Fetal Hydantoin (Phenytoin) Embryopathy

Literature and regulatory information updated: 17 October 2026.

Fetal hydantoin syndrome, within the context of the fetal anticonvulsant syndrome; with carbamazepine, phenobarbital, primidone, lamotrigine, levetiracetam, oxcarbazepine and topiramate as comparators. Fetal valproate exposure appears in the differential diagnosis.

0. Initial Reference Table

1. Introduction

Phenytoin entered therapeutic use in 1938, when Merritt and Putnam reported that sodium diphenylhydantoinate prevented electrically induced seizures in the cat and tested the compound in 200 patients with frequent, refractory seizures; in the 142 treated for two to eleven months, tonic-clonic seizures were abolished in 58% and greatly reduced in another 27%, absences were abolished in 35% and reduced in 49%, and psychic equivalents were abolished in 67%, with toxic dermatitis in 10 patients (5%) and minor manifestations of tremor, ataxia and dizziness in about 15% [32]. The drug offered seizure control without the sedation of barbiturates and became, for four decades, the most widely used antiepileptic worldwide, so its teratogenicity was discovered in a context of very wide exposure of women of childbearing age.

The first warning sign was a letter from Meadow in 1968, which linked anticonvulsants to congenital anomalies [1, 33]. Epidemiological demonstration came with two studies in 1973: Fedrick analysed, in the Oxford Record Linkage Study, 223 newborns of 168 women with epilepsy in 218 pregnancies and, after exactly matching each live birth with three controls for social class, marital status, maternal age, parity, hospital and year of delivery, found some defect of congenital origin in 13.8% of live births against 5.6% of controls (p<0.0005), unrelated to seizure frequency or to the duration of the epilepsy [34]; Monson and colleagues published in the same year the analysis of diphenylhydantoin and selected congenital malformations [35]. Speidel and Meadow in 1972 and Hill and colleagues in 1974, the latter in a prospective study of infants exposed to antiepileptics, outlined the pattern of digital and nail hypoplasia, characteristic facies and growth and intellectual deficiency [1, 36, 37]. In the same volume and month as Hill's study, Aase published a letter on anticonvulsants and congenital anomalies; the bibliography of Smith's Recognizable Patterns of Human Malformation prints the year 1970 for that letter, but indexing gives 1974 [1, 38].

The designation “fetal hydantoin syndrome” was proposed by Hanson and Smith in 1975 [39]. The following year, Hanson, Myrianthopoulos, Harvey and Smith defined it as a variable pattern of altered growth and performance that includes unusual facies, hypoplasia of the distal phalanges and other defects, and quantified it in a prospective series of 35 infants exposed in utero: 11% had features sufficient for classification as fetal hydantoin syndrome and another 31% had some of the compatible features, results supported by a case-control study of 104 infants whose mothers had received hydantoins [40]. The authors identified reduced intellectual capacity as the area of greatest concern and recommended that the treated woman be informed of the nature and magnitude of the risks before becoming pregnant [40]. The series of subsequent reviews from the same group retained the designation and discussed its nosological status [41–43].

Nomenclature subsequently shifted from the drug to the pharmacological group. Jones and colleagues described in 1989 an analogous pattern with carbamazepine, and the set of manifestations common to phenytoin, carbamazepine, primidone and phenobarbital came to be called the “anticonvulsant facies” or fetal anticonvulsant syndrome [1, 8]. Orphanet retains the two layers: the clinical group FACS (ORPHA:370068) and the drug-specific entities, among them fetal hydantoin syndrome (ORPHA:1912), fetal carbamazepine syndrome (ORPHA:370076) and phenobarbital embryopathy (ORPHA:1919) [2, 6, 9, 10]. The oldest entity in this family is fetal trimethadione syndrome (ORPHA:1913), with pre- and postnatal growth retardation, intellectual disability, developmental and speech delay and craniofacial anomalies, and with less frequent cleft palate and cardiac, urogenital and limb malformations; trimethadione was withdrawn from the European market and is little used in other countries, because of teratogenicity and adverse effects, which makes the entity a historical term of comparison [7].

The study that established attribution of causality to the drug, rather than to the epilepsy, was the prospective screening by Holmes and colleagues of 128,049 pregnant women at the time of delivery, with systematic examination of newborns for major malformations, midface and digital hypoplasia, microcephaly and short stature: the combined frequency of embryopathy was 20.6% in 223 infants exposed to one anticonvulsant against 8.5% in 508 controls (odds ratio 2.8; 95% CI 1.1–9.7) and 28.0% in 93 exposed to two or more drugs (odds ratio 4.2; 95% CI 1.1–5.1), whereas 98 infants of mothers with a history of epilepsy without treatment during pregnancy did not show an increased frequency [44]. From that work onwards, the term “anticonvulsant embryopathy” came to designate the phenotype attributable to drug exposure.

The entity changed in practical significance over the following three decades for a prescribing reason, not a reclassification. Phenytoin ceased to be a usual choice in women of childbearing age: in the European EURAP cohort, of 15,893 eligible pregnancies enrolled up to 15 November 2023, only 180 (1.4% of monotherapies) were exposed to phenytoin, against 4672 to lamotrigine (36.1%) and 2797 to carbamazepine (21.7%) [45]. Lindhout and colleagues documented, in two consecutive cohorts, that the change in prescribing policy was accompanied by a change in the pattern of malformations, which makes the epidemiology of this embryopathy a dated phenomenon [46]. The committee responsible for the British guideline expressly recorded that phenytoin and phenobarbital are not usually taken by girls and women planning a pregnancy [47]. This article therefore describes an embryopathy whose recognition remains necessary in three situations: the child already born of an exposed pregnancy, the woman who becomes pregnant while on phenytoin because alternatives have failed, and acute exposure in the setting of status epilepticus or neurosurgery, in which fosphenytoin retains an indication [5, 25].

2. Molecular Genetics and Pathophysiology

2.1. Absence of a causal gene and nature of the condition

Hydantoin embryopathy has no causal gene: it is an embryofetopathy determined by a pharmacological exposure, and Orphanet records inheritance as not applicable [2]. Phenytoin crosses the placenta and cord blood concentrations are similar to maternal blood concentrations, so fetal exposure follows maternal exposure [25]. The variability of outcome between equally exposed fetuses is the observation that organises the whole pathophysiology: most exposed fetuses are born without embryopathy, a minority show the full phenotype, and the determinants of that difference are sought in maternal and fetal metabolism and in the genetic background of the embryo [3, 8].

The classic clinical observation in favour of fetal susceptibility is the discordant expression of the syndrome in heteropaternal dizygotic twins described by Phelan, Pellock and Nance, in whom fetuses simultaneously exposed to the same maternal concentration differed in phenotype [1, 48]. Finnell and Chernoff documented variable patterns of malformation in murine models and attributed them to the genetic background of the strain [49, 50]. The general principle is the same that Fraser established for cortisone, which regularly produced cleft palate in the A/Jax strain and in only 17% of embryos of another strain: expression of the phenotype requires both drug exposure and genetic susceptibility [3]. In clinical practice, the consequence is that the severity of a teratogenic effect is partly unpredictable, because it depends on drug metabolism by the fetoplacental unit, which is partly genetically determined and varies between siblings [15].

2.2. The arene oxide and epoxide hydrolase pathway (EPHX1)

The oldest mechanistic hypothesis attributes teratogenicity to the oxidative metabolites of phenytoin. The EPHX1 gene (1q42.12; OMIM 132810; MANE transcript NM_000120.4) encodes microsomal epoxide hydrolase, an endoplasmic reticulum enzyme that hydrates arene and aliphatic epoxides into less reactive, more water-soluble dihydrodiols, detoxifying most substrates and bioactivating some [11]. The two common coding polymorphisms studied in the metabolism of phenytoin and carbamazepine are the so-called slow allele, in exon 3, and the so-called fast allele, in exon 4, whose historical protein nomenclature, Tyr113His and His139Arg, corresponds, on normalisation by VariantValidator, to NM_000120.4:c.337T>C, NP_000111.1:p.(Tyr113His), NC_000001.11:g.225831932T>C (rs1051740) and to NM_000120.4:c.416A>G, NP_000111.1:p.(His139Arg), NC_000001.11:g.225838705A>G (rs2234922), without any shift in protein numbering relative to the reference transcript [11, 13]. Buehler, Delimont, van Waes and Finnell measured enzyme activity in amniocytes from a random sample of 100 pregnant women by thin-layer chromatography and observed an apparently trimodal distribution, consistent with an enzyme regulated by a two-allele gene, in which fetuses homozygous for the low-activity allele would be at risk; in a prospective study of 19 pregnancies monitored by amniocentesis, they predicted an adverse outcome in four fetuses with activity below 30% of the standard, all of mothers on phenytoin monotherapy, and all four newborns had findings compatible with the syndrome, whereas the 15 with activity above 30% had no features of the syndrome [51]. The authors themselves described the results as preliminary [51], and the work came to feature in the bibliography of the prenatal diagnosis chapter of Smith's Recognizable Patterns of Human Malformation [52]. The group's later review maintained that fetal susceptibility correlates with the fetal level of microsomal epoxide hydrolase and that the genetics of seizures in the parents does not predict teratogenic risk [53].

