Cognitive Outcomes at Age 3-Years-Old in Children in the Maternal Outcomes and Neurodevelopmental Effects of Antiepileptic Drugs Study; a Prospective Observational Cohort Investigation

Summary

Background

Adverse neurodevelopmental effects of fetal exposure for most antiseizure medications remain uncertain. This study examines the relationship of fetal exposures to commonly used antiseizure medications on age 3-year-old neuropsychological outcomes.

Methods

This prospective, observational, multi-center, cohort investigation examined pregnancy outcomes in women with and without epilepsy enrolled during pregnancy. The primary outcome for children at 3-years-old was a blindly-assessed Verbal Index score calculated by averaging the Differential Ability Scales-II Naming Vocabulary and Verbal Comprehension subtests, Preschool Language Scale-5 Expressive Communication and Auditory Comprehension subscales, and Peabody Picture Vocabulary Test-4. Children of women with and without epilepsy were compared, and the associations of medication exposures to outcomes in exposed children were assessed.

Findings

Age 3 Verbal Index scores did not differ for children of women with epilepsy (n=284, adjusted LS Mean (95% CI)=102.7 (101.4, 103.9)) vs without epilepsy (n=87, 102.3 (99.8, 104.7)). Significant factors included maternal IQ, education, post-birth anxiety, enrollment gestational age, child sex and ethnicity. For Verbal Index scores, antiseizure medication exposure effects were not seen for maximum 3^rd-trimester blood concentrations (n=258, adjusted parameter estimate (95% CI)= −2.9 (−6.7, 1.0)). However, in secondary analyses, exposure-dependent effects were present on multiple measures, which varied by medication.

Interpretation

This prospective investigation of a large group of children with fetal exposure to newer antiseizure medications found no difference in neurodevelopmental outcomes compared to unexposed children. However, some exposure-dependent antiseizure medication effects were seen in secondary analyses. The adverse effects of maternal anxiety emphasize the importance of screening mothers during pregnancy/postpartum and implementing interventions. Additional studies are needed to clarify exposure-dependent effects.

Introduction

Our investigative group previously reported cognitive outcomes for children with fetal exposure to antiseizure medications (ASMs) commonly prescribed at that time, demonstrating impaired cognitive function in children exposed to fetal valproate compared to three other ASM monotherapies (i.e., carbamazepine, lamotrigine, phenytoin),^1–2 which contributed to FDA warning labels for valproate use in pregnancy.^3–4 Subsequently, patterns of ASM use changed, but the risks for many ASMs remain uncertain even though ASMs are among the most commonly prescribed teratogenic agents to women of childbearing potential.⁵ In this report, we describe 3-year-old cognitive outcomes for children from a new cohort with fetal exposure to ASMs commonly used now during pregnancy in the USA.

Methods

Study design and participants

The Maternal Outcomes and Neurodevelopmental Effects of Antiepileptic Drug (MONEAD) study is a prospective, observational, cohort investigation in the United States enrolling pregnant women with epilepsy and a comparison group of pregnant women without epilepsy at 20 specialty epilepsy centers (see supplementary materials). The MONEAD study evaluates both maternal and child outcomes, and the primary neurodevelopmental aims are to compare outcomes in children of women with epilepsy vs children of women without epilepsy, and to determine if there is an ASM fetal exposure-dependent relationship to cognitive outcomes in children of women with epilepsy. The MONEAD study is the first investigation of neurodevelopmental outcomes to assess ASM blood concentrations to adjust for known clearance changes in pregnancy, which might alter dose-dependent associations. This publication presents the 3-year-old cognitive outcomes.

Institutional review boards approved the study at each clinical center, and informed consent was obtained from all adult subjects. Pregnant women were recruited from epilepsy clinics and referrals from obstetricians and other physicians, as well as self-referral. Enrollment occurred December 2012 through January 2016. Inclusion criteria for pregnant women included ages 14–45 years and ≤20 weeks gestational age. For pregnant women without epilepsy, enrollment criteria for education level and race were adjusted during the study to maintain relative similarity on demographic factors to pregnant women with epilepsy. Exclusion criteria included expected IQ <70, other major medical illness including progressive cerebral disease, and history of psychogenic non-epileptic spells. For women with epilepsy, switching ASMs in current pregnancy prior to enrollment was an exclusion. Women with epilepsy were recruited regardless of ASM(s). Fathers of children and maternal relatives of pregnant women were enrolled. Data on seizures and ASMs were collected using a daily electronic diary and verified at study visits and with medical records.

