Associations between particular types of fetal malformation and antiepileptic drug exposure in utero


F. J. E. Vajda, Department of Medicine and Neurosciences, Royal Melbourne Hospital and University of Melbourne, Parkville, Vic. 3050, Australia

Tel.: 61(3) 98193056

Fax: 61(3) 93422577




To study associations between patterns of fetal malformation and individual antiepileptic drugs taken during pregnancy.


Multiple variable logistic regression and other statistical analyses of data relating to 1733 fetuses from 1703 pregnancies (147 of which were not exposed to antiepileptic drugs during pregnancy).


There were statistically significant (P < 0.05) associations between (i) valproate exposure and spina bifida, malformations of the heart and great vessels, digits, skull bones, and brain, but not hypospadias, cleft palate/lip and mouth abnormalities, (ii) topiramate exposure and hypospadias and brain maldevelopments, and (iii) carbamazepine (CBZ) exposure and renal tract abnormalities.


The valproate findings are mostly in keeping with the published literature, but the topiramate finding regarding hypospadias and the association between CBZ exposure and various renal tract abnormalities raise questions of organ specific teratogenesis. More extensive data are desirable, particularly in relation to topiramate, which is being used increasingly as a migraine prophylactic in women of childbearing potential.


There is now substantial statistical evidence that the antiepileptic drug (AED) sodium valproate (VPA) is a dose-related teratogen and some evidence that intrauterine exposure to other marketed AEDs may also be associated with teratogenesis. Part of the evidence relating to VPA has been derived from work based on the Australian Pregnancy Register. The main focus in most of the relevant literature has been the overall teratogenic risk associated with particular AEDs and whether this risk is statistically significantly increased. The extent to which particular AEDs are associated with specific patterns of fetal malformation is less clear, due to some inconsistency between the outcomes of various studies.

Syndromes associated with intrauterine AED exposure have been described, such as the fetal hydantoin, carbamazepine (CBZ), and valproate syndromes [1, 2], but their reality is open to criticism [3]. We used a statistical approach to analyze data from Australian Pregnancy Register, looking for information to indicate whether valproate, or other AEDs, may be associated with specific teratogenic patterns.

Materials and methods

The Australian Register of Antiepileptic Drugs in Pregnancy has been collecting information concerning the pregnancies and fetal outcomes for women with epilepsy, whether or not they took AEDs in at least the first trimester of pregnancy, and also for pregnant women with disorders other than epilepsy who took AEDs to treat these disorders. Recruitment has been Australia-wide and entirely voluntary.

Details of recruitment practices and of data collection, storage, confidentiality, ethics oversight, and development of the areas under investigation by the Register have been published [4-6]. The fetal malformation classification used was that of the Birth Defect Registry of Victoria [7].

The presence of fetal malformations was determined on the basis of the information available at the ends of the first postnatal month and of the first postnatal year. However, some 190 women included in the analysis were lost to follow-up during the first postpartum year, and 133 had not reached the end of that year at the time of the present data analysis. Therefore, the malformation rates here reported are likely to be a little lower than the true ones, although it is unlikely that the more significant malformations will not have been included because they will nearly always be obvious at or soon after birth.

At the time of the current analysis, the Register contains details of 1703 prospectively enrolled and completed pregnancies. Almost all (~98%) of the pregnancies were in women with epileptic seizure disorders, but 147 had taken no AEDs in at least the first trimester of pregnancy. There had been previous pregnancies in 963 of the 1703 women (56.5%). The 1703 pregnancies involved 30 sets of twins so that there were details available of 1733 pregnancy offspring. All subsequent analysis has been based on numbers of offspring of pregnancies, not numbers of pregnancies.


Malformed pregnancy rates

Fetal malformations occurred in 109 of the 1733 pregnancy offspring (6.3%). Malformations were present in five of the 147 offspring not exposed to AEDs in at least the first trimester of pregnancy (3.4%), with malformations in 104 of the remaining 1586 (6.6%) first-trimester AED-exposed pregnancies: odds ratio (OR) = 0.50 (0.2, 1.25). Further details of the malformation rates related to intrauterine exposure to the more commonly used AEDs employed in monotherapy are shown in Table 1, together with total exposures to each AED, alone or in combination with others.