The current status of this pathway is twofold. As a mechanistic hypothesis, it remains plausible and is cited in descriptions of the syndrome [3, 54]. As a clinical tool, measurement of epoxide hydrolase activity in amniocytes or lymphocytes remained a research assay, was never replicated in an independent series, was never validated as a diagnostic test and does not appear in the 2024 joint guideline of the American Academy of Neurology, the American Epilepsy Society and the Society for Maternal-Fetal Medicine or in any other guideline consulted [11, 21]. EPHX1 variants do contribute to interindividual variability in phenytoin metabolism, alongside those of CYP2C19 and CYP1A1 [14]. The two polymorphisms are pharmacogenetic susceptibility modifiers and not causal variants: the clinical pharmacogenetics consortium classifies the EPHX1–carbamazepine pair at level D, without a guideline, and there is no genotype-guided dosing recommendation for EPHX1 in any guideline or label [11]. The same gene harbours a distinct Mendelian entity, an autosomal dominant lipoatrophic diabetes due to de novo missense variants at the catalytic site described in two patients, and associations of polymorphisms with pre-eclampsia, which are not to be confused with the embryopathy [11].

2.3. IKr current blockade, embryonic arrhythmia and hypoxia

The second pathway proposes that the injury results from disturbance of the embryo's cardiac rhythm. Danielsson and colleagues demonstrated, by patch-clamp in HEK 293 cells expressing the hERG channel, that phenytoin blocks IKr tail currents with a median inhibitory concentration of 240 µM and phenobarbital with 3 mM, whereas carbamazepine produces only 20% inhibition between 250 and 500 µM; the margin between those concentrations and the free therapeutic concentration is compatible with arrhythmogenic potential [55]. The IKr current is critical for repolarisation and for regulation of embryonic cardiac rhythm [1]. In CD-1 mice, dimethadione, the active metabolite of trimethadione, caused embryonic bradycardia and arrhythmia dependent on stage, between gestational days 9 and 13, and on dose, at clinically relevant maternal plasma concentrations of 3 to 11 mM; haemorrhage in the nasopharyngeal part of the embryonic palate in the preceding 24 hours preceded cleft palate, and the free radical scavenger phenyl-N-tert-butylnitrone reduced the incidence of cleft palate from 40% to 13%, which links the arrhythmia to hypoxia and reoxygenation with generation of reactive oxygen species [56]. The same group's review showed that structurally distinct drugs that block IKr — almokalant, dofetilide, phenytoin, cisapride and astemizole — produce a similar pattern of hypoxia-related malformations, preceded by embryonic cardiac arrhythmia, and that scavengers of reactive oxygen species markedly reduce teratogenicity [57]. In culture of mouse embryos at gestational day 10, all the antiepileptics studied increased the interbeat interval and rhythm irregularity in a concentration-dependent manner, dimethadione produced an irregular rhythm already within the therapeutic range and phenytoin slightly above it, and the combination of two drugs prolonged the interbeat interval more than either alone, beyond 60 ms [1, 58].

An additional argument comes from the comparison between strains: the A/J mouse is the most susceptible to phenytoin teratogenicity and has the highest activities of superoxide dismutase and glutathione peroxidase, with catalase equal to the other strains, which points to excessive generation of reactive oxygen species linked to the arrhythmia, and not to deficient antioxidant defence, a scavenging agent having markedly reduced teratogenicity [59].

2.4. Bioactivation by peroxidases and oxidative stress

The third pathway attributes the injury to bioactivation of the drug by embryonic peroxidases. Embryonic prostaglandin H synthases and lipoxygenases convert xenobiotics, phenytoin among them, into free radicals that oxidise proteins, lipids and DNA; the embryopathy is reduced in mice lacking prostaglandin H synthase, by inhibitors of this enzyme, by antioxidant enzymes and by radical scavengers, and is aggravated in mice deficient in glucose-6-phosphate dehydrogenase, with glutathione depletion or lacking Atm or p53 [60]. The experiment of Winn and Wells quantified the asymmetry of the defences: in pregnant CD-1 mice treated with phenytoin 65 mg/kg intraperitoneally on gestational days 11 and 12 or 12 and 13, pretreatment with polyethylene glycol-conjugated superoxide dismutase alone, between 0.5 and 20 KU/kg, increased teratogenicity, whereas conjugated catalase given eight hours beforehand, between 10 and 50 KU/kg, inhibited it in a dose-dependent manner, with a maximum reduction of cleft palate, resorptions and postnatal lethality at 50 KU/kg; the embryo's baseline catalase activity is about 4% of that of the maternal liver [61].

2.5. Interference with folate and growth

A fourth line proposes inhibition of placental folic acid transport as a mechanism of growth restriction, demonstrated in the rat by Will and colleagues [62]. Phenytoin's interference with folate is documented clinically in two directions: the drug reduces folic acid concentrations, with macrocytosis and megaloblastic anaemia that respond to its replacement, and the summary of product characteristics recommends measuring serum folate at least once every six months, supplementing when necessary; conversely, folic acid reduces phenytoin concentrations [4, 25]. The folate hypothesis is taken up in the discussions of published cases, alongside defects of arene oxide detoxification [54]. The epidemiological argument in its favour comes from the case-control study that grouped folate antagonists, including phenytoin, phenobarbital, primidone and carbamazepine, and found relative risks of 2.2 (95% CI 1.4–3.5) for cardiovascular defects, 2.5 (1.5–4.2) for oral clefts and 2.5 (1.2–5.0) for urinary tract defects; multivitamins attenuated the effect of dihydrofolate reductase inhibitors, but not that of the antiepileptics, which limits interpretation of the folate pathway as a single mechanism [63].

The four pathways are not mutually exclusive and none of them has translated into a validated clinical test or a preventive intervention of demonstrated efficacy in humans [21, 51, 57, 60]. Systems biology research on the expression of chromatin genes under anticonvulsants keeps the question open at the epigenetic level [64].

2.6. Maternal pharmacogenetics: CYP2C9 and HLA-B

Phenytoin has non-linear, saturable pharmacokinetics, with autoinduction of CYP2C9, and its clearance depends on haplotypes of this gene [12, 14]. CYP2C9 is a pharmacogene and not a Mendelian gene: the clinically relevant variation consists of haplotypes designated by star alleles, to which the Clinical Pharmacogenetics Implementation Consortium assigns activity values — alleles *1 and *9 with normal function (value 1.0), alleles *2, *4, *5 and *8 with decreased function (value 0.5) and alleles *3, *6, *13 and *35 with no function (value 0.0), the function of many rare alleles remaining uncertain [12]. The sum of the two values gives the activity score of the diplotype, which corresponds to a normal metaboliser at 2.0, an intermediate metaboliser at 1.5 or 1.0 and a poor metaboliser at 0.5 or 0.0 [12]. A poor metaboliser has slower hepatic clearance and greater plasma exposure, with dose-dependent toxicity in the form of ataxia, nystagmus and sedation at conventional doses [4, 12]. The consortium's population estimates give poor metabolisers in about 2.6% of Europeans, 3.0% of Near Eastern populations and 0.6% of East Asians, and intermediate metabolisers in about 35% of Europeans, 36% of Near Eastern populations, 15% of East Asians and 24% of African American and Afro-Caribbean populations [12]. The US label expresses the same orders of magnitude as 2 to 3% poor metabolisers in white populations, 0.5 to 4% in Asian populations and less than 1% in African Americans [4]. Metaboliser status is assigned by genotype and not by biochemical assay: serum measurement measures the drug, not the enzyme [12].

In the 2020 update of the consortium guideline, genotype is read in two steps, first HLA-B*15:02 and then CYP2C9 [14]. In a patient positive for HLA-B*15:02 and without previous exposure, the strong recommendation is not to use phenytoin or fosphenytoin and to avoid carbamazepine and oxcarbazepine; if the patient has already taken phenytoin continuously for more than three months without a cutaneous reaction, use may be considered with caution, because the latency of Stevens-Johnson syndrome and toxic epidermal necrolysis is short, 4 to 28 days [14]. In a patient negative for that allele, the normal metaboliser requires no adjustment; the intermediate metaboliser with an activity score of 1.5 likewise does not, because slightly reduced metabolism does not translate into more adverse effects; the intermediate metaboliser with a score of 1.0 keeps the usual starting or loading dose and receives maintenance about 25% lower than typical; and the poor metaboliser keeps the usual loading dose with maintenance about 50% lower, always with subsequent serum level measurement [14]. The 2020 version reclassified the CYP2C9*2/*2 diplotype, with a score of 1.0, from poor to intermediate metaboliser [14]. The distinction between scores of 1.5 and 1.0 corrects the simplified formulation according to which intermediate metabolisers would not require a dose change [12, 14].

