Teratogenesis in repeated pregnancies in antiepileptic drug-treated women


Address correspondence to Frank J. E. Vajda, Department of Medicine and Neurology, University of Melbourne and Royal Melbourne Hospital, Parkville, Vic. 3050, Australia. E-mails: vajda@netspace.net.au; frank.vajda@mh.org.au


Purpose:  Considerable information is now available concerning the risk of teratogenesis in the individual pregnancy exposed to antiepileptic drugs (AEDs). However, there is comparatively little information available concerning the risk in the subsequent pregnancies of women who continue to take the AED associated with a fetal malformation in a previous pregnancy. This article addresses this matter.

Methods:  Analysis of data concerning fetal abnormalities in 1,243 women who had 2,637 pregnancies between mid-1999 and 2010 recorded in the Australian Register of Antiepileptic Drugs in Pregnancy. Of the 2,637 pregnancies, 1,114 had been completed before initial enrolment in the Register.

Key Findings:  Women taking any AED who had given birth to a malformed baby in their first enrolled pregnancy and who continue taking the same drug were at increased risk of having a malformed offspring in their next pregnancy (35.7% vs. 3.1%; odds ratio [OR] 17.6; 95% confidence interval [95% CI] 4.5–68.7). Among these women, those taking valproate (VPA) were more likely to have malformed fetuses in their next pregnancies than those who had taken VPA without fetal abnormalities (57.2% vs. 7.0%, OR 17.8; 95% CI 2.7, 119.1). There were similar although not statistically significant trends in those who had taken AEDs other than VPA. Similar, although again not statistically significant, trends were found, when considering the pairings of the most recent preenrollment pregnancy and the following one. If a woman had two or more pregnancies that resulted in AED-associated fetal malformation, the types of malformation were often different.

Significance:  Women whose last pregnancy resulted in a fetal malformation have a substantially increased risk of having further malformed fetuses if they become pregnant again while taking the same AED, particularly VPA. This suggests that maternal factors, perhaps genomic, predispose to at least VPA-associated malformations. This knowledge, together with information about the outcome of any previous pregnancy, should help in advising women with AED-treated epilepsy who plan further pregnancies.

Over nearly half a century there has been considerable interest in the question of human antiepileptic drug (AED)–related teratogenicity (Meadow, 1970; Janz, 1975; Dravet et al., 1992; Samren et al., 1997; Olafsson et al., 1998; Canger et al., 1999; Kaneko et al., 1999; Kaaja et al., 2003; Vajda et al., 2004). It has been established that valproate (VPA) possesses dose-related teratogenicity (Editorial, 1982; Omtzigt et al., 1992; Vajda & Eadie, 2005; Morrow et al., 2006; Meador et al., 2009; Tomson & Battino, 2009; Tomson et al. 2011). The hazard of teratogenicity from the remaining AEDs currently in common clinical use appears to be less. These conclusions have been reached mainly from analyses of data drawn from various registers and population-based studies that have recorded the outcomes of individual pregnancies in women who receive AEDs during pregnancy. The analyses have usually been carried out in relation to the individual pregnancy rather than considering all of the pregnancies in an individual woman. Consequently, the question of the outcomes of subsequent pregnancies in the individual woman taking AEDs has been little studied. Data on this matter would be clinically valuable to inform counseling of AED-treated women, particularly if they already had a pregnancy that resulted in an AED-associated fetal malformation. Scientifically, it could also provide evidence on whether the occurrence of AED-associated fetal malformations is at least partly genetically determined. If that were the case, a women who had one pregnancy with an AED-associated fetal malformation would be expected to be at increased risk of having further such pregnancies if the same AED continued to be taken. Two instances of such an event have already been reported by Duncan et al. (2001).

Previous analyses of the data of the Australian Register of Antiepileptic Drugs in Pregnancy (APR) had been carried out in terms of individual pregnancies rather than in terms of individual women and all their pregnancies (Eadie & Vajda, 2005; Vajda et al., 2006a,b, 2007a,b,c, 2008, 2010, 2011). A number of the women had more than one pregnancy that was included in the Register. In the study reported here, each pregnancy in the Australian Register was linked to the relevant woman. This allowed the outcomes of repeated pregnancies in the same mother to be examined.

Materials and Methods

This article is based on data collected in the APR between mid-1999 and November 2010. These data relate to pregnant women either with epilepsy (WWE) or with disorders other than epilepsy who took AEDs throughout pregnancy, or to WWE who took no AEDs in at least the first trimester of pregnancy.

