Adaptive behaviour in children exposed to topiramate in the womb: An observational cohort study

Objective: Many women with epilepsy need to continue anti-seizure medications (ASMs) throughout pregnancy. The current study investigated adaptive behaviour outcomes in children exposed to topiramate in the womb. Method: An observational, cross-sectional study was designed, recruiting mother-child-pairs from the UK Epilepsy and Pregnancy Register (UKEPR). Health, developmental histories and Vineland Adaptive Behaviour Scale-Third Edition (VABS-III) assessments were administered via telephone by a blinded researcher, supplemented with prospectively collected pregnancy and medication information. Topiramate monotherapy exposed children were compared to VABS-III normative data as recruitment was disrupted by the COVID-19 pandemic. Results: Thirty-four women with epilepsy from 135 (25%) initially agreed to participate in the study, of whom 26 women completed telephone interviews about their children ( n = 28). Children ranged from 2.5 to 17 years of age at the time of assessment. Six topiramate-exposed children were born small for gestational age, and there were significant associations between birthweight, dose and VABS-III scores. Significantly lower scores were observed in topiramate-exposed children ( n = 21) with a significant dose-response relationship established after adjustment for parental educational level. Daily mean dosage was 280.21 mg, with high dosages of topiramate associated with a 12-point reduction in VABS-III scores. Additionally, four topiramate-exposed children (19.05%) had diagnoses of Autism Spectrum Disorder, which was significantly higher than UK prevalence rates (1.1%). Conclusions: The findings of poorer adaptive behaviour, higher incidence of ASD and associations with birth weight are of concern and require further validation and replication using larger prospectively-recruited samples and comparator cohorts. Implications for research and clinical practice are discussed.


Introduction
Topiramate is primarily used for the treatment of epilepsy and migraine [1]. Topiramate may act as a physical teratogen [2,3], carrying higher levels of risk to the developing foetus with respect to major congenital anomalies and in particular, oral clefts [4][5][6][7] and has also been observed to be associated with increased rates of small for gestational age babies [7,8].
Evidence regarding foetal topiramate exposure and subsequent behavioural/ neurodevelopmental outcomes is limited, but a recent population healthcare records study from the Nordic countries demonstrated that topiramate is associated with an increased risk of autistic spectrum disorder [9]. A previous population healthcare record study from Denmark found an elevated risk of learning disability for topiramate-exposed children (n = 27) [10], which was also replicated in the larger sample from Bjørk and colleagues [9].
In observational cohort studies cohort size has been limited and this has led to conflicting outcomes. For example, in a blinded prospective cohort study, the cognitive abilities of children exposed to topiramate were no different to unexposed controls [11] (n = 27), but in contrast, Rihtman et al. [12] observed significantly poorer behavioural and cognitive outcomes compared to a control cohort, though this was based on an extremely small sample (n = 9). Data from the MOBA study reported non-significant differences in language skills [13] and autistic traits compared to controls [14] but as above, cohorts were small (n = 4 and n = 6, respectively).
Without evidence upon which to base clinical guidelines, women and prescribers alike are unable to make informed choices about topiramate treatment during pregnancy and further delineation of the risks are required.

Aims
The primary aim was to determine whether topiramate exposure in utero impacted later child adaptive behaviour skills. Subsidiary aims were to consider other developmental endpoints and potential covarying factors, their impact on main outcomes, and to explore associations between topiramate exposure and/or dose with other neurodevelopmental outcomes. It was hypothesised that: Topiramate-exposed children would show significantly poorer overall adaptive behaviour, as indicated by lower composite scores on the Vineland Adaptive Behaviour Scale 3rd Edition (VABS-III) [15], relative to the normative mean of 100 (SD 15).

Design
The study used a cross-sectional and observational design. The primary dependant outcome was adaptive behavioural skills, including communication, daily living and socialisation skills, as measured by the VABS-III. Secondary outcomes were incidence of neurodevelopmental conditions and birth outcomes. Important factors included maternal and paternal educational level, socio-economic status, child gender, alcohol exposure and nicotine exposure.