HLA-B alleles do not affect the pharmacokinetics or pharmacodynamics of phenytoin: they alter the immunological risk of severe cutaneous reaction [14, 65]. The frequency of HLA-B*15:02 carriers exceeds 15% in Hong Kong, Thailand, Malaysia and parts of the Philippines, is around 10% in Taiwan and 4% in northern China, averages between 2 and 4% in South Asia, including India, with higher values in some groups, and is below 1% in populations of European ancestry; for carbamazepine, Stevens-Johnson syndrome and toxic epidermal necrolysis occur in 1 to 6 per 10,000 new users in predominantly white populations and about ten times more often in some Asian countries [65, 66]. The US label states that limited evidence suggests HLA-B*15:02 is a risk factor for Stevens-Johnson syndrome and toxic epidermal necrolysis in patients of Asian ancestry taking phenytoin, that a genome-wide association study in Southeast Asian patients identified a risk of severe cutaneous reaction in carriers of CYP2C9*3, with decreased function, and that avoiding phenytoin as an alternative to carbamazepine should be considered in those positive for that allele or carrying CYP2C9*3, adding that genotyping never replaces clinical vigilance [4]. The British guideline translates the same datum as a population alert, by recommending awareness of the increased risk of severe cutaneous reactions in people of Han Chinese or Thai origin [47]. The European periodic reassessment procedure completed in 2021 added to section 4.4 of all phenytoin-containing medicines a warning about the increased risk of severe cutaneous reactions in carriers of the CYP2C9*3 allele and about the risk of greater toxicity in intermediate or poor metabolisers of CYP2C9 substrates [67].

This pharmacogenetics conditions maternal dose and safety. It does not predict the embryopathy: no maternal or fetal genotype is validated as a predictor of teratogenic risk, and the risk of severe cutaneous reaction is, moreover, dose-dependent [14, 21].

3. Clinical Spectrum and Manifestations by System

3.1. Frequency and definition of the phenotype in exposed children

The risk that a fetus exposed to phenytoin will develop the full fetal hydantoin syndrome was estimated at about 10% in the historical series, without a demonstrated dose-response curve and without a dose below which there is no increase in teratogenic risk [1]. The figure comes from the prospective cohort of Hanson and colleagues, in which 11% of 35 exposed infants had features sufficient for classification and another 31% had some of the compatible features [40]. A review from the same group quantified the risk attributed to phenytoin at about 10%, with about 90% of exposed infants showing no evidence of teratogenesis under good management, against a baseline risk of major congenital defect of about 3% in all pregnancies [53]. None of these values is a population prevalence: they are proportions of affected children among those exposed, obtained in an era of polytherapy and without folate supplementation [40].

The phenotype divides, in practice, into major malformations, which determine the organ prognosis, and a set of minor anomalies and alterations of growth and neurodevelopment, which determine clinical recognition. The most frequent major malformations in exposed children are precisely those that also occur frequently in unexposed children: heart defects, from septal defects to complex lesions, hypospadias, talipes and cleft lip or palate [8].

3.2. Craniofacial and orofacial

The craniofacial pattern described by Hanson and Smith comprises a wide anterior fontanelle, metopic ridge, ocular hypertelorism, broad and depressed nasal root, short nose with a bowed upper lip, broad alveolar ridge and cleft lip and palate [1]. The Orphanet definition adds epicanthic folds, malformed and low-set ears and a short neck [2]. In the description common to the anticonvulsant group, the face has a short nose with a broad root and anteverted nostrils, a thin upper lip and a broad, flat philtrum [8]. Chronic maternal medication is not the only cause of altered features: antiepileptic medication in the child or adult can itself coarsen the features and cause gingival hypertrophy, which matters when mother and child are both medicated [25, 68, 69].

In registry data, orofacial malformations with phenytoin appear with an estimated frequency of 9.7 per 1000 exposed in monotherapy (95% CI 4.4–17.2; n=904) in the systematic review of the North American guideline [21]. In the direct comparisons of the Cochrane review, orofacial and craniofacial clefts did not reach statistical significance against controls, with a relative risk of 0.67 (95% CI 0.04–12.54; N=48) against children of women without epilepsy and with no estimate possible against untreated epilepsy, because of the absence of events in 133 exposed and 530 unexposed children [17]. The discrepancy between the classic description, which lists orofacial cleft as a cardinal manifestation, and these wide confidence intervals reflects the small number of phenytoin-exposed pregnancies in recent cohorts, and not refutation of the association, which is maintained in the product information on both sides of the Atlantic [4, 20].

3.3. Limbs, nails and skeleton

Hypoplasia of the distal phalanges with small nails is the most specific feature and occurs mainly in the postaxial digits, accompanied by stiff, tapered fingers, low-arch dermatoglyphs on the hypoplastic fingertips, a “digitalised” thumb, shortened distal phalanges and metacarpals with cone-shaped epiphyses and, in some cases, hip dislocation [1]. Orphanet extends nail hypoplasia to the toes [2]. The anticonvulsant group also features joint laxity, hypoplastic and overlapping toes and talipes [8]. Phenytoin exposure is among the teratogenic causes of limb reduction defects, both through interference with morphogenesis and through vascular disruption of an already formed limb [70], and hypoplasia of the distal phalanges due to intrauterine exposure to the drug is a recognised differential of brachydactyly type B [71].

Physical examination of the digits distinguishes the three traditional antiepileptics from one another. In a study that systematically examined the digits of 115 children exposed in monotherapy, 40 to phenytoin, 34 to phenobarbital and 41 to carbamazepine, and of their parents, comparing them with 111 children matched for age and sex and with their parents, and that combined subjective assessment with objective measurements, shortening and narrowing of the fifth-finger nail and a higher frequency of arches in the dermatoglyphs were more common in those exposed to phenytoin; in those exposed to phenobarbital there was a significant reduction in nail length, but not width; and in those exposed to carbamazepine, stiffness of the interphalangeal joints predominated [72]. The three drugs therefore affect the digits differently, and the effects were most marked with phenytoin [72].

A less frequently cited skeletal manifestation is chondrodysplasia punctata. Maternal phenytoin is among the four acquired causes of chondrodysplasia punctata, alongside maternal vitamin K malabsorption, maternal warfarin and maternal systemic lupus erythematosus [73], and may produce phenocopies of the tibia-metacarpal type [74]. The punctate calcifications associated with phenytoin toxicity are non-specific and are usually confined to the tarsal bones, unlike the axial and proximal distribution of the genetic dysplasia [75].

3.4. Pre- and postnatal growth

The historical description includes mild to moderate growth deficiency, generally of prenatal onset, which may become more marked in the first months of life, and a frequent relative growth delay in the first months, of unknown cause [1]. Orphanet retains pre- and postnatal growth retardation in the definition and includes microcephaly among the less frequent manifestations [2]. The product information refers to growth abnormalities, including microcephaly and prenatal growth deficiency, as part of the syndrome [4, 20]. The Scottish guideline qualifies this point: a significant reduction in neonatal head circumference has been documented with valproate and carbamazepine, and not with phenytoin, clonazepam, lamotrigine or gabapentin, and even with the former the reduction did not amount to microcephaly defined as more than two standard deviations below the mean [24]. The same source estimates fetal exposure to antiepileptics at about one in every 250 pregnancies and attributes to antiepileptics an increase in minor malformations, namely hypertelorism, epicanthic folds and digital hypoplasia, in addition to a doubling of the risk of a newborn small for gestational age and of an Apgar score below 7 at one minute, without an increase in spontaneous abortion or stillbirth [24].

Recent prospective data do not confirm growth restriction as an effect attributable to phenytoin. The systematic review of the North American guideline estimated the frequency of newborns small for gestational age at 14.4 per 1000 exposed to phenytoin (95% CI 2.7–35.1; n=464; three class III studies; low confidence), a value lower than that for carbamazepine (75.7 per 1000; n=3033), lamotrigine (85.1; n=2597) and phenobarbital (89.3) [21]. The fetal growth analysis of the EURAP registry, published in September 2026 on 15,893 eligible pregnancies, found for phenytoin an effect in the opposite direction to that of the other drugs: the birthweight centile was 8.04 points higher than with lamotrigine (95% CI 2.81–13.27), with crude mean centile values of 45.6 (standard deviation 30.4) against 42.3 (28.6) with lamotrigine, 26.1 (24.2) with phenobarbital and 28.7 (25.9) with topiramate [45]. The absolute proportions with phenytoin were 15.1% small for gestational age (24/159), 7.5% severe forms (12/159) and 7.5% low birthweight (12/159), against 14.6%, 5.1% and 4.5% with lamotrigine and 33.0%, 14.4% and 9.0% with phenobarbital [45]. The authors noted that, in the North American registry, phenytoin was also the only antiepileptic with a risk of small for gestational age lower than that of lamotrigine, although without significance, and proposed as a possible explanation its capacity to cause hyperglycaemia [45]. None of the dose analyses for phenytoin in that registry reached significance, with a coefficient of 4.97 for the weight centile (95% CI −11.26 to 21.20) and very wide intervals for the other outcomes, which reflects the small number of exposed pregnancies [45].

The combined reading is that prenatal growth deficiency is part of the historical description of the full phenotype, obtained in children with the syndrome, but is not confirmed as an average effect of the drug in contemporary cohorts, where phenytoin is associated with higher weight centiles [1, 21, 45]. The agencies have not followed this distinction: the British review attributes fetal growth restriction to phenobarbital, topiramate and zonisamide, and not to phenytoin [22], while the product information retains the historical reference to prenatal growth deficiency [4].