Outcomes

All neuropsychological evaluations were administered and scored by assessors blinded to ASM exposures. Training and monitoring of the neuropsychological evaluations were conducted to assure quality and consistency across sites. The primary outcome for children at 3-years-old was a Verbal Index score calculated by averaging Differential Ability Scales-II (DAS-II) Naming Vocabulary and Verbal Comprehension subtests,⁶ Preschool Language Scale-5 (PLS-5) Expressive Communication and Auditory Comprehension subscales,⁷ and Peabody Picture Vocabulary Test-4 (PPVT-4).⁸ All assessments were converted to same scale (e.g., standard scores with mean=100; SD=15) prior to averaging. This measure combining standardized verbal tests was chosen because prior studies have suggested that verbal abilities might be more susceptible to fetal ASM exposure.²

Additional secondary cognitive outcomes included standard scores from DAS-II General Conceptual Ability, DAS-II Verbal Cluster, DAS-II Non-Verbal Cluster, Bayley Scales of Infant and Toddler Development-3 (BSID-III)⁹ Motor Composite, PLS-5 Total Language, PLS-5 Auditory Comprehension, PLS-5 Expressive Communication, PPVT-4, and Developmental Test of Visual-Motor Integration-6 (DTVMI-6)¹⁰, and T-scores (mean=50; SD=10) from Behavior Rating Inventory of Executive Function-Preschool (BRIEF-P)¹¹: Global Executive Composite Score, Inhibitory Self Control Index, Flexibility Index, and Emergent Metacognition Index.

Statistical analyses

Primary neurodevelopmental outcomes included: 1) Verbal Index scores in children of women with vs without epilepsy, and 2) relationship of 3^rd-trimester ASM concentrations to Verbal Index scores in children of women with epilepsy. Both analyses controlled for potential confounding factors and additional covariates associated with the outcome. No adjustment of type-1 error rate for multiple comparisons was applied since these primary outcomes address different research questions. Power calculations for each analysis are provided in supplementary materials. Based on animal models of fetal ASM and alcohol effects on the immature brain,¹² our outcome for primary analysis #2 was maximum observed 3^rd-trimester exposure. Maximum ratio ASM blood concentration and ratio defined daily dosages were calculated using similar methods as our age 2 analyses,¹³ but were computed only for mothers in whom blood levels were measured for their ASM(s). Additional derivation details are provided in supplementary materials.

Risk factors assessed in secondary analyses and as potential covariates in the adjusted models included breastfeeding, periconceptional folate use and dose (none, >0–0.4mg, >0.4–1.0mg, >1.0–4.0mg, >4.0mg), and average scores during pregnancy and/or post-birth through the child’s 3-year-old visit for anxiety (Beck Anxiety Inventory (BAI)¹⁴), depression (Beck Depression Inventory-2 (BDI-2)¹⁵), perceived stress (Perceived Stress Scale-14 (PSS)¹⁶), and sleep quality (Pittsburgh Sleep Quality Index (PSQI)¹⁷). Additional risk factors for children of women with epilepsy included 3^rd-trimester ASM dose, ASM group (monotherapy, polytherapy) and ASM category (lamotrigine monotherapy, levetiracetam monotherapy, other monotherapy, lamotrigine+levetiracetam polytherapy, other polytherapy), and specific ASM (expanding ASM category to include oxcarbazepine, carbamazepine, zonisamide, topiramate monotherapy; not included as a potential covariate due to number of levels).

Maternal IQ¹⁸ was included in all adjusted models a priori. Additional potential confounders or covariates independently associated with Verbal Index scores assessed for inclusion in the adjusted models included: maternal age, education level, enrollment employment status, household income, planned/welcomed pregnancy, major depressive episode during pregnancy, smoking, alcohol, illicit substance use, weeks gestational age at enrollment and delivery, child’s race, sex, ethnicity (Hispanic or Latino vs Non-Hispanic), birthweight, small for gestational age, major congenital malformations, and weight, height, and head circumference at time of assessment. In children of women with epilepsy additional potential covariates included mother’s epilepsy type, number of seizures, and ≥5 convulsions during pregnancy.¹⁹ In our dataset, ASM exposure was not associated with birth-related variables (major congenital malformations, birthweight, small for gestational age, weeks gestational age at delivery) or 3-year-old physical characteristics (weight, height, head circumference) when evaluated individually, so these variables would not likely be mediators. Since they may be associated with Verbal Index scores, they were assessed as potential covariates, however none were selected into the final adjusted models.

Potential covariates and risk factors were summarized and differences between mother’s study groups were assessed using the chi-square or Fisher’s exact test for categorical variables and independent t-test or Wilcoxon rank-sum test for continuous variables.

Primary Analyses.

