• Epilepsy;
  • Pregnancy;
  • Antiepileptic drugs;
  • Teratogenicity;
  • Birth defects


  1. Top of page
  2. Abstract

Summary:  A rational approach to the treatment of women of childbearing potential with epilepsy has been hampered by the lack of conclusive data on the comparative teratogenic potential of different antiepileptic drugs (AEDs). Although, several cohort studies on birth defects associated with AED use during pregnancy have been published, these have generally failed to demonstrate differences in malformation rates between AEDs, probably mainly due to insufficient power. In particular, pregnancies with new generation AEDs have been too few. In recent years, pregnancy registries have been introduced to overcome this problem—EURAP (an international collaboration), the North American, and the U.K. AED and pregnancy registries are observational studies that prospectively assess pregnancy outcome after AED exposure using slightly different methods. Each has enlisted 3–5,000 pregnancies in women with epilepsy, and the North American and the U.K. have released preliminary observations. Thus the U.K. registry reported a higher malformation rate with valproate, 5.9% (4.3–8.2%; 95% CI), than with carbamazepine, 2.3% (1.4–3.7%), and lamotrigine, 2.1% (1.0–4.0%). Most of the more recent cohort studies have also identified a nonsignificant trend toward a higher teratogenicity with valproate. These signals need to be interpreted with some caution since none of the studies to date have fully assessed the impact of possible confounders, such as type of epilepsy, family history of birth defects, etc. However, with increasing number of pregnancies it should be possible in the near future for the pregnancy registries to take such confounding factors into account and thus make more reliable assessments of the causal relationship between exposure to specific AEDs and teratogenic risks. While awaiting more conclusive results, it appears reasonable to be cautious in prescribing valproate to women considering to become pregnant if other suitable treatment alternatives, and with less teratogenic potential, are available. Any attempt to change treatment should, however, be accomplished well before conception. The importance of maintained seizure control must also be kept in mind, and the woman who needs valproate to control her seizures should not be discouraged from pregnancy, provided that counseling at the best of available knowledge is given.

Optimal medical management during pregnancy is a major health issue for women with epilepsy and a challenge to their physicians. Since the first report on an association between antiepileptic drugs (AEDs) and birth defects was published in the 1960s (1), numerous studies have confirmed the developmental toxicity of many of these agents. Potential adverse effects include intrauterine growth retardation, major malformations, dysmorphisms, and postnatal developmental delay. Although teratogenicity and other adverse developmental effects are the major concerns, such risks must be weighed against the fetal and maternal risks associated with uncontrolled seizures during pregnancy. The prevailing treatment strategy is based on the assumption that seizures, especially convulsive seizures, are more harmful to the mother and to the fetus than are the drugs, although admittedly this assumption rests on circumstantial evidence and single-case observations rather than on systematic studies. The objective of the treatment is nevertheless to maintain control of seizures, in particular tonic–clonic seizures, throughout pregnancy by using AEDs in a way that minimizes adverse effects to the mother and the fetus. Several guidelines have been published to assist physicians in fulfilling these objectives. The Commission on Genetics, Pregnancy, and the Child of the International League Against Epilepsy (ILAE) developed guidelines for the care of women with epilepsy of childbearing age in 1989 (2), and an updated version of these guidelines appeared in 1993 (3). More evidence-based practice parameters and guidelines were subsequently developed by the American College of Obstetricians and Gynecologists (4), the American Academy of Neurology (5), the Women With Epilepsy Guidelines Development Group in the United Kingdom (6), the Scottish Intercollegiate Guideline Network (SIGN) (7), and the National Institute for Clinical Excellence in the United Kingdom (NICE) (8). Several other recommendations and guidelines have been published by different organizations and as a result of workshops and symposia (9,10).

