Neurodevelopment Following Exposure to Antiseizure Medications in Utero: A Review

Abstract

Exposure in the womb to antiseizure medications and their potential impact on the brain of the developing child has long been researched. Despite this long period of interest, this review highlights that above the well-known risks associated with valproate exposure, there are more data required for conclusions regarding all other antiseizure medications. Limited experience with phenytoin and phenobarbital in monotherapy makes clearly defining the risk to later child postnatal functioning difficult, although the evidence of an impact is stronger for phenobarbital than for phenytoin. The widely prescribed lamotrigine is limited in its investigation in comparison to unexposed control children, and whilst it has been demonstrated to carry a lower risk than valproate for certain outcomes, whether it is associated with a more moderate impact on wider aspects of neurodevelopmental functioning is still to be understood. Data for levetiracetam, topiramate and oxcarbazepine are too limited to confidently draw conclusions for most neurodevelopmental outcomes. This slow accumulation of evidence impacts on the safest use of medications in pregnancy and makes counseling women regarding the risks and benefits of specific antiseizure medications difficult. Improved focus, funding, and research methodologies are urgently needed.

1. INTRODUCTION

Exposure in the womb to certain medications can lead to alterations in the development of the fetus leading to both physical anomalies, growth alterations and functional implications such as neurodevelopmental deficits [1]. Neurodevelopment is not a single outcome and the reporting of its associated outcomes cannot be reduced to a single figure. The term covers a diverse range of brain functions from intellectual abilities, language, memory and attention skills through to motor development and social functioning. It also refers to clusters of symptoms that form clinical diagnoses, such as autistic spectrum disorder and attention deficit hyperactivity disorder. The development of the brain is a protracted process that begins in utero but continues to unfold and be influenced through the postnatal years. Given this extended period of development, an impairment may not be evident early on and follow up for a number of years is required; noting that different skill types mature at different ages.

The breadth of diversity and the prolonged developmental phase may make neurodevelopmental outcomes appear more nebulous than their structural equivalents. However, if considered as a set of related but specific individual outcomes, measured by a standardized assessment, within the correct age range, the individual areas of neurodevelopment can be clearly defined and objectively and sensitively measured. These diverse functions are assessed either through direct researcher assessment, parental reports on questionnaires or, more recently, through data from routine clinical or educational practice. Neurodevelopmental outcomes from medication-exposed cohorts may be reported as risk ratios or odds ratios when the outcome is binary (e.g. the presence or absence of a diagnosis). This type of reporting mirrors that observed for other key teratogenic outcomes such as congenital anomaly prevalence. In contrast, many cognitive, motor, social and behavioural outcomes are measured on continuous scales, offering increased sensitivity as all children’s scores are included and compared on a common normally distributed scale. A deviation from the expected developmental trajectory would be evidenced by a decrease in the mean score for one group in comparison to another, and such a reduction in group mean, in turn, leads to an increase in the number of children who have abilities below the average range. It should also be considered that such reductions are a mean value for the group and for individual children, their reduction will be greater or lower than that value depending on a number of susceptibility factors.

The dose, timing and duration of exposure to certain medications is also an important consideration. Due to its diverse and protracted nature, it is not possible to identify a single “critical period” for neurodevelopment as the brain is under active development throughout the entire gestational period (and beyond) [1]. This stands in contrast to the fetus being most vulnerable to structural anomalies during the first trimester. Different areas of the brain mediate different skills, and each area has its own period of heightened susceptibility to teratogenic effects. However, a detailed discussion of these periods is beyond the scope of the current review and they have been described elsewhere [2, 3]. More pertinent to the current review is the impact of exposure dose on teratogenic effect, and so this is included where available.

Prenatal exposure to anti-seizure medications (ASMs) and their potential impact on fetal brain development and later functioning during childhood have long been discussed [4]. Despite the long period of interest, research has been slow and there are still many ASMs without evidence as to their impact on brain development in the womb. To date, a number of ASMs have been associated with altered neurodevelopmental outcomes in animal models and human studies with varying degrees of evidence. As the review below highlights, valproate has the clearest delineated risk in humans and the current state of evidence for each of the widely utilized ASMs is summarized. Given that describing children exposed to distinct monotherapy ASMs as a single unitary group is cautioned against [5], each of the commonly utilized ASM monotherapies are reviewed in turn against control group data (data from unexposed children whose mothers either have or do not have epilepsy) or other ASM comparator groups.

