Association of Maternal Antidepressant Prescription During Pregnancy With Standardized Test Scores of Danish School-aged Children

Key Points

Question  Is there an association between maternal antidepressant prescription during pregnancy and standardized test scores of Danish school-aged children?

Findings  In this population-based retrospective cohort study that included 575 369 public schoolchildren in Denmark, maternal antidepressant prescription during pregnancy was associated with significantly lower standardized test scores (range, 1-100) in mathematics (adjusted test score difference, −2.2) but no significant difference in language test scores (adjusted test score difference, −0.1) compared with children whose mothers did not fill prescriptions for antidepressants during pregnancy.

Meaning  Among Danish public schoolchildren, maternal antidepressant prescription during pregnancy was associated with statistically significantly lower standardized test scores in mathematics that were small in magnitude and no significant difference in language test scores.

Importance  Concerns exist about long-term neurodevelopmental consequences of prenatal exposure to antidepressants.

Objective  To evaluate whether maternal prescription fill for antidepressants in pregnancy was associated with performance in standardized tests among Danish schoolchildren.

Design, Setting, and Participants  Population-based retrospective cohort study of children born in Denmark between January 1, 1997, and December 31, 2009, attending public primary and lower secondary school. The children included had completed a language or mathematics test as part of the Danish National Test Program between January 1, 2010, and December 31, 2018. The age range of the eligible schoolchildren was 7 to 17 years.

Exposures  Maternal prescription fill for antidepressants during pregnancy, obtained from the Danish Prescription Register.

Main Outcomes and Measures  The difference in standardized scores between children with and without maternal prescription fill for antidepressants in mathematics and language tests (scale, 1-100; higher scores indicate better test results) was estimated using linear regression models, adjusted for relevant confounders. Ten sensitivity analyses were performed, including a sibling-controlled analysis.

Results  Among the 575 369 children included (51.1% males), 10 198 (1.8%) were born to mothers filling an antidepressant prescription during pregnancy. The mean (SD) age of children at the time of testing spanned from 8.9 (0.4) years in grade 2 to 14.9 (0.4) years in grade 8. Maternal prescription fill for antidepressants was significantly associated with a poorer performance in mathematics (mean test scores for the group exposed to maternal antidepressant fill: 52.1 [95% CI, 51.7-52.6] and for the group not exposed to maternal antidepressant fill: 57.4 [95% CI, 57.3-57.4]; adjusted difference, −2.2 [95% CI, −2.7 to −1.6]), but not in language (mean test scores for the exposed group: 53.4 [95% CI, 53.1-53.7] and for the not exposed group: 56.6 [95% CI, 56.5-56.6]; adjusted difference, −0.1 [95% CI, −0.6 to 0.3]). In the sibling-controlled analysis, the adjusted difference in mathematics (mean scores for the exposed group: 53.5 [95% CI, 52.7-54.3] and for the not exposed group: 59.0 [95% CI, 58.9-59.1]) was −2.8 (95% CI, −4.5 to −1.2) and in language (mean test scores for the exposed group: 53.9 [95% CI, 53.2-54.6] and for the not exposed group: 56.6 [95% CI, 56.5-56.7]) was −0.3 (95% CI, −1.9 to 1.2).

Conclusions and Relevance  In this study of public schoolchildren in Denmark, children whose mothers had filled prescriptions for antidepressants during pregnancy, compared with children whose mothers did not fill prescriptions for antidepressants during pregnancy, had a 2-point lower standardized test score in mathematics, a difference that was statistically significant, but had no significant difference in language test scores. The magnitude of the difference in the mathematics test score was small and of uncertain clinical importance, and the findings must be weighed against the benefits of treating maternal depression during pregnancy.

