Over the past two decades advances in the management of epilepsy and an increased armamentarium of effective antiepilepsy drugs (AEDs) have permitted persons with epilepsy greater freedom and improved quality of life. Women with epilepsy who in past generations were actively discouraged from marriage and childbearing are now marrying at rates equivalent to women without this disorder. This freedom has presented these patients and their physicians with new challenges. Women with epilepsy are unlikely to be able to discontinue medication during pregnancy and as such are at greater risk for complications of pregnancy and adverse pregnancy outcomes. An increased risk of congenital malformations is well recognized as one of these, but mounting evidence suggests that cognitive dysfunction is also seen with increased frequency.

Major efforts have been made to assess the risk of AED use in pregnancy to minimize the risk to both the mother and the child. In the last few years new data have become available on the AED effects on postnatal development. Therefore, although some critical issues remain unsolved, it is important to review the state of knowledge of reproductive and developmental hazards related to AEDs.

An increase in seizure frequency during pregnancy has been reported in approximately 30% of women, which potentiates the risk of maternal complications and perhaps increases the risk of fetal insult (resulting from bradycardia). This information alone precludes discontinuation of antiepileptic medications during pregnancy.

The first report of a malformation associated with an AED described a child exposed in utero to mephenytoin, who developed microcephaly, cleft palate, malrotation of the intestine, and a speech defect, and had an IQ of 60.¹⁸⁶ The pregnancy was also complicated by vaginal bleeding.

In 1964 Janz and Fuchs¹²⁰ performed a retrospective survey at the University of Heidelberg to evaluate the problem of AED-associated malformations. Their study included 426 pregnancies in 246 mothers with epilepsy. The rates of miscarriages and stillbirths were increased for these patients, but the malformation rate was only 2.2%, not significantly different from that of the general population of West Germany. The authors concluded that AEDs were not associated with an increased risk of malformations.

Pantarotto et al.²⁰⁸ described a neonate with aplasia of the bone marrow after phenytoin exposure in utero. Centa and Rasore-Quartino³⁴ reported the first case of congenital heart disease following intrauterine exposure to phenytoin and phenobarbital. Melchior et al.¹⁷⁶ described orofacial clefts with exposure to primidone or phenobarbital.

In a letter in Lancet in 1968, S.R. Meadow reported six cases of children with orofacial clefts, four of whom had additional abnormalities of the heart and dysmorphic facial features. All of these children had been exposed to AEDs in utero. He noted that similar abnormalities had been reported following the unsuccessful use of abortifacient folic acid antagonists. Since some AEDs act as folic acid antagonists, he postulated that this might account for AED teratogenicity, and he asked other clinicians to inform him of similar cases.¹⁷³

The first report of malformations associated with a specific AED was published in 1970.⁸⁸ Trimethadione was implicated as a teratogen in 8 of 14 pregnancies in which it was taken in the first trimester.

Dr. Meadow’s 1968 inquiry concerning AED-associated malformations resulted in the collection of 30 additional cases. This prompted a retrospective survey in which 427 pregnancies in 186 women with epilepsy were studied. This demonstrated for the first time a clear increase in the malformation rates of infants of mothers with epilepsy (IMEs). Speidel and Meadow concluded that (a) congenital malformations are twice as common in IMEs exposed to AEDs; (b) no single abnormality was specific for AED exposure; and (c) a group of these children would have a characteristic pattern of anomalies, which, at its fullest expression, consisted of trigonocephaly, microcephaly, hypertelorism, low-set ears, short neck, transverse palmar creases, and minor skeletal abnormalities.

During the last two decades, a wide variety of congenital malformations have been reported, and every anticonvulsant, whether old or new, has been implicated as a cause. No anticonvulsant drug can be considered absolutely safe in pregnancy, yet most of these drugs do not produce any specific pattern of major malformations. A possible exception to this is the association of sodium valproate with neural tube defects (NTDs). Robert and Guibaud²¹⁹ were the first to relate valproate to NTDs. Working in a birth defects registry in the Rhone Alps region of France, they reported NTDs in IMEs exposed to valproic acid in utero. Between August 1979 and August 1982, there were 72 infants with lumbosacral NTDs born in this region. Nine of the 72 (12.5%) had been exposed to valproic acid, although two of the nine also had a family history of NTDs. The mothers of five of these children took valproate monotherapy, one took valproate and phenobarbital, and one took valproate and clonazepam.

