Prenatal Carbamazepine Exposure and Academic Performance in Adolescents

Abstract

Background and Objectives

To investigate whether children born to mothers who used carbamazepine during pregnancy had worse academic performance in adolescence.

Methods

This population-based cohort study included all live-born singletons in Denmark between 1996 and 2002 who participated in the national ninth-grade exit examination (n = 370,859). Those born to mothers with prescription of antiseizure medications other than carbamazepine during pregnancy were excluded. We examined the association of in utero exposure to maternal carbamazepine redeemed during pregnancy (n = 290) with academic performance of offspring, defined by the scores in Danish and mathematics in ninth-grade exit examination. We estimated mean z-score difference with linear regression adjusted for socioeconomic factors and potential indications, including epilepsy and medication for other psychiatric disorders. Additional analyses addressing confounding by indication included comparison between in utero exposed vs past exposed and between past exposed and never exposed. In utero exposure to valproate monotherapy was used as a positive control and in utero exposure to lamotrigine as a negative control.

Results

At the age of 16.1 (SD 0.4) years, adolescents in utero exposed to maternal carbamazepine monotherapy had lower scores both in Danish and mathematics in ninth-grade exit examination (adjusted z-score difference, −0.14 [95% CI −0.24 to −0.05] and −0.17 [95% CI −0.28 to −0.07], respectively). In utero exposure to carbamazepine monotherapy was associated with lower scores than past exposure only (adjusted z-score difference, −0.24 [95% CI −0.41 to −0.06] for Danish and −0.25 [95% CI −0.44 to −0.06] for mathematics), while past exposure to carbamazepine was associated with minor decrease in offspring's academic performance (adjusted z-score difference, −0.02 [95% CI −0.09 to 0.06] for Danish and −0.07 [95% CI −0.16 to 0.01] for mathematics). The association was also observed for in utero exposure to valproate monotherapy, but not for in utero exposure to lamotrigine.

Discussion

In utero exposure to carbamazepine was associated with poorer academic performance in adolescence, as represented by lower scores in ninth-grade exit examination in Danish and mathematics. Additional studies are needed to confirm these findings because of limitations in this study and variable findings in prior studies.

Classification of Evidence

This study provides Class III evidence that academic performance, as reflected in ninth-grade exit examinations in Danish and mathematics, was worse among those exposed to carbamazepine monotherapy in utero, compared with those without in utero exposure to antiseizure medications.

The prevalence of epilepsy among women peaks during childbearing years.¹ Antiepileptic treatment during pregnancy is challenging to balance seizure control and fetal adverse effects of antiseizure medications (ASMs).² Carbamazepine is one of the most frequently prescribed ASMs for pregnant women,^3,4 which has been associated with neural tube defects in their offspring, while its teratogenicity was suggested to be milder than valproate.^5,6 Regarding the associated long-term neurodevelopmental outcomes in offspring, previous studies have yielded inconsistent results.⁷

Animal studies have shown that prenatal carbamazepine exposure could reduce hippocampal and cortical neuronal cell population and degenerative changes in cerebellar cortex, although a difference in learning behavior was not observed.^8,9 Although one study found an association between prenatal carbamazepine and intellectual disability in offspring,¹⁰ most population studies have, however, yielded inconsistent results on IQ and academic performance in offspring.^11-15 Those studies were mostly limited by insufficient power, selection bias because of nonparticipation of high-risk individuals, short follow-up time of less than 6 years, and residual confounding from other indications such as concomitant psychiatric disorders.^11-15

Primary Research Question

To investigate the association between in utero exposure to maternal carbamazepine and academic performance in offspring, defined by scores in the national ninth-grade exit examination, in a population-based cohort in Denmark, taking into consideration the effects of a number of maternal or child factors, as well as the confounding by indication.