2. SPECIFIC NEURODEVELOPMENTAL OUTCOMES FOR SPECIFIC ANTISEIZURE MEDICATION MONOTHERAPY

This section provides a review of human data for specific ASMs, where available. A Medline (OVID) search was conducted (updated March 2021) and compared to the obtained results from two previously completed systematic reviews competed by this group [5, 6].

2.1. Sodium Valproate

Valproate is thought of as a broad spectrum ASM and while the exact mechanism of action remains unclear, it is believed to act as both a sodium and calcium channel blocker, which additionally acts on GABA [7]. Valproate is an efficacious treatment for generalized epilepsy [8] but is now under strict regulatory restriction due to its impact on the developing fetus [9]. From case reports through to population database studies, there is clear evidence that valproate is associated with an increased risk of a wide range of neurodevelopmental difficulties. In infancy, attainment of early developmental skills can be delayed [5, 10-13]. In preschool and school age populations, cognitive skills such as IQ, language, attention and memory functioning are poorer [5, 14-25], rates of intellectual or learning disability are increased [26] and the need for specialized educational support greater [12, 14, 27, 28]. Social difficulties are also commonly reported [18, 29, 30] and, along with the language difficulties, leave children at a higher risk of an autistic spectrum disorder diagnosis [31-34].

The impact of valproate on the developing brain is related to the dose of the exposure, with increases in risk occurring at a daily dose as low as 800mg/d [5, 14, 16, 18, 20, 23, 25, 31]. In one study, following adjustment for other influential variables, the associated IQ reduction was 9.7 points for doses of valproate over 800mg/daily, but at doses, under this the associated reduction was 5.0 IQ points [14]. Despite doses under 800-1000mg/daily presenting a lower level of risk to the developing brain, evidence has demonstrated linear relationships between dose and outcome [14]; suggesting that even at these lower doses there may be a degree of impact on the developing brain. To date, the vast majority of women who have taken part in research have taken valproate throughout pregnancy. However, recent data from the French National Health Data System demonstrates a cumulative effect with the highest risk for certain aspects of neurodevelopment coming with chronic use throughout the three trimesters but notes that the risk from isolated 2^nd and 3^rd trimester exposure is greater than when exposure just occurs in the first trimester [35].

The impacts on the developing brain are more common than the presence of major congenital anomalies and can occur in children without major congenital anomalies [14, 22-24, 30]. Those displaying the physical risks, including facial dysmorphia, are more frequently at risk of altered cognitive and behavioral trajectories [28, 36, 37]. Recently published diagnostic and care guidelines suggest that all children exposed to valproate should have their development carefully reviewed [37].

2.2. Carbamazepine

The sodium channel blocker carbamazepine has been associated with an increased risk of neural tube anomalies [38] and a distinctive facial presentation [12, 36, 39]. Investigations have demonstrated that early developmental trajectories may also be altered in comparison to control children [11, 40, 41], but this has not been uniformly concluded [5, 10, 39, 42-45] and the totality of the data currently would suggest that in infancy, at a group level, the emergence of early skills are not severely disrupted when compared to control children.

As the children get older, there is no clear, substantial global impact of carbamazepine exposure on child cognition either. Standardized assessments of intellectual functioning conducted by blinded investigators routinely fail to demonstrate a poorer level of global intelligence in comparison to unexposed control children [14, 25, 27, 46, 47]. However, data from two linked studies did demonstrate poorer verbal reasoning abilities [14, 20], though this had not been observed in an early study with a similar methodology [46]. The intellectual abilities of children exposed to carbamazepine, however, are superior to those exposed to valproate [5, 14, 20, 25, 46, 47] and the data would suggest that their development is comparable to children exposed to levetiracetam [47] lamotrigine [47], [14, 20] and possibly phenytoin [20, 48].