Evidence-based guidelines conclude that no class I (highest level) evidence is available to guide the management of epilepsy during pregnancy (4–6), due to the absence of randomized controlled studies. However, because of the nature of the condition, it is evident that some of the most important issues in this area will never be addressed by randomized trials and will have to be investigated instead by well-designed large-scale observational studies. With respect to pharmacologic treatment, the essence of existing guidelines and recommendations can be summarized as follows:

  • • 
    Optimize treatment before conception
  • • 
    If AEDs are needed, use monotherapy
  • • 
    Chose the most effective AED for seizure type or syndrome
  • • 
    Use the lowest effective dosage

Most guidelines also include recommendations concerning prenatal diagnosis and drug-level monitoring during pregnancy, although specific recommendations vary somewhat. The same is true for vitamin K and, in particular, folate supplementation. For example, with respect to folate intake, advice ranges from the need to “secure adequate amounts of folate in the diet to the mother before conception (3) to the recommendation to “give folate supplementation of at least 0.4 mg per day” (5) or even prescribe “5 mg of folate per day” (6–8).

Although these recommendations reflect the level of knowledge at the time the guidelines were produced, they are of little help for the physician who needs to counsel a woman with epilepsy considering pregnancy. Whereas the NICE guidelines advise specific caution in the use of valproate (VPA) (8), unanswered questions still exist as to which AED is best for the individual patient, how to dose it, and what additional measures could be taken to minimize fetal and maternal risks. More specifically, physicians need to know which AED is most likely to be effective and has the least probability of causing harm. Should a risk of fetal toxicity be associated with the drug, it is important to know the type and pattern, and any relation to the dose or serum drug-level profile. It also would be useful to know to what extent the pharmacokinetics of the drug in question is affected by pregnancy and whether this could be controlled by drug-level monitoring. Pharmacokinetics in the newborn and transfer of the drug into breast milk are further issues of importance for counselling. Although some of this information is available for old AEDs, knowledge is incomplete, and even more so for the new-generation drugs. Most important, we lack conclusive data on the comparative human developmental toxicity associated with different AEDs. This is disappointing, considering that the teratogenic effects of AEDs have been the subject of intense clinical research since the 1960s, but can be explained by the methodologic shortcomings and, in particular, the insufficient power of previous studies.

Several cohort studies on birth defects associated with AED use during pregnancy have been published since the compilation of most of the guidelines quoted earlier (11–19). The majority of these are prospective (12–15,17–19) and, with the exception of two (14,17), multicenter studies. Although all studies are from the same period, reported major malformation rates with monotherapy exposure vary markedly between studies, from 3.3% to 10.5% (Table 1). This is probably a reflection of differences in study populations, outcome criteria, and assessment methods and illustrates the difficulties in comparing malformation rates for specific treatments between studies or, indeed, in pooling data. Rates of major malformations for individual AEDs also differ considerably between studies. As an example, for phenytoin (PHT), it may range from 0.7% (0.02–3.6%; 95% CI) in one study (12) to 9.1% (4.8–15.3%) in another (15), and for carbamazepine (CBZ), from 2.8% (1.3–5.0%) (17) to 7.9% (5.0–11.7%) (11).

Table 1. Malformation rates in offspring exposed to monotherapy with antiepileptic drugs in utero in some recent cohort studies
ReferenceCountryMonotherapy exposures (n)Malformation rate (%, 95% CI)
  1. CI, confidence interval.

Samrén et al., 1997Germany, Finland, Netherlands709 8.0 (6.0–10.1)
Canger et al., 1999Italy31310.5 (7.4–14.5)
Kaneko et al., 1999Canada, Italy, Japan500 7.8 (5.6–10.5)
Samrén et al., 1999Netherlands8993.3 (2.3–4.7)
Holmes et al., 2001USA2234.5 (2.2–8.1)
Kaaja et al., 2003Finland5943.3 (2.3–4.7)
Vajda et al., 2003Australia206 8.7 (5.3–13.5)
Sabers et al., 2004Denmark1093.7 (1.0–9.1)