Quiz Ref IDThe prevalence of depression and anxiety in females peaks during childbearing years.^1,2 Accordingly, antidepressants are widely used for women of fertile age, including pregnant women. Use of antidepressants in pregnancy varies between countries and within regions, with estimates ranging between 2% and 10%.^3-6Quiz Ref ID Antidepressants may be discontinued during pregnancy out of concern for the unborn child,⁷ leaving underlying maternal psychiatric conditions untreated.^3,4,8 Given the serious health implications of psychiatric disorders, particularly around the time of delivery, discontinuation of relevant, well-indicated treatment with antidepressants is not a trivial issue. Thus, it is essential to determine whether antidepressant treatment during pregnancy is safe not only with regard to the risk of congenital malformations,⁹ but also with regard to cognitive¹⁰ and neuropsychiatric development.^11,12 Numerous studies have assessed outcomes associated with prenatal exposure to antidepressants¹³; however, this is inherently difficult because randomized assessment of drug effects in pregnancy is not possible. Studies are, therefore, observational in nature and subject to varying degrees of confounding from, eg, underlying maternal psychiatric disorders.^10-12,14 Moreover, only few long-term studies have been conducted on prenatal exposure to antidepressants, especially studies focusing on cognitive outcome and school performance in the offspring.¹³

This study evaluated whether maternal prescription fill for antidepressants during pregnancy was associated with performance in standardized tests among Danish schoolchildren.

All data were analyzed at Statistics Denmark using encrypted identification numbers with no contact with individuals. By Danish law, analyses of anonymous data do not require ethical review board approval.

This was a population-based retrospective cohort study including all children born in Denmark between 1997 and 2009 with a plausible gestational age and whose parents were born in Denmark. The cohort was established using the Danish Medical Birth Register.¹⁵ Using unique personal identification numbers, parents were identified and cohort members were linked with individual-level information from other registers with national coverage.^16-18 From the cohort, we identified all children who had completed at least 1 language or mathematics test as part of the Danish National School Test Program between 2010 and 2018.

Information on maternal use of antidepressants was obtained from the Danish National Prescription Registry.¹⁷ Information included the date of dispensing and the Anatomical Therapeutic Chemical (ATC) classification code. Children were considered prenatally exposed to antidepressants if their mother had redeemed 1 or more prescriptions for drugs with the ATC code N06A from the date of the last menstrual period (LMP) minus 30 days to the day before birth. The date of the LMP was estimated by subtracting gestational age at birth (in days) from the date of birth recorded in the Medical Birth Register.¹⁵ Maternal prescription fill was considered for all antidepressants combined and for antidepressants by type (tricyclic antidepressants [TCAs], selective serotonin reuptake inhibitors [SSRIs], serotonin and norepinephrine reuptake inhibitors [SNRIs], and other antidepressants) and by individual antidepressant (the 12 most commonly used antidepressants; eTable 1 in the Supplement). The timing of exposure was subcategorized by whether prescriptions were redeemed during 4 periods: prepregnancy (LMP − 30 days to LMP − 1 day), first trimester (LMP to LMP + 97 days), second trimester (LMP + 98 days to LMP + 202 days), and third trimester (LMP + 203 days to the day before birth).

School performance was assessed using scores from standardized tests in language and mathematics completed as part of the Danish National School Test Program.^19,20 The tests were mandatory for all children attending public primary and lower secondary schools. Language ability was assessed in grades 2, 4, 6, and 8, covering 3 profile areas (language comprehension, decoding, and reading comprehension), and mathematics ability was assessed in grades 3, 6, and 8, also covering 3 profile areas (numbers and algebra, geometry, and statistics and probability). Tests were adaptive and computer-based and were initially scored according to a Rasch model within each profile area and then mapped into a single score ranging from 1 to 100 (higher scores indicated better test results).²⁰ A minimally clinically important difference was not established. The score was norm based and reflects the student’s performance as a percentile of the nationwide distribution of scores in the same test in the first year of testing (ie, in 2010 except 8th grade mathematics, which was introduced in 2018). The study included all assessments for the years 2010 to 2018.

Linear regression models were used to estimate the difference (Δ) in mean test scores between children with and without maternal prescription fill for antidepressants in pregnancy and the corresponding 95% CI. Normality was assessed by visual inspection, and generalized estimating equations (GEEs) were used to obtain robust standard errors, accounting for the lack of independence of test scores within each child. Separate models were fitted for language and mathematics performance, and examined for the association with maternal prescription fill for antidepressants overall, by type of antidepressant, and by individual antidepressant.