3.5. Major malformations by system and comparison between drugs

The following table brings together the estimates per drug in monotherapy from the three most recent data sets, with their respective denominators.

The EURAP analysis published in March 2024 is the most extensive prospective source on the subject: it brought together 10,121 monotherapy pregnancies enrolled between 1999 and October 2022, of which 9840 were exposed to the eight most used drugs, in 8483 women aged 14.1 to 55.2 years from more than 40 countries, with major malformations assessed in the first year by a committee blinded to exposure [18]. In the multivariable comparison with levetiracetam, phenytoin was the only one of the less used traditional drugs to reach significance, with an odds ratio of 2.39 (95% CI 1.08–5.31; p=0.04), against 2.02 (0.97–4.21) for topiramate, 1.26 (0.84–1.89) for lamotrigine and 1.09 (0.56–2.13) for oxcarbazepine; dose dependence was confirmed for valproate, phenobarbital and carbamazepine, and not for phenytoin [18]. Heart defects were the most frequent major malformation, especially after exposure to phenytoin, phenobarbital, topiramate and valproate [18]. In the same registry, the overall prevalence of major malformations fell from 6.1% (153/2505) in the period 1998 to 2004 to 3.7% (76/2054) in that of 2015 to 2022, a reduction of 39% that the authors attribute to the change in prescribing patterns and not to a change in the risk of each drug [18].

The divergence between the North American registries, which give values close to baseline for phenytoin, and the cohort meta-analyses, which give 5 to 6%, is the central point of interpretation. The two sources measure different things: the registry recruits on the initiative of the pregnant woman herself and compares drugs with each other, whereas the Cochrane review pools cohorts with comparison groups of women without epilepsy and of women with untreated epilepsy [17, 19]. In the comparisons with a control group, the relative risk of phenytoin against children of women without epilepsy was 3.81 (95% CI 1.91–7.57; 8 studies; 496 exposed and 1397 unexposed; I² 35%; risk difference 0.03; 0.01–0.06) and, against women with untreated epilepsy, 2.01 (1.29–3.12; 15 studies; 750 and 1588; I² 0%; risk difference 0.03; 0.01–0.05) [17]. The previous version of the same review gave relative risks of 2.38 (1.12–5.03) and 2.40 (1.42–4.08) [78]. A Bayesian network meta-analysis of 96 studies and 58,461 pregnancies estimated for phenytoin an odds ratio of 1.67 against unexposed pregnancies (95% credible interval 1.30–2.17), above carbamazepine (1.37; 1.10–1.71) and below phenobarbital (1.83; 1.35–2.47), topiramate (1.90; 1.17–2.97) and valproate (2.93; 2.36–3.69) [79].

In the direct comparisons between drugs in the Cochrane review, lamotrigine had fewer malformations than phenytoin, with a relative risk of 0.55 (95% CI 0.35–0.87; 6 studies; 4251 and 742 exposed; risk difference −0.02; −0.03 to 0.00), and levetiracetam 0.58 (0.34–0.97; 5 studies; 1018 and 687; I² 58%), with no difference in the random-effects model [17]. Phenytoin did not differ from carbamazepine (0.83; 0.62–1.11; 23 studies; 4759 and 1287), phenobarbital (0.84; 0.57–1.23), primidone (0.78; 0.39–1.56), oxcarbazepine (0.94; 0.48–1.85), topiramate (0.88; 0.48–1.61) or zonisamide (1.28; 0.42–3.93), and had fewer malformations than valproate, with a relative risk of 1.92 in favour of phenytoin (1.44–2.56; 21 studies; 2650 and 1247; risk difference 0.05) [17]. The only specific malformation with a significant difference between drugs was cardiac, less frequent with carbamazepine than with phenytoin (relative risk 0.44; 0.23–0.84; 16 studies; 4341 and 1005), with an absolute risk difference of −0.01 (−0.02 to 0.00), not significant [17].

The distribution by organ for phenytoin, according to the systematic review of the North American guideline, is 19.9 cardiac malformations per 1000 exposed (95% CI 11.6–30.3; n=955), 9.7 orofacial clefts per 1000 (4.4–17.2; n=904), 8.0 renal malformations per 1000 (2.0–18.1; n=466), 2.0 neural tube defects per 1000 (0.1–6.4; n=758) and 1.1 urogenital malformations per 1000 (0.2–6.6; n=416) [21]. Neural tube defects are the sharpest contrast with valproate, which has 14.3 per 1000, a difference of −12.3 per 1000 (−18.5 to −6.1) [21]: the risk of neural tube defect is about 5% with valproate and about 1% with carbamazepine, and is not increased with the other drugs, including phenytoin [8]. In the EURAP data, phenytoin had 4% cardiac malformations (5/125) against 1.46% for carbamazepine (28/1957), and a single neural tube defect (0.80%; 1/125) [17].

The absence of a demonstrated dose-response curve distinguishes phenytoin from the drugs with an established dose relationship. In EURAP, dose dependence was significant for carbamazepine (p=0.0140), lamotrigine (p=0.0145), phenobarbital (p=0.0390) and valproate (p<0.0001), but not for phenytoin, whose data the authors recommended interpreting with caution because of the small number of exposures [18, 76]. The analysis of the same registry devoted to dose, published in 2011, stratified 1010 pregnancies with valproate, 1402 with carbamazepine, 1280 with lamotrigine and 217 with phenobarbital and documented the gradient within each drug, with the lowest rates of major malformation for lamotrigine below 300 mg per day (2.0%; 95% CI 1.19–3.24) and for carbamazepine below 400 mg per day (3.4%; 1.11–7.71) and a progressive increase with dose for the four drugs studied; phenytoin was not included in that analysis because of insufficient exposures, which explains the absence of data by dose stratum for the drug of this article [80]. The Cochrane review recorded that most studies did not investigate dose, that the North American registry with 416 exposed pregnancies and Kaaja in 2003 with 124 exposed pregnancies found no association with dose, that Kaneko in 1999 with 132 exposed pregnancies in monotherapy found an association, and that the comparison between phenytoin at doses of 30 to 730 mg per day and lamotrigine up to 325 mg per day gave an odds ratio of 1.93 (95% CI 0.78–4.75), to be interpreted with reservation because of the range of doses pooled [17]. Polytherapy, by contrast, consistently increases the risk: 28.0% embryopathy against 20.6% in monotherapy and 8.5% in controls in the Holmes screening [44], 16.8% major malformations with polytherapy in the Meador meta-analysis cited by the Oxford Desk Reference [8] and a 48% increase in the risk of small for gestational age with polytherapy in EURAP (odds ratio 1.48; 95% CI 1.29–1.70), reaching a deficit of 14.18 weight centiles with four drugs (n=48) [45]. The contribution of polytherapy depends on the drugs combined and not on the number in itself: restricting the EURAP analysis to the 12,517 pregnancies exposed only to carbamazepine, lamotrigine, levetiracetam, phenytoin or valproate, the differences between monotherapy and polytherapy disappeared [45].

3.6. Neurodevelopment

The classic description speaks of occasional borderline to mild intellectual impairment, with childhood performance that may exceed that predicted by progress in early infancy, verbal abilities lower than non-verbal abilities and absence of a dose-response effect [1]. The two reference studies are those of the NEAD group, which in 48 children aged 3 years exposed to hydantoin gave a mean intelligence quotient of 99 and, at 6 years, of 108 [1]. The value at 3 years is an adjusted mean of 99 with a 95% confidence interval between 94 and 104, and not a range of individual values, obtained in 48 children with intelligence quotient analysis among 55 exposed to phenytoin in the cohort of 309 live births, with a median maternal dose of 398 mg per day; in the same study, the adjusted means were 101 (98–104) with lamotrigine, 98 (95–102) with carbamazepine and 92 (88–97) with valproate, and the maternal intelligence quotient of the phenytoin group was the lowest of the four, 92 (88–97) [81].

The assessment at 6 years in the NEAD study, prospective, observational, with blinded assessors, at 25 centres in the United Kingdom and the United States, included 305 mothers and 311 children, of whom 224 completed the assessment; adjusting for maternal intelligence quotient, drug, standardised dose, gestational age and periconceptional folate, the intelligence quotient was 108 with phenytoin (95% CI 104–112), 108 with lamotrigine (105–110) and 105 with carbamazepine (102–108), all higher than the 97 with valproate (94–101), with p=0.0006 for the comparison between phenytoin and valproate [82]. The systematic review of the North American guideline gave, for phenytoin, a full-scale intelligence quotient of 103.2 (95% CI 93.0–113.4; n=76; one class I and one class III study), verbal of 103.0 (95.8–110.2; n=61) and non-verbal of 106.0 (103.1–109.0; n=40), with differences against valproate of 9.29 points in the full-scale score (−1.63 to 20.21), 10.9 in the verbal score (2.0–19.8) and 4.8 in the non-verbal score (0.1–8.7) [21]. The 2014 Cochrane review found no difference between carbamazepine and phenytoin in developmental quotient (mean difference 3.02; −2.41 to 8.46; 172 and 87 children) or in intelligence quotient (−3.30; −7.91 to 1.30; 75 and 45), and found higher values with phenytoin than with valproate in developmental quotient (7.04; 0.44–13.65; p=0.04) and intelligence quotient (9.25; 4.78–13.72; p<0.0001), with no convincing evidence of a dose effect for carbamazepine, phenytoin or lamotrigine [83]. In the network meta-analysis of 29 cohorts and 5100 children, only valproate was significantly associated with cognitive delay (odds ratio 7.40; credible interval 3.00–18.46), with no significant association for phenytoin [84].