Methods

Study Population

This population-based cohort study included all live-born singletons between 1996 and 2002 in Denmark (n = 447,363) identified from the Danish Medical Birth Registry, which contains maternal and birth characteristics of every live birth and stillbirth from 1973.¹⁶ Children born in Denmark were assigned a unique civil registration number at birth, which permitted accurate linkage to various nationwide registers with nearly complete follow-up.¹⁶ Those with unrecorded sex (n = 21), born with an unrecorded (n = 12,134) or unrealistic gestational age (younger than 22 weeks or older than 45 weeks, n = 68), died (n = 2,527) or emigrated (n = 183) before age 15 years, in utero exposure to ASMs other than carbamazepine (n = 1,568), and no accessible ninth-grade examination scores on Danish and mathematics (n = 60,003) were excluded. Reasons for no accessible scores included participation to only part of the examination (n = 11,612), exempt from the examination (n = 3,967), taking an optional 10th grade examination instead (n = 5,638), and unknown (n = 38,786).

Exposure

The Danish National Prescription Registry contains all redeemed prescriptions in Denmark since 1995, from which we identified information on maternal use of ASMs.¹⁶ Information included the type of drugs coded according to the Anatomical Therapeutical Chemical (ATC) system, date of dispensing, and dose. Prescription redeemed from 30 days before the date of maternal last menstrual period (LMP) to the date of birth was defined as in utero exposure to adolescents. ASMs were identified with ATC codes (eTable 1, links.lww.com/WNL/C468).¹⁷ Although in utero exposure to carbamazepine was of interest in this study, we also used in utero exposure to valproate and lamotrigine to identify positive and negative control groups, respectively. The control groups were selected based on the established association between prenatal exposure to valproate and a poorer academic performance as well as a better safety profile of lamotrigine.¹³ Adolescents who were not in utero exposed to any ASM within the exposure time window were defined as the unexposed.

Mothers redeeming only one type of ASM within the exposure time window were defined as having monotherapy, otherwise polytherapy. The timing of exposure was defined as the date of prescription for carbamazepine and further categorized into prepregnancy (LMP− 30 days to LMP− 1 day), first (LMP to LMP+ 97 days), second (LMP+ 98 days to LMP+ 202 days), and third trimester (LMP+ 203 days to the day of childbirth). Only 37 of 290 women had carbamazepine in a single period of the 4 categories, making it impossible to compare risk estimates in each exposure window. Thus, we alternatively categorized the exposure timing into 2 periods: prepregnancy to first trimester and second to third trimesters. Children born to mothers who had prescription at both periods were categorized as being exposed at multiple periods. We also estimated mean daily dose of carbamazepine, which was calculated by the total amount of carbamazepine prescribed during the whole exposure window divided by the days of the same period.¹⁸ The exposure was further subcategorized according to the estimated mean daily dose (≤500, 501–1,000, >1,000 mg/d). We also applied 2 alternative methods to define the mean dose of ASMs for comparison: considering the length of exposed time by trimester and using the date of first prescription for carbamazepine as the start date of exposure (detailed in eAppendix 1, links.lww.com/WNL/C468).

Outcome

In Denmark, compulsory school starts at age 6 years and follows by 9 years of primary and secondary school, usually ends at age 15–16 years.¹⁹ Public and private schools are similarly structured and are completed with an exit examination at ninth grade as requested by the law.¹⁹ Written tests for Danish and mathematics are the same nationwide and each comprise several parts.¹⁹ We used the z-score to standardize the scores according to each part of both subjects (see eAppendix 2, links.lww.com/WNL/C468). Scores of other school subjects were also identified as alternate outcomes for sensitivity analysis.

