Objectives In utero exposure to antiepileptic drugs has previously been associated with adverse outcome among offspring, but evidence on longer term milestone development remains limited. We investigated the association between in utero exposure to antiepileptic drugs and learning disabilities in the first year of compulsory education among offspring and assessed which antiepileptic drugs carried the highest risk.
Methods This population-based case–cohort study used Danish nationwide register data from 2005 to 2008. Cases were offspring exposed to antiepileptic drugs in utero, and controls were unexposed offspring of mothers previously redeeming antiepileptic drug prescriptions. Offspring were followed from birth until the first year of compulsory education from 2011 to 2015. Learning disabilities were defined as mental retardation, specific developmental disorders, autism spectrum disorders, emotional/behavioural disorders or having special educational needs. Logistic regression was used to compute ORs with 95% CIs adjusted for potential confounding.
Results Of 117 475 incident singleton births, 636 cases and 434 controls were included (median age: 6.1 years, males: 55.7%). Learning disabilities were identified among 7.1% cases compared with 3.7% for controls. During any trimester, the adjusted OR of the association between in utero exposure to antiepileptic drugs and learning disabilities was 2.20 (95% CI 1.16 to 4.17). Among cases not exposed to polytherapy (n=556), in utero exposure to lamotrigine compared with another antiepileptic drug was associated with the lowest adjusted risk (OR 0.42, 95% CI 0.19 to 0.92), and valproate carried a higher risk (OR 4.67, 95% CI 1.73 to 12.59).
Conclusion In utero exposure to antiepileptic drugs was significantly associated with learning disabilities among offspring. Lamotrigine should preferentially be considered over, for example, valproate if clinically feasible.
- Antiepileptic drugs
- In utero exposure
- Learning disabilities
- Cognitive functioning
Statistics from Altmetric.com
Besides being the first-line treatment of epilepsy, antiepileptic drugs are commonly prescribed in affective disorders as well as indicated in several other conditions including generalised anxiety, neuropathic pain, migraine and even off-label treatment of obesity.1 These conditions generally require continuous treatment making discontinuation an issue of utmost importance. Consequently, most women are consigned to continue treatment prior to, during and following pregnancy, and given the wide indications for antiepileptic drugs, the total number of worldwide exposed women of childbearing age is considerable.2
Women on antiepileptic drugs during pregnancy are able to have a normal and favourable outcome.3 Some studies suggest increased risks of both congenital malformations and impaired neurodevelopment across cognitive, motor and behavioural domains among offspring exposed to antiepileptic drugs in utero.4–8 From a drug safety standpoint, valproate remains by far the most thoroughly assessed antiepileptic drug, and recent reviews imply that valproate carries the highest teratogenic risk.4 6 9 There are very limited drug safety data concerning new generation antiepileptic drugs.9 10 However, the general consensus is that especially lamotrigine and levetiracetam are relatively safe,2 5 6 which has led to lamotrigine being the antiepileptic drug of choice for women with epilepsy of childbearing age.5 Importantly, as the consequence of in utero exposure to antiepileptic drugs may not manifest until years or decades following exposure, further longer term drug safety surveillance is warranted.6 This has been addressed in a recent large-scale study observing poor school performance among sixth-grade children exposed to antiepileptic drugs in utero compared with unexposed children and children only exposed to lamotrigine in utero.11 However, further varied studies within this research field are warranted regarding more precise information on covariates likely to influence outcomes.12
Therefore, through a population-based case–cohort study using Danish nationwide registers granting access to important covariates, we investigated learning disabilities among offspring exposed to antiepileptic drugs in utero and further assessed which antiepileptic drugs carried the highest risk. As comparison, we used unexposed offspring of mothers previously redeeming antiepileptic drug prescriptions, as they represent a reliable control group sharing a similar mental vulnerability and genetic predisposition.
Study design and population
Using a register-based case–cohort study design, all births in Denmark were identified during a 4-year period from 2005 to 2008 in the Danish Medical Birth Register,13 and we followed offspring from birth until the first year of compulsory education during a 5-year period from 2011 to 2015, as done previously.14
Of all births, only offspring of mothers with an incident singleton birth were identified, after which we only included offspring of mothers exposed to antiepileptic drugs (figure 1). Offspring were assigned as cases if mothers were exposed to antiepileptic drugs within 90 days prior to conception to birth or assigned as controls if mothers were exposed to antiepileptic drugs at any time but not during pregnancy. This was chosen to eliminate any potential in utero exposure to antiepileptic drugs among controls, although this may consequently underestimate the number of cases. However, this approach reduces the chance of bias of an inaccurate comparison between mothers never exposed to antiepileptic drugs with mothers with ongoing exposure during pregnancy.
