Effects of fetal antiepileptic drug exposure

Text-only Preview

Effects of fetal antiepileptic drug exposure
Outcomes at age 4.5 years
K.J. Meador, MD
G.A. Baker, PhD
Objective: To examine outcomes at age 4.5 years and compare to earlier ages in children with
N. Browning, PhD
fetal antiepileptic drug (AED) exposure.
M.J. Cohen, EdD
Methods: The NEAD Study is an ongoing prospective observational multicenter study, which en-
R.L. Bromley, PhD
rolled pregnant women with epilepsy on AED monotherapy (1999-2004) to determine if differen-
J. Clayton-Smith, MD
tial long-term neurodevelopmental effects exist across 4 commonly used AEDs (carbamazepine,
L.A. Kalayjian, MD
lamotrigine, phenytoin, or valproate). The primary outcome is IQ at 6 years of age. Planned analy-
A. Kanner, MD
ses were conducted using Bayley Scales of Infant Development (BSID at age 2) and Differential
J.D. Liporace, MD
Ability Scale (IQ at ages 3 and 4.5).
P.B. Pennell, MD
M. Privitera, MD
Results: Multivariate intent-to-treat (n
310) and completer (n
209) analyses of age 4.5 IQ
D.W. Loring, PhD
revealed significant effects for AED group. IQ for children exposed to valproate was lower than
For the NEAD Study
each other AED. Adjusted means (95% confidence intervals) were carbamazepine 106 (102-
109), lamotrigine 106 (102-109), phenytoin 105 (102-109), valproate 96 (91-100). IQ was
negatively associated with valproate dose, but not other AEDs. Maternal IQ correlated with child
IQ for children exposed to the other AEDs, but not valproate. Age 4.5 IQ correlated with age 2
BSID and age 3 IQ. Frequency of marked intellectual impairment diminished with age except for
Correspondence & reprint
requests to Dr. Meador:
valproate (10% with IQ
70 at 4.5 years). Verbal abilities were impaired for all 4 AED groups
[email protected]
compared to nonverbal skills.
Conclusions: Adverse cognitive effects of fetal valproate exposure persist to 4.5 years and are
related to performances at earlier ages. Verbal abilities may be impaired by commonly used
AEDs. Additional research is needed. Neurology(R) 2012;78:1207-1214

antiepileptic drug; BSID
Bayley Scales of Infant Development; CI
confidence interval; DAS
Differential Ability
Scales; NART
National Adult Reading Test; NEAD
Neurodevelopmental Effects of Antiepileptic Drugs; SES
nomic status; TONI-3
Test of Nonverbal Intelligence-third edition; WASI
Wechsler Abbreviated Scale of Intelligence.
Antiepileptic drugs (AEDs) are among the most common potentially teratogenic drugs taken by
women of childbearing potential. Animal studies have demonstrated that fetal AED exposure at
doses lower than those required for anatomic malformations can produce behavior deficits.1,2 Our
group initiated a prospective study to establish if such effects can occur in humans. We have
previously reported differential effects of fetal AED exposure on cognitive outcomes in children at
age 3 years.3 Based largely on the data from this report, the Food and Drug Administration recently
released a drug safety communication reporting an increased risk of impaired cognitive development
in children whose mothers took valproate during pregnancy.4 Since deficits at age 3 years might
change with maturation, we examined the effects of fetal AED exposure on cognitive functions at
age 4.5 years and compared these findings to outcomes at ages 2 and 3 years.
Supplemental data at

From Neurology & Pediatrics (K.J.M., D.W.L.), Emory University, Atlanta, GA; Walton Centre for Neurology & Neurosurgery (G.A.B.), University
of Liverpool, Merseyside, UK; EMMES Corp. (N.B.), Rockville, MD; Medical College of Georgia (M.J.C.), Augusta; Clinical Psychology (R.L.B.),
University of Manchester, Manchester; St. Mary's Hospital (J.C.-S.), Manchester, UK; University of Southern California (L.A.K.), Los Angeles; Rush
Supplemental Data
University Medical Center (A.K.), Chicago, IL; Neurology (J.D.L.), Riddle Health Care, Media, PA; Neurology (P.B.P.), Brigham and Women's
Hospital, Boston, MA; and Neurology (M.P.), University of Cincinnati, OH.
Coinvestigators are listed on the Neurology(R) Web site at www.neurology.org.
Study funding: Supported by the National Institutes of Health (NS038455 to K.M., NS050659 to N.B.) and the United Kingdom Epilepsy Research
Foundation (RB219738 to G.B.). The study was overseen by an NIH-appointed DSMB.
Disclosure: Author disclosures are provided at the end of the article.
Copyright (c) 2012 by AAN Enterprises, Inc.

METHODS Design. The Neurodevelopmental Effects of
tion by the following calculation: 100
(observed dose
Antiepileptic Drugs (NEAD) study is an ongoing prospective
mum dose)
range of doses (i.e., maximum
observational investigation, which enrolled pregnant women
Statistical analysis. The primary analysis was intent-to-treat
with epilepsy who were on AED monotherapy (i.e., carbamaz-
310 live births including 6 twin pairs). Secondary analysis
epine, lamotrigine, phenytoin, or valproate) from October 1999
was performed on children who completed testing at age 4.5
through February 2004 across 25 epilepsy centers in the United
years (n
209 including 6 twin pairs). Secondary analyses also
States and United Kingdom. Sample sizes were estimated to de-
tect a 0.5 SD in IQ outcome. These AEDs were the most fre-
included comparison to outcomes at ages 2 and 3 years, correla-
quently prescribed during pregnancy at our centers during
tions of age 4.5 IQ to standardized doses and to maternal IQ for
enrollment. No other AEDs were employed in sufficient num-
each AED, propensity score analyses, and comparison of verbal
bers to allow adequate sample sizes for analysis. Polytherapy was
and nonverbal performances at age 4.5 years. Analyses were per-
not included because of prior reports of poorer outcomes.5,6 A
formed at the NEAD Data and Statistical Center using SAS and R.