In the basic adjusted model, covariates were included for sex, grade, and year of birth, whereas the fully adjusted models also included maternal age, firstborn status, maternal and paternal psychiatric history (any registered diagnosis [International Classification of Diseases, Eighth Revision: 290-315 or International Statistical Classification of Diseases and Related Health Problems, Tenth Revision: F00-F99] in the Psychiatric Central Research Register²¹ at the time of birth), highest level of completed maternal education²² at the time of birth, paternal income²³ in the year of birth (adjusted to 2009 prices), and maternal prescription fill for antiseizure medication (ATC: N03A), anxiolytics (ATC: N05B), and antipsychotics (ATC: N05A) during pregnancy. Children with missing information on covariates were excluded from the fully adjusted analyses. In the combined analyses of all antidepressants, the association with school performance was analyzed by prescription timing and by interactions with sex, grade, and year of birth. In addition, results were dichotomized into below average (1-35) vs average and above average (36-100),²⁰ and analyzed using a logistic regression model with GEEs, accounting for the correlation of test scores within each child.

A number of sensitivity analyses tested the robustness of the findings. First, to account for confounding by indication, analyses were stratified by maternal hospital-based diagnosis of depression at birth. Second and third, 2 negative control exposure analyses were performed: one using children of mothers with prescription fills for antidepressants before (LMP − 395 days to LMP − 31 days) but not during pregnancy as comparison, and another using children of fathers, but not mothers, with prescription fills for antidepressants during the pregnancy period as comparison. Fourth, to account for family characteristics (eg, genetics, parental intelligence, socioeconomic status, or conditions in the postnatal environment), a sibling-controlled analysis was performed based on children with a 6th grade language and a 6th grade mathematics test, using linear regression models with parental fixed effects. Fifth, an analysis was carried out using marks from the 9th grade final examinations in language and mathematics, for which different methods for evaluating school performance were used.^22,24 Sixth, whether the association with school performance differed depending on psychiatric disorder in the child was examined. Seventh, children of mothers who redeemed only 1 prescription for antidepressants during pregnancy were reclassified as unexposed. Eighth, inverse probability weighting²⁵ was used to assess whether the results were biased due to nonparticipation. Ninth, to understand the underlying confounding, the basic adjusted model was expanded by sequentially adding the confounders from the full model, one at a time. Tenth, to account for variation within schools and educational districts, terms for each school and municipality were included in the GEE model. Additional method descriptions and details of sensitivity analyses are in the eMethods in the Supplement.

A significance threshold of .05 (2-sided) was used, and subgroup analyses were considered exploratory, and hence not adjusted for multiple comparisons. Data were analyzed using SAS software, version 9.4 (SAS Institute Inc).

Among 669 884 eligible children, 94 515 (14.1%) did not have information on school performance because they attended an independent school where participation in the test program was not required (n = 80 777; 12.1%), were exempted from testing (n = 6730; 1.0%), attended a special education school (n = 1431; 0.2%), or did not participate for unknown reasons (n = 5577; 0.8%). Nonparticipating children were generally born later in the study period, had mothers with higher educational levels, and had fathers with higher income compared with children who did participate, but the children had similar levels of parental psychiatric disorders (eTable 2 in the Supplement). Among the 94 515 nonparticipating children, 1918 (2.0%) were born to mothers filling an antidepressant prescription during pregnancy.

The final study population included 575 369 children with 1 392 370 language tests and 786 553 mathematics tests. The mean (SD) age of children (51.1% males) at the time of testing spanned from 8.9 (0.4) years in grade 2 to 14.9 (0.4) years in grade 8. In total, 10 198 children (1.8%) were born to mothers filling a prescription for antidepressants during pregnancy (characteristics are available in Table 1; eTable 3 in the Supplement). A total of 2897 children (0.5%) had missing information on 1 or more covariates and were excluded from the fully adjusted analyses.

Mean test scores in mathematics were 52.1 (95% CI, 51.7-52.6) for children with maternal prescription fill for antidepressants and 57.4 (95% CI, 57.3-57.4) for children without. Mean test scores in language were 53.4 (95% CI, 53.1-53.7) for children with maternal prescription fill for antidepressants and 56.6 (95% CI, 56.5-56.6) for children without. The test scores were slightly negatively skewed (eFigure 1 in the Supplement). In the basic adjusted model (including sex, grade, and year of birth), maternal prescription fill for antidepressants was significantly associated with a poorer performance in both mathematics (Δ = −6.0 [95% CI, −6.6 to −5.5]) and language (Δ = −3.0 [95% CI, −3.5 to −2.6]). However, when accounting for potential confounders in the fully adjusted model, performance was only significantly poorer in mathematics (Δ_adj = −2.2 [95% CI, −2.7 to −1.6]), not in language (Δ_adj = −0.1 [95% CI, −0.6 to 0.3]) (Table 2).