Analyses with Verbal Index score used an imputation population. Markov chain Monte Carlo methods were employed to impute missing Verbal Index score at 3-years-old from available 2-years-old BSID-III Language Composite Score, risk factors, and variables related to outcome or likelihood of missing data. A full list of covariates included in the imputation model are provided supplementary materials. The imputation procedure generated 50 imputed datasets and fit regression models to each. Combined estimates for the least square means across the imputed datasets were obtained by applying Rubin’s Rule. Unadjusted and adjusted linear regression models were used to compare Verbal Index scores (outcome) and mother’s study group or maximum observed 3^rd-trimester ASM concentrations (main predictor), adjusting for covariates. Linear regression model assumptions were assessed separately for each model using scatter plots and Q-Q residual plots. Model estimates incorporated imputation uncertainty. Mother’s IQ and study group for primary analysis #1, and mother’s IQ and ASM concentration for #2 were included in the model a priori. Additional covariates for adjusted models were selected using a stepwise selection algorithm of the completers population (children with non-missing 3-year-old Verbal Index score) with Verbal Index score as the outcome. At each step, the model with the lowest Akaike information criterion was selected. The significance levels for covariate entry was set to p=0.10 and to remain in the model was set to p=0.15. Variables with >5% missing were excluded from consideration.

Secondary Analyses.

Secondary analyses for risk factors for Verbal Index score used the multiple imputation population and adjusted for the same variables as the primary analyses. Mother’s sleep quality scores were missing for 22% of children, and analyses for this factor were limited to only the completers population. To assess if associations between maximum 3^rd-trimester ratio ASM concentration or dose and Verbal Index scores varied across different ASMs, interaction models were run for 3^rd-trimester ASM group and ASM category.

Analyses comparing children of women with vs without epilepsy were re-run excluding women with epilepsy who were not on ASM. In addition, analyses comparing children of women with vs without epilepsy and examining women with ASM concentration and dose were re-run for all 14 cognitive measures using unadjusted and adjusted linear regression for completers, adjusting for the same variables used in the primary analyses.

The potential impact of missing data was assessed via a sensitivity analysis to compare risk factors and potential covariates between the analysis population and those excluded and with a model imputing Verbal Index scores for all children born into the study. Additional sensitivity analyses for the primary outcomes were run to assess the impact of twins, out of window visits, children with major congenital malformations or genetic defects linked to developmental delay, children with any events that may have impacted neurological development occurring prior to age 3 assessments. Primary analysis #1 was re-run excluding children of women with epilepsy not on ASM during pregnancy. Imbalances in baseline covariates were assessed using propensity scores, and primary analyses were visualized using forest plots by propensity score stratum. Additional details for sensitivity analyses and references for statistical methods are included in the supplemental materials.

Role of funding source.

The funder of the study had no role in study design, data collection, data analysis, data interpretation, writing of the report, or decision to publish.

Results

Figure S1 depicts enrollment and exclusions. There were 345 children born to women with epilepsy and 106 children to women without epilepsy. Primary analysis #1 included 284 children of women with epilepsy and 87 children of women without epilepsy. Mother’s 3^rd-trimester ASM regimen for children of women with epilepsy included 73.9% monotherapy, 22.2% polytherapy, 3.5% no ASMs throughout pregnancy, and 0.4% with no 3^rd-trimester exposure due to delivery in the 2^nd trimester. The most common ASM monotherapy regimens were lamotrigine (n=90, 42.9% of all monotherapies (n=210)) and levetiracetam (n=74, 35.2%). Additional monotherapies included: oxcarbazepine (n=14, 6.7%), carbamazepine (n=12, 5.7%), zonisamide (n=11, 5.2%), topiramate (n=3, 1.4%) and remaining monotherapies each less than 1% of all monotherapies. The most common ASM polytherapy combination was lamotrigine + levetiracetam (n=26, 41.3% of polytherapies (n=63). Mother’s 3^rd-trimester ASM regimen for each analysis as well as detailed information on the multiple other monotherapies and polytherapies are summarized in Table S1. Primary analysis #2 included 258 children of women with epilepsy with 3^rd-trimester ASM blood concentrations.

Demographic, baseline characteristics, and risk factors for children of women with vs without epilepsy for primary analysis #1 are in Table 1 and for children of women with epilepsy in primary analysis #2 in Tables S2 and S3. The median (range) maximum 3^rd trimester blood concentrations (ug/mL) for the two main ASMs were 4.3 (1.5–14.3) for lamotrigine and 20.7 (2.2–61.8) for levetiracetam; and their median (range) maximum 3^rd trimester doses (mg/day) were 600 (150–1850) for lamotrigine and 2000 (375–700) for levetiracetam. For summaries of 3^rd trimester ASM concentrations and dosages for each ASM see Tables S4 & S5, respectively.

Table 1.

Demographic, baseline characteristics, and risk factors of children (n=371) of women with epilepsy (WWE) vs women without epilepsy (WWoE) for primary analysis #1 for categorical and continuous variables.