Investigation into language and communication skills more specifically, have failed to find a clear association with reduced skills compared to control/comparator groups [10, 18, 22, 44, 47] with one exception [49]. Memory, attention and executive functioning skills have received less research attention but studies to date have not demonstrated substantial disruption to functioning in these areas [19, 20, 47], although the data is limited. Memory skills, for example, have been highlighted as an area in need of further investigation in two studies of carbamazepine exposed children [15, 17].

The application of population datasets to investigate the outcomes of children exposed in the womb to ASMs has led to the emergence of data pertaining to clinical diagnoses and educational skills such as languages and mathematics in large populations. For carbamazepine exposed children there is no clear evidence of a poorer level of educational examination results [50-52] and no increased risk of autistic spectrum disorder [31-34]. A number of studies have now investigated whether there is a risk from higher doses of carbamazepine exposure. The vast majority fail to find a dose-effect across a range of different neurodevelopmental outcomes [5, 14, 20, 31, 41, 46, 49].

2.3. Lamotrigine

Lamotrigine is today one of the most common monotherapy treatments for epilepsy with a mechanism through the sodium and calcium channels. Evidence regarding fetal risks has suggested that the risk for major congenital anomalies is lower than other ASMs and not above the background population risk [53, 54]. Global development in infancy does not appear to be below that of unexposed control children [10, 11, 44, 45], higher in comparison to those exposed to valproate [10, 55] and comparable to children exposed to carbamazepine and phenytoin [55]. The data regarding school-aged intellectual functioning is reassuring to date but is limited to three cohorts [14, 20, 47]; two of which have a degree of participant overlap [14, 20] and two not comparing to non-exposed controls [20, 47]. The two largest studies into the IQ of children exposed to lamotrigine have compared lamotrigine directly to other ASMs and demonstrated a higher level of performance than the children exposed to valproate in utero on blinded assessment [20, 47]. Comparisons to other ‘newer’ ASMs are substantially limited. In comparison to levetiracetam, a single study found no difference in intellectual ability[47]; but the levetiracetam group was small and only powered to reliably detect a large reduction in group mean scores.

Language development in children exposed to lamotrigine does not appear to be impaired in comparison to non-exposed control children [10, 20, 44, 45, 56]. Data re-purposed from the Norwegian longitudinal study MoBa shows a risk of poorer sentence formulation skills appearing at 36 months [44], however, this is based on a single question from a wider questionnaire set and difficulties with other single aspects of language development were not seen here or in a later follow up of this cohort at 5-6 years [56]. Communication abilities are not lower than age- related expectations and were higher in comparison to the children exposed to valproate [16, 18]. Early fine and gross motor skill development do not appear to be impaired following adjustment for additional influencing factors [10, 11, 21, 55] but in one study, preschool to school-aged fine and gross motor skills were poorer to a degree in the children exposed to lamotrigine [57] and this requires replication.

Whilst the risk of autistic spectrum disorder and attention deficit hyperactivity disorder are significantly increased following in utero exposure to valproate, the rate of autistic spectrum disorders diagnosis does not appear to be higher in the children exposed to lamotrigine in utero versus non exposed children [31, 33]. However, there are studies that report increased parental ratings of autistic spectrum type of symptomatology [29, 44, 58]. Based on parent completed questionnaires in preschool children (3-4 years) the data from two studies report an increased level of parental concern about aspects of their child’s development which are features of the autistic spectrum [44, 58]. A third study, with school-aged children found similar increases in parental ratings of concern, but the simultaneously completed blinded researcher ratings of behavior did not support these concerns regarding behavior [29]. However, the lamotrigine exposed children in this cohort did score lower on a task of cognitive affect recognition [29, 47]. How should these data be interpreted in light of the findings that clinically diagnosed autism are not increased? It is possible that parents are reporting sub-diagnostic threshold behavioral disruption and that these either remain sub-diagnostic or that they resolve in time. A further possibility is that there is an impact of unblinded ratings; parents who completed the questionnaires would not have been blinded to medication use or type. Further investigation is required to ensure that any risk to the social functioning of children exposed to lamotrigine is fully understood and future women are counseled appropriately.