For the dichotomized outcome, the proportion of mathematics test scores in the below-average category was 28.6% among children with maternal prescription fill for antidepressants and 21.8% among children without (fully adjusted odds ratio, 1.16 [95% CI, 1.10-1.22]). For language, the proportion of test scores in the below-average category was 25.6% among children with maternal prescription fill for antidepressants and 21.7% among children without (fully adjusted odds ratio, 1.01 [95% CI, 0.96-1.05]).

In exploratory analyses, compared with children without maternal prescription fill for antidepressants, the difference in mathematics performance of children with maternal prescription fill for SSRIs was Δ_adj = −2.2 (95% CI, −2.7 to −1.6) (mean score, 52.2 [95% CI, 51.7-52.7]), for SNRIs was Δ_adj = −2.8 (95% CI, −4.7 to −1.0) (mean score, 50.5 [95% CI, 48.9-52.1]), for TCAs was Δ_adj = −1.0 (95% CI, −2.9 to 1.0) (mean score, 52.5 [95% CI, 50.8-54.2]), and for the group of other antidepressants was Δ_adj = −2.0 (95% CI, −3.9 to −0.1) (mean score, 51.1 [95% CI, 49.4-52.7]) (Table 2). Analyses of subtypes of SSRIs and other antidepressants are shown in eTable 4 in the Supplement.

The academic performance of children of mothers filling prescriptions for antidepressants in various pregnancy periods was compared with children without any maternal prescription fill for antidepressants. Among children of mothers who filled antidepressant prescriptions in multiple trimesters (mean mathematics score: 51.8 [95% CI, 51.2-52.4]; mean language score: 53.6 [95% CI, 53.2-54.0]), the difference was Δ_adj = −3.2 (95% CI, −3.8 to −2.5) in mathematics and Δ_adj = −0.5 (95% CI, −1.1 to 0.0) in language. Among children of mothers who only filled prescriptions for antidepressants during the first trimester (mean mathematics score, 51.2 [95% CI, 50.2-52.3]; mean language score, 52.6 [95% CI, 51.9-53.4]), the difference was Δ_adj = −1.7 (95% CI, −2.9 to −0.4) in mathematics and Δ_adj = 0.4 (95% CI, −0.7 to 1.4) in language (Table 3).

For the interaction analyses of the association with mathematics performance, the lower performance associated with maternal antidepressant prescription fill was more pronounced in higher grades compared with lower grades (test for interaction with grade, P < .001) (Figure). For the interaction analyses of the association with language performance, maternal antidepressant prescription fill was significantly associated with lower performance in boys (mean test scores: exposed, 50.6 [95% CI, 50.1-51.0]; not exposed, 54.2 [95% CI, 54.1-54.2]; Δ_adj = −0.7 [95% CI, −1.3 to 0.0]) but not in girls (mean test scores: exposed, 56.4 [95% CI, 56.0-56.9]; not exposed, 59.1 [95% CI, 59.0-59.1]; Δ_adj = 0.4 [95% CI, −0.2 to 1.1]; test for interaction with sex, P = .01; Figure). For language performance, there was also an interaction with year of birth (P = .007), although no clear trend was observed (eg, increasing or decreasing by year of birth) (eFigure 2 in the Supplement).

The potential influence of residual confounding and other sources of biases were addressed in sensitivity analyses, which are summarized in Table 4. The analyses consistently confirmed that maternal prescription fill for antidepressants was significantly associated with a poorer performance in mathematics but not in language. For instance, in the sibling-controlled analysis, the adjusted difference in mathematics (mean scores: exposed to maternal prescription fill, 53.5 [95% CI, 52.7-54.3]; not exposed, 59.0 [95% CI, 58.9-59.1]) was Δ_adj = −2.8 (95% CI, −4.5 to −1.2) and in language (mean scores: exposed to maternal prescription fill, 53.9 [95% CI, 53.2-54.6]; not exposed, 56.6 [95% CI, 56.5-56.7]) was Δ_adj = −0.3 (95% CI, −1.9 to 1.2); however, in other analyses (eg, the negative-controlled analyses), the association was attenuated. Additional details of all sensitivity analyses are provided in the Supplement, including analyses addressing potential bias from confounding by indication (eTables 5-7 in the Supplement), family context and genetic predisposition (eTable 8 in the Supplement), the validity of the school performance measure (eTable 9 in the Supplement), the role of child psychiatric disorders (eTable 10 in the Supplement), misclassification of maternal antidepressant exposure (eTable 11 in the Supplement), and bias from nonparticipation (eTable 12 in the Supplement).