Recruitment for the APR was nationwide and entirely voluntary. Potentially eligible women were made aware of the Register by their treating medical practitioners, nurses, or allied health professionals, or by other women who were enrolled in the Register. Contact with the Register was exclusively by telephone. Each pregnancy received its own identification number. Relevant details of each postenrollment pregnancy were obtained from pregnant women on recruitment in the first or second trimester, at 7 months of pregnancy, within the first postnatal month and at the end of the first postnatal year. The details were entered into a database and confidentially coded. The information collected at enrollment also included details of previous pregnancies, and their outcomes. Treating doctors were contacted to confirm details of each pregnancy after it was enrolled. There was no attempt to influence the management of the pregnancy, which always remained at the discretion of the treating medical practitioner. The fetal malformation classification used was that of the Birth Defects Registry of Victoria (Riley & Halliday, 2000).

The APR database was housed initially at St Vincent’s Hospital, Melbourne, then at Monash University, and currently at the Departments of Medicine and Neurology, The Royal Melbourne Hospital, The University of Melbourne, under the ethical oversight of the ethics committees of the above institutions.

For the present analysis, the patient contact database was linked to the pregnancy details database through the individual identification numbers. This permitted identification of all of the recorded pregnancies of each woman.

The presence of fetal malformations was determined on the basis of the information available at the end of the first postnatal month and of the first postnatal year. Some women included in the analysis were lost to follow-up by the end of the first postpartum year (and others had not reached that stage at the time of data analysis). Each woman’s first APR-enrolled pregnancy has been considered her index one, and her pregnancies numbered forward and backward from the index one. In one of the analyses carried out, the last preenrollment pregnancy was considered to be the first one, and the index pregnancy considered the second one.


Study population

Data for at least the index pregnancy outcome was available for 1,243 women, all but 37 of whom had epilepsy. A total of 291 further pregnancies in 228 of these women were subsequently enrolled in the APR. Information on pregnancies that had occurred before the enrollment of the index pregnancy was collected for 596 of the 1,243 women, the other 647 index pregnancies being the women’s first ones. The 228 women with enrolled postindex pregnancies included 45 who subsequently enrolled a third pregnancy, seven who enrolled a fourth, and one who enrolled a fifth. Of the 596 women who had preindex pregnancies recorded, 288 had two preindex ones, 143 had three, 60 had four, and five had five or more. Therefore, there were a total of 2,637 pregnancies, 1,114 of them being preindex. Spontaneous abortions and abortions for maternal indications were more frequent in preindex pregnancies than in index and postindex ones, mainly because such pregnancies that ended early were unlikely to be enrolled in the APR, but such pregnancies would be considered as previous pregnancies at the time of initial enrollment.

Prospectively followed pregnancies

Table 1 shows the incidences of index and postindex pregnancy with fetal malformations in three groups of women: (1) those taking no AEDs in the first trimester of pregnancy, (2) those taking AEDs other than VPA, and (3) those taking VPA, whether alone or in combination with other AEDs.

Table 1.   Malformation rates in pregnancies of prospectively studied women
Pregnancy numberNo. of womenTreatmentNo. in each subgroupNo. with malformation%
  1. Pregnancy 0 is the index pregnancy. The numbers shown appear not to reconcile completely in some instances because of the occurrence of some spontaneous abortions. The numbers in this table for the second pregnancies do not correspond exactly with those of Fig. 1, because some women’s AEDs were changed between the two pregnancies and these women were not included in the second pregnancy data in Fig. 1.

01,243No AEDs11754.3
Non-VPA AEDs789354.4
+1288No AEDs1616.3
Non-VPA AEDs16042.5
+245No AEDs000
Non-VPA AEDs3213.1
VPA 13 00

Where there were sufficient numbers of women to permit valid statistical analysis, in both the index and first postindex pregnancies (Table 1), the fetal malformation rates were significantly higher in pregnancies exposed to VPA than in pregnancies exposed to AEDs other than VPA (index pregnancies 13.1% vs. 4.4%: odds ratio [OR] 3.24; 95% confidence interval [95% CI] 2.03–5.15: first postindex pregnancies 13.5% vs. 2.5%; OR 6.07; 95% CI = 1.7–21.7). These findings are consistent with the findings from previous reported analyses from the APR in which all pregnancies were regarded as independent events (Vajda et al., 2004; Vajda & Eadie, 2005; Vajda et al., 2007a,b,c, 2011). The rates for AEDs other than VPA were not appreciably different from the rates when no AEDs were involved.