nonexposed control group was not included at the direction of
For the primary analysis, linear regression models were used
the NIH review panel. This is a planned interim analysis.
to examine group differences in IQ adjusting for maternal IQ,
standardized AED dose, maternal and gestational age at delivery,
Standard protocol approvals, registrations, and patient
years maternal education, race/ethnicity (self-reported), and al-
consents. Institutional review boards at each center approved
cohol use during pregnancy. Additional covariates examined
the study, and written informed consent was obtained prior to
were epilepsy/seizure types, seizure frequency during pregnancy,
enrollment according to the Declaration of Helsinki. Study con-
employment, socioeconomic status, US/UK site, tobacco use,
duct and patient safety were monitored by an NIH appointed
birthweight, unwanted pregnancy, breastfeeding, prior preg-
Data Safety Management Board. The study is registered at clini-
nancy birth defects and complications, present pregnancy com-
caltrials.gov as NCT00021866.
plications, AED compliance, and other medications used (most
Participants. Pregnant women with epilepsy on the 4 mono-
common in descending order were vitamins, non-narcotic anal-
therapies were enrolled. Since maternal IQ is the major predictor
gesics, narcotics, antibiotics, iron, antiemetics, antacids, and lo-
of child IQ in population studies,7 mothers with IQ
70 were
cal anesthetics). Our a priori hypothesis was that specific AED,
excluded to avoid floor effects. Other exclusion criteria included
dose, and maternal IQ were important covariates, so these were
positive syphilis or HIV serology, progressive cerebral disease,
included as predictors in a linear model with child IQ as out-
other major disease (e.g., diabetes), exposure to teratogenic
come. Other covariates were added individually to the model
agents other than AEDs, poor AED compliance, drug abuse in
and were included if found to be significant ( p
0.05), or if
the prior year, or drug abuse sequelae.
they did not exhibit collinearity with existing predictors. When
added to the model, none of the baseline variables substantively
Procedures. Information was collected on potentially con-
changed inferences regarding AED group differences.
founding variables, including maternal IQ, age, education, em-
For the intent-to-treat analysis, Markov Chain Monte Carlo
ployment, race/ethnicity, seizure/epilepsy types and frequency,
methods were used to impute missing age 4.5 outcomes from avail-
AED dosages, compliance, other medicines, socioeconomic sta-
able age 2 and 3 outcomes and from baseline variables related to
tus,8 UK/US site, preconception folate use, use of alcohol, to-
outcome or missingness.14,15 Baseline variables in the imputation
bacco, or other drugs during pregnancy, unwanted pregnancy,
model included AED, dose, maternal IQ and age, gestational age at
abnormalities and complications in present or prior pregnancies,
delivery, years of maternal education, alcohol use, race/ethnicity,
enrollment and birth gestational age, birthweight, breastfeeding,
and socioeconomic status (SES). Mothers of children with missing
and childhood medical diseases. Cognitive outcomes were evalu-
age 4.5 outcomes differed significantly from the rest on maternal
ated by assessors (blinded to AED) using the Differential Ability
IQ, years of maternal education, and SES ( p
0.05, t test for
Scales (DAS)9 at ages 3 and 4.5, and Bayley Scales of Infant
continuous variables, Fisher exact test for categorical variables).
Development (BSID)10 at age 2. Testing for age 4.5 outcomes
ranged from 51 to 61 months of age. Standardized scores were
They had lower IQ, fewer years of education, and lower SES. These
calculated. Separate investigations with very similar designs in
variables were included in the imputation model. Standard errors
the United States and United Kingdom were merged after initi-
and confidence intervals of parameter estimates incorporated impu-
ation. Maternal IQs were determined by different measures due
tation uncertainty. Least squares mean IQs were estimated for each
to later merger: Test of Nonverbal Intelligence-third edition
group adjusting for maternal IQ, maternal age, dose, gestational age,
(TONI-3)11 in 267 mothers (67 UK), Wechsler Abbreviated
years maternal education, alcohol use, and race/ethnicity. Similar
Scale of Intelligence (WASI)12 in 20 (all UK), and National
secondary analyses, without imputations for missing cognitive data,
Adult Reading Test (NART)13 in 17 (all UK). Training and
were conducted for the completer sample (i.e., children with cogni-
monitoring of neuropsychological evaluations were conducted to
tive testing at 4.5 years).
assure quality and consistency. Workshops were conducted for
To investigate whether baseline differences in seizure type or
neuropsychological test batteries annually, and each assessor was
other characteristics explain the association of valproate with
required to identify errors and provide appropriate correction for
poorer cognitive outcomes, post hoc subgroup analyses were con-
videotaped testing sessions containing errors in administration
ducted and forest plots were created. Subgroups were defined by 1)
and scoring. Assessors also had to submit videotape of their prac-
seizure type and 2) propensity scores.16 Further description of our
tice test session with record forms to the Neuropsychology Core
propensity analysis approach is available in our prior report.3
for review, feedback, and approval. If assessors failed, they sub-
To test the probability that the proportion of children with
mitted additional video assessments for approval.
marked impairment of cognition (i.e., IQ
70) is unchanged
Therapeutic dosages (mg/day) vary across AEDs. In order to
across time, Cochran Q statistic was used to test the null hypoth-
allow comparisons across AEDs, average AED dose during preg-
esis using the sample of children with cognitive testing at all 3
nancy was standardized, as in our prior report,3 relative to ranges
ages (i.e., 2, 3, and 4.5 years). To test for differences across AEDs
observed within each AED group in the intent-to-treat popula-
in % IQ
70 at age 4.5, Fisher exact test was used.
