Association Between Maternal Hormonal Contraception Use and Central Nervous System Tumors in Children

Key Points

Question  Are children born after maternal use of hormonal contraception at an increased risk of central nervous system (CNS) tumors?

Findings  In this Danish cohort study that included 1 185 063 children, the adjusted incidence of CNS tumors among children of mothers with recent hormonal contraceptive use vs mothers without any hormonal contraceptive use was 5.0 vs 5.3 per 100 000 person-years; this difference was not statistically significant.

Meaning  There was no significant association between any maternal hormonal contraceptive use and CNS tumors among children.

Importance  The incidence of central nervous system (CNS) tumors in children appears to be increasing, yet few risk factors are established. There is limited information regarding whether maternal hormonal contraception use increases this risk.

Objective  To examine the association between maternal hormonal contraception use and CNS tumors in children (<20 years).

Design, Setting, and Participants  In this nationwide cohort study based on population-based registry data, 1 185 063 children born in Denmark between January 1, 1996, and December 31, 2014, were followed up for a diagnosis of a CNS tumor (final follow-up on December 31, 2018).

Exposures  Maternal hormonal contraception use was analyzed according to any use, regimen (combined/progestin only), and route of administration (oral/nonoral), categorized as recent use (≤3 months before start and during pregnancy), previous use (>3 months before start of pregnancy), and no use. For injections, implants, and intrauterine devices that are used for a different time period, the categorization was appropriately altered.

Main Outcomes and Measures  Hazard ratio (HR) and incidence rate difference (IRD) of CNS tumors diagnosed at younger than 20 years.

Results  After 15 335 990 person-years of follow-up (mean follow-up, 12.9 years), 725 children were diagnosed with a CNS tumor. The mean age at diagnosis was 7 years, and 342 (47.2%) of the diagnosed children were female. The adjusted incidence rate of CNS tumors per 100 000 person-years was 5.0 for children born to mothers with recent hormonal contraception use (n = 136 022), 4.5 for children born to mothers with previous use (n = 778 843), and 5.3 for children born to mothers with no use (n = 270 198). The corresponding HRs were 0.95 ([95% CI, 0.74-1.23]; 84 children with CNS tumors; IRD, −0.3 [95% CI, −1.6 to 1.0]) for recent use and 0.86 ([95% CI, 0.72-1.02]; 421 children with CNS tumors; IRD, −0.8 [95% CI, −1.7 to 0.0]) for previous use, compared with no use. No statistically significant associations were found for recent or previous use of oral combined, nonoral combined, oral progestin only, or nonoral products compared with no use of hormonal contraception.

Conclusions and Relevance  Among Danish children, there was no statistically significant association between any maternal hormonal contraception use and CNS tumor risk.

Childhood cancer is the leading cause of disease-related deaths in children in the western world,¹ and central nervous system (CNS) tumors are some of the most lethal and common types.² The incidence of CNS tumors in children appears to have been increasing.^3,4 In Denmark, the annual incidence rate (IR) of CNS tumors in children increased from about 3.5 to 5.0 per 100 000 person-years from 1977 to 2014, an increase not likely explained by changes in registration and improved diagnostic methods only.⁵ However, to date, established risk factors remain limited to a few genetic syndromes and ionizing radiation.⁶

Sex hormones are considered potent human carcinogens,⁷ and hormonal exposure in utero is recognized to cause cancer in exposed offspring.⁸ To date, 5 studies have reported risk estimates for the association between maternal use of hormonal contraception and childhood CNS tumor risk.^9-13 Four of these were case-control studies and found no association,^9,10,12,13 whereas a case-cohort study based on registry data found an increased risk.¹¹ None of these studies included information on more contemporary types of hormonal contraceptives (ie, nonoral types) and all were based on self-reported exposure information. Using detailed registry-based information on hormonal contraception, the aim of the present study was to assess the association between maternal hormonal contraception use and the risk of childhood CNS tumors (age 0-19 years based on World Health Organization guidance) in a nationwide cohort of all live-born children in Denmark from 1996 to 2014.