Figure 1 shows the fetal outcomes of the index and first postindex pregnancies in the 1,243 women studied, but only women who took the same AEDs in both pregnancies are included in the postindex pregnancies whose outcomes are considered. There was a significantly higher rate of fetal malformations in the first postindex pregnancy in women taking any AED when the index pregnancy had resulted in a fetal malformation (35.7% vs. 3.1%; OR 17.6; 95% CI 4.5–68.7). Within this set of pregnancies, women taking AEDs other than VPA had a higher rate of fetal malformations in the first postindex pregnancy when the index pregnancy had resulted in a fetal malformation (14.3% vs. 1.96%), but the difference was not statistically significant (OR 8.33, 95% CI 0.75–92.4), but would have been significant at a 90% CI. However, there was a significantly higher rate of fetal malformations in the first postindex pregnancy in women taking VPA when the index pregnancy had resulted in a fetal malformation (57.1% vs. 7.0%; OR 17.8; 95% CI 2.7, 119.1). Mean VPA doses were higher in the index pregnancy group with malformed fetuses than in index pregnancies with nonmalformed fetuses (1,448, standard deviation [SD] 882 vs. 882, SD 576 mg/day; difference 566 mg/day, 95% CI 361–770 mg/day). In the first postindex pregnancies, the mean dosage difference was not statistically significant (1,057, SD 588 vs. 803, SD 541 mg/day; difference 254 mg/day, 95% CI −212, 720 mg/day). Although only small numbers were involved, the risk of a first postindex pregnancy outcome being malformed was higher, but not statistically significantly so, when VPA exposure was involved, than when exposure only to other AEDs had occurred (57.1% vs. 14.3%: OR 8.0, 95% CI 0.6–06.9).

Figure 1.

Showing the reproductive outcomes in consecutive pregnancies in women enrolled in the Register. The upper half of the figure contains data for index and first postindex pregnancies, and the lower half data for most recent preindex pregnancy and the corresponding index one. The right-hand two columns contain data only for pregnancies in which the same AEDs were taken in both pregnancies.

Retrospective pregnancies

Spontaneous or induced abortions occurred in 44.3% of the pregnancies of the 596 women who had been pregnant at least once before initial enrollment in the Register. Of the remaining pregnancies, fetal malformations had occurred in 4 (2.8%) of the 145 not exposed to AEDs, in 12 (10.1%) of the 119 exposed to AEDs but not VPA, and in 11 (16.2%) of the 68 whose AED exposure included VPA. Relative to the malformation rate in AED-unexposed pregnancies, there was a statistically significant increase in rates in all AED-exposed pregnancies (OR 4.94, 95% CI 1.7–14.6), in pregnancies exposed to AEDs other than VPA (OR 3.95, 95% CI 1.24–12.6), and in pregnancies exposed to VPA (OR 6.80, 95% CI 2.08–22.2). However, the apparently higher malformation rate in VPA-exposed pregnancies compared with those exposed only to other AEDs was not statistically significant (OR 1.72, 95% CI 0.71–4.14). The outcomes of the subsequent pregnancies (index ones) were followed for the immediate preindex pregnancies whose AED exposure involved the same agents in both pregnancies. The outcomes are shown in the lower half of Fig. 1. As in the corresponding situation for index and postindex pregnancies in women taking the same AEDs in both pregnancies (upper half of Fig. 1), malformation rates appeared higher in the second pregnancy when the first one had resulted in birth of a baby with a malformation (21.43% vs. 6.25% when the first pregnancy outcome was not malformed, OR 3.42, 95% CI 0.81–14.43). For VPA-exposed pregnancies compared with those exposed to other AEDs, the rates were 25.0% vs. 11.1%; OR 2.67, 95% CI 0.19–26.8). Neither of these higher rates was statistically significant.

Women with malformations in more than one pregnancy

There were 15 women who had two or more pregnancies that resulted in fetuses with malformations. In four, all the pregnancies were preindex; in six, one pregnancy was preindex and the other(s) index or postindex; in the remaining five, the pregnancies were all index or postindex. AED exposure was involved in all except two of the first preenrollment pregnancies in the individuals concerned. With these exceptions, and one other, there were no changes in the AEDs involved in each woman’s pregnancies. Data for the malformation patterns in the affected offspring of each woman involved are shown in Table 2. In most instances exposure to the same drug in a given woman’s pregnancies was not associated with the occurrence of the same type of malformation in her subsequent affected offspring.

Table 2.   Fetal malformations in women who had more than one pregnancy with fetal malformations
 Pregnancy no.First pregnancy malformationPregnancy no.Second pregnancy malformationPregnancy no.Third pregnancy malformationAED involved
  1. The pregnancy number is the number of the individual woman’s pregnancy, beginning with her first pregnancy ever, not her index pregnancy.

  2. CBZ, carbamazepine; LTG, lamotrigine; VPA, valproate.

  3. aSubject I also had three prior pregnancies, all terminated for “maternal indications.”