Educational attainments of children exposed to lamotrigine in two population cohorts do not appear to be unusually low [51, 52] and rates of learning disability are in line with background populations in Denmark [26]. Finally, no association between dose of lamotrigine and outcome has been identified for early development, intelligence, language functioning and other skills [5, 14, 18, 20, 47, 51, 55-57, 59]; although larger groups of children exposed to lamotrigine are still required to fully answer this important question.

2.4. Levetiracetam

A congenital anomaly rate comparable to the background population rate has been demonstrated for levetiracetam [54, 60]. This medication is believed to act by modulating synaptic neurotransmitter release through binding to the synaptic vesicle protein SV2A in the brain [7]. Early developmental milestone attainment is reported to be comparable [13, 61] and intellectual abilities in school-aged children exposed to levetiracetam are not reduced in comparison to unexposed control children [62]. However, the majority of evidence comes from a single longitudinal series. There is limited data on other cognitive areas in comparison to typically developing control children, but the areas of language, memory and attention are reported in one study to be consistent with age expectations [62]. A number of studies have included levetiracetam exposed children but the monotherapy group size is too small for drawing conclusions [26, 56].

In comparison to valproate there is evidence of advanced early development [13, 61], higher school-aged IQ [47, 62], and superior language, memory and attention scores [47, 62]. Additionally, there are fewer social development difficulties [47] and exposure to levetiracetam does not appear to be associated with the increased risk of autistic spectrum disorder that valproate is [31]. Few studies have directly compared the neurodevelopment of children exposed to levetiracetam to other newer ASMs, including lamotrigine. No differences between levetiracetam and lamotrigine school aged cognitive functioning is reported in one study; but the size of the levetiracetam group in this study was relatively small and would only be reliable for a large effect on cognition [47]. Further, young children with levetiracetam exposure do not appear at risk of clinical diagnoses [31], but this requires replication in an older cohort which is past the average age of diagnosis for many of these conditions. There is limited investigation into the potential risk posed by higher doses. Research to date has not shown a dose-effect [13, 31, 61, 62], but larger group sizes are required to be able to adequately address this question.

2.5. Oxcarbazepine

In contrast to the other ASMs, the majority of evidence regarding the longer-term neurodevelopment of children exposed to the sodium channel blocker oxcarbazepine comes from secondary use of routine population data. Studies using Danish population register data demonstrated that the rate of autistic spectrum disorder or attention deficit hyperactivity disorder is not found to be increased in comparison to unexposed controls [33, 63] and recent replication of this came from a French national cohort [31]. In terms of academic skills, Danish language levels were comparable to control children’s results, but mathematical skills were lower [51]. Data from other groups using directly collected research data have been too small to draw conclusions against unexposed children [56, 59, 64], demonstrating the usefulness of the national population registers in terms of cohort size. Comparison to outcomes of other ASM exposed cohorts is limited to lamotrigine, where the limited outcomes assessed show no differences [31, 33].

2.6. Phenobarbital

Barbiturate exposure in the later stages of pregnancy or around delivery is associated with an alteration in neonatal behaviors [65]. Additionally, plasma levels of phenobarbital in the neonate are present for a number of days postnatally [66] and neonatal withdrawal symptoms are observed in some [67]. Phenobarbital is a stimulator of GABAA and its use as an ASM has declined over the past decades. Phenobarbital is capable of disrupting fetal brain development inducing long term neurodevelopmental difficulties in rats [68], but its human investigations are limited by the era of its use. Due to the association with an increased risk of congenital anomalies [69] and early reports of disrupted development, there is a tendency to conclude that phenobarbital is associated with altered fetal brain development. The picture, however, is much more opaque and limited by the number of children exposed to monotherapy for whom data is available.

In infancy early global development does not appear to be altered [12, 70] but the use of blinded, standardized assessment of skills with adequate cohort sizes are few. The Kerala Pregnancy Register found comparable development in infancy and in the preschool years for the children exposed to phenobarbital [42], but found poorer intellectual abilities in middle childhood [48]. Despite a relatively small group size (n=22), the effect difference was relatively large at 8 IQ points in comparison to control children and was replicated when phenobarbital use in polytherapy (n=69) was considered [48]. Further, recent work from this group has documented enduring language difficulties in children exposed to monotherapy and polytherapy phenobarbital [71].