Covariates

Birth year, sex, parity, and maternal age at birth were retrieved from the Danish Medical Birth Registry.¹⁶ Adolescents of Danish origin was defined as having at least 1 parent who is both a Danish citizen and born in Denmark and was retrieved from the Civil Registration System.²⁰ The maternal level of education and family income were identified from the Danish Integrated Database for Longitudinal Labor Market Research.¹⁶ The Danish National Patient Register and the Danish Psychiatric Central Research Register were used to identify diagnosis of epilepsy and any psychiatric disorders with specific ICD codes (eTable 2, links.lww.com/WNL/C468).^16,21 The maternal history of epilepsy/any psychiatric disorder was defined as a corresponding diagnosis before birth. Maternal redeemed prescriptions for antidepressants, antipsychotics, or anxiolytics in the exposure window were identified from the Prescription Registry (eTable 2, links.lww.com/WNL/C468). In addition, children diagnosed with epilepsy before the ninth-grade examination were also identified. A major congenital anomaly, defined by the European Surveillance of Congenital Anomalies classification system, was identified from the Patient Registry with an established algorithm (eTable 2, links.lww.com/WNL/C468).²² For adolescents, psychiatric disorders additionally included attention deficit and hyperactivity disorder (ADHD), which was defined by either diagnosis in the Psychiatric Registry or ADHD-specific medication in the Prescription Registry (eTable 2, links.lww.com/WNL/C468).

Statistical Analyses

Linear regression was used to estimate the z-score difference in Danish and mathematics, respectively, between adolescents in utero exposed to carbamazepine and adolescents without in utero exposure to any ASM. We adjusted for sex (male or female) and year of examination (each year as 1 category) in model 1. Fully adjusted model (model 2) additionally included maternal level of education (0–9, 10–14, or 15 years or older), family income (categorized by quartiles in the full population), maternal age at childbirth (younger than 20, 20–24, 25–29, 30–34, or 35 years or older), parity (1, 2, or ≥3), whether the adolescent was of Danish origin (yes or no), maternal history of epilepsy (yes or no), maternal history of any psychiatric disorder (yes or no), in utero exposure to antidepressant (yes or no), antipsychotics (yes or no), and anxiolytics (yes or no). Missing data existed only in maternal education (0.5%) and family income (<0.1%) with very low proportions for which missing indicator method was applied. Based on model 1, we additionally adjusted for potential confounders in model 2, one at a time, to compare risk estimates with model 1 and investigate the effects of major confounders in the association.

We performed the following analyses to take confounding by indication into consideration: (1) the risk estimates for in utero exposure to carbamazepine was compared with in utero exposure to valproate and lamotrigine, respectively. (2) In utero exposure was compared with past exposure (more than 30 days before LMP) to carbamazepine. In addition, past exposure was alternately defined by more than 1 year before LMP for comparison. (3) Past exposure was compared with never exposure (no prescription before childbirth) to carbamazepine to assess the association between the indications with outcomes. (4) The analyses were also performed stratified by the maternal history of epilepsy.

To assess the potential mediation role of the following variables, the analyses were performed restricted to adolescents with no corresponding traits, one each time, including major congenital anomaly, any psychiatric disorder, epilepsy, preterm birth, and low birth weight. We assessed the differences in the risk estimates regarding timing of exposure, dose of exposure, and exposure to carbamazepine polytherapy. Additional analyses were performed by restricting on adolescents with no missing covariates, defining exposure as 2 or more prescriptions, using inverse probability weighting to balance characteristics with adolescents not participated in the examination (detailed in eAppendix 3, links.lww.com/WNL/C468), and using scores in other subjects in ninth-grade examination as outcomes. Subgroup analyses were performed stratified by sex and birth year. Adolescents in utero exposed to other ASM monotherapy were also identified for comparison. To consider potential confounding effect by severity of anxiety and depression during pregnancy, maternal hospital admission for mood and anxiety disorders (diagnostic codes in eTable 2, links.lww.com/WNL/C468) during the exposure window was used as a proxy and additionally adjusted. In addition, we used paternal prescription for carbamazepine in the exposure window as a negative control. Sibling-matched analysis was not applicable because of limited sample size (#outcome <20 in the exposed group). SAS 9.4 (SAS Institute, Cary, NC), Stata 15 (StatCorp, College Station, TX), and R version 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria) were used for the analyses.