Neurology 78 April 17, 2012

Table 1
Demographics and IQ results for mothers of 209 children with IQ at 4.5 years
Antiepileptic drug
Mothers, n
Mean maternal IQs (95% CI)
100 (95, 104)
103 (99, 107)
93 (88, 98)
99 (94, 104)
100 (97, 102)
95 (91, 98)
Mean maternal ages, y (95% CI)
31 (29, 32)
31 (30, 32)
30 (28, 32)
28 (26, 30)
30 (30, 31)
29 (28, 30)
Mean dose, mg/d (95% CI)c
798 (687, 908)
486 (429, 543)
387 (346, 428)
992 (833, 1150)
Standardized dose (95% CI)d
33 (28, 38)
37 (33, 42)
47 (41, 53)
25 (20, 30)
36 (33, 38)
33 (28, 37)
Gestational age at birth, wk
39 (38, 39)
39 (39, 40)
39 (38, 39)
39 (39, 40)
39 (39, 39)
39 (38, 39)
(95% CI)
Years of maternal education
14 (14, 15)
15 (14, 16)
13 (12, 14)
14 (13, 14)
14 (14, 15)
13 (13, 14)
(95% CI)
Preconception folate, n (%)
29 (55)
47 (65)
18 (45)
26 (68)
120 (59)
53 (52)
Alcohol use, n (%)e
5 (9)
5 (7)
1 (3)
3 (8)
14 (7)
10 (10)
Epilepsy typesf
Localization related, n (%)
46 (87)
38 (53)
30 (75)
6 (16)
120 (59)
63 (62)
Idiopathic generalized, n (%)
5 (9)
26 (36)
6 (15)
29 (76)
66 (33)
31 (31)
GTCS, n (%)f
2 (4)
8 (11)
4 (10)
3 (8)
17 (8)
7 (7)
Nonconvulsive seizures, n (%)
16 (33)
25 (40)
13 (35)
10 (27)
64 (35)
22 (24)
None, n (%)
41 (85)
49 (78)
29 (78)
30 (81)
149 (81)
75 (82)
>5 Convulsions, n (%)
2 (4)
0 (0)
2 (4)
1 (3)
5 (3)
3 (3)
45 (85)
63 (88)
23 (58)
32 (84)
163 (80)
81 (80)
3 (6)
1 (1)
3 (8)
1 (3)
8 (4)
6 (6)
3 (6)
3 (4)
13 (33)
2 (5)
21 (10)
10 (10)
2 (4)
5 (7)
1 (3)
3 (8)
11 (5)
4 (4)
Abbreviations: CI
confidence interval; GTCS
generalized tonic-clonic seizures; NA
not applicable.
a Total across antiepileptic drugs for mothers whose children had age 4.5 cognitive outcomes. Note that there were 6 twins for a total of 209 children.
b Missing
mothers for whom their children's age 4.5 year cognitive outcomes are missing. Data were available for seizure frequency in 276 mothers
(91%), breastfeeding in 245 (81%), antiepileptic drug compliance in 238 (78%), and infant birthweight in 307 children (99%). Data for remaining covari-
ates were complete.
c Average dose for pregnancy.
d See Methods for description of how antiepileptic drug dosages during pregnancy were standardized.
e Any alcohol use during pregnancy (yes/no).
f Epilepsy types: localization related (includes cryptogenic and symptomatic); idiopathic generalized (includes absence, juvenile myoclonic, genetic, and
other idiopathic generalized not otherwise classified); GTCS (unknown if generalized or secondary generalized).
g Convulsions
n (%) of mothers without convulsions or
5 during pregnancy. Seizure frequency during pregnancy not reported for 17 mothers.
RESULTS The primary intent-to-treat analysis in-
with maternal TONI measure yielded results similar to
cluded 304 mothers and 310 live births (6 sets of twins).
the original analyses.
Mean gestational age at enrollment was 18 weeks (range
Adjusted mean age 4.5 IQ scores along with 95%
4 -37), which did not differ across drugs. The second-
confidence intervals (CIs) for each AED group are
ary completer analysis included 203 mothers and 209
listed in table 3 for the intent-to-treat and completer
children (6 sets of twins). Baseline maternal characteris-
analyses. Children exposed in utero to valproate had
tics for the intent-to-treat sample were given in our
significantly lower mean IQ than each of the other
prior publication,3 and for age 4.5 completers are de-
AEDs. Scatterplots and correlations (figure 1) depict
picted in table 1. Statistical results for the primary
the relationship of age 4.5 IQ with standardized
intent-to-treat and completer analyses with age 4.5 IQ
dose. Dose-dependent effects were present for val-
as the dependent variable in the linear regression analy-
proate but not for other AEDs. Similar results were
ses are in table 2. Because different measures were used
found for separate analyses using first and third tri-
to estimate maternal IQ, a covariate for type of maternal
mester dosages. Child and maternal IQ were signifi-
IQ measure was added to the primary model; this cova-
cantly related for each AED except valproate (figure e-1
riate was nonsignificant (p
0.39). In addition, reanal-
on the Neurology(R) Web site at www.neurology.org).