In the study by Dessens and colleagues [72], adult age cognitive functioning amongst those exposed to ASMs as fetuses was investigated in 65 phenobarbital monotherapy exposed adults. Blinded assessments found reduced verbal reasoning skills and an increased rate of learning difficulties and learning disability (termed mental retardation in the paper) with rates of 14% and 13% for the monotherapy phenobarbital and the phenobarbital+ phenytoin polytherapy group in comparison to the matched control level of just 1%. Similarly, Reinisch and colleagues demonstrated the impact of prenatal phenobarbital exposure in a non-epilepsy cohort [73]. Uniquely, using two independent samples the first of their studies demonstrated reduced verbal IQ to a magnitude of 7 IQ points and highlighted that dose of the phenobarbital exposure and the trimester of exposure was associated with a more negative outcome. In their replication study, which included 81 men, a reduction in IQ was found to a similar level and replication of the impact of dose and timing of exposure were obtained [73].

2.7. Phenytoin

Similar to phenobarbital, there is evidence that prenatal exposure to phenytoin (sodium channel blocker) is associated with postnatal behavioral disruption in rodent models [74] but human studies remain limited. In the 1970’s evidence surfaced which suggested that there may well be an effect on intelligence in children with the physical symptoms of Fetal Hydantoin Syndrome [75, 76]. Scolnik and colleagues found poorer cognitive development scores in 34 phenytoin exposed children following blinded research assessments and adjustment for the influence of maternal IQ and other factors [77]; most notable, was the high rate of cognitive development below the average range. One other research group also noted an association between phenytoin exposure and child IQ [78], but other studies have found normal patterns of early global development in phenytoin exposed children [39, 43, 79-81]. A previous meta-analysis of prospective study evidence of infant global cognitive development in comparison to control children did not find reduced infant performance [5]. Further, the mean intellectual abilities of school-aged children exposed to phenytoin have not been reported to be reduced in comparison to control children [25, 48], but are limited in the number of studies and by study cohort sizes for monotherapy phenytoin.

When compared to children exposed to other ASMs, the children exposed to phenytoin have been found to have higher levels of intellectual and other cognitive skills than children exposed to valproate [20, 55]. A meta-analysis of prospective study evidence reported higher intellectual abilities in comparison to the children exposed to valproate (MD 7 IQ points), but with no detectable performance difference in comparison to the children exposed to carbamazepine[5]. In the NEAD study, the largest monotherapy cohorts of phenytoin exposed children (n=51) were followed longitudinally until 6 years of age. Intellectual functioning was comparable to those exposed to lamotrigine or carbamazepine and higher than the children exposed to valproate [20, 55], but a dose-dependent effect on adaptive daily living skills was observed [16]. Further, there were increased rates of below-average development in this phenytoin exposed cohort when they were young [20], which could possibly indicate a small group of children with heightened susceptibility.

Investigations into other cognitive skills such as language, memory or attentional abilities are even further limited, and no firm conclusions can be reliably drawn. Scolnik and colleagues found reduced early language development [49, 77] whilst Wide and colleagues reported a subtle impact on motor development in preschool children [43]. Dean et al. [12] reported elevated rates of nonverbal developmental delays. There are no studies on whether phenytoin exposure is associated with an increased risk of clinical diagnoses such as autism spectrum disorder or attention deficit hyperactivity disorder; but the era of its widespread use corresponds to a time when only very severe cases would be diagnosed with these conditions.

2.8. Topiramate

Despite being in use for a number of decades, there is very limited experience with regards to the longer-term outcomes of children exposed to Topiramate (sodium channel blocker, stimulation of GABA-A receptors) in the womb. Recent work has delineated an increased risk to the physical development of the fetus in terms of small for gestational age (SGA) [78] and increased risks of orofacial clefts [79].