Registers, covariates and definitions
At birth, all Danish citizens are assigned a unique personal identification number registered in the Civil Registration System,15 which is used in all healthcare, educational and social contacts facilitating confidential linkage between registers in Denmark. From the Danish Medical Birth Register, we obtained data on birth, child sex, gestational age (GA) based on estimates of last menstrual period and ultrasound measures, preterm birth (GA <37+0), maternal smoking status (never, active and former smoker) and maternal age at birth. Apgar score at 5 min was also obtained from the Danish Medical Birth Register and divided into 0–3, 4–6 and 7–10 according to standardised definitions to reflect severe, moderate and mild to no distress, respectively.16 Birth weight relative to GA was further derived from the Danish Medical Birth Register and classified according to small for GA (SGA), appropriate for GA (AGA) and large for GA (LGA). This was obtained by determining the mean birth weight and SD of all births at GA 40+0, where no risk factors (smoking, hypertension and preterm birth) for having an SGA infant were evident, as described previously.17–19 Accordingly, 10th and 90th percentiles of birth weights for each GA were determined forming the lower and upper limits of AGA infants, with SGA and LGA infants under 10th and above 90th percentiles, respectively.
Both the Danish National Patient Register20 and the Danish Psychiatric Central Research Register21 were used to obtain data on maternal comorbidities using International Classification of Diseases, 10th revision (ICD-10) codes. The Danish National Prescription Register22 was used to obtain data on redeemed prescriptions using Anatomical Therapeutic Chemical (ATC) codes. Prior to conception, we identified maternal epilepsy, affective disorders (mania, bipolar disorder and depression), other psychiatric disorders, diabetes and hypertension with complications as well as redeemed antidiabetic and antihypertensive drug prescriptions. During pregnancy, we further identified gestational diabetes, hypertension without oedema and/or proteinuria, oedema and proteinuria without hypertension and pre-eclampsia as well as redeemed antidiabetic and antihypertensive drug prescriptions. For details of ICD-10 and ATC codes, see table S1 in the online supplementary file 1.
The highest achieved maternal education, as a proxy for socioeconomic status, was assessed at conception using data from the Danish Population Education Register.23 In accordance with the International Standard Classification of Education, known as ISCED, we defined three socioeconomic groups: group 0–2 including primary to lower secondary education, group 3–5 including upper secondary to short-cycle tertiary education and group 6–8 including bachelor’s to doctoral degree, as described previously.14 24
In utero exposure
Using the Danish National Prescription Register, we identified antiepileptic drugs (ATC: N03A) and the time at which the prescription was redeemed. Accordingly, we defined three in utero exposure periods during first trimester including exposure from 90 days prior to conception to day 84 of pregnancy (last day of the first trimester), second trimester including exposure from day 85 of pregnancy to day 196 of pregnancy (last day of the second trimester) and third trimester including exposure from day 197 to birth.
Our main outcome was learning disabilities among offspring in the first year of compulsory education. This was defined as offspring being diagnosed with mental retardation (ICD-10: F7x), specific developmental disorders (ICD-10: F80–83), autism spectrum disorders (ICD-10: F84), emotional/behavioural disorders (ICD-10: F9x) or having special educational needs. We defined special educational needs as either receiving special class lessons, additional time for a special education programme or having been classified with a learning disabilities obtained from the Special Education Register as part of the Danish Population Education Register.
Data on mortality were retrieved from the Danish Cause of Death Register25 to assess that mortality was rare and that it was not necessary to account for this as a competing event among offspring.
Categorical variables were reported as counts with percentages and continuous variables as medians with 25th and 75th percentiles. Differences were compared with χ2 and Mann-Whitney tests as appropriate.
First, we investigated the association between in utero exposure to any antiepileptic drug during any trimester and learning disabilities in a logistic regression model. Both univariable and multivariable analyses were performed, and ORs with 95% CIs were computed. Adjustments were made for maternal age at birth, educational level, smoking status, epilepsy, affective disorders, other psychiatric disorders, birth year, preterm birth, Apgar score and birth weight relative to GA.