Categorical Variables
		Children of WWE, N = 284		Children of WWoE, N = 87
		n (%)		n (%)		P-Valuea
Child’s Sex - Male		134 (47)		47 (54)		0.26
Child’s Race						0.0045
White		231 (81)		58 (67)
Black or African American		14 (5)		12 (14)
Other/Unknown		39 (14)		17 (20)
Child’s Ethnicity - Hispanic or Latino		57 (20)		21 (24)		0.42
Twin		20 (7)		6 (7)		0.96
Small for Gestation Age		17 (6)		9 (10)		0.16
Major Congenital Malformations		16 (6)		1 (1)		0.14
Any Breastfeeding		215 (76)		77 (89)		0.0107
Periconceptional Folate Use		249 (88)		57 (66)		<.0001
Periconceptional Folate Dose						<.0001
0 mg		35 (12)		30 (34)
> 0 – 0.4 mg		25 (9)		13 (15)
> 0.4 – 1 mg		53 (19)		31 (36)
> 1 – 4 mg		150 (53)		13 (15)
> 4 mg		21 (7)		0 (0)
Pregnancy Planned/Welcomed						0.83
Planned		200 (70)		59 (68)
Unplanned/Welcome		76 (27)		26 (30)
Unplanned/Unwelcome		8 (3)		2 (2)
Depression During Pregnancy b		11 (4)		3 (3)		1.00
Mother’s Education						0.29
College Degree (Advanced)		71 (25)		29 (33)
College Degree (Not Advanced)		131 (46)		34 (39)
No College Degree		82 (29)		24 (28)
Mother’s Employment						0.42
Employed full-time		157 (55)		49 (56)
Employed part-time		42 (15)		17 (20)
Unemployedc		85 (30)		21 (24)
Household Income						0.82
Subject desires not to answer		13 (5)		6 (7)
$24,999 or less		51 (18)		14 (16)
$25,000–49,999		31 (11)		8 (9)
$50,000–74,999		44 (15)		10 (11)
$75,000–99,999		44 (15)		16 (18)
$100,000 and up		101 (36)		33 (38)
Smoking d		18 (6)		5 (6)		0.84
Alcohol Use d		65 (23)		29 (33)		0.0500
Illicit Substance Use d		10 (4)		4 (5)		0.75
Continuous Variables
	Children of WWE, N = 284		Children of WWoE, N = 87
	N	Mean (SD)	N		Mean (SD)	P-Valuee
WGA at Birth	284	38.4 (2.1)	87		38.7 (1.5)	0.26
WGA at Enrollment	284	13.6 (4.6)	87		15.4 (3.9)	0.0007
Birthweight (kg)	284	3.20 (0.60)	87		3.27 (0.59)	0.35
Mother’s Age at Enrollment	284	30.7 (5.1)	87		29.9 (5.0)	0.17
Mother’s IQ	284	97.7 (12.9)	87		104.7 (12.9)	<.0001
Father’s IQ f	195	104.1 (13.4)	50		104.4 (14.0)	0.90
Maternal Relative’s IQ f	107	99.9 (16.3)	29		102.7 (11.2)	0.38
Pittsburgh Sleep Quality Index f,g
Pregnancy+Postpartum	221	5.71 (2.60)	67		5.40 (2.28)	0.39
Pregnancy	260	5.69 (3.03)	78		4.83 (2.48)	0.0215
Postpartum	224	5.71 (2.71)	67		5.84 (2.69)	0.72
Child’s 3YO Weight (Z-Score) h,i	279	0.26 (1.07)	83		0.27 (0.86)	0.95
Child’s 3YO Height (Z-Score) h,i	282	0.32 (1.13)	86		0.43 (0.94)	0.41
Child’s 3YO Head Circumference h,j	284	4.4 (1.8)	87		4.6 (1.8)	0.30
	N	Median (IQR)	N		Median (IQR)	P-Valuee
Beck Depression Inventory-2 k
Enrollment - 3YO Visit	284	5.3 (2.7, 7.9)	87		4.4 (2.2, 6.7)	0.09
Pregnancy	284	5.7 (3.3, 9.0)	87		5.0 (3.0, 8.3)	0.22
Post-Birth	284	4.5 (2.0, 7.9)	87		3.7 (1.3, 6.4)	0.0408
Beck Anxiety Inventory k
Enrollment - 3YO Visit	284	3.6 (2.1, 7.1)	87		2.5 (1.2, 4.7)	0.0006
Pregnancy	284	4.7 (2.3, 8.0)	87		3.5 (2.0, 7.0)	0.09
Post-Birth	284	3.0 (1.3, 5.7)	87		1.7 (0.4, 3.3)	<.0001
Perceived Stress Scale-14 k
Enrollment - 3YO Visit	284	17.8 (13.0, 22.1)	87		16.9 (12.5, 20.6)	0.23
Pregnancy	284	17.7 (13.0, 22.4)	87		16.0 (12.3, 21.0)	0.12
Post-Birth	284	18.2 (13.1, 22.3)	87		16.8 (12.8, 21.2)	0.31

Notes: WWE = women with epilepsy; WWoE = women without epilepsy; WGA = weeks gestational age. SD = Standard deviation; IQR = Interquartile range. Two pairs of children of WWE and WWoE had mothers who were sisters and one pair of children of WWE had mothers who were twin sisters.