Recruitment and participants
Mother-child-pairs were identified from the United Kingdom Epilepsy and Pregnancy Register (UKEPR), a prospective research study established in 1996 to investigate major congenital malformation prevalence following in-utero exposure to ASMs [16]. Enrolment onto the register takes place via self-referral or by a health professional within the first/second trimester of pregnancy. Mother-child-pairs were eligible for inclusion in this follow up study if the child had been a live birth and was up to 17 years of age at the time of participation and mothers with epilepsy were taking topiramate monotherapy during pregnancy or were untreated during pregnancy. Families were not invited to participate in cases where conditions associated with neurodevelopment impairment were suspected (e.g., maternal learning disability).
Although registration with the UKEPR is prospective, recruitment into the current follow-up study was retrospective. Potentially eligible participants were invited to participate via and on receipt of a positive response, mothers were contacted and screened for eligibility before formal enrolment into the study.
Ethical approvals were obtained from the Health Research Authority (HRA, study reference: 19/NW/0299), with local approvals also granted by Manchester University Hospitals NHS Foundation Trust and Belfast Health and Social Care Trust which host the UKEPR. All participants provided informed written consent.

Procedure and measures
Data collection took place via telephone interviews lasting 40 -60 minutes, with assessments conducted by a blinded researcher during 2019-2020. A brief, semi-structured health and background interview was undertaken with mothers, followed by a parent-rated measure of adaptive behaviour. Adaptive behaviour was the primary outcome, measured using the VABS-III [15]. The VABS-III is widely used, has strong psychometric properties [17] and has been used in studies of ASM exposure during pregnancy [18,19]. The measure provides overall estimates of adaptive behaviour skills (ABC) as well as domain-specific estimates of communication, daily living and socialisation skills ( Table 1).
On the VABS-III, domain and ABC scores are standardised, with lower scores conferring poorer adaptive behaviour skills. Scores can be classified using qualitative descriptors, based on deviation from the expected mean (100, SD = 15), with scores of 85 or lower classified as below the average range.
Prospectively collected data on additional factors, such as exposure status, dosage and pregnancy/birth outcomes, were obtained via the UKEPR database. Individual dosage information was available and represented dose of topiramate around the time of enrolment on the register. UKEPR data on birth weight and gestational age were inputted into UK World Health Organisation growth charts [20] in order to calculate the birth weight for gestational age centiles for each study child, identifying those children falling below the 10th centile and therefore classed as small for gestational age [21]. Estimates of socio-economic status were generated by inputting postcodes into nationally-held and freely available statistics on deprivation indices [22,23].
A semi-structured interview format was used to gather health and background information that was not available on the UKEPR database. This included information about incidence of health and neurodevelopmental conditions within the study children and parental factors including educational attainment and family history of special educational needs, illnesses and neurodevelopmental conditions. Participants were debriefed, with ethical standards adhered to throughout. VABS-III data were double scored and data entry was checked by a second researcher to reduce scoring and data entry errors.

Data analysis
Although a comparison between a topiramate-exposed cohort and a 'no medication' unexposed control cohort was initially planned, an adequately-sized control cohort was not obtained due to recruitment difficulties and the COVID-19 pandemic disrupting the final months of recruitment and interviews. Thus, comparisons to the VABS-III normative sample (n = 2560) were made, an approach previously utilised in this area [18,24].
After tests for normality, the primary analysis was a mean comparison of adaptive behaviour skills, as assessed by the VABS-III, to normative sample data [15]. Potential confounding variables were explored by assessing their relationships with VABS-III scores, using mean difference and correlational analyses with the conventional significance value of 0.05 adopted. Variables explored were socio-economic status, parental educational attainment, parental age at birth, employment status, maternal epilepsy type, seizure exposure, other maternal health conditions, folate status, breastfeeding status, alcohol exposure, nicotine exposure, gestational age at birth, birth Table 1 Overview of Vineland adaptive behaviour scale-third edition (VABS-III). N.B. Additional domains of motor skills and maladaptive behaviour were not included due to unavailability of normative estimates across the age range under study.
weight, child gender, child age at assessment and other child health factors. Cut-off values for low and high dosages (≤200 mg/day and >200 mg/day, respectively) were informed by the British National Formulary [25] guidance for topiramate.
Correlational and mean difference analyses were completed to examine the impact of topiramate exposure/dose on other outcomes, including gestational weight at birth, birth weight centile, incidence of malformations, incidence of neurodevelopmental conditions and the presence of special educational needs. All data analysis were performed using IBM SPSS Statistics 25.