The Parental Exposures and Child Health cohort, which was initially established to assess the association between maternal hormonal contraception use and the risk of childhood cancer,¹⁴ follows a national cohort of all live-born children born in Denmark from January 1, 1995, to December 31, 2014. To establish the Parental Exposures and Child Health cohort, the unique personal identification number given to all residents in Denmark was used to link individual-level data from several population-based nationwide registries (eTable 1 in the Supplement). Because information on maternal hormonal contraception was available only from 1995 onwards, we excluded children born before 1996, children for whom length of gestation was implausible or missing, and children with missing information on maternal age from the present study. The project is registered with the Danish Cancer Society in agreement with the General Data Protection Regulation. According to Danish legislation, it is not necessary to seek ethics approval or informed consent for studies based on registry data.

Maternal-redeemed prescriptions for hormonal contraception were identified using the Danish National Prescription Registry (eTable 2 in the Supplement).¹⁵ The registry is reimbursement-driven with automated bar code–based data entry, which provides data of high validity and completeness. Because most contraceptives are prescribed for 3 months, exposure was categorized as recent use (≤3 months before start of and during pregnancy), previous use (>3 months before start of pregnancy), and no use (reference). The categorization was altered appropriately for the types of contraception that are used for a different time period (eTable 3 in the Supplement). Recent use was the primary exposure of interest, with previous use serving as the contrast. Regimen of maternal use of hormonal contraception was defined as estrogen-progestin combined or progestin only and route of administration was defined as oral or nonoral. Nonoral included injections, implants, intrauterine devices, patches, and vaginal rings. Emergency contraception was considered a separate category. Pregnancy start was estimated by subtracting gestational age from date of birth. For 87% of records, information on gestational age is accurate within 1 week.¹⁶ In Denmark, gestational age is based on ultrasonography and, in few cases, on the date of onset of the last menstrual period.¹⁷

The Danish Cancer Registry, estimated to be 95% to 98% complete,¹⁸ was used to identify CNS tumors, which were classified according to the International Classification of Childhood Cancer, 3rd Edition.¹⁹ The children were followed up from their date of birth until diagnosis of a CNS tumor, censoring (death, emigration, other cancer, or 20th birthday), or December 31, 2018, whichever occurred first.

Hazard ratios (HRs) and 95% CIs were calculated using the Cox proportional hazards model. The variance of the estimates was estimated by a robust variance estimate adjusting for within-cluster correlation to account for correlation between siblings.²⁰ Proportional hazard tests were based on weighted residuals and fulfilled for all models. IRs (cases divided by person-years) and IR differences (IRDs; IRs in exposed children minus IRs in unexposed children) with 95% CIs were calculated by Poisson regression. No adjustment was made for multiple comparisons. Because of the potential for type 1 error due to multiple comparisons, findings for analyses of secondary end points and subgroups should be interpreted as exploratory.

Children’s risk of CNS tumors was evaluated according to timing of maternal hormonal contraception use (recent and previous), regimen, route of administration, and emergency contraception, compared with no use. Besides attained age, which was used as the underlying timescale, year of birth was included as a potential confounder in all models a priori. Other potential confounders, including origin (Danish or descendant of immigrant), birth order (1 or ≥2), maternal and paternal age at birth, and maternal and paternal education (basic, vocational, or higher), were investigated for their effect on the association between any maternal hormonal contraceptive use and children’s risk of any CNS tumor by complete case analysis.²¹ Only year of birth (with no missing values) was included in further analyses because none of the other factors changed the risk estimate by more than 10%. Year of birth and parental age at birth were tested for linearity. Because no deviation from linearity was found, these variables were included as continuous variables. Further adjustment for maternal smoking was conducted for a subset of children born from 1998 onwards, because information on maternal smoking was available from 1997.

Sensitivity analyses were performed to investigate whether the choice of reference group affected the results. First, a preplanned analysis was performed using children born to mothers with previous or no use of hormonal contraception as the reference group instead of children of those with no use only. Second, for the association between maternal use of nonoral progestin-only hormonal contraception and CNS tumors in children, a post hoc analysis using a reference group of children of mothers who had used nonoral progestin-only products after the date of birth of the child was conducted to address the potential for unmeasured confounding. Third, 2 post hoc analyses including information on CNS tumor types and specific drugs types (ie, injections, implants, and intrauterine devices) were carried out for the association between maternal use of nonoral progestin-only contraception and CNS tumors. Fourth, a likelihood test was performed post hoc to assess whether the 3 nonoral progestin-only products were statistically significantly different. For all analyses, a significance level of 5% was applied with 2-sided P values. Analyses were performed with the statistical software Stata, version 11.2 (StataCorp).

