Abstract
Objective: To investigate the effects of prenatal exposure to monotherapy levetiracetam, topiramate, and valproate on child cognitive functioning.
Methods: This was a cross-sectional observational study. Children exposed to monotherapy levetiracetam (n = 42), topiramate (n = 27), or valproate (n = 47) and a group of children born to women who had untreated epilepsy (n = 55) were enrolled retrospectively from the UK Epilepsy and Pregnancy Register. Assessor-blinded neuropsychological assessments were conducted between 5 and 9 years of age. Information was collected on demographic and health variables and adjusted for in multiple regression analyses.
Results: In the adjusted analyses, prenatal exposure to levetiracetam and topiramate were not found to be associated with reductions in child cognitive abilities, and adverse outcomes were not associated with increasing dose. Increasing dose of valproate, however, was associated with poorer full-scale IQ (−10.6, 95% confidence interval [CI] −16.3 to −5.0, p < 0.001), verbal abilities (−11.2, 95% CI −16.8 to −5.5, p < 0.001), nonverbal abilities (−11.1, 95% CI −17.3 to −4.9, p < 0.001), and expressive language ability (−2.3, 95% CI −3.4 to −1.6, p < 0.001). Comparisons across medications revealed poorer performance for children exposed to higher doses of valproate in comparison to children exposed to higher doses of levetiracetam or topiramate.
Conclusions: Preconception counseling should include discussion of neurodevelopmental outcomes for specific treatments and their doses and women should be made aware of the limited nature of the evidence base for newer antiepileptic drugs.
GLOSSARY
- AED=
- antiepileptic drug;
- CI=
- confidence interval;
- FSIQ=
- full-scale IQ;
- UK-EPR=
- United Kingdom Epilepsy and Pregnancy Register
Concern about the use of valproate in women of childbearing age has led to a shift in prescribing practices toward newer antiepileptic drugs (AEDs), especially levetiracetam, lamotrigine, and topiramate.1,–,3 Prenatal exposure to lamotrigine has been demonstrated to be associated with significantly higher neuropsychological functioning than children exposed to valproate, both in infancy4,5 and at school age.6,7 However, there is extremely limited evidence regarding the risks that may be associated with exposure to levetiracetam or topiramate.8 For infants exposed to levetiracetam, consistent neurodevelopment with control infants and superior development in comparison to infants exposed to valproate at 1 and 3 years of age has been reported.9,10 Only a single study has reported on the neurodevelopment of children exposed to topiramate and although this documents an association between prenatal exposure and reduced neurodevelopmental outcome, its findings are substantially limited because of the topiramate cohort size (n = 9).11
This study aimed to delineate the cognitive abilities of school-age children exposed prenatally to monotherapy levetiracetam or topiramate in comparison to children born to women with untreated epilepsy and children exposed to valproate. This study had a directional hypothesis, which stated that children exposed to levetiracetam or topiramate would not differ from control children in their cognitive abilities but would have performed significantly higher than the children exposed to valproate on measures of cognitive ability. Child IQ was the primary outcome variable with other, more specific cognitive abilities, investigated as secondary outcome variables.
METHODS
The study was a cross-sectional observational study. Mother–infant pairs were identified from the United Kingdom Epilepsy and Pregnancy Register (UK-EPR), a national pregnancy register that investigates the prevalence of major congenital malformations following exposure to AEDs. Detailed information about the register and its methodology has been reported previously.12 Women with epilepsy were enrolled in the UK-EPR through self-referral or referral by their health professional. Recruitment occurs within the first or second trimester facilitating prospective documentation about health and well-being during the pregnancy. Within 3 months of birth, details about the birth and health of the child are reported to the register by local health care services. Mother–infant pairs were eligible for inclusion in this neurodevelopment follow-up study if the infant had been a live birth between September 2004 and May 2007 and mothers either were taking levetiracetam, topiramate, or valproate monotherapy or they were untreated during their pregnancy. Recruitment was national across the UK. Families were not invited to participate if their child had a genetic condition associated with neurodevelopmental impairment. Initially, this study also aimed to investigate the neurodevelopment of children exposed to gabapentin in utero; however, the numbers enrolled in the UK-EPR were low. Means and SDs for the gabapentin data (n = 14) are reported in table format for information.