Results
One hundred and thirty-five invitations were sent out to potentially eligible mother-child-pairs (n = 106 invitations to the topiramate group and n = 29 invitations to the 'no medication' group). Two mothers declined participation (1.5%), four invitations were returned to sender (3.0%) and ninety-one invitations received no response (67.4%). Positive responses for thirty-five children (including n = 3 sibling pairs) were received (25.9%), of whom thirty-two were enroled into the study. Six mothers (n = 7 children including n = 1 sibling pair) were either not enroled or did not provide data due to being not contactable (n = 3), due to not attending the arranged appointment (n = 2) or due to a change in circumstances (n = 2). Therefore, of those enroled into the study, data was provided by 26 women for 28 children, corresponding to a 81.2% participant completion rate. Please see Fig. 1 below.

Participants
A total of 26 mothers took part in the study and provided information about 28 children (including two sibling pairs). The monotherapy topiramate-exposed and 'no medication' groups were unequal in size (n = 25 and n = 3, correspondingly) and not appropriate for statistical comparison. Thus, the alternative analytic approach was taken whereby the topiramate-exposed group were compared to the test normative group. Cohort demographics for the sample recruited can be seen in Table 2.
In the topiramate-exposed group, dose of topiramate ranged from 100 to 800 mg total daily dose, with a mean daily dose of 280.21 mg. Twenty-three mothers took topiramate throughout the duration of pregnancy. One mother stopped topiramate at six weeks gestation; this child was included in the topiramate cohort. Nine children (36%) were exposed to maternal seizures in utero, with exposure to convulsive seizures in four cases (16.0%). Six children (24%) were breastfed and almost all children (n = 22, 92%) were exposed to pre/peri-conceptual folate. There was no family history of special educational or developmental conditions. Two children who were siblings (8%) had a paternal family history of malformations. At the time of assessment, 84% (n = 21) of mothers were in employment.
Adopting a conservative approach, four children were excluded from the main analysis due to the presence of factors that may impact neurodevelopment (e.g., acquired brain injury, genetic conditions, neurological conditions). Therefore, of the 24 children entered the analysis, 21 were prenatally exposed to topiramate and three were not exposed to any medication.

Main results
Adaptive behaviour data, as ascertained via the VABS-III were analysed for 21 topiramate-exposed children. Unadjusted means, standard deviations and rates of performance below average adaptive level for those included in the study are presented in Table 3. An equivalent table summarising the outcomes for the children excluded from the analysis (n = 4) is presented in supplementary e-Table 1.
Children exposed to topiramate had poorer levels of adaptive behaviour (see Table 4 and

Influence of non-exposure variables
Demographic and clinical variables were investigated for their influence on ABC and domain scores, as per the method. As it was planned that significantly influencing variables would be adjusted for in subsequent regression analyses, Bonferroni corrections were not applied.
Parental educational attainment yielded significant differences on adaptive behaviour outcomes. Topiramate-exposed children with at least one parent who attended education beyond compulsory requirements had substantially higher ABC scores (MD = 20.29, p = .005), communication skills domain scores (MD = 30.07, p <0.001) and socialisation skills domain scores (MD = 18.43, p = .020) than topiramate-exposed children with no parents having attended higher education. Alcohol exposure also led to significantly different Daily Living Skills domain standard scores; however, on review of alcoholexposed cases, alcohol consumption was very low and infrequent and was thus considered a chance finding. No other significant influences on VABS-III scores were identified.