Standard Protocol Approvals, Registrations, and Patient Consents

This study was approved by the Danish Data Protection Agency (Record No. 2013-41-2569). By Danish law, no informed consent is required for a registry-based study using anonymized data.

Data Availability

All data are stored at the secure platform of Denmark Statistics, which is the central authority on Danish statistics with the mission to collect, compile, and publish statistics on the Danish society. Owing to restrictions related to Danish law and protecting patient privacy, the combined set of data as used in this study can only be made available through a trusted third party, Statistics Denmark (dst.dk/en/kontakt). This state organization holds the data used for this study. University-based Danish scientific organizations can be authorized to work with data within Statistics Denmark. Researchers can apply for access to these data when the request is approved by the Danish Data Protection Agency: datatilsynet.dk, the email address for the Danish Data Protection Agency is dt@datatilsynet.dk. Requests for data may be sent to Statistics Denmark: dst.dk/en/OmDS/organisation/TelefonbogOrg.aspx?kontor=13&tlfbogsort=sektion or the Danish Data Protection Agency: datatilsynet.dk.

Results

Of the 370,859 adolescents who participated in the ninth-grade final examination, 290 were in utero exposed to carbamazepine monotherapy. Exposed children were more likely to be a second-born child, to have a mother at lower level of education and family income, to have maternal diagnosis of a psychiatric disorder, and to be in utero exposed to antidepressants, antipsychotics, or anxiolytics (Table 1).

Table 1.

Baseline Characteristics of Study Population

The mean (SD) age of adolescents at the time of examination was 16.1 (0.4) years. Adolescents in utero exposed to carbamazepine monotherapy had lower scores both in Danish and mathematics in ninth-grade exit examination (mean z-score, −0.26 [95% CI −0.36 to −0.17] and −0.34 [95% CI −0.44 to −0.24], respectively). In both basic and adjusted models, in utero exposure to carbamazepine monotherapy was associated with a poorer performance in Danish and mathematics (z-score difference of the full model, −0.14 [95% CI −0.24 to −0.05] and −0.17 [95% CI −0.28 to −0.07], respectively; Figure 1). Sequential adjustment for each confounder showed that maternal education and maternal epilepsy were among the most important potential confounders (eTable 3, links.lww.com/WNL/C468).

Figure 1. Association of In Utero Exposure to Carbamazepine, Valproate, and Lamotrigine With Scores in Danish and Mathematics in Ninth-Grade Exit Examination.

Basic adjusted model: adjusted for sex and year of examination. Fully adjusted model: adjusted for sex, year of examination, Danish origin, maternal parity, maternal age at delivery, maternal education, family income, maternal diagnosis of epilepsy, maternal diagnosis of psychiatric disorders, and maternal fill of antidepressants, antipsychotics, and anxiolytics during pregnancy. ASM = antiseizure medication.

To account for confounding by indication, we performed the following 4 analyses: First, when using in utero exposure to valproate monotherapy as a positive control, we observed that the magnitude of the association between in utero exposure to carbamazepine monotherapy and academic performance was comparable with that of valproate (Figure 1). As expected, we did not observe an association of in utero exposure to lamotrigine, a negative control, with academic performance (Figure 1). Second, the academic performance of offspring of mothers with gestational exposure to carbamazepine monotherapy was poorer than with only past exposure (adjusted z-score difference, −0.21 [95% CI −0.37 to −0.04] for Danish and −0.21 [95% CI −0.39 to −0.03] for mathematics; Table 2). The results were consistent when defining past exposure as 1 year before last LMP (eTable 4, links.lww.com/WNL/C468). Third, compared with women never exposed to any ASM, past exposure to carbamazepine was associated with only minor decrease in offspring's academic performance (adjusted z-score difference, −0.02 [95% CI −0.09 to 0.06] for Danish and −0.07 [95% CI −0.16 to 0.01] for mathematics; Table 2). Fourth, among adolescents with no maternal hospital diagnosis of epilepsy, in utero exposure to carbamazepine monotherapy was associated with worse academic performance (adjusted z-score difference, −0.36 [95% CI −0.54 to −0.17] for Danish and −0.40 [95% CI −0.61 to −0.18] for mathematics; Figure 2). Among adolescents born to mother diagnosed with epilepsy, the association attenuated (adjusted z-score difference, −0.08 [95% CI −0.18 to 0.03] for Danish and −0.11 [95% CI −0.23 to 0.01] for mathematics; Figure 2); still, carbamazepine-exposed adolescents had poorer academic performance, compared with lamotrigine-exposed adolescents (adjusted z-score difference, −0.16 [95% CI −0.33 to 0.02] for Danish and −0.28 [95% CI −0.48 to −0.09] for mathematics).

