Prenatal Exposure to Antiseizure Medications and Risk of Epilepsy in Children of Mothers With Epilepsy

Key Points Question Is use of antiseizure medication in pregnancy among mothers with epilepsy associated with epilepsy risk in their children? Findings In this cohort study of 38 663 children of mothers with epilepsy, there was a significant increase in the risk of epilepsy associated with prenatal exposure to valproate and certain other antiseizure medications, but these associations did not persist in sensitivity analyses (eg, in sibling analyses). Meaning The finding that there is no association between prenatal exposure to valproate and other antiseizure medications and epilepsy risk in children of mothers with epilepsy in analyses more robust to confounding suggests that prenatal exposure to these medications does not add to the child’s risk of epilepsy beyond that associated with the maternal epilepsy itself.


Introduction
Valproate is one of the most efficacious antiseizure medications (ASMs) especially for generalized epilepsy, 1 but there is concern over the adverse-effect profile of the drug when used in pregnancy. 2[18] Other ASMs may also impact the development and functioning of the brain, including phenobarbital and benzodiazepines such as diazepam and clonazepam, 9,[19][20][21] with neuropsychological effects having been described in human children. 22,23Preclinical animal data for other ASMs are fewer, but topiramate, 24,25 carbamazepine, 26 oxcarbazepine, 26 and vigabatrin 9 exposure in utero have all been associated with alterations in neuronal developmental processes.
Aberrations in the development of the fetal brain may increase risk of epilepsy. 27Given the alterations in fetal brain development experimentally induced by valproate and other ASMs in rodent models, 28 it could be hypothesized that changes in neuron connectivity and excitability 5,8 or from cytoarchitecture changes and/or migration abnormalities may lead to epileptogenic cortical areas. 6,260][31] Examining whether use of ASMs among pregnant mothers with epilepsy may explain a maternal effect is nevertheless not straightforward in observational human studies.Genetic risk of epilepsy varies according to the type of maternal epilepsy, which informs the selection of a specific ASM.Valproate is likely to be used for generalized epilepsies, 1,32 which have higher heritability than focal epilepsies, while sodium channel blockers, such as carbamazepine and oxcarbazepine, are used for focal epilepsies. 1,32With these considerations in mind, we aimed to examine whether use of valproate and other ASMs among pregnant mothers with epilepsy was associated with epilepsy in their children.

Study Design, Setting, and Population
This is a prospective, population-based cohort study carried out within the Nordic register-based study of antiepileptic drugs in pregnancy (SCAN-AED) project. 33The cohort consisted of pooled and harmonized data from nationwide health and social registers from Denmark, Finland, Iceland, Norway, and Sweden. 13,14Using the unique personal identification number assigned to all persons living in the country, we ensured accurate linkage of individual-level information across national registries.The relevant ethical and/or data-protection authorities in all participating countries approved the project.According to legislation in the Nordic countries, informed consent was waived because it is not required for register-based studies using pseudonymized data.We followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
In the present study, we included all live-born singletons of mothers with epilepsy from January

Prenatal ASM Exposure
Information on maternal use of ASMs was retrieved from the national prescription registers, which hold information on all redeemed prescriptions, including the date of dispensing, the Anatomical Therapeutic Chemical (ATC) classification code, and the number of dose units per pack.Children were considered prenatally exposed to an ASM if their mother had redeemed any prescription for medications with ATC codes N03A, N05BA09, or S01EC01 during the exposure window, which was defined as 30 days before the date of the last menstrual period (estimated using gestational age in days at birth) until birth.Monotherapy was defined as pregnancies in which the mothers had redeemed prescriptions for only 1 type of ASM within the exposure window and polytherapy as 1 or more prescriptions for 1 or more different ASMs during the exposure window.Polytherapy was divided into combinations with and without valproate.We calculated the mean daily dose of an ASM for each monotherapy as the sum of the defined daily doses from all prescriptions filled in the exposure window divided by the number of days in that period and created cutoffs at 50% and 100% of the defined daily dose (see the eMethods in Supplement 1).Since the duration of the ASM use might differ substantially, we also analyzed cumulative dose in the same period in a sensitivity analysis.