Dose investigations
Associations between topiramate dosage (low dose = ≤200 mg/d, high dose = >200 mg/day) and neurodevelopmental outcomes were analysed using hierarchical multiple regression, limited to the children exposed to topiramate. Following adjustment for parental higher education, there was a significant negative association between topiramate dose and ABC scores (Fig. 4, β = − 0.405, 95% CI [− 22.347 -− 3.006], p = .013), whereby children exposed to high dose topiramate scored over two thirds of a standard deviation below children exposed to low dose topiramate (B = − 12.678). Thus, once the expected influence of parental education was adjusted for, higher doses of topiramate were associated with poorer communication, socialisation and global adaptive behaviour scores (Fig. 3).  ≤85 would therefore be classified as below average adaptive levels.

Table 4
Results of comparisons against normative sample.

Secondary outcomes
As part of the study's subsidiary aims, further health and neurodevelopmental outcomes were explored in relation to topiramate exposure. Within the topiramate-exposed cohort (n = 21), four children (16.0%) were born with major congenital malformations. Six children (28.6%) were born small for gestational age and when compared with UK population estimates (10%, NHS, 2016) using a binomial test, this proportion was significantly greater than expected (p = .014). A significant negative relationship was identified between birthweight centile and dose of topiramate (see Fig. 4, r(24) = − 0.407, p = .048). Furthermore, birthweight centile was significantly correlated with VABS-III outcomes, with lower centile children obtaining poorer for overall ABC scores (Fig. 5, r(21) = 0.479, p = .028) and socialisation skills scores (r (21) = 0.504, p = .020).
Maternal interview identified five children (23.8%) that did not meet their early developmental milestones on time. Six children (28.6%) were reported as having difficulties with learning at school and three children (14.3%) were in receipt of formal supported learning provision. Six children (28.6%) were reported as having difficulties with social interaction.
Amongst the total sample recruited (n = 28), there were six cases of children with existing diagnoses of Autism Spectrum Disorder (ASD) made via routine clinical services and independent of the study. All six children had been exposed to topiramate; however, two children represented those excluded from analyses due to the presence of other possibly influencing conditions. Therefore, four children (19.05%) in the included topiramate-exposed cohort had formal diagnoses of ASD with no other risk factor which is higher than the prevalence within the general UK population. A summary of the demographics and outcomes for the four topiramate-exposed children diagnosed with ASD can be seen in Table 5.
For all four children with diagnoses of ASD, there was no family history of congenital malformations, special educational needs, ASD or other neurodevelopmental conditions and none had major congenital malformations themselves. The mean daily topiramate exposure dose was 285.00 mg/daily, with three children (60%) exposed to high-dose topiramate (>200 mg/d) in-utero. While there were no cases of major congenital malformation in those with ASD diagnoses, one child was born with a congenital heart problem at birth that self-resolved. Adaptive behaviour skills were below average levels for all children (100%) on all outcomes except daily living skills, where 75% of children (n = 3) fell below average levels. Three children (75%) had formal learning support in place. All four children were born smaller than average for gestational age at birth (≤25th centile), with one child below the 10th centile.