Table 2.

Association of Gestational Exposure, Past Exposure, and Never Exposure to Carbamazepine With Academic Performance in Offspring

Figure 2. Association of In Utero Exposure to Carbamazepine, Valproate, and Lamotrigine With Scores in Danish and Mathematics in Ninth-Grade Exit Examination, Stratified by Maternal Diagnosis of Epilepsy.

Basic adjusted model: adjusted for sex and year of examination. Fully adjusted model: adjusted for sex, year of examination, Danish origin, maternal parity, maternal age at delivery, maternal education, family income, maternal diagnosis of psychiatric disorders, and maternal fill of antidepressants, antipsychotics, and anxiolytics during pregnancy. ASM = antiseizure medication.

Adolescents born to mothers filling prescriptions for carbamazepine seemed to have a higher rate of major congenial anomaly compared with unexposed adolescents (4.5% vs 2.6%). After excluding adolescents diagnosed with congenial anomaly, in utero carbamazepine-exposed children still had worse academic performance than those unexposed (eTable 5, links.lww.com/WNL/C468). Carbamazepine-exposed adolescents were more frequently to be diagnosed as epilepsy (6.9% vs 1.9%, diagnosed before age 15 years). Excluding these adolescents did not alter the risk estimates substantially (eTable 5, links.lww.com/WNL/C468). Similarly, analyses restricted to adolescents with no diagnosis of psychiatric disorder before age 15 years showed similar results as previously (eTable 5, links.lww.com/WNL/C468).

Among the 290 adolescents in utero exposed to carbamazepine monotherapy, 239 (82.4%) were exposed in both early and late period of pregnancy. We observed no clear trend in the association according to different timing of prescription because of large variance in the estimation (Table 3). The estimated median (first and third quartiles) dose of carbamazepine exposure was 586 mg/d (387–795 mg/d). In utero exposure to various doses of carbamazepine was consistently associated with poorer academic performance in both Danish and mathematics, although an obvious dose trend was not observed (Table 3). We additionally identified 50 adolescents in utero exposed to carbamazepine polytherapy, mostly dual therapy (N = 40) and in combination with vigabatrin (N = 14) and lamotrigine (N = 12). Carbamazepine polytherapy was associated with a more prominent decrease in academic performance in both Danish and mathematics (Table 3).

Table 3.

Polytherapy, Time of Exposure, and Dose of Exposure in the Association of In Utero Exposure to Carbamazepine With Academic Performance in Offspring

The association between in utero exposure to carbamazepine monotherapy and worse academic performance was consistent when restricting the analyses to adolescents with no missing covariates, excluding adolescents born with preterm birth and low birth weight, using repeated maternal prescription of carbamazepine in pregnancy as the exposure, using inverse probability weight to account for bias from nonparticipation, using scores in other subjects as outcomes, and in sex-specific analyses (eTables 6–7, links.lww.com/WNL/C468). Stratified by birth year, the risk estimates tended to be lower in adolescents born in more recent years (eFigure 1, links.lww.com/WNL/C468). Risk estimates for other ASMs are also presented in eTable 8, links.lww.com/WNL/C468. Additional adjustment for maternal hospital admission for mood and anxiety disorders did not alter the estimates (eTable 6, links.lww.com/WNL/C468). Considering the length of exposed time by trimester or using the date of first prescription for carbamazepine as the start date of exposure yielded similar results to those in our main analysis (eTable 9, links.lww.com/WNL/C468). As expected, parental using of carbamazepine in the same exposure window, as a negative control, was not associated with adolescents' academic performance (eTable 10, links.lww.com/WNL/C468).