Performing sequential adjustment for each of the confounders suggested that maternal education and maternal psychiatric disorders were the 2 most important confounders (eTable 13 in the Supplement). Accounting for the variation within schools and educational districts did not change the results (mathematics Δ_adj = −2.2 [95% CI, −2.7 to −1.6]; language Δ_adj = −0.1 [95% CI, −0.6 to 0.3]).

Quiz Ref IDIn this study of more than 575 000 Danish school-aged children, maternal antidepressant prescription fill in pregnancy was consistently and significantly associated with a lower test score in mathematics but not in language tests after adjusting for multiple potential confounders, including socioeconomic factors. Maternal prescription fill for antidepressants was associated with a statistically significant higher odds ratio of scores in the category below average in mathematics, but the observed difference in mean test scores was small (2 points out of 100 points) and unlikely represents a clinically relevant difference in mathematical skills. The significant association with mathematics performance persisted in sensitivity analyses, but most of these sensitivity analyses attenuated the association. Because the magnitude of the difference was small and because the difference was attenuated in the sensitivity analyses, the finding must be weighed against the benefits of treating maternal depression during pregnancy.

The different antidepressant medication classes assessed in this study all cross the placenta and thus reach the fetus.²⁶ Except for norepinephrine and dopamine reuptake inhibitors (bupropion), and to a lesser degree also TCAs, the other antidepressants (SSRIs, SNRIs, and monoamine oxidase inhibitors) all influence serotonergic (5-HT) neuronal synaptic transmission.²⁷ Serotonin and serotonergic neurons play a key role in neurodevelopment and regulate cell growth.²⁶ Serotonin levels during early brain growth are central to the regulation of social/emotional and cognitive development.^26,28 TCAs have multiple mechanisms of action,²⁹ and bupropion acts selectively to inhibit the noradrenergic and dopaminergic reuptake transporter pump systems.²⁹ Prenatal exposure to maternal fill for TCAs and bupropion was not significantly associated with lower test scores in mathematics. However, no conclusive evidence with regard to potential mechanism can be provided from an observational study, and the power of analyses of antidepressants by subtype was limited for some subgroups.

To our knowledge, no previous study has examined the association between maternal antidepressant use and language and mathematics abilities in school tests.^13,30 Others have assessed associations with preschool speech, language, special education needs, and delayed school entry.^31-33 A Finnish study showed that preschool children of mothers purchasing SSRIs during pregnancy had a 37% higher risk of being diagnosed with a speech or language disorder than unexposed children.³¹ In a study of 178 mother-child dyads (child age, 2.5-5.5 years), the authors described an association between prenatal SSRI exposure and expressive language and behavior problems.³² A Danish study found that in utero SSRI exposure in all 3 trimesters was associated with delayed elementary school entry but not with special education needs.³³Quiz Ref ID In the present study, the crude estimates showed that children with maternal antidepressant prescription fill had lower language test scores, but this was explained by the covariates included in the fully adjusted model, suggesting that the lower language scores were related to maternal characteristics associated with antidepressant use and not to prenatal antidepressant exposure per se. Thus, the findings of this study do not support an association between prenatal antidepressant use and specific developmental disorders of speech and language.³¹