Discussion
The results demonstrated that prenatal exposure to topiramate was associated with poorer daily adaptive behaviour outcomes. In a crosssectional and retrospectively-recruited sample, our comparisons of parent-rated adaptive behaviour with normative sample data indicated that children exposed to topiramate had significantly poorer skills in their daily living skills, socialisation skills and global adaptive behaviour. Despite interrupted recruitment, our sample (n = 21) provided adequate power (90%, d = 0.6, VABS-III score MD = 10) to detect the large effect size observed, with mean scores falling over half a standard deviation from the norm. Whilst direct comparisons are not possible, it is of note that the mean scores generated here are below those reported by   R. Knight et al. studies using a similar methodological approach in the investigation of other ASMs such as lamotrigine, where global adaptive behaviour scores were M = 103 [18]. Findings were both statistically and clinically significant, with 42.9% of the exposed cohort falling below the average range for global adaptive behaviour. Significant dose-response associations were observed, with lower adaptive behaviour scores in cases of high-dose (>200 mg/d) topiramate exposure. Additional associations were also identified between birthweight centile and adaptive behaviour outcomes and between topiramate dose and birthweight centile, in line with previous research [7,8]. Further work is needed to understand the adaptive behaviour in larger groups exposed to topiramate, in comparison to other ASM exposed groups, including other cognitive and behavioural outcomes still to be comprehensively delineated.
After the cautious exclusion of children with other conditions or a positive family history linked with neurodevelopmental outcomes, four of the 21 children exposed to topiramate had a formal diagnosis of ASD. It is of note that, had all invited families participated in the study, the rate of ASD observed would remain above typical UK estimates [26] at 3.7%, suggesting that opt-in bias cannot fully explain this finding. Bjørk and colleagues [9] recently demonstrated in a large population sample an association between topiramate and ASD diagnoses and observed that doses over 100 mg/d carried the highest risk. All our four cases were above this dose threshold. Further, gestational birth weight was at or below the 25th centile. Whilst another ASM valproate is associated with an increased ASD risk [27][28][29], research is needed to examine whether the topiramate exposure's association with ASD diagnoses is direct or mediated through its documented impact on foetal growth. Further, the social skills of the children exposed to topiramate were, on average, nine points lower. It may be that challenges in aspects of social functioning for children exposed to topiramate fall on a continuum, with other children demonstrating below diagnostic levels of social difficulty. Further work should investigate this as a priority so that families are provided with evidence-based risk counselling and to ensure the risks to children of mothers with epilepsy are reduced as far as possible.

Strengths and limitations
Strengths included the blinded administration of a standardised measure with proven ability to detect functional deficits associated with teratogenic exposures [18,19], undertaken by a small number of outcome assessors (n = 2). Double scoring of outcome measures, data entry checks and the prospective collection of pregnancy and exposure information based on individual medical records further increased the reliability of our findings. Although, after relevant exclusions, our exposed cohort was limited to 21 children, the sample achieved was comparable to existing topiramate research [30]. The significant findings with large effect sizes indicated adequate power. Furthermore, the capacity to exclude children with conditions associated with altered neurodevelopmental outcomes, ensured a purer investigation the impact of topiramate exposure. Attention to key covariates, including maternal, paternal and child variables, was another factor that strengthened the methodological quality of the study.
The main weakness of the study was the poor recruitment rate which may have lead to a sample more skewed towards children with poorer neurodevelopmental functioning. This could have contributed to the large effect sizes observed; mothers with concerns about their child's functioning might have been more eager to participate. A further limitation was the absence of an equally sized control group. Whilst this was planned in the original study protocol, the COVID-19 pandemic led to a halt on recruitment and main analyses were limited to contrasts against normative data, limiting the strength of our comparisons due to possible confounding via baseline differences between groups. Although potential covariates were assessed and associations with adaptive behaviour outcomes were carried out, we were not able to adjust for all baseline characteristics. We did not have information about maternal marijuana use or use of other narcotics. Furthermore, planned regression analyses were not sufficiently powered to input several covariates at once. It is possible that our findings could have been confounded by multiple covarying factors that affected adaptive behaviour outcomes in a cumulative fashion. Further limitations were the use of a parental report measure where reporting mothers were unblinded to their child's exposure status which could influence responses.

Implications and future directions
Due to the limitations discussed above, the current findings are preliminary and require further replication and study in order to be fully understood. Decades following its approval in 1995, the neurodevelopmental trajectory of children exposed to topiramate in the womb is far from understood as demonstrated recently [9]. Our findings demonstrate a need for routine and systematic study of newly approved ASMs and better systems for pharmacovigilance [31,32]. If lessons are to be learned from previous failures to act upon harm evidence regarding valproate exposure, then this must proceed as a matter of urgency.

Conclusions
This study presents important preliminary results regarding poorer adaptive behaviour outcomes and higher incidence of ASD diagnoses for children exposed to topiramate in-utero. Alongside evidence of a doseresponse relationship, these findings suggest that topiramate could be associated with altered neurodevelopmental trajectories and highlight the need for further investigations. Future research should be undertaken as a matter of urgency to elucidate whether children exposed to topiramate are at an increased risk of poorer neurodevelopmental outcomes.

Table 5
Summary of included topiramate-exposed children with diagnosed ASD.