A1Overlapping toes2Hand-polydactyly  VPA, VPA
B2Spina bifida4Hypospadias  None, VPA
C1Transposed viscera; tracheoesophageal fistula; cardiac septal defects2Atrial septal defect  CBZ and LTG, VPA
D2Anencephaly6Down syndrome  CBZ, CBZ
E1Renal malformation4Plagiocephaly  VPA, VPA
F1Spina bifida5Club feet6Microcephaly, plagiocephalyVPA, VPA, VPA
G1Hypospadias; hypertelorism2Spina bifida; hydrocephalus  VPA, VPA
H1Bulbus cordis; ventricular septal defect2Cleft palate and lip  VPA, VPA
I4aTrigger thumb5Cleft palate; Pierre Robin malformation  VPA, VPA
J1Coronal synostosis2Cranial synostoses  CBZ, CBZ
K1Polydactyly2Polydactyly  VPA, VPA
L1Brachial cyst2Axillary sinus  CBZ, CBZ
M1Down syndrome; heart malformation4Cleft palate  VPA, VPA
N1Neural tube defect3Spina bifida  None, VPA
O1Fetal valproate syndrome2Termination for defects ?nature  VPA, VPA


The finding of the present analysis that there was a higher incidence of fetal malformation associated with intrauterine VPA exposure was not unexpected, having been found previously in many studies where each pregnancy was analyzed independently of knowledge of the mother’s previous pregnancy record, as in our previous publications (Vajda et al., 2004; Vajda & Eadie, 2005; Vajda et al., 2007a,b,c, 2011). However, the focus of this article was on the outcomes of serial pregnancies in women taking AED. The outcome of the analysis raises several important findings that warrant discussion.

First, women taking VPA in their index pregnancies that resulted in fetuses with malformations had a statistically significantly greater risk of having a malformed fetus in their next pregnancies if they continued to take VPA. The data for immediate preindex pregnancies followed into index ones showed a similar but not statistically significant trend. The latter data may be less reliable than the former, the preindex pregnancy information being obtained retrospectively and often being less easy to confirm from professional sources. Combining the two datasets did not greatly increase overall pregnancy numbers, if the same index pregnancy was not to be counted twice. Although the available numbers were small, the analysis of the data for the prospectively followed pregnancy pairs indicated that, if a woman taking VPA in one pregnancy gave birth to a malformed fetus, her absolute risk of having a malformed fetus in her next pregnancy if she continued to take the drug was >50%, with an odds ratio of >17, as compared with women continuing to take the drug whose index pregnancies produced nonmalformed fetuses. This increased risk was unlikely to be explained by VPA dose, as the mean dosage taken in the postindex pregnancy was not significantly different in women who had a previous fetal malformation that those who did not. There was also a trend to an increased risk of a malformed fetus in women who had a malformed fetus in previous pregnancies in which they took an AED other than VPA, although this trend was not as strong and did not attain statistical significance. Folic acid, being taken in 97% of the pregnancies resulting in malformed fetuses, is unlikely to have been a relevant factor in the outcomes. Of interest, if in two pregnancies, fetuses with malformations were born to the same mother, the patterns of malformation in the two fetuses often were not identical. These observations might be interpreted as indicating that there is an inherent individual vulnerability of some women to a nonspecific pro-teratogenic effect of AEDs, in particular VPA. This raises the possibility that there may be specific genetic characteristics in women, or in their partners, that render such women more likely to have fetuses with malformations if they are taking AEDs. This possibility was mentioned by Duncan et al. (2001) and warrants further specific investigation with pharmacogenomic studies.

The results of this study provide information that is relevant in the counseling of women with AED-treated epilepsy who are planning further pregnancies and indicate the importance of knowing the fetal outcomes of their previous pregnancies before providing advice. The present analysis indicates that women who have already had malformed fetuses while taking VPA are at substantial risk of having further malformed infants if they continue to take this drug. The risk may be less, although still present, if only AEDs apart from VPA continue to be taken in consecutive pregnancies, if the first has resulted in a malformed fetus. Even if the malformation in the earlier pregnancy was a relatively minor one, the experience of the present study indicates that any subsequent malformation may not necessarily also be minor.


We wish to acknowledge the help of our medical and nonmedical colleagues, both in referring patients and in increasing patient awareness of the Register. We thank the Scientific Advisory Board and the Ethical Research Committees of St. Vincent’s Hospital, Monash Medical Centre, the Royal Melbourne Hospital, and other institutions for their ethics assessments of the study. The Australian Register is grateful for support from the Epilepsy Society of Australia, Royal Melbourne Hospital Neuroscience Foundation, The Victorian Epilepsy Foundation, Epilepsy Australia, National Health and Medical Research Council linkage grant, and also for financial support from a number of pharmaceutical companies, including Sanofi-Aventis, UCB Pharma, Janssen-Cilag, Novartis, Sci-Gen, and Pfizer, as well as for past support from Glaxo.


None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.