Epilepsy in Children
We retrieved data on children with epilepsy from patient registers, which contained diagnostic information from inpatient admissions and outpatient visits in specialist care.We considered the children to be affected by epilepsy if they were registered with a diagnosis of epilepsy (ICD-10 codes G40-G41).The date of epilepsy onset was defined as the admission date of the child's first hospital contact with any registered diagnosis of epilepsy.

Statistical Analysis
The children were followed up from birth until onset of epilepsy, emigration, death, or the end of follow-up on December 31, 2017, whichever came first.Data analysis was performed from October 2022 to December 2023.Using Cox proportional hazards regression, we estimated adjusted hazard ratios (AHRs) and corresponding 95% CIs for epilepsy, according to prenatal ASM exposure and ASM dose.In all models, we used the age of the child as the underlying time scale and allowed for separate baseline hazards for each country and birth year (as stratum) to account for country differences in diagnostic rates and calendar-period effects.In the basic adjusted model, a covariate was included for the sex of the child, whereas the fully adjusted models also included smoking in pregnancy, use of antidepressants (ATC code N06A) in pregnancy, and maternal characteristics assessed at the time of birth (age, parity, highest level of completed education, and psychiatric comorbidity).There was missing information on maternal smoking status (range, 6%-8%), maternal educational level (range, 2%-5%), and maternal parity (range, 0%-1%), and we included a separate missing category for these variables. 34Proportionality of hazards was assessed visually using log minus log plots.The threshold for statistical significance was set to 2-tailed .05,and there was no adjustment for multiple testing.
We estimated the cumulative incidence of epilepsy using a nonparametric approach based on the subdistribution hazards, with death and emigration as competing events.To better account for differences in maternal epilepsy type, we carried out a series of additional analyses.First, we restricted the population to children of mothers with active epilepsy (ie, those fulfilling epilepsy criteria from 1 year before pregnancy and until birth).Second, we performed a negative control exposure analysis using children whose mothers used an ASM in the year before (365 to 30 days before the last menstrual period) but not in pregnancy as the reference group (ie, assuming that the indication for treatment was similar in those who discontinued and continued a certain ASM).This analysis did not include data from Finland, as medication information was not available for the year prior to pregnancy.Third, we performed a sibling-controlled analysis based on siblings discordant for prenatal valproate exposure to better account for maternal characteristics, including epilepsy type and genetic characteristics.In these models, the maternal identification number was used as the stratum variable in the Cox proportional hazards regression analysis, and we included covariates for sex of the child and maternal characteristics that may have differed between pregnancies.Finally, we used ASD (ICD-10 code F84 excluding F84.2-F84.4)and major congenital malformations 35 as positive control outcomes for prenatal valproate exposure to evaluate the sensitivity of the algorithm used to estimate ASM dose and for the sibling design in detecting adverse effects.Log binomial regression models with similar adjustment variables were used to estimate adjusted relative risks (ARRs) of congenital malformations.Statistical analyses were performed using Stata, version 18 (StataCorp LLC).