ysis of the completer group analysis using only those
Verbal abilities were significantly lower than nonver-
Neurology 78 April 17, 2012

0.0037) as children became older. From age 3 to 4.5,
Table 2
Results for linear regression analysis with age 4.5 IQ as dependent
IQ improved for carbamazepine ( p
0.008), lam-
variable for intent-to-treat analysis (n
310 children) and
completer analysis (n
209 children)
otrigine ( p
0.0001), and phenytoin ( p
but not valproate ( p
0.57). At age 4.5, 10% of
F value
Coefficient (CI)
p Value
children exposed to valproate had marked intellec-
Age 4.5 intent-to-treat analysisa

41%; average R2 over all datasets)
tual impairment compared to 0%- 4% for other
AEDs ( p
0.0064); this AED difference was also
AED (4 groups)
present in the subset of children tested at all 3 time
Maternal IQb
0.31 (0.17, 0.45)
points ( p
Maternal education, y
1.10 (0.28, 1.91)
Propensity score analysis demonstrates that the
Maternal age, yc
0.63 (0.26, 0.99)
results are not due to differences in baseline variables
Gestational age
1.28 (0.41, 2.15)
related to either the child IQ outcome or chances of
belonging to the valproate group. The forest plot dis-
Alcohol usef
7.40 ( 14.73,
playing means by seizure type and AED group shows
Standardized doseg
0.15 ( 0.26,
that AED group differences cannot be explained by
Age 4.5 completer analysish (R2
imbalances in underlying seizure type (figure e-2). A
AED (4 groups)
subanalysis of patients with juvenile myoclonic epilepsy
Maternal IQ
0.28 (0.14, 0.43)
did not alter the results. The analysis examining sensi-
Maternal education, y
1.28 (0.43, 2.12)
tivity of results to missing data demonstrated that the
Maternal age, yc
0.68 (0.29, 1.07)
results cannot be explained by incomplete data.
Gestational age
0.91 ( 0.03, 1.85)
DISCUSSION The present age 4.5 year results are
consistent with our prior findings,3 and demonstrate
Alcohol use
10.94 ( 18.94,
reduced cognitive abilities in children exposed in
Standardized doseg
0.12 ( 0.23,
utero to valproate monotherapy. Similar to our age 3
Abbreviations: AED
antiepileptic drug; CI
confidence interval; NA
not applicable.
year findings,3 valproate dose was associated with
a Cognitive outcomes were available for at least one test age (i.e., 2, 3, 4.5 years) in 270
lower child IQ at age 4.5, but the dose of the other
children (87% of intent-to-treat sample).
b Higher maternal IQ was associated with higher child IQ scores (r
0.51; p
AEDs was not. In addition, maternal IQ was asso-
c Maternal age was associated with higher child scores (r
0.33; p
0.0001); lower child
ciated with child IQ at age 4.5 for each AED ex-
scores seen for youngest mothers.
cept valproate, similar to our age 3 finding.3 Our
Race/ethnicity has 4 categories (Caucasian non-Hispanic, Black non-Hispanic, Hispanic,
and other).
findings are consistent with other studies report-
e Child IQ scores were higher for Caucasian than other race/ethnicity categories ( p
ing impaired cognition in children with fetal val-
0.015), and scores were higher for children whose mothers completed more years of edu-
proate exposure.17-22
cation (r
0.43; p
f Scores were lower in children whose mothers reported alcohol use ( p
0.05); mean (95%
The present results add confidence to our prior
CI) child IQ score adjusted for factors in the primary analysis model was 97 (90:102) for
results, and also add new findings. Since age 4.5 out-
children exposed to alcohol in utero vs 104 (102:106) in children not exposed.
comes are related significantly to age 2 and 3 out-
See Methods for description of how AED dosages during pregnancy were standardized,
and see Results for details of dose effects. Median dosages for the whole pregnancy aver-
comes, impairments may be detected early
age were carbamazepine 800 mg, lamotrigine 483 mg, phenytoin 390 mg, valproate 1,000
facilitating appropriate intervention programs. Cog-
mg. Means (ranges) of whole pregnancy average dosages in mg were carbamazepine 797
nitive improvements occurred as children aged. It is
mg (33-1,800), lamotrigine 486 mg (50-1,217), phenytoin 387 mg (67-780), valproate
possible that this improvement could be due at least
992 mg (200-2,250).
h Age 4.5 outcomes were available in 209 children (67%). Reasons for absence of age 4.5
in part to practice effects. However, a different test
test results included early terminations and missing scheduled testing within assessment
was use at age 2 vs 3 and 4.5. Further, improvement
did not appear to occur in valproate exposed chil-
bal abilities across all drugs combined and for each
dren, in which 10% of the children had marked in-
AED (table 4).
tellectual impairment range (i.e.,
70 IQ) at age 4.5.
Age 4.5 IQ correlated with age 2 BSID (r
If practice effects are occurring, they do not appear to
0.0001) and age 3 DAS IQ (r
0.77, p
be present in the valproate group.
0.0001). Adjusted mean cognitive scores (BSID and
Higher maternal IQ was associated with higher
DAS IQ) and % marked intellectual impairment
child IQ overall and for carbamazepine, lamotrigine,
( 70 standard score) are in table e-1 for each AED
and phenytoin individually, but not for valproate.