In an attempt to clarify this observation further, Robert et al.²²⁰ identified 141 pregnancies in women with epilepsy using a combination of questionnaires and electroencephalogram (EEG) registries. This unselected cohort had a malformation rate of 17.7% in the offspring, including four cases of spina bifida. In all cases, there was intrauterine exposure to valproic
acid, although only one woman was on valproate monotherapy.

Subsequent analysis of these exposures identified spina bifida aperta as the specific NTD associated with the valproic acid exposure.¹⁵⁸ Methodologic problems make frequency estimates imprecise, because most published data are case reports, case series, or very small cohorts from registries that were not designed to evaluate pregnancy outcomes. The prevalence of spina bifida with valproate exposure is approximately 1% to 2%,¹⁵⁷ and with carbamazepine 0.5%.^103,223 However, a prospective study in the Netherlands found that IMEs exposed to valproate had a 5.4% prevalence rate of spina bifida. Average daily valproate doses were higher in the IMEs with spina bifida (1,640 ± 136 mg/day) than in the unaffected IMEs (941 ± 48 mg/day). Another group of investigators found that valproate doses of 1,000 mg/day or plasma concentrations of 70 μg/mL or less are unlikely to cause malformations.¹²⁸ Both groups recommend that valproate dosage be reduced whenever valproate must be used in pregnancy.^128,204

Overall, data available in the literature suggest that IMEs who have intrauterine exposure to anticonvulsant drugs are twice as likely to develop birth defects as infants not exposed to these drugs. Table 1 shows the incidence of malformations reported by historical cohort and prospective and retrospective controlled studies.

There are a number of factors that could account for the increased rates of malformations seen in IMEs. These include maternal seizures during pregnancy, a genetic predisposition related to having epilepsy, falls and injuries from seizures, and lower socioeconomic status and its attendant limited access to prenatal care. There are, however, a number of observations that strongly implicate AEDs as the cause of such teratogenicity seen. Comparisons of malformation rates in the offspring of mothers with epilepsy treated with AEDs compared to those without AED treatment consistently reveal higher rates in the children of the treated group (Table 2).

The precise mechanisms by which AEDs cause birth defects are not fully known. Susceptibility to AEDs and the expression of teratologic outcome may be under genetic control, as suggested by the variability in the frequency and pattern of malformations found in humans and in experiments with inbred mouse strains.^{30,46,73,159,217,246,265}

One of the earliest hypotheses proposed that the teratogenic effects of AEDs are related to their interference with folate metabolism. The rationale is that, in humans, low maternal levels of folate and vitamin B₁ and high levels of homocysteine are associated with an increased risk of NTDs.^{38,137,179,279} These observations, together with the finding that periconceptional folate supplementation significantly reduces the risk of birth defects and, in particular, that of NTDs,^29,36,100 suggest that folic acid antagonists may increase the risk of congenital malformations. Low folate levels have been found consistently in pregnant women treated with AEDs.¹⁰² In addition, the risk of adverse outcome increases with rising AED blood concentrations and decreases with folate levels.^44,45,199

Details of the relationships between AEDs and NTDs and between AEDs and folic acid are not yet clear, as it is difficult to explain (a) why the use of multivitamin supplements containing folic acid reduces the risk of malformations caused by some folic acid antagonists but not that due to AEDs¹⁰⁰ and (b) why the risk of NTDs is associated with valproate rather than with phenytoin and phenobarbital, two drugs that decrease folate levels the most. Nevertheless, the available experimental data support the idea that folate metabolism is likely to be involved in some particular aspects of valproic acid teratogenesis.^{5,60,62,107,129,241,254,255,270}

Bioactivation of AEDs by the embryonic cytochrome P450 (CYP) system to arene oxide reactive intermediates, which are thought to be teratogenic, has been a prevailing hypothesis for many years.^{6,20,30,74,75,153,155,169,208,215,225,246} This conclusion was consistent with the particular high risk associated with combinations of carbamazepine, phenobarbital, and valproate reported in humans,¹⁵⁵ and interpreted as being the result of the cumulative effects of inducing AEDs and valproate inhibition of epoxide-hydrolase responsible for epoxide elimination. The use of extended-release forms of carbamazepine and valproate may lower the risk factor as peaks and valleys are eliminated in these preparations. Further studies are needed to assess this concept.