Second, during any trimester, we calculated ORs of learning disabilities among cases exposed to a specific antiepileptic drug in utero compared with controls and compared with cases exposed to another antiepileptic drug in utero.
A two-sided p value <0.05 was considered statistically significant. Data management and statistical analyses were performed using SAS V.9.4 and R V.3.5.0 (R Core Team (2018). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/).
Of 117 475 incident singleton births in Denmark from 2005 to 2008, 636 were identified as offspring exposed to antiepileptic drugs in utero (cases) and 434 as unexposed offspring of mothers previously redeeming antiepileptic drug prescriptions (controls) (figure 1).
The median age of offspring was 6.1 years, with 55.7% being males and 44.3% females. The remaining demographics were generally well balanced between cases and controls (table 1). Notably, more births occurred in earlier years (2005 rather than in 2008) among cases compared with controls, and there was a difference between cases and controls regarding birth weight relative to GA, with more cases being SGA or AGA rather than LGA. Maternal epilepsy was more frequently identified among cases (37.9%) compared with controls (11.5%), which was the opposite for other maternal psychiatric disorders (cases: 21.4%, controls: 27.4%).
Figure 2 displays the three most redeemed antiepileptic drugs for both cases and controls: lamotrigine (cases: 54.2%, controls: 27.9%), valproate (cases: 10.5%, controls: 6.9%) and clonazepam (cases: 9.9%; controls: 12.9%). Table S2 in the online supplementary file 1 shows the distribution of antiepileptic drugs among cases and controls with epilepsy. Notably, most cases and controls with epilepsy tended to have lamotrigine prescribed.
As relatively few cases (12.6%) were exposed to antiepileptic polytherapy during any trimester, we were not able to differentiate between monotherapy and polytherapy in the main analysis. Table S3 in the online supplementary file 1 shows the distribution of which antiepileptic drugs that were combined among cases exposed to polytherapy in utero. Notably, lamotrigine was the most frequently used antiepileptic drug in combination with another one.
Offspring learning disabilities
During follow-up, learning disabilities were identified among 45 cases (7.1%) and 16 controls (3.7%) in the first year of compulsory education. Table S4 in the online supplementary file 1 shows the distribution of the number of each diagnosis and special educational needs among cases and controls. Notably, a higher proportion of offspring with a diagnosis of learning disabilities tended to have special educational needs.
The crude OR of the association between in utero exposure to antiepileptic drugs and learning disabilities was 1.99 (95% CI 1.11 to 3.57) during any trimester (figure 3A). After adjustment for potential confounding, the OR remained similar (2.20, 95% CI 1.16 to 4.17).
Among cases not exposed to polytherapy (n=556), we first investigated the risk of learning disabilities with each antiepileptic drug compared with no in utero exposure. Here, valproate, topiramate, levetiracetam and phenobarbital were associated with an increased risk, whereas the remaining antiepileptic drugs including lamotrigine were not associated with evidence of a differential risk (figure 3B).
Second, we investigated the risk of learning disabilities with each antiepileptic drug compared with in utero exposure to another antiepileptic drug (figure 3C). In utero exposure to lamotrigine compared with another antiepileptic drug during any trimester was associated with the lowest adjusted OR of learning disabilities of 0.42 (95% CI 0.19 to 0.92), and valproate was associated with a higher adjusted OR of 4.67 (95% CI 1.73 to 12.59). In addition, phenobarbital was associated with the highest adjusted OR of 12.61 (95% CI 1.98 to 80.15). However, this result along with results from the remaining 13 antiepileptic drugs in monotherapy was characterised by wide 95% CIs and insignificant p values making it difficult to draw firm conclusions.
To test the robustness and consistency of our results, several sensitivity analyses were performed.
First, in utero exposure during all trimesters was investigated, and ORs were generally comparable with those of the main analysis (Figure S1 in the online supplementary file 1).
Second, when limiting in utero exposure to a specific trimester as well as investigating different trimester combinations, we observed similar ORs to those in the main analysis (Figure S2A–F in the online supplementary file 1).
Third, when in utero exposure was limited to any trimester excluding exposure during 6 months breastfeeding period following birth, the crude OR was similar to the main analysis; however, the adjusted OR was not associated with evidence of a differential risk (Figure S3 in the online supplementary file 1).
Fourth, when excluding cases exposed to valproate in utero and controls of mothers previously redeeming valproate prescriptions, crude and adjusted ORs were not associated with evidence of a differential risk (Figure S4 in the online supplementary file 1).