across ages 2, 3, and 4.5 years for all children tested at
Greater maternal education and age were both asso-
each age and for those tested at all 3 ages. Overall,
ciated with higher child IQ; lower child IQ was seen
mean IQ scores improved ( p
0.0001) and %
in very young mothers. Lower gestational age and
marked intellectual impairment decreased ( p
maternal alcohol use were both associated with lower
Neurology 78 April 17, 2012

equal based on normed data in healthy children. Per-
Table 3
Adjusteda mean IQs, 95% CIs, DCIs, and p valuesb for comparisons to
formance IQ is more susceptible to practice effects
valproate, and sample sizes for each AED at age 4.5 years for ITT (n
310) and completer (n

209) analyses for each antiepileptic drug
than verbal IQ.24,25 Although practice effects could
affect the cognitive outcomes, it is unlikely that such
an effect is a major factor in our findings. The 30-day
practice effects on the DAS for the verbal and non-
Mean IQ
verbal cluster scores are modest,9 and our test-retest
102, 109
102, 109
102, 111
91, 100
interval was 1.5 years. Further, practice effect does
5, 16
5, 16
4, 18
not explain the verbal/nonverbal split at age 3 on the
0 .0008
first exposure to DAS, nor the fact that the split does
not widen across AEDs between age 3 and 4.5. In
addition, several studies have found verbal impair-
ments in children exposed to valproate,17,18,22 and
Mean IQ
one other study reported that language functions
103, 110
102, 109
102, 111
92, 101
were impaired in children with antenatal AED expo-
3, 17
2, 16
2, 18
sure.26 Nevertheless, we caution that this finding re-
quires replication in separate cohorts to determine if
language abilities are particularly susceptible to fetal
Abbreviations: AED
antiepileptic drug; CI
confidence interval; DCI
difference confi-
AED exposure. If this verbal impairment from fetal
dence interval; ITT
intent-to-treat; NA
not applicable.
AED exposure is confirmed, it has important practi-
a Mean age 4.5 years IQ scores adjusted for maternal IQ, maternal education, maternal age,
cal and theoretical implications.
dose, race/ethnicity, and alcohol use during pregnancy. ITT analysis includes imputed IQ
data. Unadjusted IQ means for the ITT analysis were carbamazepine
106, lamotrigine
Alcohol can induce widespread neuronal apopto-
108, phenytoin
101, valproate
95. Unadjusted means for the completer analysis were
sis in the immature brain,27 raising concern that
107, lamotrigine
109, phenytoin
102, valproate
AEDs might produce similar effects. Some AEDs in-
b p Values adjusted for 3 pairwise comparisons to valproate using Dunnett correction.
duce neuronal apoptosis in the immature brain when
child IQ. Child IQ was higher in Caucasians. Similar
given in monotherapy, and many AEDs can enhance
results were found in our completer analysis.
apoptosis in polytherapy.28-33 The effect is dose de-
The prospective design, blinded cognitive assess-
pendent, occurs at therapeutically relevant blood lev-
ments using standardized measures, and detailed
els, and requires only relatively brief exposure.
monitoring of multiple potential confounding fac-
Valproate appears to induce the apoptosis at dosages
tors are strengths of our study. Limitations include a
relatively lower in terms of effective therapeutic dose
relatively small non-population-based sample, loss of
compared to other AEDs.28,29 However, many AEDs
enrolled subjects to analysis, lack of randomization,
have not been tested in this model. In regards to the
lack of AED blood levels, and absence of an unex-
induced cognitive deficits, the neuronal apoptosis
posed control group during pregnancy. Another
likely only contributes in part, and the more impor-
weakness of our study is that different measures were
tant factor may be impaired physiology in the surviv-
used to estimate maternal IQ. However, the maternal
ing neurons, similar to alcohol.34 Although it is
IQ measures used in our study correlate well with a
possible that widespread neuronal apoptosis or dys-
standard adult IQ measure (i.e., full-scale IQ of the
function might differentially affect verbal abilities,
Wechsler Adult Intelligence Scale).11-13 Further, re-
we have hypothesized that the greater verbal impair-
gression analyses controlling for type of maternal IQ
ment might result from altered cerebral lateralization
measure and analysis of the completer sample limited
from fetal AED exposure.23 Additional research is
to mothers receiving the TONI both produce similar
needed to test this hypothesis and fully delineate the
results. A potential confounding issue for observa-
mechanisms of behavioral teratogenesis.
tional studies is that differences in baseline character-
The treatment of epilepsy in women of childbear-
istics might affect child IQ and alter results.
ing potential requires balancing risks that seizures
However, the observed associations persist in analy-
pose to the mother and child vs risks of teratogenesis
ses adjusted for baseline characteristics, including
(i.e., congenital malformations and cognitive defi-
propensity scores and seizure type.
cits) in the child. A discussion of these risks should
We previously reported that all 4 of these com-
occur prior to pregnancy in all women of childbear-
monly used AEDs impair verbal abilities compared
ing potential treated with AEDs. Based on the cur-
to nonverbal abilities at age 3.23 Here, we find that
rent literature, valproate poses the most consistent,
this verbal deficit persists for all 4 AEDs at age 4.5
frequent, and severe risks for both anatomic and be-
years. The verbal and nonverbal measures are ex-
havioral teratogenesis.3,6 We contend that valproate
pressed as standard scores, so these scores should be
is a poor first choice AED for most women of child-
Neurology 78 April 17, 2012

Figure 1
Scatterplots of age 4.5 IQ vs standardized dose for each antiepileptic drug during pregnancy
bearing potential. Some women with generalized ep-
risk factor for damage to the immature brain,36 but
ilepsy can only be controlled by valproate.35
there are inadequate human data to determine if
However, if other AEDs are ineffective in seizure
multiple daily doses or extended release formulations
control, then valproate can be used subsequently.
would reduce risk.