These data contrast with the position of the regulators. The British committee's review concluded that, for phenobarbital and phenytoin, although clinical studies show inconsistent results, the totality of the data shows the possibility of adverse neurodevelopmental effects [22], and the harmonised European text states that the studies are contradictory but that a risk cannot be excluded [20, 67]. The difference in reading is explained by the evidence base: the values favourable to phenytoin come from samples of 40 to 76 children, in a single research programme, whereas the regulatory assessment also integrates older reports and studies of heterogeneous quality [21, 22, 82]. The North American guideline issues no specific recommendation for phenytoin: the drug is neither among those of lowest risk, which are lamotrigine, levetiracetam and oxcarbazepine, nor among those to be avoided because of a given outcome, and appears mainly as a comparator in the recommendations on neurodevelopment [21].

The MONEAD programme, conducted at 20 North American epilepsy centres with 456 pregnant women enrolled between December 2012 and January 2016, shifted the question from comparison between drugs to plasma concentration: adaptive function at 4.5 years did not differ between children of women with epilepsy (n=302) and without epilepsy (n=84), with an estimate of 0.4 (95% CI −2.5 to 3.4), but decreased with the maximum antiepileptic concentration in the third trimester (−7.8; −12.6 to −3.1), an effect evident for levetiracetam (−18.9) and for lamotrigine (−12.0) [85]. In the same cohort, creativity did not differ between groups and executive function showed an exposure-dependent effect (7.0; 2.9–11.2), more marked with levetiracetam (12.9) [86]. Phenytoin was not analysed separately in either of these assessments, because of the small size of the subgroup [85, 86]. One consistent finding links the dysmorphic phenotype to cognitive prognosis. In 80 children exposed to phenytoin, phenobarbital and carbamazepine, in monotherapy and polytherapy, recruited for reassessment, the three with microcephaly had a deficit of 23.7 points of full-scale intelligence quotient compared with those of normal head circumference; excluding microcephaly and with Bonferroni correction, the presence of midface or digital hypoplasia correlated with deficits of 12.7 points in the verbal quotient, 10.0 in the performance quotient and 12.8 in the full-scale quotient (p=0.0061), without any reduction associated with major malformations [87]. The Oxford Desk Reference attributes to antiepileptic exposure in general a risk of neurodevelopmental disorders about four times higher, but the two studies it cites as the basis, a cohort of 415 children and a Danish registry study, both refer exclusively to valproate, with an adjusted odds ratio of 6.05 for valproate monotherapy in the first and adjusted risks of 2.9 for autism spectrum and 5.2 for childhood autism in the second, and neither includes a phenytoin-exposed group, so the figure does not apply to antiepileptics in general or to this drug [8, 88, 89].

The data specific to phenytoin point in a favourable direction in three recent data sets. The North American population cohort built from public and private insurance databases between 2000 and 2021 assessed exposure in the second half of pregnancy, the period of synaptogenesis, using as reference pregnant women with epilepsy without antiepileptics (8887 unexposed children) and with exposures ranging between 219 and 5261 children per drug: valproate and zonisamide were associated with several outcomes, with adjusted hazard ratios between 1.26 and 4.50, carbamazepine and oxcarbazepine with moderate increases in attention deficit hyperactivity disorder and behavioural disorder (1.23 to 1.40), and levetiracetam and phenytoin were not associated with an increased risk of any of the outcomes [90]. The matched cohort of 34 children exposed to phenobarbital, 40 to phenytoin and 41 to carbamazepine in monotherapy, with controls matched for age, sex and maternal characteristics and analysis adjusted for maternal intelligence quotient, found a significant effect of exposure on the overall verbal quotient, but not on the full-scale or performance quotient; only the phenobarbital group had reduced verbal and full-scale quotients relative to controls, and in the comparison between drugs phenobarbital was inferior to phenytoin in both, with no difference for carbamazepine against the others [91]. In the Scottish retrospective cohort, developmental delay occurred in 8 of 24 children exposed to phenytoin aged over 21 months (33%), or 7 of 24 (29%) excluding those with a family history of delay, against 14 of 58 (24%) and 11 of 58 (19%) with carbamazepine, 6 of 60 (10%) and 5 of 60 (8.3%) with phenobarbital, 13 of 35 (37%) and 7 of 35 (20%) with valproate and 16 of 43 (38%) and 12 of 43 (28%) with polytherapy, with the authors themselves noting that interpretation of the phenytoin group is difficult because it rests on eight children with delay [92].

In the Indian Kerala registry, assessment of 446 adolescents aged 13 to 21 years, of whom 272 were exposed to monotherapy, 133 to polytherapy and 41 unexposed, with 73 cases of phenytoin exposure, gave a full-scale intelligence quotient of 89.2 (standard deviation 21.5; n=146) in the exposed against 96.9 (8.8; n=11) in the unexposed (p=0.03), with lower language scores in the exposed and, compared with state averages, more delayed schooling (2.6% against 0.1%), more school dropout (1% against 0.11%), greater need for examination support (4.4% against 0.1%) and less university entry (19.9% against 37%); the study does not present values separated by drug and the unexposed group is very small [93]. In the large Nordic registry study, with 4,546,605 children born between 1996 and 2017, of whom 38,661 were children of mothers with epilepsy and 16,458 were exposed, valproate was associated with an adjusted risk of 1.80 (95% CI 1.60–2.03) for the combined psychiatric outcome, topiramate with attention deficit hyperactivity disorder (1.78 to 2.38) and levetiracetam with anxiety (2.17), with no data reported for phenytoin [94].

3.7. Perinatal manifestations and neonatal haemorrhage

Two perinatal events are specific to exposure to enzyme inducers. The first is neonatal withdrawal: tremor, irritability, seizures and feeding difficulty, described in 20% of those exposed to antiepileptics in the retrospective Grampian cohort [8, 92]. The second is haemorrhagic disease, potentially fatal, due to a reduction in vitamin K-dependent clotting factors in newborns exposed in utero to phenytoin, which the US label describes as a drug-induced condition, preventable by administration of vitamin K to the mother before delivery and to the newborn after birth [4]. The European summary of product characteristics uses the same wording, recommending vitamin K to the mother during the last month of gestation and to the newborn after delivery [25]. The two product information documents do not state a dose, but the British and Scottish obstetric guidelines do: 1 mg of vitamin K intramuscularly to all newborns of women with epilepsy taking enzyme-inducing antiepileptics, a group in which phenytoin is expressly included [23, 24]. The primidone label gives the historical basis for this practice, by describing neonatal haemorrhage with a coagulation defect similar to that of vitamin K deficiency in newborns of mothers taking primidone and other anticonvulsants [95]. Placental transfer of phenytoin is virtually complete: the ratio between the cord concentration and the maternal serum concentration is close to 1, as it is for phenobarbital, carbamazepine, lamotrigine, oxcarbazepine and topiramate [23]. The most recent verification of this ratio, in the MONEAD programme, had 305 pairs of maternal and cord concentrations in 199 women, but only one pair with total phenytoin and one with free phenytoin, so it produced no estimate of its own for the drug [96].

3.8. Other manifestations and associations

The list of minor anomalies includes short neck, rib anomalies, small and widely spaced nipples, umbilical and inguinal hernias, pilonidal sinus, coarse and abundant hair, hirsutism, low hairline, abnormal palmar crease and strabismus [1]. Occasional manifestations include microcephaly, brachycephaly, positional foot deformities, coloboma, ptosis, upslanting palpebral fissures, webbed neck, pulmonary or aortic valve stenosis, coarctation of the aorta, patent ductus arteriosus, septal defects, single umbilical artery, pyloric stenosis, duodenal atresia, anal atresia, renal malformation, hypospadias, micropenis, ambiguous genitalia, cryptorchidism, symphalangism, syndactyly, terminal transverse limb defect, split hand and holoprosencephaly [1]. Hypertrichosis is also among the direct pharmacological effects of phenytoin [25, 97], and optic nerve hypoplasia is among the described consequences of teratogenic exposure to anticonvulsants [98].

The association with neoplasms rests on case reports and is not quantified. The US label mentions several reported cases of neoplasms, including neuroblastoma [4], and the European summary adds that the role of the drugs relative to other factors has not been determined [25]. The report by Allen and colleagues described the triad of fetal hydantoin syndrome, neuroblastoma and haemorrhagic disease in a newborn, noting that it was the fourth neural crest-derived tumour reported after intrauterine exposure to phenytoin, which would suggest transplacental carcinogenesis [99, 100]. Experimentally, the carcinoma-promoting effect is unequivocal in rodents for phenobarbital and for phenytoin, which caused lymphoid and hepatic tumours in the rat, but the causal relationship in humans remains under debate [101].