Enrollment into this follow-up study was retrospective. There were approximately 3 times more eligible participants for the valproate-exposed group and the no-medication group, than for the other exposure groups and therefore each third mother identified was included in the recruitment list for these 2 groups. Recruitment letters and information sheets were posted out to those identified. A follow-up letter was issued if no response had been received. Mother–infant pairs who returned a positive response were formally enrolled in the study and informed consent was obtained.
Pregnancy details and details about the mother's epilepsy, including AED dose and seizure information, were collected from the prospectively collected records of the UK-EPR. No seizure diaries or frequent monitoring of seizure activity had been taken and therefore seizure exposure was dichotomized as present or absent. Details of the mother and father's educational history and employment were collected through a semistructured interview at the time of the assessment. Alcohol, nicotine, and concomitant medication use for the second and third trimesters, which is not routinely collected by the UK-EPR, were collected through maternal report retrospectively. Maternal intellectual functioning was measured with the Test of Nonverbal Intelligence.13
Sixty percent of children exposed to levetiracetam and 20% of children exposed to valproate enrolled in this study were previously assessed at 3 years of age as part of an ongoing study9 but were not part of the infant cohort reported by this study group.10
Neuropsychological assessments were conducted blinded by authors R.B., R.C., C.R., or R.S. either in the child's home or school. The assessment battery included the Wechsler Intelligence Scale for Children–Fourth Edition14 or the Wechsler Preschool and Primary Scale of Intelligence–Third Edition15 if the child was 5 years of age. The primary outcome measures were the full-scale IQ (FSIQ), verbal index, nonverbal index, and the processing speed index. Analysis of these outcomes was adjusted for administered assessment test version (Wechsler Intelligence Scale for Children–Fourth Edition or Wechsler Preschool and Primary Scale of Intelligence–Third Edition). Specific cognitive domains were assessed utilizing subtests from the NEPSY: A Developmental Neuropsychological Assessment, 2nd edition16 and the Clinical Evaluation of Language Fundamentals–Fourth Edition,17 with parental rating of behavior collected using the Behavior Assessment System for Children, Second Edition.18 Assessments were double-scored and data entry double-checked to minimize errors. Feedback was provided to the family on the outcome of the assessments.
The data were analyzed using multiple linear regression. The following covariates were considered: maternal epilepsy type, treatment group, dose, professional employment, maternal IQ, maternal age, gestational age of child at birth, sex, age, and exposure to seizures, tobacco, or alcohol. Inverse probability weighting19 was used to account for the influence of missing outcomes. The analysis of AED doses utilized dose recorded in the UK-EPR at enrollment, which represents dose around the time of conception. For the purposes of comparison between the AED types, doses were standardized by dividing dose by the median dose for each respective AED type. The median was used because the doses had skewed distributions. To take into account that 3 components of child IQ were investigated, multiple comparisons were considered when interpreting the results of the primary analysis. The secondary analyses, which involve a large number of statistical models (23 different models) have been added for completeness but should be regarded as merely exploratory. In a separate analysis, comparisons across AEDs were made at different dose levels (0.5 times median dose, median dose, 1.5 times median dose, and 2 times median dose) for the primary outcome. Data analysis was performed using the statistical packages MLwiN 2.16 and R i386 3.1.1.
Standard protocol approvals, registrations, and patient consents.
Approval was obtained from the North West Regional Ethics Committee, UK, and Belfast Health and Social Care Trust who host the UK-EPR. All participants provided informed written consent.
RESULTS
Four hundred forty-nine invitations to participate were sent out with 201 positive responses received (45%). Forty cases declined participation (9%), while the majority did not respond (46%). Of the 201 positive responses, 16 cases (8%) required exclusion because of conditions likely to effect cognitive functioning (e.g., brain injury, meningitis) or were too old (>9 years) by the time of assessment. Therefore, 185 of the 449 who were sent letters completed the assessment (41%). The percentage completing the study from the available sample on the UK-EPR varied by treatment group (no medication 35% of those eligible, topiramate 53%, levetiracetam 67%, gabapentin 58%, and valproate 31%).
The groups were comparable across the majority of demographic variables (table 1). However, differences were found in frequency of seizures, with those exposed to levetiracetam being exposed to the highest number of seizures (43%). The mothers of children exposed to valproate were older and they had the highest mean maternal IQ and level of folate supplementation (table 1).