An argument against the foregoing proposal is that the content of CYP in the embryo is low, making bioactivation of AEDs unlikely. Other information against a bioactivation mechanism for teratogenesis is that the most teratogenic AEDs, such as trimethadione and ethotoin, which cause the same pattern of birth defects as phenytoin, are not metabolized through an
epoxide metabolite.^72,271 Potentiation of phenytoin teratogenesis by drugs inhibiting CYP-mediated metabolism²⁵⁰ further argues against a role for the CYP system in teratogenic bioactivation of phenytoin.

Another postulated bioactivating pathway is AED co-oxidation to free radical intermediates. After interaction with molecular oxygen, these compounds liberate reactive oxygen species (ROS), which may be teratogenic. Deficiency of free radical scavenging enzymes, responsible for eliminating ROS, has been associated with malformations in the offspring of mothers with epilepsy taking AEDs.^210,272,273 A recent hypothesis proposes that reactive intermediates may be produced by AED-induced embryonic bradycardia/arrhythmia. Such arrhythmias cause periods of interrupted oxygen supply and generation of highly toxic ROS in the embryonic tissues during the reoxygenation/reperfusion phase.^12,39

Other postulated mechanisms are tissue damage due to fetal hemorrhage, possibly involving vitamin K deficiency,¹¹⁴ or interference with placental carnitine transport.²⁷⁷

Recently, mechanisms involving homeobox (HOX) genes have been proposed as an explanation for AED teratogenicity.¹⁹⁴ Retinoic acid signaling, involved in regulating transcription of genes that are critical to early brain development, may be affected by teratogens.²²¹ Such effects include alteration of the expression of retinoic acid receptor⁸⁷ or valproate inhibition of histone deacetylases,^{23,61,91,94,212} a key element in the regulation of many genes playing important roles in cellular proliferation and differentiation.^69,133,170 Phosphodiesterase-mediated inhibition of cyclic adenosine monophosphate,⁸⁵ increase of ROS levels,¹⁹⁰ disruption of normal pH within the embryonic milieu,²³³ or inhibition of the detoxifying epoxide hydrolase could also be relevant to the teratogenic effect of valproic acid.¹¹⁶

In summary, the mechanisms behind developmental toxicity of AEDs are presently far from completely understood, are likely to be multiple, and undoubtedly differ between individual AEDs.

NTDs are uncommon malformations occurring in 6 in 10,000 pregnancies. Spina bifida and anencephaly are the most commonly reported NTDs; they occur in approximately 4,000 pregnancies annually, affecting 2,500 to 3,000 births in the United States each year.^113,187 The types of NTDs associated with AED exposure are primarily myelomeningocele and anencephaly, which result from abnormal neural tube closure between the third and fourth weeks of gestational age.

Previous thinking about NTDs visualized the fusion of the neural tube as one in which the lateral edges met in the middle and fused both rostrally and caudally, similar to a bidirectional zipper. Recent studies have suggested that there are multiple sites for neural tube closure,^90,259 and that different etiologies may result in different types of abnormality. There are four separate sites along the neural tube where neurulation develops. The first is midcervical; the second is at the cranial junction of the prosencephalon and mesencephalon; the third, which fuses in a caudal direction only, is at the site of the future mouth or stomadeum; and the fourth is the region over the rhombencephalon between the second and third regions.²⁵⁹ The timing of closure differs among the various sites.

The majority of human NTDs can be explained by failure of one or more closure sites. Anencephaly with frontal and parietal defects is due to failure at the second closure site. Holoacrania, which also involves defects of the posterior cranium to the foramen magnum, is due to failure of closure of the second and fourth sites. Lumbar spina bifida results from failure of the first closure site. The development of closure sites is under genetic control but is also affected by environmental factors. In twins, concordance rates are only 56% for anencephaly and 71% for spina bifida. In Great Britain there is a male preponderance of lumbar spina bifida and female preponderance of holoacrania and anencephaly. Even valproic acid appears to have species differential effects being associated with spina bifida in humans and exencephaly in mice.²³⁵ A number of risk factors are associated with neural tube defects.