When valproate is necessary, we recommend em-
Our present knowledge remains inadequate on
ploying the lowest dose possible. Recent animal data
many issues to direct care of women with epilepsy
suggest that peak drug levels may be an important
who might become pregnant.6,37,38 If teratogenetic is-
sues alone are considered, fetal risks are not fully de-
lineated for most AEDs in regards to malformations
Table 4
Verbal and nonverbal cluster means and their mean differences at
or cognitive outcomes. Investigations of both ana-
age 4.5 years across AEDsa
tomic and behavioral teratogenesis in humans are
Verbal cluster
Nonverbal cluster
based on observational studies. Thus, a signal needs
(95% CI)a
(95% CI)a
(95% CI)a,b
p Valuec
to be observed in multiple cohorts, as is the case for
102 (98, 105)
108 (104, 112)
6.2 (2.0, 10.3)
valproate. The recent EURAP results show dose-
99 (96, 102)
108 (105, 112)
9.1 (5.4, 12.7)
dependent malformation risks for not only valproate,
101 (96, 105)
107 (102, 112)
6.5 (1.5, 11.5)
but also carbamazepine, lamotrigine, and phenobarbi-
91 (86, 95)
102 (96, 106)
10.6 (5.5, 15.7)
tal.39 Teratogens act in a dose-dependent manner, and
All AEDs combined
99 (96, 101)
107 (104, 109)
8.1 (5.7, 10.4)
delineation of dose effects for both anatomic and func-
Abbreviations: AED
antiepileptic drug; CI
confidence interval.
tional deficits in other AEDs may require larger sample
a Mean verbal and nonverbal cluster scores, and their mean differences using a linear model
sizes than valproate. Further, the relative risks of specific
controlling for maternal IQ, dose, age of mother, gestational age, alcohol use, race/ethnicity,
polytherapy combinations are unknown. Understand-
and years of maternal education.
ing these effects is critical to management of women
Difference of nonverbal cluster minus verbal cluster.
c p Values for paired t tests.
with epilepsy who are of childbearing potential.
Neurology 78 April 17, 2012

The risk for malformations is primarily due to
and Treatment (Lippincott Williams & Wilkins, 2006) and Controversial
exposure early in the first trimester, but the risk for
Issues in Psychiatric Aspects of Epilepsy (Elsevier, 2008); has served on
speakers' bureaus for GlaxoSmithKline, UCB, and Pfizer Inc; and has
cognitive deficits appears to be in the third trimester,
received research support from GlaxoSmithKline, Novartis, and Pfizer
similar to alcohol. While dose might be a reasonable
Inc. Dr. Liporace serves on the speakers' bureau for and has received
surrogate for AED exposure early in pregnancy, it is
speaker honoraria from UCB; receives publishing royalties for Crash
Course Neurology
(Elsevier, 2006); and serves on the Professional Board of
not late in pregnancy due to considerable variability
Epilepsy Foundation of Eastern Pennsylvania. Dr. Pennell serves as a con-
in AED clearance during pregnancy across AEDs and
tributing editor to Epilepsy Currents and on the editorial board of Epilep-
across individual patients.40 To date, no study of out-
sia; has received research support from the NIH, the Milken Family
comes in children of women with epilepsy have con-
Foundation, the Epilepsy Foundation, UCB, and Marinus Pharmaceuti-
cals; has received travel reimbursement from the NIH, the Milken Family
trolled for this factor by assessing AED blood levels,
Foundation, the Epilepsy Foundation, American Epilepsy Society, and
which would be a much better measure of intrauter-
the National EpiFellows Foundation; and serves as a volunteer member of
ine AED exposure. Further, it is critical that our ob-
the American Epilepsy Society Board of Directors and as chair of the
servation of language deficits in 4 commonly used
Professional Advisory Board for the Epilepsy Foundation. Dr. Privitera
has served on scientific advisory boards or as a consult for Ortho-McNeil-
AEDs be examined in other cohorts and in addi-
Janssen Pharmaceuticals, Inc., UCB, Johnson & Johnson, and the Epifel-
tional AEDs. Given the evolving nature of our evi-
lows Foundation; has received funding for travel and speaker honoraria
dence base, clinicians and patients should remain
from Ortho-McNeil-Janssen Pharmaceuticals, Inc., Pfizer Inc., Glaxo-
SmithKline, Janssen, and UCB; has served on speakers' bureaus for UCB,
attentive to future information and its implications.
Pfizer Inc., GlaxoSmithKline, and Ortho-McNeil-Janssen Pharmaceuti-
cals, Inc.; and has received research support from UCB, Ortho-McNeil-
Janssen Pharmaceuticals, Inc., the NIH (K01-DA020485 [co-I], K23
Dr. Meador drafted the manuscript and supervised the study. All other
NS052468 [co-Mentor], 1K23NS02170 - 01 [co-I], 2R01-NS38455
authors reviewed and revised the manuscript. Drs. Meador, Baker, Co-
[site PI]), the American Epilepsy Society, and the Shor Foundation for
hen, and Loring contributed to study design. Drs. Meador and Baker
Epilepsy Research. Dr. Loring serves on scientific advisory boards for the
obtained funding. Dr. Browning conducted the statistical analyses and
Epilepsy Foundation; serves on the Steering Committee for the NINDS
interpreted the data with Drs. Meador, Cohen, and Loring. All authors
Common Data Element Project, as Consulting Editor for the Journal of
were involved in acquisition of data.
Clinical and Experimental Neuropsychology and for Epilepsy Research, and
on the editorial boards of Epilepsia, Journal of Pediatric Epilepsy, and Neu-
ropsychology Review; serves as a consultant for NeuroPace, Inc. and UCB;
receives publishing royalties for Neuropsychological Assessment, 4th ed. (Ox-
The investigators thank the children and families who gave their time to
ford University Press, 2004) and INS Dictionary of Neuropsychology (Ox-
participate in the NEAD Study.
ford University Press, 1999); estimates that 50% of his clinical effort
involves neuropsychological testing including Wada testing; receives re-
search support from the NIH/ NINDS, Epilepsy Foundation, and UCB;
Dr. Meador serves on the editorial boards of Neurology(R), Behavior &
and receives travel reimbursement from the NIH.