Although each study represents a major effort and provides valuable data, each lacks the power to demonstrate differences in teratogenic potential between the various treatments. Given the large number of different AEDs and the numerous different drug combinations, the number of patients on each individual treatment regimen is far too small for a meaningful internal comparison. Hence it is not surprising that, although the studies demonstrate beyond doubt an increased risk for birth defects with use of AEDs in general, they all fail to detect differences in birth-defect rates among different treatment regimens. Wide et al. (20) used a different method in a recently published study from Sweden. This nationwide population-based survey used the Swedish Medical Birth Registry to identify all infants exposed to AEDs in utero and born between July 1995 and December 2001. The occurrence of congenital malformations was assessed by record linkage to other registries, including the Swedish Registry of Congenital Malformations. The rate of more severe malformations among 1,256 monotherapy exposures was 5.4% (4.2–6.8%) and higher after exposure to VPA, 9.7% (6.4–13.9%) compared with CBZ, 4.0% (2.7–5.7%), the two most frequently used AEDs. The odds ratio for severe malformations with VPA versus CBZ was 2.6 (1.4–4.7). Whereas the Swedish Medical Birth Registry has the advantage of being population based, it unfortunately does not include details on drug dosage or information on the indication for treatment, such as type of epilepsy.

The complexity and importance of this issue has been reinforced by the introduction of the new-generation AEDs and the need to assess their teratogenic potential. To address these concerns, pregnancy registries have been established for the prospective assessment of birth defects associated with specific AEDs. Some pharmaceutical companies have set up registries collecting information on the company's own products (21). Although important, the value of these registries is limited by the lack of an internal control, which is best provided by comparison with other AEDs. In contrast, non–company-based registries include information on all AEDs in use. The North American AED Pregnancy Registry (NAREP) (22) and the United Kingdom Epilepsy and Pregnancy Registry (23) were launched in 1996 to 1997. EURAP (24), an international registry that originated in Europe in 1999, now includes collaborators from 40 countries from Europe, Asia, Australia, and South America. With somewhat different methods, these registries continue to enroll pregnancies at increasing rates and have each enlisted 3–5,000 pregnancies in women with epilepsy. The North American and the U.K. registries have already published some preliminary observations in abstract form. NAREP discloses malformation rates associated with specific treatments as soon as they are found to differ significantly from the rate in an unexposed control group, and thus far, a significant elevation in risk has been reported with exposure to phenobarbital (PB; 12.5% malformation rate; relative risk, 5.6, 2.1–11.0, n = 40) (25) or VPA (8.9%; relative risk, 6.0, 3.5–10.2; n = 123) (26), although no comparisons were made between individual AEDs. The U.K. registry reported major malformation rates of 2.3% (1.4–3.7%) for CBZ (n = 700), 5.9% (4.3–8.2%) for VPA (n = 572), and 2.1% (1.0–4.0%) for lamotrigine (LTG; n = 390), all in monotherapy (23). The preliminary reports from the U.K. and from the Swedish registries both suggest a higher risk for birth defects with VPA compared with that with CBZ. At this stage, these results should still be interpreted with caution and assessed against the background of a long series of studies that failed to demonstrate differences among AEDs. It should, however, be pointed out that most of the recent large-scale prospective studies, despite their relatively low power, did identify a trend toward a higher teratogenicity of VPA over other commonly used AEDs (12,14,15,17,20) (Fig. 1). Overwhelming signals point toward a greater teratogenicity risk with AED polytherapies compared with single-AED therapy (12,13,15–17,20), although in principle, each specific drug combination should be assessed individually, because they are likely to differ in teratogenic potential.


Figure 1. A within-study comparison of malformation rates for different antiepileptic drugs (AEDs) in monotherapy, based on data from seven published studies. Odds ratio and 95% confidence intervals for malformations were calculated for individual AEDs by using the malformation rate reported for carbamazepine in each study as reference (odds ratio, 1). Valproate is marked with open symbols. PB, Phenobarbital; PHT, phenytoin; VPA, valproate; LTG, lamotrigine.

Download figure to PowerPoint

A major limitation of all studies reported to date, including registry data, is that none allowed assessment of the impact of possible confounders, such as type of epilepsy, seizure frequency, family history of birth defects, and exposure to additional risk factors. For example, evidence has been presented that the teratogenicity of VPA can be under genetic control, and that at least in some VPA-treated women, the history of a pregnancy resulting in a neural tube defect may predict a huge risk of further birth defects in subsequent pregnancies (27,28). What impact the removal of pregnancies with a previous history of birth defects would have on VPA-associated teratogenic risk, as reported preliminarily in registry-based studies, is unknown. In observational studies, women are not randomized to different AEDs, and the selection of a particular AED and its dosage will depend on individual variables such as type of epilepsy and seizure frequency, which theoretically could be linked to a higher or lower risk of malformations.

