. For example, the prevalence of antidepressant use during pregnancy, has been estimated in one study that used health plan databases to ascertain antidepressant-use to be 1.5% in 1996 increasing to 6.2% in 2005,44 compared to the 0.1% prevalence we observed among subcohort members. However, despite the limitations of using birth records as a data source for exposure assessment, the key advantages of our study design is that exposures were collected prior to NB diagnosis and the comparison population was randomly sampled from the general population (as discussed earlier). Our study was also limited because we did not have information on NB stage or MYCN amplification to address whether risk factors vary by these characteristics as indicated by some studies.9, 27 Another limitation of our study is the lack of follow-up of subcohort members and on cases diagnosed prior to 1988 (the year of the MCSS inception). Missed subcohort diagnoses after 1988 would most likely have been due to out-migration because MCSS is estimated to capture over 99.9% of Minnesota cancer cases.45 Given the rarity of NB, it is unlikely that missed cases could have caused a large bias in our results. Under the extreme assumption that the entire subcohort moved out of Minnesota after birth, we would have missed at most 2 cases assuming that ∼1/4,800 children will develop NB prior to age 15.1, 46 Finally, the small size of our study makes it is possible that these findings may be the result of chance and multiple comparisons testing.

In conclusion, our study does not provide evidence for a strong link between most perinatal characteristics and NB, implying that other environmental or genetic factors may be more etiologically relevant. These data suggest, however, that small size for gestational age, a history of maternal fetal loss, maternal drug-use andmaternal hypertension and young maternal age in infants maybe related to NB. Although based on small numbers, the finding regarding drug-use is intriguing and worthy of further investigation.