Table 1. Exposure to individual AEDs, occurring in >10 pregnancies, and associated malformation rates for the drugs used in monotherapy
AEDAED total exposuresAED in monotherapyMalformations% Malformed
  1. AED, antiepileptic drug; CBZ, carbamazepine; CZP, clonazepam; ETHO, ethosuximide; GPT, gabapentin; LEV, levetiracetam; LTG, lamotrigine; OxCBZ, oxcarbazepine; PB/PMD, phenobarbitone, primidone; PHT, phenytoin; TPM, topiramate; VGT, vigabatrin; VPA, valproate.

No AEDs14714753.4

The malformation rates for all the drugs collectively, used in monotherapy (72.3% of all AED exposures) and in AED combinations (27.7% of all AED exposures), were relatively similar, being for monotherapy 72 of 1147 = 6.28% and for polytherapy 32 of 439 = 7.29%; OR = 0.85 (0.56, 1.31). The malformation rate associated with VPA exposure in monotherapy was statistically significantly higher that associated with all the other AEDs in monotherapy, 37 of 271 = 13.65% vs 35 of 876 = 4.00%; OR = 3.80 (2.34, 6.17). In contrast, the malformation rate associated with exposure to all non-VPA AEDs in monotherapy (4.00%) was not statistically significantly greater than the rate in pregnancies unexposed to AEDs in at least the first trimester, viz. 5 of 147 = 3.40%; OR = 0.85 (0.33, 2.20).

Relationship between various parameters and malformed fetus rates

Because restricting analysis to pregnancy offspring exposed to only a single AED appreciably reduced the amount of information available, multivariate logistic regression was applied to the combination of AED monotherapy and polytherapy data in subsequent analyses in this article.

The contributions of various recorded data items to the risk of a malformed fetus occurring were assessed, Table 2 showing the various regression partial coefficients and their P values. The only statistically significant regression coefficients (P < 0.05) were those for having (i) siblings with malformations and (ii) exposure during pregnancy to (a) VPA and (b) topiramate (TPM). The association with the presence of malformations in siblings, but not with malformations in other family members, is probably related to maternal AED exposure in earlier pregnancies in the same woman and not necessarily to genetic factors.

Table 2. Coefficients of a multiple variable logistic regression for risk of a malformed fetus on various data items and the corresponding P values. Statistically significant P values are shown in bold type
 Value P
  1. AED, antiepileptic drug; CBZ, carbamazepine; CZP, clonazepam; ETHO, ethosuximide; GPT, gabapentin; LEV, levetiracetam; LTG, lamotrigine; OxCBZ, oxcarbazepine; PB/PMD, phenobarbitone, primidone; PHT, phenytoin; TPM, topiramate; VGT, vigabatrin; VPA, valproate.

Family history of malformations, excluding siblings−0.6669710.5048
Siblings with malformations+1.006052 0.0144
Exposure to CBZ+0.2179830.4275
Exposure to LTG−0.1483830.5622
Exposure to VPA+1.173438 <0.0001
Exposure to LEV−0.3950080.4105
Exposure to TPM+0.746239 0.0341
Exposure to PHT+0.3082610.5338
Exposure to CZP−0.5452520.3021
Exposure to GPT−15.2848230.9918
Exposure to OxCBZ−14.7578670.9931
Exposure to PB or PMD+0.2451570.8153
Exposure to ETHO−14.9614950.9933
Exposure to VGT−14.7478800.9942
Folate intake before conception−0.1327180.5573

In view of the relative paucity of data and high P values, exposure to the less frequently used AEDs was not pursued further.

Malformations associated with particular AEDs

The recorded malformations in the 104 malformed fetuses among the 1586 exposed to AEDs during pregnancy, and in the 147 not so exposed, are set down in Table 3. Because multiple malformations were often present in the same fetus, the total of individual malformations exceeds the total number of fetuses. In the Table, the malformations are often classified in terms of the affected regions of the body—dealing with the matter at the level of the numerous individual malformations within regions would have often resulted in numbers too small to be useful. Drugs not associated with malformations are not included in the Table.