The strongest association for NTDs (relative risk [RR] 3% to 8%) is a previous pregnancy with NTD.¹⁸⁰ There are strong ethnic/geographic associations with NTDs. Rates per 1,000 are 0.22 for Caucasians, 0.58 for persons of Hispanic descent, and 0.08 for persons of African descent. The incidence of NTDs in Mexico is 3.26 per 1,000; for Mexican-born persons living in California, 1.6 per 1,000; and for U.S.-born persons of Mexican descent, 0.68 per 1,000.⁹⁷ Diabetic mothers have 7.9 times the rates of NTDs in their offspring.¹⁷ Deficiencies of glutathione, folate, vitamin C, riboflavin, zinc, cyanocobalamin, and selenium and excessive exposure to vitamin A have been linked to NTDs. Higher rates are seen in children of farmers, cleaning women, and nurses.^28,171 Pre-pregnancy weight also seems to be a factor. The RR for NTDs in women weighing 80 to 89 kg is 1.9, and 4.0 for those weighing over 110 kg. AED exposure may be a necessary but not sufficient risk factor for the development of NTDs. Women with epilepsy, as all women
of childbearing age, should be supplemented with folate. The dose recommended by the Centers for Disease Control and Prevention of 400 μg/day may not be high enough for many women who do not metabolize folate effectively. For those individuals with a history of NTDs, 4 mg of folic acid daily is recommended. Even with folate supplementation, women taking valproate should avail themselves of prenatal diagnostic ultrasound to detect NTDs.

A primary question in evaluating the relation between maternal epilepsy and congenital malformations is whether the association is further confounded by the genetics of epilepsy itself. Genetic susceptibility to teratogenic effects of AEDs is suggested by family studies.^{58,64,86,130,144,160,165,206,211,261,268} Several case-control series reported a higher proportion of relatives with epilepsy in patients with cleft palate or lip^{2,57,65,78,136} or NTDs.^156,219 Greenberg et al.⁹³ found extensive use of phenobarbital among 414 children with cleft palate or lip. However, the statistical significance of the association disappeared after exclusion of cases with a positive history of malformations. Other cohort studies have confirmed familial occurrence of fetal abnormalities.^{7,9,32,43,48,63,128,138,174,184,192,193,200,205,243,269} Parental epilepsy was also found to increase birth defects,^{49,53,128,136,205} although the majority of such studies reported conflicting findings.^{18,110,123,156,200,204,228}

Seizure occurrence during pregnancy does not increase malformation rates according to most^{9,56,71,76,110,141,192,218,228,237,240,243,244,281} but not all the authors.^{126,158,164,192,203} The issue of whether epilepsy per se increases the risk of congenital malformations was first raised by a large American and Finnish study²³⁷ that reported higher rates of malformations in children of treated mothers and fathers with epilepsy. Later studies confirmed these findings and added that even untreated mothers with epilepsy had a higher risk of having a child with birth defects.^{18,79,139,141,163,177,216} Other investigators, however, found no increased risk of congenital malformations among children of fathers with epilepsy.^42,80,157

A recent meta-analysis concluded that malformation rates among offspring of women with untreated epilepsy were similar to that of nonepileptic controls.⁷⁷ However, these results should be interpreted with caution because of differences in ascertaining cases and controls, as well as the small number of studies and cases considered. Data from the U.K. Registry, the largest prospective study available today, indicate that among the 3,607 enrolled pregnancies, the malformation rates for 239 women with epilepsy who had not taken AEDs during pregnancy (3.3%) were similar to those of 2,598 women with monotherapy exposures (3.7%).¹⁸⁵ Data from 31 studies including exposed offspring, nonexposed offspring, and healthy controls confirm that while treatment of pregnant women with AEDs undoubtedly increases the risk of birth defects, the increased risk cannot be entirely ascribed to AEDs (Table 2).

The more common malformations found in the offspring of women with epilepsy are heart defects, cleft lip or palate, hypospadias, and limb defects. There is fairly solid evidence of a specific association between NTDs and valproic acid^{10,21,32,101,125,157,204,219,231} and, to a lesser extent, carbamazepine.^10,101,223 The effect of valproic acid appears to be restricted almost entirely to the posterior neural cord.^158,160 Some studies suggest that the risk of congenital heart disease is increased in pregnant women exposed to barbiturates.^{8,10,32,56,192} This association emerges also from data of the U.S. Registry,¹¹² although the authors neglected to point out that 4 of 77 children exposed to phenobarbital were born with congenital heart defects (5.2%). The associations between valproic acid and limb reduction defects^10,222 or hypospadias^10,231 and between barbiturates and cleft palate/lip^10,125,192 need to be confirmed.