From 1996 through 2018, a total of 15 335 990 person-years and 725 incident CNS tumors had accumulated in the 1 185 063 children included in the study (Figure 1) during a mean follow-up of 12.9 years (range, 0-19 years). The mean age at CNS tumor diagnosis was 7 years, and 342 (47.2%) of the diagnosed children were female. In the total cohort of children, 136 022 (11.5%) were born to mothers who had used hormonal contraception less than or equal to 3 months before or during pregnancy (recent use), 778 843 (65.7%) were born to mothers who had stopped using hormonal contraception more than 3 months before pregnancy start (previous use), and 270 198 (22.8%) were born to mothers who had never used hormonal contraception. Characteristics of the study population by maternal use of contraception are listed in the Table. Compared with children born to women with recent use of hormonal contraception, children born to women with no use were more often born in earlier calendar years, as descendants of immigrants, of higher birth order, with higher parental age, and to mothers with a diagnosis of infertility.

For the complete case analysis, 102 689 children with missing information for any of the potential confounders were excluded. Maternal recent use (83 children with CNS tumor; IR, 5.0 per 100 000 person-years; IRD, −0.3 [95% CI, −1.6 to 1.0]; HR, 0.95 [95% CI, 0.74-1.23]) or previous use (421 children with CNS tumor; IR, 4.5 per 100 000 person-years; IRD, −0.8 [95% CI, −1.7 to 0.0]; HR, 0.86 [95% CI, 0.72-1.02]) of any hormonal contraception product before birth was not statistically significantly associated with childhood CNS tumor risk compared with no use (Figure 2). For the most commonly used products (ie, the oral combined products), no statistically significant association was found with maternal recent or previous use compared with no use. Also, no statistically significant associations were found with recent or previous maternal use of nonoral combined products, oral progestin-only products, nonoral progestin-only products, and emergency contraception compared with no use. Of the 3 subgroups of nonoral progestin-only products examined, recent use of implants and intrauterine devices showed no significant association. Although recent use of injections (n = 885) was associated with an increased CNS tumor risk (<5 children with a CNS tumor; IR, 34.6 per 100 000 person-years; IRD, 29.3 [95% CI, −9.9 to 68.5]; HR, 6.74 [95% CI, 2.15-21.15]), compared with no use (Figure 3), the likelihood ratio test result was not statistically significant (P = .1035), indicating that the nonoral progestin-only products did not have different effects on CNS tumor risk.

Only minor differences were observed when taking into account maternal smoking (eTable 4 in the Supplement). When using children born to women with previous or no use of hormonal contraception as the reference group instead of no use, results were similar, except for recent use of nonoral progestin-only products, which became statistically significantly associated with an increased CNS tumor risk in children (eTable 5 in the Supplement). The association was still statistically significant when children born to women who used nonoral progestin-only products after birth was used as the reference instead of children born to mothers with no use before birth (eTable 6 in the Supplement). For all main types of CNS tumors except “other gliomas,” increased HRs were observed for maternal use of nonoral progestin-only products compared with no use, although none were statistically significant (eTable 7 in the Supplement).

In this large, Danish, nationwide, register-based cohort study, no statistically significant association was found between maternal use of any type of hormonal contraception and CNS tumors in the children. In agreement with the study findings, 4 case-control studies found no association between maternal use of oral hormonal contraceptives and childhood CNS tumors.^9,10,12,13 In contrast, the only other study based on registry data from 1973 to 1989 reported an odds ratio of 1.6 (95% CI, 1.0-2.9) for CNS tumor in children born to women who had used oral hormonal contraception up to pregnancy.¹¹ Because hormonal content in oral contraceptive pills has changed over time (eg, from higher doses to lower), this could explain the different findings, because the present study was based on a more recent data. Furthermore, the different results may also be due to differences in exposure measurement, because the information on hormonal contraceptive use in the former study was self-reported to the registry (ie, disclosed at antenatal visits) as opposed to the exposure information in the present study, which was based on automatic barcode registration at the date of redemption.