3.9. Compared entities of the same group

Fetal carbamazepine syndrome (ORPHA:370076) has facial dysmorphism with similarities to that of valproate, including epicanthic folds, upslanting palpebral fissures, short nose, micrognathia and malar hypoplasia, with nail dysplasia and major malformations of cleft lip or palate, neural tube defects and heart defects; combined exposure to carbamazepine and valproate is associated with significant developmental delay, especially of verbal intelligence, and with a high rate of congenital anomalies [9]. In the dysmorphological description, the face is less obvious than with valproate, with a short nose, full cheeks, small mouth and chin and a philtrum that is not flat, and nail hypoplasia is specifically noted after carbamazepine [8]. The age at onset includes, in addition to the antenatal and neonatal periods, childhood [9].

Phenobarbital embryopathy (ORPHA:1919) results from first-trimester exposure and has the particularity that infants are generally asymptomatic, with rare reports of an increased risk of intellectual disability, tetralogy of Fallot, unilateral cleft lip and mitral valve hypoplasia, and of minor anomalies of hypertelorism, epicanthus, hypoplastic nose, low-set ears, prognathism, digital hypoplasia, brachydactyly and hypospadias [10]. In quantitative data, phenobarbital has the highest prevalence of major malformations among the traditional antiepileptics after valproate, of 60.3 per 1000 (95% CI 47.1–75.0; n=1116), with cardiac malformations in 4.4% and orofacial clefts in 2.2% [21], and is the drug with the clearest dose-response relationship after valproate: in EURAP, 2.7% up to 80 mg per day (95% CI 0.3–9.5), 6.2% between 80 and 130 mg per day (3.0–11.1) and 11.7% above 130 mg per day (4.8–22.6), with an odds ratio of 5.41 between the highest and the lowest stratum (1.05–27.89) [17]. In fetal growth, phenobarbital is among the three drugs with the greatest adverse effect, with a deficit of 8.08 weight centiles against lamotrigine (95% CI −11.89 to −4.26) and 33.0% of newborns small for gestational age [45]. Primidone is dealt with in the same group because it is converted to phenobarbital, and its estimated prevalence of major malformations is the highest in the series of the North American guideline, of 101.5 per 1000, with a very wide interval because of the small number of exposures (95% CI 50.4–167.7; n=99) [1, 21].

The more recently introduced comparators complete the picture. Lamotrigine is the drug with the largest number of documented pregnancies and without an increase in major malformations at maintenance doses, with more than 12,000 pregnancies assessed by the British committee [22], 2.1% malformations in the North American registry (52/2461) [19] and 2.3% in the United Kingdom and Ireland registries (95% CI 1.8–3.1; n=2198) [30]. Levetiracetam has 2.0% (26/1283) in the North American registry and more than 1800 pregnancies without an increase in malformations in the British review [19, 22]. Oxcarbazepine has 1.5% (5/327) malformations, but is one of the drugs with the greatest effect on fetal growth, with a deficit of 5.10 weight centiles against lamotrigine and a risk of small for gestational age 3.5 times higher above 1700 mg per day than at doses below 1100 mg per day [19, 45]. Topiramate is the comparator with the least favourable profile among the recent drugs: 5.1% major malformations in the North American registry (26/510), a specific association with cleft lip, a relative risk of 4.07 against children of women without epilepsy (95% CI 1.64–10.14) in the Cochrane review, the greatest growth deficit in EURAP, with 11.93 weight centiles below lamotrigine (95% CI −16.72 to −7.15) and 30.6% small for gestational age, and an association with attention deficit hyperactivity disorder in the Nordic registries [17, 19, 45, 94].

4. Clinical Recognition and Warning Signs

4.1. When to suspect in the newborn and infant

Suspicion arises from the combination of a phenotype and a history of exposure, and either of the two may be the starting point. In the newborn, the combination that should prompt questions about maternal medication is hypoplasia of the distal phalanges and nails, especially in the postaxial digits, with a broad, depressed nasal root, short nose and bowed upper lip [1, 2]. Nail and digital hypoplasia is of particular value because it is uncommon in unexposed children and because it correlates with cognitive prognosis [87]. Cleft lip or palate, congenital heart disease, hypospadias and talipes are common malformations in the general population and do not in themselves raise suspicion; they do so when they coexist with the minor anomalies or with a history of exposure [8].

In situations where the child is assessed already with developmental delay, the pregnancy history should systematically include anticonvulsant medication and corticosteroid therapy in the presence of cleft lip and palate [102], medication during pregnancy in the presence of hypospadias [103] and the use of anticonvulsants in the presence of optic nerve hypoplasia [98]. Teratogenic exposure is one of the aetiological categories to consider in any dysmorphology consultation, alongside genetic causes [104].

4.2. Targeted history

The points to cover are the three-generation pedigree; the exact dose and type of antiepileptic taken during pregnancy, which may require requesting the obstetric records; other teratogenic exposures, including alcohol and maternal illness; maternal seizures during pregnancy, delivery and the perinatal period; birthweight; symptoms of neonatal withdrawal, such as tremor, seizures and feeding difficulty; and a detailed developmental and behavioural history [8]. Documentation of the dose matters doubly: for assessment of the case and for counselling about a subsequent pregnancy, in which continuation of the same drug at the same dose is the determinant of risk [15].

4.3. Targeted examination

Examination follows a defined sequence: height, weight and head circumference; eyes, looking for prominence of the metopic ridge, hypertelorism or telecanthus, epicanthic folds and infraorbital groove; mouth and nose, assessing nasal length, the length and smoothness of the philtrum, the thickness of the upper lip, eversion of the lower lip and the presence of cleft lip or palate; hands and feet, looking for joint laxity, nail hypoplasia, stiffness of the interphalangeal joints, long overlapping fingers, hypoplastic or overlapping toes and talipes; cardiac auscultation; genital inspection for hypospadias; and inspection of the back for signs of spina bifida occulta [8]. In examination of the hands, measurement of the length and width of the fifth-finger nail and observation of the dermatoglyphs add discriminating information, because shortening and narrowing of that nail and the frequency of arches are more marked with phenytoin exposure, whereas phenobarbital shortens the nail without narrowing it and carbamazepine mainly produces interphalangeal stiffness [72].

4.4. Limits of recognition by facies

The facial gestalt is not sufficient for diagnosis. In the retrospective Liverpool and Manchester study, 375 children of 219 mothers with epilepsy, from 6 months to 16 years of age, of whom 274 were exposed, including 26 to phenytoin in monotherapy, underwent physical examination, a neuropsychological battery and blinded scoring of photographs by a panel of dysmorphologists: 47% of exposed children were correctly identified as having been exposed, but 45% of the unexposed had some of the facial features associated with exposure, which led the authors to conclude that the diagnosis of fetal anticonvulsant syndrome is difficult to establish on the basis of facies alone [105]. In that study, the correlation between verbal intelligence quotient and facial dysmorphism was significant only in children exposed to valproate [105]. Moreover, the anticonvulsant facies is common to phenytoin, carbamazepine, primidone and phenobarbital, so the face does not identify the drug [1].

4.5. Warning signs and emergencies

Three situations require immediate action. In the exposed newborn, unexplained haemorrhage or bruising requires consideration of haemorrhagic disease due to deficiency of vitamin K-dependent factors, which is potentially fatal and preventable with 1 mg of vitamin K intramuscularly at birth [4, 23, 25, 99]. The British obstetric guideline adds surveillance of the newborn for lethargy, feeding difficulty, sedation and signs of withdrawal, with serum measurement of the antiepileptic in the infant when appropriate, especially in the preterm infant [23]. In the newborn with a murmur or cyanosis, echocardiography is necessary because of the frequency of congenital heart disease with this exposure, which in EURAP was 4% [17, 21]. In the exposed child who has microcephaly or midface or digital hypoplasia, systematic developmental assessment is a formal indication, because of the association with deficits of 10 to 24 intelligence quotient points [87].

On the maternal side, the warning sign is the woman of childbearing age on phenytoin without documented preconception counselling or without effective contraception, a situation in which enzyme induction may cause hormonal contraceptives to fail and produce an unplanned pregnancy under exposure [25, 47, 67]. A woman planning pregnancy while on phenytoin should be referred urgently to the epilepsy team [22, 47].

4.6. Newborn screening

There is no population newborn screening for this condition, nor a biochemical marker that identifies it [8, 21]. What does exist, and is recommended, is targeted antenatal surveillance: detailed anatomy ultrasound in all pregnant women treated with any antiepileptic, and specialist antenatal monitoring directed at the described malformations when phenytoin is continued [21, 25].

5. Diagnosis: Criteria, Tests and Algorithms

5.1. Absence of formal criteria and of a confirmatory test

There are no formal diagnostic criteria published by a scientific society for hydantoin embryopathy, nor a test that confirms it [8, 21]. Diagnosis is clinical and rests on three pillars: documentation of phenytoin exposure in the first trimester, with the type and dose of the drug; identification of a compatible phenotype by targeted examination; and exclusion of genetic and metabolic causes that would explain the findings [8]. The general rule of teratogenic dysmorphology is that the diagnosis of a teratogenic syndrome is made after a genetic or other cause for the anomaly has been excluded, and that caution is required when data on the drug in pregnancy are very limited or when the proposed phenotype rests on few cases with insufficient genetic investigation [15].