Group demographic information by antiepileptic exposure groups
Children exposed to valproate had the lowest unadjusted mean scores for FSIQ and verbal reasoning (table 2) and across a number of other cognitive and behavioral domains (table 3).
Unadjusted means, standard errors, and rates below average performance by group for primary cognitive outcomes
Unadjusted means and SD by group for the specific cognitive and behavioral outcomes
In the adjusted analyses, exposure to levetiracetam was not found to be associated with reductions in FSIQ, verbal abilities, nonverbal abilities, or processing speed (table 4), and dose of levetiracetam was not predictive of poorer outcome. Consistently, being exposed to levetiracetam was not associated with poorer outcomes on language, memory, attention, and executive functioning or behavioral variables when outcomes were adjusted for covariates (table 5 and table e-1 at Neurolgy.org). Similarly for topiramate, no association with prenatal exposure was found for FSIQ, verbal abilities, nonverbal abilities, or processing speed, and dose of topiramate was not predictive of poorer outcomes (table 4). Regarding specific cognitive and behavioral outcomes, prenatal exposure to topiramate was not associated with poorer outcomes across the domains but had higher scores (better performance) on one aspect of attention and executive functioning in comparison to the control children (table 5 and table e-1). A dose effect was observed for valproate, with higher doses of valproate associated with poorer FSIQ, verbal and nonverbal abilities. In particular, an increment of 800 mg (median value) of valproate was significantly associated with a 10.6-point reduction in FSIQ, an 11.2-point reduction in verbal abilities, and an 11.1-point reduction in nonverbal abilities (table 4). Secondary analyses suggested that valproate may also be associated with poorer outcomes on the expressive naming index as well as the behavioral variables of withdrawal, adaptability, and daily living skills but not other measures of language, memory, attention and executive functioning, or behavioral outcomes (table 5 and table e-1).
Multivariate outcomes for the primary outcome IQ and its indexes
Results of multivariate analyses for the specific cognitive outcomes
AED comparisons by dose.
At half the median dose of valproate (400 mg/d), no significant differences were found in comparison to children exposed to half the median doses for levetiracetam (750 mg/d) or topiramate (100 mg/d) or in comparison to no-medication controls for FSIQ, verbal or nonverbal reasoning (table e-2). Comparison at the median doses suggests differences in verbal reasoning between exposures to valproate (800 mg/d) in comparison to no-medication controls and topiramate (200 mg/d). However, as dose increased, valproate was associated with reductions in FSIQ, verbal and nonverbal reasoning. For example, at 2 times the median dose of valproate (1,600 mg/d) when compared to the no-medication group and exposure to 2 times the median doses of topiramate (400 mg/d) and levetiracetam (3,000 mg/d), the reduction in FSIQ ranged from 11 to 16 IQ points, reduction in verbal reasoning ranged from 17 to 21 points, and reduction in nonverbal reasoning ranged from 6 to 14 (table e-2). Despite not adjusting for multiple testing when comparisons at different dose levels are made, the results show consistently that as valproate dose increases, the reductions in FSIQ, verbal and nonverbal components are more pronounced.
Influence of nonexposure variables.
Regarding confounders, maternal epilepsy type was extensively investigated both as a covariate and as an interaction term with treatment type and no association with child IQ or specific cognitive abilities was demonstrated (data not shown, but available on request). Exposure to seizures was also not associated with outcome. Consistent with the child developmental literature, higher maternal IQ was associated with improved child performance on FSIQ, verbal abilities, and nonverbal abilities as well as language and memory abilities and aspects of attention and executive skills (tables 4 and 5). Socioeconomic status was also associated with child performance, with those who had parents in professional employment scoring 6.1 points higher in FSIQ with similar associations for verbal abilities and aspects of attention and executive functioning (table 4). Tables 4 and 5 and table e-1 include further information pertaining to the influence of covariates on specific outcome variables.