Classification of Evidence

This study provides Class III evidence that academic performance, as reflected in ninth-grade exit examinations in Danish and mathematics, was worse among those exposed to carbamazepine monotherapy in utero, compared with those without in utero exposure to ASMs.

Discussion

In this large population-based cohort study, we observed that in utero exposure to carbamazepine was associated with poorer academic performance in adolescents, represented by lower scores in ninth-grade exit examination in Danish and mathematics, respectively. The association was consistent in the analyses addressing confounding by indication, nonparticipation, misclassification bias of exposure, and information bias of outcomes, although some of the analyses showed attenuated risk estimates. This association was unlikely to be explained by higher incidence of major congenital anomaly, epilepsy, and psychiatric disorders in the exposed adolescents before the examination. In addition, carbamazepine polytherapy was associated with greater pronounced decrease in academic performance.

Previous studies on the association of in utero exposure to carbamazepine with IQ or academic performance in offspring have yielded conflicting results, mostly because of methodologic limitations such as selection bias and small sample size.^12,14,23-28 A population-based study involving 160 exposed children found that prenatal exposure to carbamazepine was associated with a lower probability of achieving “excellence” in English, Swedish, and mathematics at age 16 years.¹⁴ The Neurodevelopmental Effects of Antiepileptic Drugs study reported an association between prenatal exposure to carbamazepine and lower IQ at age 6 years,²⁹ although a dose-response effect was not observed.^23,24 Other studies, each involving less than 100 exposed offspring, reported no association between prenatal carbamazepine and lower IQ younger than 6 years.^12,26,27 A record-linkage study performed in a Danish population found no difference in academic tests of Danish and mathematics at second to eight grade between children exposed and not exposed to prenatal carbamazepine monotherapy.¹³ In our study, we used the score of exit examination at ninth grade with higher participation rate (82.9% of all eligible children, compared with only 20% children followed to the 8th grade in the previous Danish report), which could reduce selection bias from nonparticipation and resulted in a larger sample size (N = 290 vs 121 for exposed children with eight grade scores in the previous Danish report). In addition, we were able to take into consideration of a number of potential confounders including socioeconomic factors and other indications for carbamazepine.

To be noted, regarding other neurodevelopmental outcomes of offspring, including developmental quotient, autism, language impairment, and intellectual disability, a consistent negative association was observed with in utero exposure to valproate while the results were inconsistent with in utero exposure to carbamazepine.^18,28,30-32 In our study, in utero exposure to carbamazepine and valproate showed similar magnitudes of association on academic performance in Danish and mathematics at age 16 years. Studies assessing neurodevelopmental outcomes in adolescence were scarce, and our findings might indicate a long-term effect of in utero exposure to carbamazepine. On the other hand, our findings should be interpreted in the context of previous knowledge,^7,13,18,28 which suggests potential overestimation of the effect of carbamazepine or underestimation of the effect of valproate in our findings. Selection bias from nonparticipation could be an explanation because most heavily affected children (by prenatal exposure to valproate) would require special education and not participate the exit examination.¹⁹ In addition, despite a consistent association between prenatal carbamazepine and poorer academic performance in most analyses in this study, we should note that when the analyses were restricted to adolescents born to mothers with epilepsy, the difference was not statistically significant. A dose-dependent association between prenatal carbamazepine and academic performance was not observed either. Again, both analyses were limited by small sample sizes, which prevented us from yielding informative findings. Future studies are critically needed to validate these findings.

