Pediatric Germ Cell Tumors and Maternal Vitamin Supplementation: a Children's Oncology Group Study

Pediatric germ cell tumors (GCT) are rare malignancies affecting ∼340 children <15 years of age every year in the United States (1, 2).⁵

GCTs resemble tissues of normal early human development and are classified into two major histologic types, seminomas/dysgerminomas and nonseminomas, which can occur at either extragonadal or gonadal locations. Seminomas/dysgerminomas are rare before puberty and are thought to result from direct transformation of primordial germ cells (4, 5). In contrast, nonseminomas, which are composed of teratomas, yolk sac tumors, and choriocarcinomas, are most common during early childhood and are thought to arise from spontaneous differentiation of embryonal carcinomas that develop from primordial germ cells. Teratomas have differentiated along somatic cell lineages and often contain all three germ cell layers, whereas yolk sac tumors (endodermal sinus tumors) and choriocarcinomas have differentiated along extraembryonic cell lineages and contain cells resembling those of the endoderm and placental trophoblast lineage, respectively (4, 6).

Molecular studies suggest that aberrant DNA methylation may be involved in the evolution of the various GCT histologic subtypes (4, 7-11). The developmental time period may be especially vulnerable for abnormal DNA methylation to occur due to the epigenetic reprogramming of the genome that takes place (12). We hypothesized that maternal vitamin supplementation, a source of the methyl donor folic acid (13), during or around pregnancy may decrease the risk of GCTs in the offspring. In a case-control study in the Children's Oncology Group (COG), we evaluated whether maternal vitamin supplementation was related to GCT development in children.

Complete study details have been previously published (14) and are summarized briefly below. Individuals diagnosed with GCTs (n = 278) from birth to 14 years during January 1st, 1993 to December 31st, 2001 were identified through participating COG institutions in the United States and Canada. Cases with GCTs at all sites except the brain and with the following histologies: dysgerminoma/seminoma/germinoma, embryonal carcinoma, yolk sac tumor, choriocarcinoma, mature teratoma, and mixed GCTs were eligible for inclusion. Controls (n = 423) were sampled from the population through random digit dialing and frequency matched to cases on sex and birth year within 1 year, at ratios of approximately 1:2 for males and 1:1 for females. Both case and control subjects were excluded if they did not have a telephone in their residence, if their mother did not speak English, or if she was not available for interview. The response rates for case and control subjects were 81% and 66%, respectively, as reported previously (14). Institutional Review Boards at the University of Minnesota and each case's COG institution approved all protocols for this study.

Information on sociodemographic characteristics, family history, and prenatal and postnatal medical conditions and exposures was collected from participant mothers through both a self-administered questionnaire and telephone interview. Less than 2% of individuals had missing data on all of the variables included in this analysis. Mothers provided the following information about vitamin supplementation during pregnancy on the self-administered questionnaire: (a) whether they took vitamins in the 6 mo before pregnancy, during pregnancy, or while nursing (yes, no, do not know); (b) the specific time point when they were taken (6 mo before pregnancy, trimester1, trimester 2, trimester 3, and/or while nursing), (c) the type of vitamin; and (d) whether a doctor prescribed them. Because the majority of participants who took vitamins took prenatal vitamins that were prescribed by a physician or multivitamins (99%), we did not consider specific vitamin formulations separately in our analyses.

We used SAS version 9.1 to conduct all analyses. The association between maternal vitamin use and GCTs was modeled using unconditional logistic regression. Potential confounders of the association between maternal vitamin use and pediatric GCTs were examined using a forward selection process (15) where each potential confounder (maternal education: less than high school, high school, college (no degree), college degree, and/or graduate school; annual household income: <20,000, 20,000-30,000, 30,001-50,000, >50,000; maternal race: White, other; maternal age; and child's birth year) was added to models that included the matching variables sex and child's age one at a time to determine their effect on the parameter estimate. If they changed the parameter estimate by >10%, they were retained in the model.

The majority (58%) of male cases had tumors located in the gonads, whereas in females, the percentage of tumors located in the ovaries (51%) was similar to those located at extragonadal locations (47%). The most common tumor histology in both sexes was yolk sac tumor. Most male cases were diagnosed before the age of 5 years, whereas the majority of female cases were diagnosed after this age (Table 1).

Slightly more cases than controls were born before 37 weeks of gestation (Table 2). Cases had a slightly lower mean birth weight than controls, although more cases than controls had birth weights above 4,000 grams. The mean maternal age was similar in cases and controls, whereas control mothers were more likely to report having a college degree (37.9% versus 30.9%), White race (84.2% versus 76.6%), and household income of ≥20,000 per year (79.2% versus 68.0%)

Any vitamin use during pregnancy was associated with a slight reduction in the risk of GCTs [odds ratio (OR), 0.7; 95% confidence interval (CI), 0.4-1.2] in models adjusted for sex, child's age, maternal education, and household income (Table 3). Maternal vitamin use during the 6 months before conception and first trimester (periconception), or first trimester specifically, was not associated with GCTs. In analyses that excluded subjects who did not take any vitamins before or during their pregnancy or while nursing, no associations were observed with respect to timing (data not shown).