The operational definition closest to a criterion is that of the Holmes screening, which classified as anticonvulsant embryopathy the presence of a major malformation, midface hypoplasia, digital hypoplasia, microcephaly or short stature in the exposed newborn, assessed by systematic examination [44]. This definition was designed for epidemiological research and not for individual diagnosis; applied in isolation, it would classify 8.5% of unexposed newborns as affected [44].

5.2. Documentation of exposure

Exposure is documented through the obstetric records and information from the prescriber, with the name of the drug, the daily dose, the gestational period covered and any polytherapy [8]. The critical window is first-trimester organogenesis, which is the exposure period considered in all pregnancy registries [19, 76]. Maternal serum concentration values recorded during pregnancy, when available, add information on actual exposure, given that phenytoin clearance increases over the course of gestation [4, 106].

5.3. Exclusion of genetic causes

Genetic investigation does not serve to confirm the embryopathy: it serves to make it the more plausible explanation, by excluding alternatives. In a child with developmental delay accompanied by dysmorphism or with a family history of learning difficulty, the Oxford Desk Reference recommends genomic array, fragile X testing and measurement of urinary amino acids and organic acids, with other investigations as clinically indicated [8]. The array has a specific justification here: it may identify microdeletions that predispose to epilepsy and that may have been inherited from the mother herself, namely at 15q11.2, 15q13.3 and 16p13.11, a scenario in which the child would have both a genetic cause and an exposure [8]. The probability is not negligible, since about 15% of cases of epilepsy with associated manifestations have a pathogenic or possibly pathogenic copy number variation, and a similar proportion has a de novo variant in a candidate gene in epilepsy with developmental delay or epileptic encephalopathy [107].

5.4. Nomenclature of variants found during investigation

The variants relevant in this context are not causal of the embryopathy, and their nomenclature follows the general rules. Variants identified on array or on sequencing of the proband should be reported with the gene, the transcript and its version, the HGVS designation at the DNA and protein level, zygosity and the class according to the criteria of the American College of Medical Genetics and the Association for Molecular Pathology [16]. In pharmacogenes, the nomenclature is different and uses haplotypes designated by star alleles, such as CYP2C9*1, *2, *3, *8 and *9, and human leucocyte antigen alleles with four-digit notation, such as HLA-B*15:02, to which an activity value and a metaboliser phenotype correspond, and not a pathogenicity classification [12, 14]. The conversion between a diplotype and the metaboliser phenotype is tabulated and was changed in the 2020 update, in which the CYP2C9*2/*2 diplotype moved from poor to intermediate metaboliser [14]. The HLA-B gene has more than 7000 described alleles, which makes full notation of the tested allele indispensable [14].

5.5. Tests without clinical validation

Measurement of epoxide hydrolase activity in amniocytes, proposed in 1990 as prenatal prediction of risk, rested on a series of 19 pregnancies, was never independently replicated, was presented by the authors themselves as preliminary and does not appear in any current guideline [11, 21, 51]. It should not be offered as a risk test. Genotyping of CYP2C9 and HLA-B is indicated for maternal safety and dose, and not for predicting the embryopathy; a negative result for HLA-B*15:02 does not eliminate the risk of Stevens-Johnson syndrome or toxic epidermal necrolysis, and genotyping does not replace clinical vigilance [4, 14].

5.6. Prenatal and preimplantation diagnosis

There is no prenatal diagnosis of the embryopathy as an entity, because there is no variant to test for: genetic prenatal diagnosis is only available for conditions with a known mutation or cytogenetic abnormality [107]. What does exist is morphological surveillance. The North American guideline recommends detailed anatomy ultrasound in all pregnant women treated with any antiepileptic, and fetal cardiac screening specifically with phenobarbital exposure, because of the 4.4% frequency of cardiac malformations with this drug [21]. The European summary of product characteristics requires that, if phenytoin is continued in pregnancy, specialist antenatal monitoring directed at the described malformations be instituted [25]. The Oxford Desk Reference includes the offer of prenatal diagnosis among the recommendations for follow-up of the pregnant woman on an antiepileptic, in the sense of detailed morphological assessment [8]. The British obstetric guideline sets out the schedule: anomaly ultrasound between 18 weeks and 0 days and 20 weeks and 6 days, according to the national screening programme, able to detect major heart defects and neural tube defects, and with no studies supporting routine fetal echocardiography; adding maternal serum alpha-fetoprotein to ultrasound raises detection of neural tube defects to 94 to 100% [23]. The performance values of each test alone, obtained in unselected populations, are 98% detection for anencephaly and 75% for spina bifida by ultrasound, and 82% for spina bifida with maternal serum alpha-fetoprotein at a threshold of 2.5 multiples of the median at 16 weeks, with 1.9% false positives; anencephaly is detectable from 11 to 12 weeks, surveillance for spina bifida in an at-risk pregnancy is carried out with serial ultrasound from 16 weeks, and alpha-fetoprotein does not detect closed defects and is less sensitive in women taking valproate, which makes it complementary and not an alternative to ultrasound [108]. Amniocentesis is no longer used for this purpose alone [108]. Preimplantation genetic diagnosis has no application in this condition, for the same reason: there is no genotype to select, and prevention is achieved by modifying the exposure before conception [21, 22].

5.7. Algorithm

For a child with a compatible phenotype, the practical sequence is: confirm the exposure with dose and period from the obstetric records; complete the dysmorphological examination with measurement of weight, height and head circumference; request echocardiography in the presence of a murmur, cyanosis or significant dysmorphism; request genomic array, fragile X testing and urinary metabolic screening when there is developmental delay, dysmorphism or a family history of learning difficulty; formally assess development if there is delay, microcephaly or midface or digital hypoplasia; and screen vision and hearing [8, 87]. When genetic investigation identifies an alternative cause, the diagnosis of embryopathy does not stand, even if the exposure is documented; when it does not, the diagnosis remains clinical and probabilistic, and should be communicated as such [15, 105].

5.8. Exposure in an ongoing pregnancy

For individual risk assessment of an ongoing exposure, teratology information services exist. In the United Kingdom, the national service is free for health professionals, with a dedicated line and accessible reviews, and should be contacted both for the assessment and for recording the exposure and the outcome, for surveillance purposes [15]. In Europe there is the European network of teratology information services, and in the United States the equivalent service [15]. Enrolment in pregnancy registries is recommended by the labels themselves: the US label states that enrolment in the antiepileptic pregnancy registry should be done by the patient herself through a dedicated telephone line [4, 29].

5.9. Content of the report

When genetic investigation is performed, the report includes the gene, the transcript with its version, the HGVS designation at the DNA and protein level, zygosity and the pathogenicity class assigned according to the criteria of the American College of Medical Genetics and the Association for Molecular Pathology, with explicit statement of the limitations of the technique used, namely the inability of the array to detect single-nucleotide variants and that of exome sequencing to detect repeat expansions and deep intronic variants [8, 16]. A variant of uncertain significance neither establishes nor excludes an alternative diagnosis, and its interpretation benefits from segregation studies in the parents [16]. In the pharmacogenetic report, the content is the diplotype, the inferred metaboliser phenotype, the human leucocyte antigen allele tested and the corresponding dosing recommendation [14].

6. Differential Diagnosis

6.1. Comparative table

6.2. The central problem: distinguishing within the group

The dominant difficulty is not distinguishing hydantoin embryopathy from a genetic syndrome, but distinguishing the antiepileptic embryopathies from one another. The anticonvulsant facies is common to phenytoin, carbamazepine, primidone and phenobarbital, and the face does not identify the drug [1]. There are, nonetheless, orienting markers: the risk of neural tube defect is about 5% with valproate and about 1% with carbamazepine, and is not increased with the other drugs [8, 21]; the face is more marked with valproate, with trigonocephaly, infraorbital grooves, a small downturned mouth and an everted lower lip [8]; nail hypoplasia is specifically noted after carbamazepine, just as it is cardinal with phenytoin [1, 8]; cardiac malformations are more frequent with phenytoin than with carbamazepine (relative risk 0.44 in favour of carbamazepine; 95% CI 0.23–0.84) and are the characteristic finding with phenobarbital, at 4.4% [17, 21]; and orofacial cleft is more frequent with phenobarbital and topiramate, at 2.2% and 1.4% [21]. To this set of markers has been added, since 2024, a morphometric criterion: measurement of the nails and reading of the dermatoglyphs separate the three traditional drugs, with shortening and narrowing of the fifth-finger nail and a higher frequency of arches with phenytoin, isolated nail shortening with phenobarbital and interphalangeal stiffness with carbamazepine [72]. Definitive distinction depends on the prescribing record, and with polytherapy it may not be possible to attribute the phenotype to one drug [8, 44].