Carbamazepine, similar to valproate, was found to be associated with an increased risk in folate-sensitive birth defects, such as neural tube defects and orofacial clefts, indicating impairments of folate absorption in fetus.^5,33 Folate-related pathways play a pivot role in fetal neurodevelopment, which could affect long-term behavioral and cognitive outcomes, as manifested in prenatal exposure to valproate.^13,18,34 Prenatal folate supplementation might improve cognitive outcomes in offspring exposed to ASMs, including carbamazepine and valproate.³⁵ The Danish Health and Medicines Authority has recommended daily folic acid supplementation for all pregnant women and high dose (5 mg/d) for women with epilepsy.³⁶ A Danish cohort of 1996–2003 reported more than 75% of pregnant women with epilepsy taking folic acid during pregnancy,³⁷ and the adherence to folic acid showed an increasing trend.³⁸ This might partially explain a decreasing trend in the magnitude of the association between in utero exposure to carbamazepine and academic performance over recent years.

In the subgroup of adolescents born to mothers with epilepsy, the difference in academic performance attenuated between carbamazepine-exposed and unexposed adolescents. Using lamotrigine-exposed adolescents as the comparison group, carbamazepine-exposed adolescents showed consistently lower scores in this subgroup. Unexpectedly, among adolescents with maternal epilepsy, lamotrigine-exposed adolescents showed better academic performance than the unexposed adolescents, which might indicate selection bias that could result in underestimation of the association with carbamazepine. Women diagnosed with epilepsy who did not take ASMs might have poorer adherence to treatment^39,40 that could be due to lower education level and difficulty in accessing health care, which could contribute to poor academic performance in offspring. However, the observed risk estimates, which were comparable with risk estimates for prenatal exposure to valproate, were not likely to be fully explained by selection bias alone.

The estimated effect size of in utero exposure to carbamazepine on scores in ninth-grade exit examination was ∼0.15 (Cohen⁴¹ D; 0.14 for Danish and 0.17 for mathematics), which could be considered of small clinical significance at an individual level. In other words, there was a probability of superiority of 54.2% that an adolescent picked at random from the unexposed group would have a higher score than an adolescent picked at random from the carbamazepine-exposed group. In comparison, the maternal low education level (0–9 years) was associated with an effect size of 0.37 (Cohen D), compared with the middle education level (10–14 years), corresponding to a probability of superiority of 60.3%. However, academic performance in adolescents is a predictor of future education and employment,^42,43 which thus remains of importance at a population level.

Consistent with previous studies, in utero exposure to lamotrigine was associated with a relatively low risk, suggesting that it might be a more promising candidate for epilepsy control during pregnancy.^13,23 However, when lamotrigine failed to control epilepsy, other ASMs could not be avoided, that is, for those with epilepsy and severe psychiatric disorders. ASM use in pregnancy needs to consider the benefits of maintaining seizure control (or other diseases) vs risk of ASM exposure to the fetus. Given the fact that information on safety in pregnancy is available for a small minority of ASMs, clinicians frequently face with using ASMs with conflicting (such as carbamazepine), inadequate, or no data to guide their decisions. Thus, future validation of our findings, as well as investigations of safety profile of other ASMs, is critical. Carbamazepine polytherapy was found to be associated with a more prominent decrease in academic performance, although the number of adolescents exposed was limited. This could reflect either change of medication during pregnancy or using multiple types of ASMs.

Our study has several strengths. First, the nationwide registry system in Denmark permitted a population-based design and nearly complete follow-up in a long time span of 16 years, which minimized potential selection bias and recall bias. Second, the registers provided comprehensive data on potential confounders and other characteristics for sensitivity analysis addressing confounding by indication, nonparticipation, misclassification bias of exposure, and information bias of outcomes. Third, the universally compulsory exit examination had a high participation rate, resulting in less selection bias. Fourth, validity of scores in national examination was high, which have been widely used in previous studies.⁴⁴