Neurology, Epilepsy and Behavior, Epilepsy Currents, Epilepsy.com, and the
Journal of Clinical Neurophysiology and on the Professional Advisory Board
Received August 15, 2011. Accepted in final form December 1, 2011.
for the Epilepsy Foundation; has received travel support from sanofi-
aventis; and received research support from GlaxoSmithKline, Eisai Inc.,
Marinus Pharmaceuticals, Inc., Myriad Genetics, Inc., NeuroPace, Inc.,
Pfizer Inc, SAM Technology Inc., Schwartz Pharma (UCB), the NIH/
Fisher JE, Vorhees CV. Developmental toxicity of antiepi-
NINDS, and the Epilepsy Foundation. Dr. Baker has served on a scien-
tific advisory board for sanofi-aventis; serves on the editorial board of
leptic drugs: relationship to postnatal dysfunction. Phar-
Epilepsy and Behaviour; has received speaker honoraria from Eisai Inc.; has
macol Res 1992;26:207-221.
received educational support from UCB, sanofi-aventis, and Pfizer Inc;
Gaily E, Meador KJ. Neurodevelopmental effects. In: En-
and receives research support from Epilepsy Research UK, Medical Re-
gel J, Pedley TA, eds. Epilepsy: A Comprehensive Text-
search Council UK, and Epilepsy Action UK. Dr. Browning serves on a
book, 2nd ed. Philadelphia: Lippincott Williams &
safety monitoring committee for Ligocyte Pharmaceuticals; receives re-
Wilkins; 2007:1225-1233.
search support from the NIH (NINDS, NIAID); and holds stock in Hu-
Meador KJ, Baker GA, Browning N, et al. Cognitive func-
man Genome Sciences, Inc. Dr. Cohen serves on the editorial board of
tion at 3 years of age after fetal exposure to antiepileptic
Developmental Neuropsychology; receives publishing royalties for Children's
drugs. N Engl J Med 2009;360:1597-1605.
Memory Scale (Pearson, 1997); and receives research support from the
FDA Drug Safety Communication: Children born to moth-
NIH. Dr. Bromley has served on a scientific advisory board for sanofi-
aventis; has received funding for travel from UCB; has served as an expert
ers who took Valproate products while pregnant may have
witness in a medico-legal case; and receives research support from UCB
impaired cognitive development (6/30/2011). Available at:
and sanofi-aventis. Dr. Clayton-Smith serves as Co-Editor for Clinical
Dysmorphology and on the editorial board of European Journal of Human
Accessed March 1, 2012.
Genetics; has served as an expert witness in medico-legal cases; and receives
Kluger BM, Meador KJ. Teratogenicity of antiepileptic
research support from the NIH (NICDR), NHS Foundation Trust, and
medications. Semin Neurol 2008;28:328 -335.
Action Medical. Dr. Kalayjian has received speaker honoraria from Glaxo-
Harden CL, Meador KJ, Pennell PB, et al. Practice Param-
SmithKline and Ortho-McNeil, and research support from Marinus
eter update: management issues for women with epilepsy:
Pharmaceuticals, Inc. and Medical College of Georgia. Dr. Kanner has
focus on pregnancy (an evidence-based review): teratogen-
served on scientific advisory boards or as a consultant for GlaxoSmithK-
esis and perinatal outcomes: report of the Quality Stan-
line, Ortho McNeill, Pfizer Inc, Valeant Pharmaceuticals International,
and UCB; has received speaker honoraria from GlaxoSmithKline, Ortho-
dards Subcommittee and Therapeutics and Technology
McNeil, Pfizer Inc, and UCB; serves on the editorial boards of Epilepsy &
Subcommittee of the American Academy of Neurology
Behavior, Epilepsia, and CNS Spectrums; receives publishing royalties for
and American Epilepsy Society. Neurology 2009;73:133-
Psychiatric Issues in Epilepsy. Second Edition: A Practical Guide to Diagnosis
Neurology 78 April 17, 2012

Sattler JM. Assessment of Children, 3rd ed. San Diego:
Ikonomidou C, Bittigau P, Ishimaru MJ, et al. Ethanol-
Jerome M. Sattler; 1992.
induced apoptotic neurodegeneration and fetal alcohol
Hollingshead AB. Four Factor Index of Social Status. New
syndrome. Science 2000;287:1056 -1060.
Haven, CT: Yale University; 1975.
Bittigau P, Sifringer M, Genz K, et al. Antiepileptic drugs
Elliott CD. Differential Abilities Scales, 2nd ed. San Anto-
and apoptotic neurodegeneration in the developing brain.
nio, TX: The Psychological Corporation; 1990.
Proc Natl Acad Sci USA 2002;99:15089 -15094.
Bayley N. Bayley Scales of Infant Development, 2nd ed.
Bittigau P, Sifringer M, Ikonomidou C. Antiepileptic
San Antonio, TX: The Psychological Corporation; 1993.
drugs and apoptosis in the developing brain. Ann NY Acad
Brown L, Sherbenou RJ, Johnsen SK. Test of Non-verbal
Sci 2003;993:103-114.
Intelligence (TONI-3), 3rd ed. Austin, TX: Pro-Ed; 1996.
Glier C, Dzietko M, Bittigau P, Jarosz B, Korobowicz E,
Wechsler D. Wechsler Abbreviated Scale of Intelligence
Ikonomidou C. Therapeutic doses of topiramate are not
(WASI). San Antonio, TX: Psychological Corporation;
toxic to the developing rat brain. Exp Neurol 2004;187:
403- 409.