Discussion
This is the first study, to our knowledge, examining whether prenatal exposure to valproate and other ASMs may be associated with epilepsy risk in children of mothers with epilepsy.In this cohort study of singletons in Denmark, Finland, Iceland, Norway, and Sweden, we found that children of mothers with epilepsy who used valproate, topiramate, or clonazepam in pregnancy were more likely to develop epilepsy compared with children of mothers with epilepsy not using any ASM in pregnancy.
However, sensitivity analyses in siblings and children of mothers who discontinued ASM treatment prior to pregnancy suggest that the associations with valproate and topiramate use found in the initial analyses were most likely explained by differences in some underlying factors, such as the In our cohort, nearly 1 in 10 children of mothers with epilepsy using valproate in pregnancy had developed epilepsy by the age of 15 years.However, risk of epilepsy did not differ according to whether the mother had used high, medium, or low doses of valproate, and risk of epilepsy was similar in siblings exposed and unexposed to valproate.These analyses suggest that the increased risk of epilepsy found in children with prenatal valproate exposure was likely confounded by underlying factors associated with both maternal valproate use and the child's risk of epilepsy (eg, the type of maternal epilepsy).Although due to sample size, our possibilities to perform in-depth analyses were more limited for children with prenatal topiramate and clonazepam exposure, we speculate that similar mechanisms (ie, confounding) may account for the associations with these ASMs.However, the analyses of clonazepam were especially limited by low numbers, and since the mechanism of action (ie, γ-aminobutyric acid receptor activation) 20 is different from that of topiramate and valproate, we cannot rule out that the association with clonazepam could reflect a true association.Another interesting finding that relates to the broader question of the teratogenicity of valproate stems from the sibling analyses of malformations.These analyses showed that 6.1% of siblings of children exposed to valproate who were themselves unexposed were diagnosed with major malformations, which is significantly higher than the approximately 3% of children unexposed to ASMs in the general population. 35This has, to our knowledge, not been reported before and gives rise to questions regarding the role of genetic, epigenetic, or other environmental factors in risk of Abbreviations: AHR, adjusted hazard ratio; ASM, antiseizure medication; NA, not analyzed due to low numbers or insufficient follow-up time.
a Based on 4 Nordic countries (Denmark, Iceland, Norway, and Sweden; medication information was not available from Finland for the year prior to pregnancy) from 1996 to 2017.
b Adjusted for year of birth, sex of the child, and country of birth.
c Additionally adjusted for maternal age, parity, educational level, smoking in pregnancy, use of antidepressants in pregnancy, and maternal psychiatric comorbidity.adverse offspring outcomes in women with epilepsies previously treated or treatable with valproate.
It is also possible that the siblings were not truly unexposed (eg, if the mother had stockpiled medication before pregnancy and therefore did not fill new prescriptions in the exposure window).
31]37 These studies have found that risk of epilepsy in children of parents with epilepsy depended on the type of parental epilepsy. 29,30,37For instance, in the Rochester Epidemiology Project, the 40-year cumulative risk of epilepsy was reported to be 7.3% for children of parents with generalized epilepsy vs 2.9% in children of parents with focal epilepsy. 30rthermore, while Ottman et al 31 did not report cumulative risks of epilepsy in children by maternal use of specific ASMs, they did compare any with no use and found no association with offspring seizure risk. 31Thus, it is plausible that the variation in epilepsy risk according to maternal ASM use found in our study may reflect variation in genetic risk for specific epilepsy types that are associated with the use of specific ASMs.

Limitations
This study has limitations.We relied on register-based identification of epilepsy in children and their mothers, and some misclassification of their epilepsy status was possible. 38,39In addition, classification of the subtype of maternal epilepsy (a key confounder in this study) was challenging due to the low validity of ICD-10 codes for epilepsy subclassification, 39 and sufficient adjustment for the subtype of maternal epilepsy was consequently difficult.For this reason, we performed several sensitivity analyses using approaches that were more robust to confounding by indication, such as the sibling analysis and the use of the discontinuers as the reference.However, these sensitivity analyses were based on smaller groups, which limited the statistical power to detect between-group differences.Another limitation was the use of maternal prescription fills as a proxy for prenatal ASM exposure, and although high levels of adherence to these medications have been reported, 40 some misclassification was possible.Nevertheless, the quality of the data in the national prescription registers was considered to be high.The reimbursement structure in the Nordic health care systems provided a strong economic incentive for recording all dispensed drugs and ensured high sensitivity in capturing medication exposures. 41Furthermore, we estimated mean daily dose by dividing the cumulative amount dispensed by the length of the entire pregnancy.This may have resulted in underestimation of daily dose for mothers who discontinued early in pregnancy, which could have further diluted dose-response associations.We did, however, find similar patterns when considering the cumulative dose, and we validated the sensitivity of our algorithm to detect adverse effects of prenatal valproate exposure by demonstrating dose dependency with known valproate-associated risks (ASD and major malformations), suggesting that we should have been able to detect a dose association with epilepsy as well had there been one.Finally, while it would have been of interest to also study the association in a population without epilepsy (eg, children of mothers using valproate for bipolar disorder or migraine), this was not possible with our data, since the number of children who developed epilepsy in these groups was too low for any meaningful analysis.