The study results should be interpreted in light of the following limitations. First, the indication for prescriptions was not recorded. For women taking carbamazepine who had no hospital contact for epilepsy, the indication could not be determined. In this study, we adjusted for drugs for psychiatric disorders as a proxy for other potential indications. Still, there might be misclassification due to incompleteness of either the diagnosis of epilepsy or drug use as well as other indications that were unable to identify from the registers, such as neuropathic pain. Second, the Patient Register records no data on the severity of diseases. In this study, women having carbamazepine prescriptions during pregnancy were mostly prescribed more than once. These women might represent a subgroup of patients with high severity, which might introduce confounding by indication and result in overestimation. However, multiple prescriptions for ASMs during pregnancy were also observed in other ASMs, including lamotrigine, making this explanation less likely.⁴⁵ Third, this study could not involve individuals that did not enter primary school at all, which might be due to special conditions.¹⁹ We applied an inverse probability weighting analysis to account for nonparticipation in a subanalysis, and we found similar results. In addition, in the analyses restricted to adolescents without carbamazepine exposure–related conditions, such as major congenital anomaly, the results were consistent. Fourth, a prescription for ASMs did not guarantee drug use, which might cause misclassification. A previous study reported high adherence in pregnant women with epilepsy⁴⁶; therefore, misclassification should not be a major problem. In addition, in the subanalysis defining exposure as 2 or more prescriptions for reducing potential misclassification, similar results were observed. Fifth, carbamazepine could induce life-threatening allergy in individual carrying specific human leukocyte antigen (HLA) alleles, most commonly HLA-B*57:01, which could be found in 7% of the European population.⁴⁷ A Danish study reported a potential association between HLA-B*57:01 and intellectual disability.⁴⁸ Women taking carbamazepine as a routine treatment are less likely to carry these HLA alleles, which might be a source of selection bias and resulted in underestimation. Although we had no data on genetics, the relatively low effect size of the alleles should have limited influence on our conclusions. Sixth, we had no data on several potential confounders and effect modifiers, such as maternal IQ,⁴⁹ frequency and type of seizures during pregnancy,²¹ severity of anxiety and depression during pregnancy, blood levels of ASMs during pregnancy,⁵⁰ and folate use during pregnancy.³⁵ In a sensitivity analysis, we used maternal hospital admission (outpatient and inpatient) for depression and anxiety during the exposure window as a proxy for severe cases of anxiety and depression during pregnancy. Additional adjustment for this proxy did not change the results essentially. Seventh, the analyses on the association between other types of ASMs and academic performance in offspring could be underpowered and thus not very informative. Future studies with larger sample size are warranted.

In utero exposure to carbamazepine was associated with poorer academic performance in adolescent, represented by lower scores in ninth-grade exit examination in Danish and mathematics. Additional studies are needed to confirm these findings because of limitations in this study and variable findings in prior studies.

Glossary

ADHD

attention deficit and hyperactivity disorder

ASMs

antiseizure medications

ATC

Anatomical Therapeutical Chemical

HLA

human leukocyte antigen

ICD

International Classification of Diseases

LMP

last menstrual period

Appendix. Authors

Study Funding

This study was supported by grants from the Danish Council for Independent Research (DFF-6110-00019B, DFF-9039-00010B, and DFF-1030-00012B), the Nordic Cancer Union (R275-A15770 and R278-A15877), the Karen Elise Jensens Fond (2016), Novo Nordisk Fonden (NNF18OC0052029), the National Nature Science Foundation of China (82073570, 82125032, 81930095, and 81761128035), the Science and Technology Commission of Shanghai Municipality (19410713500 and 2018SHZDZX01), the Shanghai Municipal Commission of Health and Family Planning (GWV-10.1-XK07, 2020CXJQ01, and 2018YJRC03), the Shanghai Clinical Key Subject Construction Project (shslczdzk02902), the Guangdong Key Project (2018B030335001), and Innovative Research Team of High-level Local Universities in Shanghai. The funders had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclosure

All authors report no relevant disclosures. Go to Neurology.org/N for full disclosures.