Data which was powered only to detect a large effect size, found comparable cognitive and behavioral outcomes in comparison to control children on blinded assessments of school aged children [62]. In contrast, a population-based study found that rates of children with a learning disability were increased [26]; but again, numbers were limited (n=27) here for an outcome where a high degree of variance is likely. In the largest cohort of topiramate exposed children, Blotière and colleagues [31] reported that there was no increase in the prevalence of diagnostic codes for autism, learning disability, or attention deficits hyperactivity disorder or visits to a speech and language therapist for 477 children exposed to topiramate monotherapy in utero. Whilst this cohort is impressive in terms of its size, the majority of children were below school age and therefore below the average age of diagnosis for many of these conditions and therefore these results must be interpreted with caution. Other work to date included small cohorts of topiramate exposed children and have reported mixed results [56-59]. There is currently no reliable evidence as to whether there is a dose associated risk to brain development and functioning from topiramate exposure in utero.

3. DISCUSSION

The overwhelming conclusion of this review is that there is too large a latency between the onset of a new ASMs use and conclusions regarding documentation of either fetal risk or relative safety. It is difficult then to provide women with clear risk-benefit information about specific ASM treatment options.

The available data highlight increased levels of risk above the background for valproate and phenobarbital exposure. Carbamazepine does not appear to carry a clear and significant risk to intellectual development in exposed children. However, it still remains to be seen as to whether there are more specific areas of deficits or whether there are subgroups within the exposed children (e.g. those with dysmorphic features) that are at an increased risk of neurodevelopmental difficulties. For monotherapy phenytoin exposure in utero, the picture maybe not be aa bleak as once thought. However, there are limits to the data available and a risk to fetal brain development cannot yet be ruled out. Animal and human data, from both prenatal and neonatal use, demonstrate an impact on brain development and functioning; but drawing firm conclusions and quantifying the risk is difficult due to the era in which the majority of investigations occurred and its frequent use in combination with phenobarbital. Lamotrigine has become the treatment of choice in many countries. However current data is limited when it comes to comparisons to non-exposed control children and follow-up remains into only the early school years; therefore, a risk to later maturing cognitive skills cannot yet be ruled out. Of particular note regarding lamotrigine are the parental-rated concerns regarding social and interpersonal functioning, which require further investigation and delineation. Similar are the cases of levetiracetam and oxcarbazepine, whereby investigations are limited by cohort size or the length of follow-up.

Considering the importance of this issue, the fact that data is so limited across the ASMs, with the exception of valproate, places both the mother and her fetus at an increased risk of poorer treatment outcomes. A lack of rigorous evidence on which to base treatment decisions undoubtedly has, and maybe continues, to lead to more children being exposed to development-altering medications than is necessary. Conversely, a woman may avoid what would be an optimal treatment for her epilepsy due to a lack of evidence regarding a medication’s safety in pregnancy. Valproate very clearly highlights how severe the problem can be. In this case, medication was used for decades prior to a clear picture of significant risk, after which then followed a substantial reduction in use. There is no clear estimate of the number of children and young adults affected around the world by valproate exposure but the widespread usage and the rate of the associated difficulties would suggests is tens of thousands over the last 40 years.

The slow movement of research in this area is multifactorial, but continued failure to prioritize the impact on the developing brain is central to this latency. Research groups, clinical guidelines [82] and regulators [83, 84] primarily focus on the presence or absence of major congenital anomalies, with longer term child brain functioning being a secondary consideration. A recent survey by the International League Against Epilepsy (ILAE) Pregnancy and Women’s Task Force found that only 25% of member countries had guidelines that mention ASM specific risks to neurodevelopment [82] for example. Research funding, as for many topics, is challenging to come by [85], but the perceived nebulous nature of neurodevelopment undoubtedly also impacts on the amount of research undertaken. Whilst there are numerous aspects of how the brain functions which fall under the term ‘neurodevelopment’ and a longer period of follow up is required, a team with the correct knowledge regarding child brain development, measurement and interpretation can develop methodologies capable of reliably assessing longer term brain outcomes; able to detect deviations from normal developmental patterns as early as the infant years.