^aP-Values from chi-square test except for Major Congenital Malformations, Depression During Pregnancy, and Illicit Drug Use which used Fisher’s exact test.

^bDepression based on SCID Mood Module at any depression assessment during pregnancy.

^cUnemployed includes stay at home parent without outside job, unemployed due to disability, full-time student and looking for work.

^dSelf-reported smoking, alcohol use, or illicit substance use (including marijuana) at any time during pregnancy.

^eP-Values from independent t-test (WGA at Birth, WGA at Enrollment, Birthweight, Age, IQ Scores, Sleep Scores, Weight, Height, Head Circumference) or Wilcoxon Rank-Sum Test for non-normally distributed data (assessed by Q-Q Plots) (Depression, Anxiety, and Stress Scores).

^fFather’s IQ, Maternal Relative’s IQ, and sleep scores not assessed for inclusion in imputation model due to extensive missing data.

^gWeighted monthly average of PSQI assessments completed during time period. At most one PSQI score per day was used in the calculation. Pregnancy: Enrollment through the day before delivery; Postpartum: 6 weeks through 9 months post-delivery; Pregnancy+Postpartum: Assessments completed during either period (only women with scores in both periods included).

^hMissing 3YO measurement replaced with 2YO measurement.

ⁱWeight-for-age and height-for-age z-scores derived using the least square means method based on CDC growth charts.

^jHead circumference-for age ordinal categories derived using USA head circumference growth chart percentiles. 1= 03-rd, 2= >3rd – 10th, 3= >10th-25th, 4= >25th-50th, 5= >50th-75th, 6= >75th-90th, 7= >90th-97th, 8= >97^th (see Supplement for reference by Rollins et al, 2010).

^kAverage of all assessments completed during time period. Enrollment - 3YO Visit: enrollment through 3YO visit; Pregnancy: enrollment through day of delivery; Post-Birth: post-delivery through 3YO visit.

Primary Analysis of Children of Women With vs Without Epilepsy:

No significant differences were seen in Verbal Index scores between children of women with vs without epilepsy; unadjusted LS Mean (95% CI)=102.4 (100.8, 103.9) vs without epilepsy (103.4 (100.6, 106.2)); adjusted mean difference (95% CI): 0.4 (−2.4, 3.2) (Figure 1). Significant factors in the full model included maternal IQ, education level, post-birth maternal anxiety, weeks gestational age at enrollment, and child’s sex/ethnicity (Table 2).

Figure 1.

Plots of child Verbal Index score and mother IQ for women with epilepsy (WWE) and women without epilepsy (WWoE) groups. Imputation population (children with imputed scores excluded from figure), n = 334.

Table 2.

Full model summaries for Age 3 Verbal Index score in children of women with epilepsy (WWE) vs women without epilepsy (WWoE) and in children of women with epilepsy as a function of 3^rd-trimester antiseizure medication (ASM) blood concentrations.