Nelson H, Willison J. National Adult Reading Test
Manthey D, Asimiadou S, Stefovska V, et al. Sulthiame
(NART). Oxford, UK: NFER-Nelson Publishing; 1991.
but not levetiracetam exerts neurotoxic effect in the devel-
Li KH. Imputation using Markov chains. J Stat Comput
oping rat brain. Exp Neurol 2005;193:497-503.
Simul 1988;30:57-79.
Katz I, Kim J, Gale K, Kondratyev A. Effects of lam-
Little RJA, Rubin DB. Statistical Analysis with Missing
otrigine alone and in combination with MK-801, pheno-
Data, 2nd ed. New York: John Wiley & Sons, Inc.; 2002.
barbital, or phenytoin on cell death in the neonatal rat
Rosenbaum P, Rubin DB. The central role of the propen-
brain. J Pharmacol Exp Ther 2007;322:494 -500.
sity score in observational studies for causal effects. Bi-
Kim JS, Kondratyev A, Tomita Y, Gale K. Neurodevelop-
ometrika 1983;70:41-55.
mental impact of antiepileptic drugs and seizures in the
immature brain. Epilepsia 2007;48(suppl 5):19 -26.
Adab N, Kini U, Vinten J, et al. The longer term outcome
Medina AE, Krahe TE, Coppola DM, Ramoa AS. Neona-
of children born to mothers with epilepsy. J Neurol Neu-
tal alcohol exposure induces long-lasting impairment of
rosurg Psychiatry 2004;75:1575-1583.
visual cortical plasticity in ferrets. J Neurosci 2003;23:
Gaily E, Kantola-Sorsa E, Hiilesmaa V, et al. Normal in-
telligence in children with prenatal exposure to carbamaz-
Marson AG, Al-Kharusi AM, Alwaidh M, et al. The
epine. Neurology 2004;62:28 -32.
SANAD study of effectiveness of valproate, lamotrigine, or
Thomas SV, Ajaykumar B, Sindhu K, Nair MK, George
topiramate for generalised and unclassifiable epilepsy: an
A, Sarma PS. Motor and mental development of infants
unblinded randomised controlled trial. Lancet 2007;369:
exposed to antiepileptic drugs in utero. Epilepsy Behav
1016 -1026.
2008;13:229 -236.
Kondratyev A, Gale K, Tsukerman S. Phenobarbital-
Bromley RL, Mawer G, Love J, et al, Liverpool and Man-
induced apoptosis in the neonatal rat brain occurs in a
chester Neurodevelopment Group [LMNDG]. Early cog-
limited compartment of neurons: American Epilepsy Soci-
nitive development in children born to women with
ety 2010. abstract 2.156. Available at: http://www.aesnet.
epilepsy: a prospective report. Epilepsia 2010;51:2058 -
display/st/gale/sy/2010/sb/All/id/12750. Accessed March
Shallcross R, Bromley RL, Irwin B, Bonnett LJ, Morrow J,
1, 2012.
Baker GA, Liverpool Manchester Neurodevelopment
Harden CL, Hopp J, Ting TY, et al. Practice Parameter
Group, UK Epilepsy and Pregnancy Register. Child devel-
update: management issues for women with epilepsy: focus
opment following in utero exposure: levetiracetam vs so-
on pregnancy (an evidence-based review): obstetrical com-
dium valproate. Neurology 2011;76:383-389.
plications and change in seizure frequency: report of the
Nadebaum C, Anderson V, Vajda F, Reutens D, Barton S,
Quality Standards Subcommittee and Therapeutics and
Wood A. The Australian brain and cognition and antiepi-
Technology Assessment Subcommittee of the American
leptic drugs study: IQ in school-aged children exposed to
Academy of Neurology and American Epilepsy Society.
sodium valproate and polytherapy. J Int Neuropsychol Soc
Neurology 2009;73:126 -132.
Harden CL, Pennell PB, Koppel BS, et al. Practice Param-
Meador KJ, Baker GA, Browning N, et al. NEAD Study
eter update: Management issues for women with epilepsy:
Group. Foetal antiepileptic drug exposure and verbal ver-
focus on pregnancy (an evidence-based review): vitamin K,
sus non-verbal abilities at three years of age. Brain 2011;
folic acid, blood levels, and breastfeeding: report of the
134:396 - 404.
Quality Standards Subcommittee and Therapeutics and
Sawrie SM, Chelune GJ, Naugle RI, Luders HO. Empiri-
Technology Assessment Subcommittee of the American
cal methods for assessing meaningful neuropsychological
Academy of Neurology and American Epilepsy Society.
change following epilepsy surgery. JINS 1996;2:556 -564.
Neurology 2009;73:142-149.
Hermann BP, Seidenberg M, Schoenfeld J, Peterson J, Le-
Tomson T, Battino D, Bonizzoni E, et al, for the EURAP
veroni C, Wyler AR. Empirical techniques for determining
study group. Dose-dependent risk of malformations with
the reliability, magnitude, and pattern of neuropsycholog-
antiepileptic drugs: an analysis of data from the EURAP
ical change after epilepsy surgery. Epilepsia 1996;37:942-
epilepsy and pregnancy registry. Lancet Neurol 2011;10:
609 - 617.
Thomas SV, Sukumaran S, Lukose N, George A, Sarma
Pennell PB, Hovinga CA. Antiepileptic drug therapy in
PS. Intellectual and language functions in children of
pregnancy I: gestation-induced effects on AED pharmaco-
mothers with epilepsy. Epilepsia 2007;48:2234 -2240.
kinetics. Int Rev Neurobiol 2008;83:227-240.
Neurology 78 April 17, 2012