There is also work to be done to improve research methodologies and interpretation. The reporting of single ASM drug groups or single ‘monotherapy’ or ‘polytherapy’ groups has undoubtedly limited conclusions and in light of clear differential outcomes across the specific ASM drugs. These reporting practices should cease. The relatively recent emergence of population databases to investigate neurodevelopmental outcomes presents an enormous opportunity to acquire data from large cohorts and will inevitably speed up the availability of evidence for certain outcomes such as clinical diagnoses (e.g. autistic spectrum disorder or attention deficit hyperactivity disorder). However, these are only able to provide data on a subset of the exposed population who obtained a diagnosis and not those who were not referred or those whose symptoms sat just below the boundary for diagnosis but who experience considerable impact on their daily functioning. Academic results provide a proxy measurement of cognitive functioning and future work is required to understand the sensitivity of such data for the investigation of prenatal exposure outcomes. For neurodevelopmental outcomes which sit primarily outside of health diagnoses, innovation is still required to speed up the collection of neurodevelopmental outcome data. Disease-specific pregnancy registers such as the pregnancy and epilepsy registers around the world [86], whilst primarily focused on congenital anomalies, have been used in an ad hoc manner to follow up cohorts of ASM exposed children using blinded assessment. The extension of these to include neurodevelopmental outcomes as a routine part of their work, as done in the Kerala Register [42, 48], would improve the detection of medications that carry an increased risk to later postnatal child development. With carefully observed outcomes on a continuous scale of measurement the number of subjects required to achieve adequate power are far fewer for than needed for relatively rare binary outcomes such as major congenital anomalies. Small effect sizes or more moderate levels of impact on the developing brain will require larger groups but are still likely to be obtainable when recruitment is over a number of hospitals or regions.

The adapted principles of teratology [1, 2] also need central consideration in future investigations. Timing, duration and dose of the exposure are important and will likely influence the type and severity of the outcomes observed. The susceptibility of a particular subgroup of children should also be considered. Whilst the normative curve for IQ data appears to remain symmetrical but shifted to the left of the scale [14], it is possible that a bimodal distribution may be possible following some exposures. In these cases, children with a particular susceptibility factor or factors may demonstrate an increased risk of neurodevelopmental difficulties, which is not seen in others within the exposure cohort. Efforts, therefore should be made to check continuous data for bimodal distributions and consider rates of below-average cognitive or social scores and increased rates of diagnoses within specific subgroups. An example of this would be for topiramate, where work is needed regarding the outcomes of the children born SGA as they may have different neurodevelopmental risks than those exposed to topiramate who were not SGA.

Finally, the work reviewed here pertains to evidence of each medication when used as monotherapy. The evidence for polytherapy is difficult to interpret, as single polytherapy groups contain many different combinations of exposures and the make-up of each cohort varies from study to study and there are too few similarities in combinations and methodologies to provide definitive review. Polytherapy combinations which include valproate carry a higher risk [13], but even here, more is required to understand the doses used in combination therapies and whether the dose in fact drives the additional risk. Detailed work looking at specific polytherapy exposures is urgently required in addition to work extending monotherapy exposure investigations, as polytherapy treatment is a reality for many women and their children.

CONCLUSION

There remain clear limitations to our current knowledge regarding in utero exposure to ASMs which undermine decision-making regarding treatment for women requiring ASM. Whilst this review focuses on the outcomes in the fetus, maternal disease control is also important and ASMs have different efficacy rates for the different epilepsies [7, 87] and careful decision-making is required to optimise both maternal and fetal health. Importantly, a lack of data for specific ASMs should not be misinterpreted as evidence of safety, and many of the newer ASMs are without data. This undermines the ability to provide informed preconceptual counseling. Comparisons of ASMs directly are important but should be conducted with an additional comparison to an unexposed control group to ensure more moderate effects on brain development can be understood. This review highlights the need for further robust studies to be carried out in order to provide both patients and healthcare practitioners with the information they need to make clear, informed treatment decisions.

CONSENT FOR PUBLICATION

Not applicable.

FUNDING

Dr. Bromley was supported in part for this work by the National Institute for Health Research during the period of this work (PDF-2013-06-041). The views expressed in this publication are those of the author(s) and not necessarily those of the NHS, the National Institute for Health Research, or the UK Department of Health. Dr. Bromley was also part-funded for this work by Epilepsy Research UK (research grant 1703).

CONFLICT OF INTEREST

The authors declare no conflict of interest, financial or otherwise.