We assessed whether the association between any maternal vitamin use and GCTs varied by sex, age at diagnosis, tumor location, and tumor histology (data not shown). Nonsignificant reduced risks were observed with any maternal vitamin use in males (OR, 0.6; 95% CI, 0.2-1.7) and females (OR, 0.8; 95% CI, 0.4-1.4) and in children ≤2 (OR, 0.8; 95% CI, 0.4-2.0) and >2 years of age (OR, 0.7; 95% CI, 0.4-1.4). Any maternal vitamin use was also associated with a reduced risk for both gonadal (OR, 0.6; 95% CI, 0.3-1.1) and extragonadal (OR, 0.8; 95% CI, 0.4-1.6) tumors. Directions of associations between maternal vitamin use and GCTs varied by histology with inverse associations for dysgerminoma/seminoma (OR, 0.5; 95% CI, 0.2-1.3) and teratomas (OR, 0.5; 95% CI, 0.2-0.9) but not yolk sac tumors (OR, 1.1; 95% CI, 0.5-2.3). Neither periconception nor first trimester use was significantly associated with GCTs in any of these subgroup analyses (data not shown).

The results from this study provide a pattern, albeit weak, that is consistent with the a priori hypothesis that maternal vitamin supplementation during the perigestational period reduces the risk of GCTs. We observed some variation by histology but many of the observed inverse associations were modest and failed to reach statistical significance. Moreover, analyses evaluating the timing of vitamin exposure before and during pregnancy did not indicate any relationship to GCT risk.

One previous case-control study that included 105 pediatric GCTs cases and 639 controls reported no association with maternal vitamin use (OR, 1.1; 95% CI, 0.6-2.0; ref. 16). Case-control studies of several other childhood tumors have generally indicated that prenatal vitamin use and/or folate supplementation reduces risk. A recent review by Goh and Koren (17) reported reduced risks of childhood cancers in association with maternal vitamin supplementation in six of seven studies of brain tumors, four of four studies of acute lymphoblastic leukemia, two of two studies of neuroblastoma, and one of one study of retinoblastoma. Some of the reviewed studies suggested that vitamin supplemention during the periconception period specifically is associated with a reduction in risk. However, we did not find any evidence for a relation between timing of maternal vitamin use and GCTs.

Molecular abnormalities in GCTs vary by sex, diagnosis age, and histologic subtype and include aneuploidy, chromosome abnormalities (most commonly involving 12p), and abnormal DNA methylation patterns (4, 7, 9, 10). Aberrant methylation patterns have commonly been detected in pediatric GCTs in several genes, including the imprinted genes IGF2 and H19, testis/cancer-associated genes, and tumor suppressor genes (4, 7-9, 11). In addition, expression of the p40 protein of the long interspersed nucleotide element-1 (Line-1) transposable element, which is normally silenced through epigenetic mechanisms (18), has been reported in pediatric malignant GCTS (9). A biologically plausible hypothesis is that maternal vitamin deficiency during and around pregnancy alters GCT risk in offspring by affecting DNA methylation patterns in germ cells. The developmental time period may have increased susceptibility to DNA methylation errors because of the two major waves of epigenetic reprogramming that occur in the blastocyt and germ cells (12).

A major strength of this study is the relatively larger number of subjects compared with previous studies. To our knowledge, only five previous case-control studies of childhood GCTs (case numbers ranging from 41-105) have investigated etiologic factors in GCT development (16, 19-22). The size of our study also allowed for a more detailed examination than previous studies of the association between GCTs and etiologic factors by subject and tumor characteristics.

Our study had several limitations. Although our study included more cases than prior studies, all studies of this type of tumor, including ours, have limited statistical power to detect modest effect sizes. Another potential limitation is selection bias. Selection bias could have occurred if control participants were not a representative sample of the nondiseased population from which cases arose (15). Given that the response rate among potential controls was 66% (23) and more control than case mothers reported being of White race and having higher levels of education, characteristics associated with vitamin supplementation during pregnancy (24), we cannot exclude that selection bias affected our risk estimates. In addition, recall bias, where a mother's ability to accurately recall her exposures during pregnancy depends on whether her child is a case or control, may lead to differential exposure misclassification and consequent bias in the risk estimate (15).

In conclusion, these data are consistent with results from studies of other childhood cancers suggesting a reduced risk of malignancy in association with maternal vitamin supplementation. However, limitations regarding statistical power and biases that affect observational studies preclude firm conclusions.

No potential conflicts of interest were disclosed.