Table 3. Fetal malformations and the associated AEDs. Numbers of AEDs associated with individual malformations may exceed the numbers of the malformation because a malformation may occur in a fetus exposed to more than one AED
  1. AED, antiepileptic drug; CBZ, carbamazepine; CZP, clonazepam; ETHO, ethosuximide; GPT, gabapentin; LEV, levetiracetam; LTG, lamotrigine; OxCBZ, oxcarbazepine; PB/PMD, phenobarbitone, primidone; PHT, phenytoin; TPM, topiramate; VGT, vigabatrin; VPA, valproate.

Total exposed158651954244715111784147
Total with malformations10428275451155
Spina bifida102180100
Sacral groove42020000
Heart/great vessels2639151211
Urinary tract118132200

Logistic regression was used to explore relationships between exposure to individual drugs and the more frequently encountered individual malformations (Table 4). There appeared to be statistically significant associations between (i) VPA and spina bifida, malformations of the heart and great vessels, digits, skull bones, and brain; (ii) CBZ and a number of different urinary tract abnormalities; and (iii) TPM and hypospadias and various abnormalities of the brain. While spina bifida and hypospadias are discrete abnormalities, the other categories of abnormality that appeared associated with particular AEDs involved a mix of individual types of malformation, none occurring in any great number. Among the 15 fetuses with malformations of the heart or great vessels associated with VPA exposure, there were 12 instances of defects in the interatrial or interventricular septa, but logistic regression (details not shown) found the association with the drug was significant only at a = 0.0693 level.

Table 4. Logistic regressions for individual malformations of the form: logit risk = a + b1 + ….. bn, where b values represent exposures to particular AEDs. Statistically significant P values are shown in bold type
  1. AED, antiepileptic drug; CBZ, carbamazepine; CZP, clonazepam; ETHO, ethosuximide; GPT, gabapentin; LEV, levetiracetam; LTG, lamotrigine; OxCBZ, oxcarbazepine; PB/PMD, phenobarbitone, primidone; PHT, phenytoin; TPM, topiramate; VGT, vigabatrin; VPA, valproate.

Spina bifida
Coefficient−6.328607+0.46478−1.10845+2.511267−18.160115+1.013949−17.605609−17. 87458
P value<0.00010.62190.3013 0.0075 0.99880.34980.99900.9986
P value<0.00010.22880.8615 0.0142 0.47890.68480.91020.7386
P value<0.00010.29250.19290.63980.5264 0.0186 0.92300.9291
Urinary tract
P value<0.0001 0.0045 0.58960.12120.19260.07060.99840.9980
P value<0.00010.26530.09390.01840.86570.99920.99930.9991
Skull bones
P value<0.00010.29170.1806 0.0079 0.87060.77160.37180.9303
P value<0.00010.64190.7949 0.0060 0.9034 0.0297 0.99640.998
P value<0.00010.99940.45860.15470.99950.99950.99960.9994
P value<0.00010.99960.99960.44070.99960.99970.41550.9996

For the VPA-associated abnormalities of digits (= 12), skull bones (= 8), brain (= 10), and face (= 5), the most common ones were, respectively, extra digits (= 7), plagiocephaly (= 4), and hydrocephalus and Arnold-Chiari malformations (= 5).

No single renal tract abnormality (= 8) was associated with CBZ more than twice, and the three brain abnormalities associated with TPM exposure were all different. One may suspect that, with more data, statistically significant evidence of an association between VPA and defects in the cardiac septa might emerge, but with the arguable exception of hypospadias, there are too few data to warrant undue confidence in accepting associations between CBZ and individual renal tract malformations and between TPM and the various brain malformations found.