Mathematics abilities may be associated with IQ, but 5 studies examining IQ in the offspring of women who had used antidepressants in pregnancy found no association with offspring IQ after adjusting for maternal disorders.¹³ In contrast, this study identified an association between maternal prescription fill for antidepressants and poorer results in mathematics. The association with mathematics performance was unexpected; yet, the amount of sensitivity analyses supporting the finding suggests that the result was not attributable to unmeasured confounding or lack of validity of the school performance measure. An alternative explanation for the finding may be that the educational system was more proficient in compensating for language deficiencies given that the minimum hours of language education across the years of schooling exceed hours of mathematics education by more than 60%.³⁴Quiz Ref ID This hypothesis is supported by the finding that the difference in mathematics ability between children with and without maternal prescription fill for antidepressants increased with increasing school grade. However, mathematical skill acquisition is hierarchical and each step of increased set of mathematical skills builds on knowledge of lower levels,³⁵ and requires engagement of more complex neurocognitive networks in the brain.³⁵ The greater difference in mathematical tests at higher grades could reflect that an association between prenatal exposure to antidepressants and mathematics becomes more evident in higher grades as larger and more widespread functions of the brain are required.³⁵

The present study offers several advantages over previous studies. First, it was based on systematically collected long-term data with no recall bias. Second, the study was large, which yielded great precision in the school performance estimates. Third, using the many Danish nationwide registers enabled adjustment for a comprehensive range of potential confounders that are likely important given that the crude differences in school performance were greatly attenuated in the fully adjusted models. Fourth, the prescription register has complete coverage because the reimbursement structure provides a strong economic incentive for recording all dispensed drugs. Fifth, the school tests have been demonstrated to have good test properties (eg, the simulation-based test-retest reliability and the Person Separation Index score for reliability was greater than 0.8 in all profile areas of language and mathematics) and they were standardized across the entire nation, thereby limiting any teacher bias.³⁶

This study has several limitations. First, not all children attended public schools, and therefore the cohort was subject to selection. However, more than 85% of all children participated in 1 or more school tests and the proportion of children with maternal antidepressant prescription fill was only slightly different among participating and nonparticipating children. Inverse probability weighting was used to account for differences in characteristics, such as maternal education and paternal income, between children who did and did not participate in the national tests. However, this analysis did not suggest that the poorer test results in mathematics were explained by nonparticipation.

Second, although the analyses included adjustment for hospital-based diagnoses of maternal and paternal psychiatric disorders, this may not have fully accounted for maternal depression if the depression did not rise to the level of requiring hospitalization. However, it is not clear how this limitation would apply differently to language and mathematics skills, and this limitation did not apply to, eg, the sibling analyses.

Third, it was assumed that the pregnant women who redeemed a prescription for antidepressants also ingested the medication. One Danish study reported 80% adherence for antidepressant medication prescribed to pregnant women,³⁷ suggesting that some children who were considered exposed were, in fact, not. However, analyses restricting the exposure definition to at least 2 prescriptions (where such misclassification was reduced) were in line with the main findings.

Fourth, the pattern of use of antidepressants in Denmark^3-6 and potential confounding factors, such as maternal education and income, may differ from other countries. This may limit the generalizability, but also provides an opportunity to evaluate the role of these biasing structures in future studies.

In this study of public schoolchildren in Denmark, children whose mothers had filled prescriptions for antidepressants during pregnancy, compared with children whose mothers did not fill prescriptions for antidepressants during pregnancy, had a 2-point lower standardized test score in mathematics, a difference that was statistically significant, but had no significant difference in language test scores. The magnitude of the difference in the mathematics test score was small and of uncertain clinical importance, and the findings must be weighed against the benefits of treating maternal depression during pregnancy.

Corresponding Author: Jakob Christensen, DrMedSci, PhD, Department of Neurology, Aarhus University Hospital, Palle Juul-Jensens Blvd 165, DK-8200 Aarhus N, Denmark (jakob@farm.au.dk).

Accepted for Publication: September 15, 2021.

Author Contributions: Drs Trabjerg and Dreier had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Christensen, Dreier.

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

Drafting of the manuscript: Christensen, Dreier.

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

Statistical analysis: Christensen, Trabjerg.

Obtained funding: Christensen.

Administrative, technical, or material support: Christensen.

Supervision: Christensen, Dreier.

Conflict of Interest Disclosures: Dr Christensen reported receiving personal fees from UCB Nordic and Eisai AB during the conduct of the study. No other disclosures were reported.

Funding/Support: The study was funded in part by the Central Denmark Region and the Novo Nordisk Foundation (NNF16OC0019126).

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.

Data Sharing Statement: Data were based on Danish national registers, and individual-level data cannot be shared. However, summary statistics, in addition to the results provided in the Results section and Supplement, may be provided on request.