Fifth, we investigated the risk of learning disabilities among cases exposed to valproate monotherapy in utero compared with in utero exposure to lamotrigine monotherapy as reference, and results conferred an increased risk. When using in utero exposure to valproate monotherapy as reference, lamotrigine conferred a decreased risk (Figure S5 in the online supplementary file 1).
Sixth, we further investigated the dose–response relationship between in utero exposure to lamotrigine ≤100 mg and >100 mg compared with controls and compared with cases exposed to another antiepileptic drug in utero. A similar analysis was carried out for valproate ≤600 mg and >600 mg. Again, results were generally comparable with those of the main analysis (Figure S6 in the online supplementary file 1).
Seventh, when only investigating cases and controls with Apgar score 7–10 reflecting mild to no distress, ORs were slightly higher compared with the main analysis (Figure S7 in the online supplementary file).
In this Danish nationwide register-based case–cohort study, we found that of 117 475 incident singleton births, 636 offspring were exposed to antiepileptic drugs in utero, and 434 were unexposed offspring of mothers previously redeeming antiepileptic drug prescriptions. Our key finding was that in utero exposure to any antiepileptic drug during any trimester was associated with a more than twofold increased adjusted risk of learning disabilities in the first year of compulsory education among offspring. In addition, the majority of the redeemed prescriptions were for lamotrigine (more than 50% and 25% for cases and controls, respectively), and when compared with another antiepileptic drug among cases not exposed to polytherapy, lamotrigine was associated with a 0.42-fold decreased adjusted risk of learning disabilities, whereas valproate carried a higher adjusted risk of 4.67. Results for remaining antiepileptic drugs were more inconclusive due to limited event rates.
In clinical practice, our results are relevant as a considerable number of women of childbearing age are exposed to antiepileptic drugs worldwide,2 and it is important that these women receive comprehensive patient education and counselling. For some women, discontinuation of antiepileptic drugs during pregnancy is not always the appropriate choice, and it is crucial that treatment is not ceased without consultation with a clinician due to the risk of seizure relapse or other serious consequences.26 In other women who are prescribed antiepileptic drugs for chronic pain management or psychiatric disorders, clinicians should reconsider whether continuing treatment is safe and responsible.27 Our results rather imply a more cautious approach towards continuing antiepileptic drugs during pregnancy if the indication is doubtful, and clinicians must outweigh risks associated with in utero exposure to antiepileptic drugs with risks associated with untreated and uncontrolled maternal symptoms.27 As mentioned, large-scale studies on milestone development including early educational skills among offspring exposed to antiepileptic drugs in utero, especially lamotrigine, are limited. However, our study suggests that if antiepileptic treatment is warranted during pregnancy, lamotrigine should preferentially be prescribed if clinically feasible. While valproate carried a high risk of learning disabilities, lamotrigine carried the lowest risk, whereas results from remaining antiepileptic drugs were limited due to sample size limitations emphasising the need for more research regarding drug safety surveillance of antiepileptic drugs among women of childbearing age. In addition, we investigated the risk of learning disabilities in regards to lamotrigine and valproate doses and found that high doses of lamotrigine of >100mg decreased the risk compared with another antiepileptic drug. This was the opposite for valproate as a dose of >600 mg was associated with a nearly eightfold increased risk of learning disabilities, though keeping in mind the wide 95% CIs.
Most of the recent reviews4 6 are based on studies mainly investigating overall neurodevelopment and IQ scores among offspring exposed to antiepileptic drugs in utero. Only few studies28–31 have investigated special educational needs and reported insignificantly increased ORs. Although consistently observing that offspring exposed to valproate in utero had increased special educational needs, these studies were limited to questionnaire-based data, missing information on covariates, heterogeneous case and control groups, and small sample sizes, which is a general problem within this research field.32 However, our study adds to the growing body of evidence to suggest that the effect of in utero exposure to antiepileptic drugs on neurodevelopment among offspring is long term and not necessarily mitigated over time.12
Interestingly, Baker et al 28 found an association between in utero exposure to valproate and poor IQ among offspring at age 6 years. In addition, another recent study by Elkjær et al 11 found an association between in utero exposure to antiepileptic drugs and poor scores on national school tests among sixth-grade children. Our study generally supports these results. Although we assessed a similar age group, we used a substantial different outcome. One could argue that our composite outcome of mental retardation, specific developmental disorders, autism spectrum disorders, emotional/behavioural disorders or special educational needs is a rather vague proxy for assessing the effect of in utero exposure to antiepileptic drugs compared with a more specific proxy for learning disabilities like a low IQ score33 or poor performance on national school tests.11 As such, that we observed an increased risk of learning disabilities is somewhat expected given the effect of in utero exposure to antiepileptic drugs, especially valproate, on IQ score.34 However, our composite outcome may have allowed for a more wide detection of subtle neurodevelopmental manifestations of in utero exposure to antiepileptic drugs that in the longer run may result in learning disabilities and impaired early educational skills. Therefore, the novelty of our study is that in utero exposure to antiepileptic drugs are associated with learning disabilities among offspring, though the detection of these requires a multidisciplinary approach.