The statistically significant association between TPM exposure and hypospadias in the full population studied may partly explain why there were 54 male offspring, but only 41 female offspring with malformations (the sexes of some aborted fetuses were not recorded, although the natures of their malformations were). Logistic regression analysis was applied only to male offspring of pregnancy (i.e. a baby boy), relating the presence of hypospadias to (i) exposure to the individual commonly used AEDs, and (ii) exposure to the other AEDs but with the actual TPM dosages employed (Table 5; Fig. 1). There were statistically significant relationships between the occurrence of hypospadias and exposure to TPM, and also to TPM dosage (150, 200, 300, and 400 mg per day in the affected pregnancies), the P value being smaller in case of the latter association (0.0192 vs 0.0105).

Table 5. Logistic regressions for presence of hypospadias of the form: logit risk = a + b1 + …, where b values represent exposures to particular AEDs in the upper equation, but TPM dose (mg per day) rather than TPM exposure is shown in the lower equation. Statistically significant P values are shown in bold type
  1. AED, antiepileptic drug; CBZ, carbamazepine; CZP, clonazepam; ETHO, ethosuximide; GPT, gabapentin; LEV, levetiracetam; LTG, lamotrigine; OxCBZ, oxcarbazepine; PB/PMD, phenobarbitone, primidone; PHT, phenytoin; TPM, topiramate; VGT, vigabatrin; VPA, valproate.

Coefficient−3.65903−0.70698−0.970817+0.381125−0.905181+1.478844 −0.152876−0.1899946
P value<0.00010.32240.14870.52260.3938 0.0192  0.88740.8568
Coefficient−3.53772−0.839868−1.072059+0.312377−1.10612 +0.005135−0.248607−0.147308
P value<0.00010.23730.11380.59960.3161  0.0105 0.81850.8888
Figure 1.

Calculated regression for risk of hypospadias relative to topiramate dose, with the upper 95% confidence limit of the regression line.

If the data are approached in another way, there were four instances of hypospadias among the 60 male infants exposed to TPM during pregnancy (6.67%) and 10 in the 681 males exposed to AEDs apart from TPM (1.47%; OR = 4.79; 95% CI = 1.46, 15.77). However, the rates of occurrence of hypospadias in TPM-exposed fetuses (6.67%) were not statistically significantly higher than those in the males not exposed to AEDs (3 of 65, i.e. 4.62%: OR = 1.48, 95% CI = 0.32, 6.89). It has been shown that if a woman taking AEDs gives birth to a malformed fetus, she is at increased risk of having malformed fetuses in subsequent pregnancies [8, 9]. Because of the possibility that this effect could have influenced the hypospadias findings, the comparisons made immediately above were repeated after excluding pregnancies in which there was a history of fetal malformations in siblings or other family members. The hypospadias rates were then three in 39 (7.70%) for the TPM-exposed pregnancies and eight in 503 (1.59%) for pregnancies exposed to other AEDs (OR = 5.16; 95% CI = 1.31, 20.3), and for the pregnancies not exposed to AEDs, three in 65, that is, 4.62% (OR = 1.72; 95% CI = 0.38, 8.99).


The value of conclusions drawn from the present analysis is inevitably limited by the relatively small numbers of instances of individual malformations that were available for study, and studies of larger populations might produce different outcomes. The difficulty of obtaining well-matched control populations and limitations of size of malformed fetal collections beset much of the published work in the area. There is a peculiarity in the AED-unexposed control data in the present study, in that three of the four male fetuses with malformations had hypospadias. This is a surprisingly high proportion and leads to concern about using these particular data for comparison purposes. Multivariable logistic regression has been used in the present study to increase the number of subjects involved by including fetuses exposed to AED polytherapy as well as AED monotherapy. Only statistically significant (P < 0.05) findings from the investigation are discussed below.