For the subgroups of nonoral progestin-only hormonal contraception assessed, recent use of implants and intrauterine devices showed no statistically significant associations. In contrast, injections were significantly associated with an increased risk, compared with no use. In addition, recent use of the main group of nonoral progestin-only products was significantly associated with CNS tumors in all sensitivity analyses, although these were post hoc. However, the result for injections was based on a small number of cases and the result of the likelihood test was null, indicating that the nonoral progestin-only products may not have different effects on CNS tumor risk. Also, multiple comparisons were not adjusted for. Even if the results for this subgroup were confirmed, because CNS tumors in children are uncommon, the high relative risk estimates would translate to low absolute risk increases. Some studies have found maternal use of fertility treatment, in which nonoral progesterone (natural counterpart of progestin) is commonly used, to be associated with childhood CNS tumors.^22,23 Furthermore, although the timing of use is different, in young women, progestin-only hormonal contraception was found to increase glioma risk,²⁴ supporting a role of progestogens in the development of CNS tumors.

Although the causal association between in utero exposure to diethylstilbestrol (an estrogen analogue) and risk for adenocarcinoma of the vagina is widely recognized,²⁵ the mechanism by which maternal use of hormones may increase cancer risk in children is not clear. Sex hormones are considered potent carcinogens and may affect children both before and after conception by inducing epigenetic changes and/or by affecting cell differentiation, cell division, and number of cells at risk.^20,26 An association between maternal use of injectable contraceptives and increased risk of chromosomal anomalies and major malformations in children has previously been reported²⁷; however, the tumorigenic properties of different progestogens have not been systematically studied, and their effect on the proliferation of neural stem cells or neural progenitors is currently unknown.²⁸ However, the wide distribution of progesterone receptors in the brain suggests that progestogens may play a role.^28,29

Strengths of this cohort study include the individual-level linkage to a number of nationwide population-based registries with complete mother-child linkages and less than 2% missing information on variables needed for establishing the cohort (gestational and maternal age) and no missing information on variables included in the final adjusted models (children’s age and calendar year of birth). Hence, the study is based on detailed prospective data, limiting concerns about selection and recall bias, loss to follow-up, and temporality. The large number of person-years and cancers increases the statistical precision, and the population-based nationwide design increases the generalizability of the results. Exposure and outcome data are likely highly valid because information on maternal hormonal contraceptive use is based on automatic transfer of information from bar codes from the Danish National Prescription Registry, and the Danish Cancer Registry is virtually complete with correct diagnoses.

This study has several limitations. First, because CNS tumors in children are uncommon, few cases were available for subgroup analyses, limiting the statistical precision of certain estimates. Second, some women may not have used the hormonal contraception redeemed or used them at a different time, resulting in some nondifferential misclassification of the exposure and a possible underestimation of the associations. Third, information on the 2 established risk factors for CNS tumors in children (ie, genetic syndromes and ionizing radiation) was not available. Genetic syndromes should not be adjusted for as they would be on the causal pathway. However, if ionizing radiation was associated with maternal hormonal contraception use, this could have affected the results. Fourth, the possibility of residual or unknown confounding cannot be excluded due to the observational design of the study. Fifth, women who use nonoral progestin-only products, such as implants and injections, may differ from other women because this type of contraception is more often used by women with certain disabilities.³⁰ However, to our knowledge, there is no study linking maternal disability to CNS tumor risk in children. Sixth, due to the high number of comparisons, the potential for type 1 error cannot be excluded; therefore findings in subgroups should be interpreted as exploratory.

Among Danish children, there was no statistically significant association between any maternal hormonal contraception use and increased risk for CNS tumors.

Corresponding Author: Marie Hargreave, PhD, Virus, Lifestyle and Genes, Danish Cancer Society Research Center, Strandboulevarden 49, DK-2100 Copenhagen, Denmark (mariehar@cancer.dk).

Accepted for Publication: December 6, 2021.

Author Contributions: Dr Hargreave had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Hargreave, Mørch, Kjaer.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Hargreave.

Critical revision of the manuscript for important intellectual content: All authors.

Statistical analysis: Hargreave.

Obtained funding: Hargreave.

Supervision: Hargreave, Kjaer.

Conflict of Interest Disclosures: Dr Mørch reported receiving personal fees from Novo Nordisk as an employee from 2017 to 2019 and grants from Novo Nordisk for a collaborative research project outside the submitted work. No other disclosures were reported.

Funding/Support: The study was supported by the Danish Cancer Research Foundation, the Arvid Nilssons Foundation, the Gangsted Foundation, the Harboe Foundation, and the Johannes Clemmesens Foundation.

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.