DISCUSSION
This cross-sectional study indicated that there was no association between child cognitive functioning during middle childhood and exposure to levetiracetam in utero. This group was adequately powered to detect a medium effect size with 89% power (GPower 3.1.3)20 and therefore smaller levels of discrepancy cannot be ruled out. A lack of an association is consistent with our group's earlier finding in infants younger than 24 months of age10 and in later infancy,9 although there was group overlap with this later cohort. Dose is a key concept in terms of teratogenicity, and increasing doses of levetiracetam were not found to be associated with poorer outcomes in comparison to control children and were documented to have superior outcomes to the children exposed to higher doses of valproate. The lack of significant dose effect for levetiracetam also supports preclinical studies in which it is reported, even at high doses, to be exempt from the increased neuronal apoptosis seen following administration of other AEDs in rodents.21 Levetiracetam is reportedly becoming an encouraging alternative to valproate in the treatment of generalized epilepsy22 because of its lack of association with major congenital malformations.2,23
Topiramate has been associated with an increased rate of major congenital malformation (4.2%),2 with a specific association to orofacial clefts24 and therefore should be considered teratogenic. The group of children included exposed to topiramate was small and therefore only large differences in neurodevelopment would have reliably been detected and the results must be interpreted with caution. No significant level of difference between the children exposed to topiramate or control children were documented in the adjusted model and increasing doses of topiramate were not associated with poorer outcomes. Furthermore, children exposed to higher doses of topiramate were associated with more favorable cognitive outcomes than those exposed to higher levels of valproate. The reported findings here are in conflict with the small group (n = 9) reported previously.11 However, the cohort in this present study was larger, comprised children from a narrower age range, utilized blinded assessments, and undertook statistical control for a wider number of likely confounding factors, which may account for the difference in findings. A lack of an association between topiramate exposure and impaired neuropsychological function would support preclinical data, which suggests that, at therapeutic ranges, administration of topiramate in isolation is not associated with increased levels of apoptosis in rodent models.21,25
In relation to valproate, the findings of this study are consistent with those of other cohorts in demonstrating that prenatal exposure to valproate was associated with significant risks to child cognitive ability in a dose-dependent manner.6,–,8,26,–,29 The mean IQ point reduction ranged from 10 to 11 points, which is consistent with the associated decreases reported in a recent meta-analysis.8 Reduction in group means leads to increased numbers of children falling below the average range, which is demonstrated for the valproate group with 19% falling over 1 SD from the mean for FSIQ in comparison to 6% of control children. Having an FSIQ below the average range has implications for educational outcomes in the teenage years.30
The median dose in this valproate-exposed group was relatively low regarding the treatment of seizures (800 mg daily) but consistent with the reported mean dose from other cohorts.7,26 The data here highlighted the substantial effect higher doses of valproate have on cognitive functioning in comparison to higher doses of topiramate and levetiracetam.
No significant influence of maternal epilepsy type or exposure to seizures was found to be associated with child cognitive functioning, supporting previous results.6,7,29 However, as discussed below, limited information was available on seizure exposure.
A major strength of this study is its utilization of a national register to reduce the latency between AED monotherapy license and teratology risk information. Valproate, for example, was licensed for monotherapy use in the 1970s but the associated risks of prenatal exposure to later child cognition were only delineated within the last decade. Such a delay in information undermines clinical risk–benefit decision-making, and the utilization of participants collected in malformation registers may reduce the time taken to provide the first evidence. Further methodologic strengths of this study included its prospective collection of pregnancy details, which reduces recall bias, the utilization of blinded standardized neuropsychological assessment, the collection and control for a number of influential covariates, assessment of school-age child IQ, and adequate power to detect large effect sizes. Finally, the consideration of AED dose is important in view of the principles of teratology.31
There are a number of limitations to the present study. Pregnancy registers represent only a small proportion of women with epilepsy in the community and therefore sampling bias is possible. The low rate of recruitment highlights the challenges of using malformation registers to collect neurodevelopmental outcomes, but is consistent with the challenges experienced by others using this model of recruitment.28 The higher rate of participation in the levetiracetam group is believed to be linked to 60% of this group being assessed as part of an earlier study.9 Higher recruitment/retention rates are reported for longitudinal studies with multiple follow-ups across the early childhood years,6,7 but such methodologies are costly and may not facilitate collection of data for less frequently prescribed AEDs. Retrospective recruitment and collection of certain data (i.e., maternal alcohol use) may lead biased results. The 4 groups, however, were recruited and assessed identically and therefore it was not thought that the differential outcomes were attributable to recruitment bias. Of relevance, a recent Cochrane review found comparable outcomes from prospective longitudinal studies and retrospectively recruited cohorts from prospective pregnancy registers.8 The use of data pertaining to seizure exposure and AED dose at time of enrollment in UK-EPR means that no information was available on seizure exposure or AED dose alterations toward the end of the pregnancy and the results should be viewed in light of this. From a statistical viewpoint, although multiple comparisons were considered to interpret the results of the primary analysis for the IQ outcome, the secondary analyses (which involved 23 different statistical models) are merely exploratory, as well as the comparisons across AEDs that were made at different dose levels. Finally, consideration of the age of these children at assessment is required. In middle childhood, cognitive development is still dynamic and as age-appropriate cognitive skills become more complex, differential results might be seen.