Conclusions
This cohort study found associations between prenatal exposure to valproate and certain other ASMs with epilepsy in children of mothers with epilepsy.However, these associations attenuated upon further analyses of discordant siblings and children of mothers who discontinued ASM treatment prior to pregnancy.These findings suggest that prenatal ASM exposure may not increase epilepsy risk in children of mothers with epilepsy and indicate that differences were more likely associated with other underlying factors (eg, possibly the heritability of the maternal epilepsy).

Table 4 .
In these analyses, the AHR was 1.69 (95% CI, 0.91-3.16)for risk of epilepsy in children of mothers who used valproate in pregnancy compared with children of mothers

Table 1 .
Characteristics of Children of Mothers With Epilepsy by Prenatal Valproate Exposure discontinued valproate before pregnancy.There was no difference in epilepsy risk between children of mothers who used topiramate in pregnancy compared with children of mothers who discontinued topiramate before pregnancy (AHR, 1.19; 95% CI, 0.26-5.44).For most other ASMs, data were too limited for analyses.Among 25 608 mothers with epilepsy included in this study, we identified 258 (1.0%) who used valproate in at least 1 pregnancy and no ASM in at least 1 other pregnancy.In these sibling sets, we observed no difference in epilepsy risk between children with prenatal valproate exposure and their unexposed sibling (AHR, 0.58; 95% CI, 0.23-1.46)(Table5).When considering 418 sibling sets in which the mother used valproate in at least 1 pregnancy and no valproate in at least 1 other pregnancy (ie, including mothers using other ASMs than valproate, or no ASMs), we still observed no difference (AHR, 0.95; 95% CI, 0.50-1.82).In sensitivity analyses based on the 418 sibling sets of mothers using valproate in 1 pregnancy and no valproate in another pregnancy, we found a higher risk of ASD (eTable 4 in Supplement 1) and major malformations (eTable 5 in Supplement 1) in the sibling exposed to valproate the unexposed sibling (ASD: AHR, 6.41 [95% CI, 2.00-20.58]and major malformations: 6.1% of unexposed siblings vs 9.3% of exposed siblings; ARR, 1.66 [95% CI, 1.03-2.67]).Due to limited numbers, we were unable to undertake sibling analyses for topiramate and clonazepam. who

Table 2 .
Association of Prenatal Exposure to Valproate and Other ASMs With Childhood Epilepsy aheritability of the maternal epilepsy.For clonazepam, the analyses were limited by low numbers, and it remains unclear to which extent the association with epilepsy could be explained by similar factors or by the medication itself.

Table 3 .
Association of Different Doses of Prenatal Exposure to Valproate and Other ASMs With Childhood Epilepsy a Abbreviations: AHR, adjusted hazard ratio; ASM, antiseizure medication; NA, not analyzed due to low numbers or insufficient follow-up time.a Based on 5 Nordic countries (Denmark, Finland, Iceland, Norway, and Sweden) from 1996 to 2017.b Adjusted for year of birth, sex of the child, and country of birth.c Additionally adjusted for maternal age, parity, educational level, smoking in pregnancy, use of antidepressants in pregnancy, and maternal psychiatric comorbidity.d Rounded for data privacy reasons.JAMA Network Open | Neurology Prenatal Exposure to Antiseizure Medications and Risk of Epilepsy JAMA Network Open.2024;7(2):e2356425.doi:10.1001/jamanetworkopen.2023.56425(Reprinted) February 26, 2024 7/13 Downloaded from jamanetwork.comby guest on 03/06/2024

Table 4 .
Association of Prenatal Exposure to Valproate and Other ASMs With Childhood Epilepsy Among Children Whose Mothers Discontinued ASMs in the Year Before Pregnancy a

Table 5 .
Sibling Analyses of the Association of Prenatal Valproate Exposure With Childhood Epilepsy a Adjusted for sex of the child, maternal age, parity, smoking in pregnancy, and use of antidepressants in pregnancy.