Ongoing pregnancy registries, as indeed most previous studies, have focused on the risk of major congenital malformations. In recent years, attention has been increasingly paid to the possibility that AED exposure during pregnancy also may affect adversely postnatal psychomotor development of the offspring, and the suggestion has been made that this effect may not be fully apparent until school age. This issue has been addressed in a number of small-scale studies with pronounced differences in methods and outcome, and with only a few being population based (29–31). A recent retrospective questionnaire-based survey attracted particular attention because of the concerns it raised (32). This study reported that children exposed to VPA in utero encountered learning problems at school more often than did children exposed to other AEDs such as CBZ. However, as indeed underlined by the authors, these results should be interpreted with caution in view of the relatively low response rate (only 57% of the women returned adequately completed questionnaires) and the retrospective method. Unfortunately, prospective studies have included very few children exposed to VPA: the most recent prospective, controlled, population-based study of intelligence after prenatal exposure to AEDs included only 13 children exposed to VPA monotherapy compared with 86 children exposed to CBZ (31).

The major problem in interpreting observational studies on postnatal neurobehavioral development as well as structural birth defects is to evaluate causality. Although linking adverse outcomes to a specific AED requires exclusion of alternative explanations, many confounding factors could influence outcome of pregnancy in a woman with epilepsy. Maternal age, socioeconomic conditions, parental educational status, nutritional status, genetic predisposition for birth defects or developmental disorders, comorbidities, type of epilepsy, seizure control during pregnancy, and exposure to other potential teratogens are examples of such factors. It is difficult to dissect the contribution of drug treatment if, as may well be the case, the choice of a particular AED is selectively influenced by any of these independent potential risk factors. Large numbers of pregnancies, as well as prospective and reliable recording of critical information, are necessary to control for major confounding variables. This is why signals on possible associations between exposure to a specific AED and adverse outcome must be confirmed in independent studies and analyzed in depth, taking into account other potential explanations for the observed associations. Although differences in methods hamper the possibility to pool data from different registries, the existence of independent studies provides an opportunity to confirm or refute such signals.

Despite the signals discussed, more data are necessary to draw evidence-based recommendations for selection of AEDs for women with epilepsy of childbearing potential. With continued support to the established pregnancy registries, more solid comparative data on the most frequently used monotherapy regimens will surely become available within a few years, possibly including comparisons of risks for specific birth defects rather than general malformation rates. Data emerging from these registries may also shed some light on risks associated with specific AED combinations and on the influence of other potentially important prognostic variables, including folate supplementation. Some of the registries, as well as independent investigators, also initiated extended prospective follow-up of exposed children, which eventually will provide better comparative data on potential adverse effects on psychomotor development.

The critical question at present is how patients and their physicians should react to the available signals while awaiting more conclusive results. Signals of potential teratogenic drug effects tend to attract the interest of the media and may therefore be brought to the attention of the general public, often with nonscientific journalistic style, which indeed has been the case with VPA in some countries. The problem may be compounded, both with the lay public and the medical profession, by dissemination of unbalanced messages associated with drug promotion in an increasingly competitive marketplace. All this attention will increase the pressure on the woman as well as on her physician and may result in irrational decisions and potentially harmful abrupt drug discontinuations or treatment changes during pregnancy. The importance of maintained seizure control for the well-being of women with epilepsy, as well as for their unborn children, must be kept in mind. The woman who needs VPA to control her seizures should thus continue taking the drug and not necessarily be discouraged from pregnancy, provided that counseling with the best of available knowledge is given. If other suitable treatment alternatives for the patient's epilepsy are available, and with less teratogenic potential, they should be considered. Any attempt to change treatment should, however, be accomplished well before conception and never after pregnancy has started.


  1. Top of page
  2. Abstract