Taken together, when assessing potential neurodevelopmental manifestations of in utero exposure to antiepileptic drugs among offspring, clinicians have to be aware that a single figure like an IQ score cannot necessarily quantify learning disabilities. Therefore, offspring should be assessed on several neurodevelopmental domains including cognition, motor function and behaviour.6 As such, our study underscores that susceptible offspring should be more closely monitored throughout childhood allowing for early detection of signs and symptoms of impaired neurodevelopment and subsequent intervention. This notion was also supported by a previous study, where an association was observed between in utero exposure to antiepileptic drugs and impaired fine motor skills, but not impaired social skills or temperament, among offspring.35 Furthermore, the study did not identify any harmful effects of breast feeding when mothers were exposed to antiepileptic drugs, which is generally in line with our results as well as another study.36
Most of the limitations were related to the observational design of our study, which could not rule out unaccounted or residual confounding. Importantly, we did not have access to data on fathers, and we could consequently not adjust for potential confounding related to this. Furthermore, we could not adjust for potential learning disabilities of the parents. In addition, assignment of antiepileptic drugs was not randomised, and consequently the largest proportion of mothers had epilepsy, thus potentially reflecting confounding by indication as well as selection bias. It is possible that mothers with refractory symptoms of epilepsy more often received the most potent and unfortunately teratogenic antiepileptic drug valproate, which has been associated with autism spectrum disorders among offspring exposed in utero.37 Furthermore, as antiepileptic drugs have various indications, it is likely that some underlying diseases like bipolar disorder affect the mother–child attachment to a greater extent than epilepsy among others causing a more unstable mother–child attachment and neglect of care.
The Danish Prescription Register includes approximately 97.5% of all redeemed prescriptions, but the adherence to and indication for treatment remain unknown.38 However, a previous study showed high compliance rates between redeemed prescriptions for antiepileptic drugs and maternal use during pregnancy in Denmark.39 Finally, we could not account for the effect of other possible confounding factors including disease severity, seizure frequency, dosage and serum levels of antiepileptic drug, prescribing practices, exposure to other potential teratogens or use of, for example, folic acid supplement during pregnancy. Recently, the latter has been associated with a reduced risk of autism spectrum disorders among offspring exposed to antiepileptic drugs in utero.40
In utero exposure to antiepileptic drugs was significantly associated with offspring learning disabilities. Importantly, caution should be exerted when prescribing antiepileptic drugs to women of childbearing age, especially when in doubt of indication. If discontinuation during pregnancy is not warranted, lamotrigine should preferentially be prescribed if clinically feasible due to a decreased risk of learning disabilities compared with other antiepileptic drugs including valproate. As such, more surveillance of in utero exposed offspring should be initiated. For women of childbearing potential on antiepileptic drugs, these findings must be balanced against the treatment benefits for women who require valproate for epilepsy control.
Contributors LFB: study design, data management, data analysis, drafting of manuscript and manuscript preparation; CP: study design, data management, data analysis and manuscript revision; KK: study design, data management and manuscript revision; MPA: study design, data management and manuscript revision; CR, CT-P, JN and SH: manuscript revision.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests CP reported receiving speaker honorarium from Lundbeck. KK reported receiving speaker honorarium from Novartis and research grants from the Laerdal Foundation. CT-P reported receiving research grants from Bayer. JN reported receiving research grants from Lundbeck and Pfizer and speaker honoraria from Bristol-Myers Squibb, Lundbeck and Sunovion.
Patient consent Not required.
Ethics approval Statistics Denmark and the Danish Health Data Authority approved the use of data for this study. By Danish law, ethics approval is not required for register-based studies.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.