Model Parameter	Parameter Estimate (95% CI)	P-value
Children of WWE and WWoE, Imputation Analysis (N = 371)
Mother’s Study Group: WWE vs WWoE	0.4 (−2.4, 3.2)	0.77
Mother’s IQ	0.3 (0.2, 0.4)	<.0001
Child’s Sex: Male vs Female	−4.9 (−7.1, −2.6)	<.0001
Child’s Ethnicity: Hispanic or Latino vs Non-Hispanic	−5.5 (−8.5, −2.6)	0.0003
Education Level		0.0005
College Degree (Advanced)	Ref	-
College Degree (Not Advanced)	−1.7 (−4.6, 1.1)	0.22
No College Degree	−7.0 (−10.7, −3.3)	0.0002
Post-Birth Average Beck Anxiety Inventory score	−0.4 (−0.7, −0.2)	0.0014
Weeks Gestational Age at Enrollment	−0.3 (−0.5, 0.0)	0.0462
Mother’s Age at Enrollment	0.2 (0.0, 0.5)	0.06
Children of WWE with 3 rd -Trimester Blood Concentrations, Imputation Analysis (N = 258)
3rd-Trimester Max Observed Ratio ASM Blood Concentrationsa	−2.9 (−6.7, 1.0)	0.15
Mother’s IQ	0.1 (0.0, 0.3)	0.0235
Child’s Sex: Male vs Female	−3.7 (−6.5, −1.0)	0.0072
Post-Birth Average Beck Anxiety Inventory score	−0.4 (−0.7, −0.1)	0.0071
Education Level		0.0098
College Degree (Advanced)	Ref	-
College Degree (Not Advanced)	−2.5 (−6.0, 0.9)	0.15
No College Degree	−8.0 (−13.1, −2.9)	0.0022
Child’s Ethnicity: Hispanic or Latino vs Non-Hispanic	−4.8 (−8.6, −0.9)	0.0156
Weeks Gestational Age at Enrollment	−0.3 (−0.6, 0.0)	0.0471
Household Income		0.11
$24,999 or less	Ref	-
$25,000–49,999	2.7 (−3.1, 8.5)	0.37
$50,000–74,999	5.0 (−0.8, 10.7)	0.09
$75,000–99,999	6.2 (0.1, 12.2)	0.044
$100,000 and up	5.9 (0.3, 11.6)	0.0379
Subject desires not to answer	9.9 (2.5, 17.3)	0.0089

Variables selected into model using a stepwise selection algorithm with the completers population. Mother’s study group and IQ included in the first model a priori. Mother’s 3^rd-trimester ratio ASM blood concentration and IQ included in the second model a priori. At each step, the model with the lowest Akaike information criterion was selected. Significance level for covariate entry was set to p = 0.10 and significance level to remain in the model was set to p = 0.15.

^aRatio maximum observed ASM concentration, calculated as ratio upper limit for therapeutic range. For mothers on polytherapy, ratio ASM concentration calculated by summing the ratio ASM concentration for each ASM. Maximum observed value recorded during 3^rd-trimester, including day of deliver

Primary Analysis of Children of Women with Epilepsy for 3^rd-Trimester ASM Exposure:

In children of women with epilepsy, a negative association between Verbal Index score was and ratio maximum observed 3^rd-trimester ASM blood concentrations were observed in unadjusted analysis (parameter estimate (95% CI)= −7.5 (−12.1, −2.9), p=0.0013), but was not significant in adjusted analysis (−2.9 (−6.7, 1.0), p=0.15) (Table 2, Figure S2). Significant factors in the full model were the same as primary analysis #1.

Secondary Analyses of Children of Women With vs Without Epilepsy.

Completer analyses for the 14 cognitive measures were not statistically significant (Table S6). Measures of maternal depression, anxiety and perceived stress demonstrated an inverse relationship to Verbal Index scores in univariate analyses, and adverse effects were seen for maternal depression and anxiety in adjusted models (Tables S7). However, only post-birth anxiety was selected into the full model (Table 2, Figure 2). Note that these findings did not differ across children of women with vs without epilepsy although women with epilepsy had more symptoms of anxiety and depression in the post-birth period (Table 1).²⁰ The relation of breastfeeding, periconceptional folate use, dose, and maternal sleep quality to Verbal Index scores were significant in univariate but not adjusted models (Table S8, Figure S3, Table S9).

Figure 2.

Age 3 Verbal Index Score vs. maternal post-birth average Beck Anxiety Inventory (BAI) score in children of women with epilepsy (WWE) and women without epilepsy (WWoE). Imputation population (children with imputed scores excluded from figure), n = 334.

Secondary Analyses for ASM Exposure in Children of Women with Epilepsy.

Analyses of the relationship of other cognitive variables to the ratio maximum observed 3^rd-trimester ASM blood concentrations were significant for multiple measures in unadjusted analysis but were significant in adjusted analysis only for PPVT-4 (Table S10). Associations between ASM group, ASM category and Specific ASM with Verbal Index scores were significant in univariate but not adjusted models (Table S11, Figure 3). In adjusted, imputation analyses of 3^rd-trimester ASM concentrations stratified by ASM categories, Verbal Index scores were adversely associated with higher levetiracetam concentrations (parameter estimate (95% CI)= −9.0 (−17.2, −0.7), p=0.0330) (Figure 4). Significant adverse associations for levetiracetam concentration were seen in adjusted analyses for 5 (DAS-II General Conceptual Ability Standard score, DAS-II Nonverbal Cluster, PLS-5 Auditory Comprehension Standard Score, PPVT-4 Standard Score, and DTVMI-6 Standard scores population) of the other 13 cognitive measures among the completers, although not significant for Verbal Index Score (adjusted parameter estimate (95% CIs)= −8.3 (−17.0, 0.3); p=0.06) (Table S12).

Figure 3.