The present analysis has provided evidence that intrauterine exposure to VPA is associated by more than chance with certain maldevelopments in various parts of the fetal body. The association between VPA and spina bifida was noted in the early 1980s [10], but the subsequent literature has tended to focus on overall rates of fetal maldevelopment, rather than on particular types of malformations. Associations between the drug and various malformations have been reported in some, but not all, studies, and the outcomes of a number of the larger studies have been pooled in a meta-analysis by Jentink et al. [11]. That study concluded that there was statistically significant evidence of associations between VPA exposure and spina bifida, atrial septal defect, cleft palate, hypospadias, polydactyly, and craniosynostosis. Although some of the malformation categories were classified differently, the present study also found similar associations except in relation to cleft palate and hypospadias. It appears likely that VPA-related teratogenesis may involve a degree of organ- or tissue-related selectivity, though not the more limited organ or tissue specificity associated with thalidomide exposure. VPA forms numerous biotransformation products in the human body, by virtue of various metabolic pathways, for example, glucuronidation, ß-oxidation, ώ and ώ-1 oxidations, etc. [12]. At least one metabolite, 4-en-VPA, produces reactive, potentially toxic derivatives [13]. It is possible that some of the various VPA metabolites are teratogenic, with specificities for particular tissue elements, and that this explains the range of teratogenicities related to maternal VPA intake, rather than VPA itself being a rather tissue-promiscuous teratogen.

The associations found between TPM exposure and the occurrence of hypospadias and various brain maldevelopments were not anticipated on the basis of the monotherapy data, which suggest that TPM monotherapy is associated with a low rate of overall teratogenesis. With the larger population exposed to the drug in mono- and polytherapy, statistically significant associations emerged, and the association in relation to hypospadias appeared to be dose-related. Some previous studies have failed to find statistically significant evidence of TPM teratogenesis overall [14, 15]. Hunt et al. [16], in an analysis of the UK Pregnancy Register data, also found statistically significant evidence that TPM exposure was associated with increased fetal malformation rates, but their rates for oral clefts (2.2%) did not quite reach a P < 0.05 value [a point brought out by Fountain [17]]. However, subsequent publications have shown statistically significant associations between cleft lip/palate and TPM exposure [18, 19]. Hunt et al. [16] had found a 5.1% hypospadias rate associated with TPM exposure in their study. While this was not statistically significant, they pointed out that the rate was some 14 times the expected background rate of 1:300 and that the then available topiramate product information mentioned that instances of hypospadias had been reported to the drug company marketing the agent. Hernandez-Diaz et al. [20] noted a hypospadias occurrence rate of 1.1% in TPM monotherapy–exposed pregnancies, which was not statistically significant. Although the available evidence for an association between intrauterine TPM exposure and hypospadias is based on relatively small numbers of instances and needs to be confirmed in larger studies before it can be regarded as persuasive, it does raise concerns about the possible use of the drug during pregnancy, particularly when it is becoming more widely employed as a migraine preventative. In relation to migraine, Green et al. [21] went into the matter of the risk of oral clefts and other major congenital malformations and concluded that TPM was no worse than other AEDs in these regards. The data of Fig. 1, if supported by larger studies, would suggest that the hazard of hypospadias at the relatively low doses used in migraine prophylaxis, often 50 mg a day, may not necessarily produce prohibitive hazards of hypospadias, but with the additional emerging evidence of associations between TPM and oral clefts, those who treat migraine in women of childbearing potential prescribers need to be aware of the possibilities.

While there is published evidence that CBZ is a human teratogen [22, 23], no statistically significant evidence of a relationship between exposure to the drug and fetal malformations overall was obtained in the present study. The statistically significant association between CBZ and various renal tract abnormalities that was found was therefore unexpected. It should be noted that numbers of the individual abnormalities recorded were small.

The present investigation lends support to previously published data showing that VPA is a teratogen with a rather wide but not ubiquitous range of tissue-affecting capacity. The study also suggests that TPM is a human teratogen, with a possible proclivity to produce hypospadias, although larger-scale investigations are needed before this matter can be taken as established.


We wish to acknowledge the support of our colleagues, medical and non-medical, both in referring patients and 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 assessments of the study. The Australian Register is indebted for support to the Epilepsy Society of Australia, The Victorian Epilepsy Foundation, Epilepsy Australia, and Royal Melbourne Hospital Neuroscience Foundation and also for generous financial support from the pharmaceutical industry, including Sanofi-Aventis, UCB Pharma, Janssen-Cilag, Novartis, and SciGen.

Conflict of interest

Only the corresponding author received travel grants from sanofi and SciGen. All other co authors have declared no conflict of interest.