In conclusion, the potential risks to neurodevelopment posed by prenatal exposure to valproate at higher doses can be substantial and should be a central aspect of preconceptual counseling. The documentation that the IQ of school-age children exposed to levetiracetam and topiramate is similar to nonexposed control children is reassuring but replication and extension are required.
AUTHOR CONTRIBUTIONS
Dr. Bromley contributed to the acquisition of funding, the conception and design of the study, data collection, study coordination, analysis and interpretation of data, drafting the article, and final approval. Dr. Bromley accepts full responsibility for the finished article, had access to any data, and controlled the decision to publish. Ms. Calderbank contributed to data collection, interpretation of data, drafting of the article, and final approval. Dr. Cheyne conducted the analysis and contributed to the interpretation of data, drafting the article, and final approval. Ms. Rooney contributed to data collection, interpretation of data, drafting of the article, and final approval. Dr. Trayner contributed to data collection, interpretation of data, drafting of the article, and final approval. Prof. Clayton-Smith contributed to the conception and design of the study, analysis and interpretation of data, drafting the article, and final approval. Dr. García-Fiñana supervised the data analysis and contributed to the interpretation of results, drafting of the article, and final approval. Ms. Irwin contributed to data collection, interpretation of data, drafting of the article, and final approval. Dr. Morrow contributed to the acquisition of funding, conception and design of the study, analysis and interpretation of data, drafting the article, and final approval. Dr. Shallcross contributed to data collection, interpretation of data, drafting of the article, and final approval. Prof. Baker contributed to acquisition of funding, conception and design of the study, analysis and interpretation of data, drafting the article, and final approval.
STUDY FUNDING
The study was solely funded by Epilepsy Research UK (P0902). During the write-up period of this work, Dr. Bromley was funded by the National Institute for Health Research (NIHR) (PDF-2013-06-041).
DISCLOSURE
R. Bromley has received lecture fees from Sanofi-Aventis (2 occasions), received conference travel support from UCB Pharma, and provided expert testimony pertaining to fetal anticonvulsant syndrome. R. Calderbank, C. Cheyne, C. Rooney, and P. Trayner report no disclosures relevant to the manuscript. J. Clayton-Smith has given expert testimony pertaining to fetal anticonvulsant syndrome. M. García-Fiñana reports no disclosures relevant to the manuscript. B. Irwin received sponsorship to attend meetings and honoraria for presentations from Eisai, UCB, and Sanofi-Aventis. J. Morrow has received unrestricted educational grants from Eisai, GlaxoSmithKline, Novartis, Sanofi-Aventis, Pfizer, and UCB for the running of the UK Epilepsy and Pregnancy Register. R. Shallcross has attended conferences with the support of UCB Pharma and has received honorarium for lectures. G. Baker has received educational grants from Sanofi-Aventis to support this research directly; he has received educational grants from UCB Pharma and lecture speaker fees from Sanofi-Aventis, UCB Pharma, and GSK. Prof. Baker has given expert testimony on fetal anticonvulsant syndrome. Go to Neurology.org for full disclosures
ACKNOWLEDGMENT
The authors thank all of the participants who contributed to the study.
Footnotes
UK Epilepsy and Pregnancy Register coinvestigators are listed at Neurology.org.
Go to Neurology.org for full disclosures. Funding information and disclosures deemed relevant by the authors, if any, are provided at the end of the article.
Supplemental data at Neurology.org
- Received July 29, 2015.
- Accepted in final form November 17, 2015.
- © 2016 American Academy of Neurology
REFERENCES
Disputes & Debates: Rapid online correspondence
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within the last 8 weeks.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.