Age 3 Verbal Index score by mother’s ASM group, ASM category and specific ASM, children of women with epilepsy on ASM with 3rd trimester blood levels. Box plot whiskers extend to 1.5 times the IQR, points outside the whiskers are more extreme than 1.5 times the IQR. Imputation population (children with imputed scores and those with missing covariates excluded from figure) (n = 235).

Figure 4.

Age 3 Verbal Index Score vs 3^rd-trimester maximum observed ratio ASM blood concentrations by mother’s ASM category in children of women with epilepsy. Imputation population (children with imputed scores excluded from figure), n = 235.

Analyses of maximum 3^rd-trimester dosages across all ASMs with a larger sample size (n=260) revealed an inverse association with Verbal Index scores (adjusted parameter estimate (95% CI)= −0.97 (−1.77, −0.16); p=0.0192) as well as across all pregnancy doses (Table S13; Figure S4). When dose analyses were restricted to children in the ASM concentration analyses (i.e., those with 3^rd-trimester ASM concentrations), 3^rd-trimester dose was associated with lower Verbal Index score in unadjusted analyses (Parameter Estimate (95% CI)= −1.8 (−3.0, −0.7), p=0.0020), but was not in adjusted analyses (−0.7 (−1.7, 0.3), p=0.15), similar to findings for ASM concentrations on the same cohort.

In adjusted analyses of the completer population across 14 cognitive measures by maximum 3^rd-trimester dosages across all ASMs, significant adverse effects were seen for Verbal Index score, DAS-II General Conceptual Ability, PLS-5 Auditory Comprehension and PPVT-4 Standard scores, and BRIEF-P Emergent Metacognition Index T-score (Table S14). In adjusted analyses of the completer population for 14 cognitive measures by 3^rd-trimester ASM dosages by ASM categories, significant adverse associations were seen for levetiracetam on DAS-II General Conceptual Ability Standard score, DAS-II Nonverbal Cluster, DTVMI-6 Standard scores, and BRIEF-P Emergent Metacognition Index T-score (Table S15).

Sensitivity Analyses.

Comparisons of included and excluded children are given in Tables S16–S21 and imputation analyses including all children are given in Tables S22–S23. Other sensitivity analyses are in Tables S24–S37 and Figures S5–S6. All were consistent with primary analyses.

Discussion

The three major findings in our present analyses include: 1) cognitive outcomes in 3-year-old children of women with epilepsy did not differ from children of women without epilepsy; 2) signals for fetal exposure-dependent effects were seen for multiple measures, especially for levetiracetam; and 3) post-birth anxiety in mothers with and without epilepsy adversely affected child cognitive outcomes.

Some prior studies have shown differences between children of women with epilepsy vs women without epilepsy. The main ASMs in our study lamotrigine and levetiracetam have not shown adverse effects. All human studies of fetal ASM effects are observational, and thus require replication to be certain of findings. The unique aspect of our investigation includes the prospective collection of a large database on many aspects of cognition and potential confounding factors, which adds credence to the findings.

In our study, significant factors associated with better verbal abilities were higher maternal IQ and education, less maternal post-birth anxiety, child female or non-Hispanic, and marginally lower weeks gestational age at enrollment. Most of these factors have been noted previously as influencing neurodevelopment. Women with epilepsy had more symptoms of anxiety and depression, but this did not lead to differences in cognitive outcomes in their children. However, the impact across all children of maternal anxiety and mood on neurodevelopment²¹ suggests that routine screening and interventions should be conducted.

Although there was no association of Verbal Index scores to 3^rd-trimester ASM blood concentrations across all ASMs, secondary analyses revealed negative effects on one measure (PPVT-4) across all ASMs, and for levetiracetam on 5 of the 14 cognitive measures. Higher 3^rd-trimester dosages were associated with worse cognitive performance across all ASMs for 5 of 14 cognitive measures including Verbal Index score, and for levetiracetam on 4 of the 14 cognitive measures. Thus, exposure-dependent associations were more apparent for levetiracetam. ASM concentrations were within therapeutic ranges for the vast majority of women, so exposure-dependent effects cannot be explained by excessively high exposures. However, these secondary analyses need to be interpreted with caution, especially given multiple comparisons. Thus, these findings require replication in a separate cohort.

Nevertheless, findings of ASM exposure-dependent effects should not be surprising. Teratogens are known to act in an exposure-dependent manner upon susceptible genotype. All ASMs are potential teratogens, and exposure-dependent effects have been seen even for lamotrigine for malformations.²² Thus, decision-making for ASMs in pregnancy needs to balance benefits of maintaining an individual target concentration to lower risk of seizures²³ against unnecessary ASM over-exposure to the developing fetus.

Levetiracetam is one of the first-line ASMs for women of childbearing age given the low rates of major congenital malformations, neonatal complications, and adverse neurodevelopmental outcomes.^{22, 24} In the present study, it is important to remember that children exposed to levetiracetam did not differ overall from children of mothers without epilepsy. So, levetiracetam can continue to be used safely in women of childbearing potential with consideration of the level of dosing during pregnancy like other ASMs.

The present results are similar to our 2-years-old results in the same cohort.¹³ However, cognitive assessment at 2-years-old is less precise and predictive of school performance and ultimate adult cognitive abilities than 3-years-old neuropsychological assessments, which are more detailed across additional domains. Although the overall results are similar, the exposure-dependent signal for all ASMs and for levetiracetam is stronger in the present study.

The lack of effects on cognitive outcomes from breastfeeding while taking ASM is similar to our findings at 3-years-old in our prior cohort¹ and to findings by a Norwegian study,²⁵ although we found beneficial effects at age 6 in our prior cohort.²⁶ In our present cohort, we have demonstrated that ASM blood levels are generally low in children who are breastfeeding.²⁷ Given the multiple benefits of breastfeeding to both the mother and child,²⁸ women with epilepsy should be encouraged to breastfeed.

We did not find a significant effect of periconceptual folate on cognitive outcomes as has been previously reported.^2,29 However, in our prior cohort, folate’s effect was marginal at age 3,¹ but was robust at 6-years-old.²

Strengths of our study include the prospective design with detailed observational data to control for multiple potentially confounding factors and additional covariates associated with the outcome combined with formal objective assessment of cognitive abilities in children and their mothers using standardized measures. The MONEAD study is the first investigation to assess the association of ASM exposure with child outcomes using ASM blood concentrations.

There are limitations of our study including the lack of ASM randomization due to ethical and practical issues in investigations of pregnant women with epilepsy, which may introduce selection bias. Like all observational studies, there may be residual confounding including measurement error and additional unmeasured confounding due to factors such as genetics, environmental exposure, and missing data on collected covariates such as sleep quality, thus requiring replication across studies. Additionally, the use of statistically-based variable selection may not have identified the full set of confounders. Neuropsychological assessments at 3-years-old may not detect associations seen at older ages. Beyond lamotrigine and levetiracetam, sample sizes for many ASMs in our study were small, precluding individual ASM evaluations. The MONEAD study enrolled pregnant women with epilepsy irrespective of ASM, so the ASM distribution likely reflects current prescribing patterns at US epilepsy centers, but may not reflect ASM use in the general population. However, the two most common ASMs in MONEAD were the same as a recent large USA database study.³⁰ The use of ASMs for non-epilepsy indications has expanded and now compose the majority of ASM exposures in pregnancy. The teratogenic effects of most ASMs remain unknown,⁵ which is an important area for future research along with delineation of genotypic risk factors for ASM teratogenicity.

Contributors

KJM contributed to study design and conceptualisation, data collection, data interpretation, statistical analysis, figures and tables creation, literature search, writing, and critical approval of the final paper. MJC contributed to study design and conceptualization, data collection, data interpretation, writing and critical approval of the final paper. DWL contributed to the study design, data interpretation, and review of the study manuscript. AGM contributed to development of statistical methods, interpretation of results, and critical review of manuscript and supplement. CB contributed to development of statistical methods, literature review, data analysis and interpretation, figure and table creation, drafting and critical review of manuscript and supplement. CPR contributed to data interpretation, statistical analysis, figures and tables creation, literature search, writing, and critical approval of the final paper. AKB contributed to study design, data collection and analysis, interpretation, and critical approval of the final paper. PEV contributed to data collection and critical approval of the final paper. LAK contributed to data collection and critical approval of the final paper. EEG contributed to data collection and critical approval of the final paper. ERG contributed to data collection and critical approval of the final paper. JH contributed to data collection and critical approval of the final paper. JC contributed to acquisition, analysis, and interpretation of data and critical revision of the manuscript. MS contributed to acquisition, analysis, and interpretation of data and critical revision of the manuscript. JAF contributed to data collection, data interpretation, and critical review and approval of the final paper. SH contributed to data acquisition, review and approval of the final paper. AMP contributed to data collection and critical approval of the final paper. PBP contributed to study design and conceptualization, data collection, data interpretation, statistical analysis, and critical approval of the final paper. All site PIs verified the underlying local site data, which was double checked by KJM, PBP, AGM, CB, CPR, and other members of the database center. All authors had full access to all data in the study and agreed to submit for publication.

Data Sharing

Data collected for the study included deidentified individual participant data, a data dictionary defining each field in the set, related data sets, study protocol, statistical analysis plan, and informed consent form will be available upon request to qualified investigators when the age 6 year-old analyses have been published.