6.3. Genetic causes of nail and distal phalanx hypoplasia

The GeneReviews chapter on Coffin–Siris syndrome expressly lists fetal hydantoin syndrome among the differentials of small nails with hypoplasia of the distal phalanges, describing it as associated with facial dysmorphism, digitalised thumbs, low hairline, short or webbed neck, growth restriction and cognitive deficits [16]. The practical distinction from Coffin–Siris rests on two axes: in the latter, hypoplasia is concentrated in the fifth finger and toe, the face is coarse with a wide mouth and full, everted lips, there is hypertrichosis with sparse scalp hair and intellectual disability is almost always moderate to severe, with severely affected speech, whereas in hydantoin embryopathy the cognitive impairment is borderline to mild [1, 111]. Brachymorphism-onychodysplasia-dysphalangism syndrome comes closer to the embryopathy through its mild impairment, but is accompanied by marked short stature and short fifth fingers [16]. The remaining entities of the same group — 4q deletion, DOORS syndrome and Nicolaides–Baraitser syndrome — have their own markers, respectively the curved volar nail of the fifth finger, deafness with osteodystrophy and seizures, and sparse hair with prominent phalanges [16].

6.4. Embryopathies with epiphyseal stippling and short phalanges

Warfarin embryopathy is the closest differential on radiological grounds. Exposure to coumarins between 6 and 9 weeks produces nasal hypoplasia with a depressed root and a deep groove between the alae and the nasal tip, stippling of uncalcified epiphyses with axial distribution, in the proximal femora and the calcanei, which disappears after the first year, hypoplastic distal phalanges with an inverted triangle configuration, low birthweight with catch-up in most cases and, occasionally, choanal atresia, clefts, heart disease, upper airway obstruction and cervical instability, with normal cognition in most cases; exposure between 14 and 20 weeks produces, instead, abnormalities of the central nervous system and of the eye [110]. The risk is dose-dependent, with about 80% poor outcome above 5 mg daily against 3 to 10% below that dose, and the embryopathy occurs in 6 to 30% of fetuses exposed between 6 and 12 weeks [15]. Maternal vitamin K malabsorption produces a clinically indistinguishable picture, and congenital vitamin K epoxide reductase deficiency gives rise to a pseudo-warfarin embryopathy [73]. Maternal phenytoin appears in the same list of causes of acquired chondrodysplasia punctata, so the overlap is real and not merely conceptual [73]; in phenytoin phenocopies, the stippling is usually confined to the tarsal bones, which helps to distinguish it from the axial and proximal distribution of the genetic dysplasias [75].

6.5. Other embryopathies

Fetal alcohol spectrum disorders share with hydantoin embryopathy small nails, pre- and postnatal growth restriction, facial dysmorphism and cognitive deficits, and are expressly listed in the same differential by GeneReviews [16]. The distinction is made by the history of maternal consumption and by targeted dysmorphological assessment [104]. Maternal diabetic embryofetopathy enters the differential through the combination of heart defects, growth abnormalities and optic nerve hypoplasia, and is distinguished by documentation of periconceptional maternal glycaemic control [98, 104]. Valproate exposure is distinguished by a neural tube defect risk of about 5%, by the more marked facies and by the dose-dependent cognitive effect [8, 82].

6.6. Coexistence of exposure and a genetic cause

Coexistence is the clinically most demanding situation, and it is predictable: maternal epilepsy may have a heritable genetic basis, including microdeletions at 15q11.2, 15q13.3 and 16p13.11, in which case the child inherits the predisposition and is simultaneously exposed to the drug [8]. Documentation of the exposure therefore does not remove the need for genetic investigation in a child with developmental delay or dysmorphism, and identification of a genetic cause does not exclude that the exposure contributed to specific malformations [8, 15].

7. Treatments and Risk Management Strategies

7.1. Principle: there is no treatment of the embryopathy, there is management of the exposure

Hydantoin embryopathy has no specific treatment, pharmacological or otherwise, and no intervention reverses established malformations [21]. What exists divides into two fronts: prevention, which is played out in the choice and dose of the antiepileptic before conception and in periconceptional supplementation, and treatment of the consequences in the child, which is surgical for correctable malformations and consists of rehabilitation and surveillance for neurodevelopment [8, 21]. A search of ClinicalTrials.gov identified no interventional trials on phenytoin embryopathy or on prevention of the teratogenicity of antiepileptics; the relevant studies are observational, namely the MONEAD programme and the NEAD programme, both completed [112]. A search without restriction on study type confirmed the absence of any registered study on the entity, either under the designation of fetal hydantoin syndrome or under that of fetal anticonvulsant syndrome, and identified as relevant records only observational studies, among them MONEAD (565 participants, completed in December 2022), NEAD (331 participants, completed in 2012), the pregnancy registries of two manufacturers, a European congenital anomaly study and a physiologically based pharmacokinetics study in pregnancy and lactation recruiting until 2028; among the studies with phenytoin or fosphenytoin as the intervention, none has pregnancy as its setting, and among those terminated or withdrawn none related to teratogenicity [113]. The absence of registered trials in this search does not demonstrate the absence of research in the area.

The therapeutic decision has to weigh both sides of the risk. The fetal risk of exposure is set against the morbidity and mortality of seizures for mother and fetus: accidents, social consequences such as loss of the driving licence, sudden unexpected death in epilepsy, documented in maternal mortality statistics, fetal trauma from maternal falls, status epilepticus in pregnancy and prolonged seizures in labour, with 1 to 2% of women with epilepsy having a seizure during labour [8]. The joint North American guideline translates this balance into two recommendations of opposite and complementary direction: minimise convulsive, generalised tonic-clonic and focal-to-bilateral seizures during pregnancy (level A), and exercise caution in withdrawing or switching a drug effective against tonic-clonic seizures in an established pregnancy, even if that drug is not optimal for the fetus (level B) [21]. The Oxford Desk Reference states the same rule more directly: avoid withdrawal or change of antiepileptic once pregnancy is established, because the risks generally outweigh the possible gains, and the mother should not stop or alter medication without discussing it with the neurologist [8].

7.2. Phenytoin: indications, formulations and doses

Phenytoin is indicated for the control of tonic-clonic and partial seizures, or of a combination of both, and for the prevention and treatment of seizures during or after neurosurgery or severe head injury, with second-line use in trigeminal neuralgia when carbamazepine is ineffective or not tolerated [25]. In the United States, the registered indications for the chewable formulation are the treatment of generalised tonic-clonic and complex partial seizures and the prevention and treatment of seizures during or after neurosurgery [4]. Fosphenytoin, an injectable prodrug, is indicated for the treatment of generalised tonic-clonic status epilepticus and for the prevention and treatment of seizures during neurosurgery, in all paediatric age groups [5].

The available formulations comprise phenytoin sodium capsules, oral suspension and chewable tablets of phenytoin base, and fosphenytoin sodium solution for injection [4, 5, 25]. Equivalence between salts requires attention: 100 mg of phenytoin sodium corresponds to 92 mg of phenytoin on a molecular weight basis, but these molecular equivalents are not necessarily biological equivalents, so a change of formulation requires care [25].

The dose must be individualised, because there is wide interindividual variability in serum concentrations with equivalent doses; introduction is with small doses and gradual increments until seizure control or the appearance of toxic effects, and maintenance should be the lowest dose compatible with seizure control [25]. Pharmacokinetics are non-linear and saturable, with autoinduction of CYP2C9, which explains why small dose increases can produce large increases in concentration [14]. Phenytoin is highly protein bound, usually 90% in adults, with a volume of distribution between 0.52 and 1.19 L/kg, and variability in albumin binding confounds monitoring when the free fraction is not measured [14, 25]. Intravenous injection of fosphenytoin carries a boxed warning about the cardiovascular risk associated with rapid infusion rates, with severe hypotension and arrhythmias that have also been reported at rates equal to or lower than that recommended, requiring careful cardiac monitoring during and after administration [5].

The adverse effects relevant to a woman on long-term treatment, independently of pregnancy, include gingival hyperplasia, more frequent in childhood and with poor oral hygiene; hirsutism and hypertrichosis; coarsening of the features and enlargement of the lips; severe cutaneous reactions, with acute generalised exanthematous pustulosis, Stevens-Johnson syndrome, toxic epidermal necrolysis and drug reaction with eosinophilia and systemic symptoms, the last sometimes fatal; haematopoietic complications, including thrombocytopenia, agranulocytosis, pancytopenia and aplastic anaemia, as well as macrocytosis and megaloblastic anaemia that respond to folic acid; pure red cell aplasia, added to the European product information in 2021 with unknown frequency; reduced bone mineral density, with osteopenia, osteoporosis, osteomalacia and fractures, through induction of cytochrome P450 and increased metabolism of vitamin D3, with a fracture risk 2 to 6 times higher with chronic exposure to antiepileptics and an almost doubled annual bone loss in phenytoin users, mediated by accelerated vitamin D catabolism, reduced calcium absorption, secondary hyperparathyroidism and increased resorption, with a possible contribution from reduced vitamin K [114]; and the suicidal ideation and behaviour described with antiepileptics in general [25, 67]. The summary of product characteristics recommends measuring serum folate at least every six months and supplementing when necessary, and the British guideline recommends considering vitamin D and calcium in long-term treatment with carbamazepine, phenytoin, primidone and valproate because of reduced bone mineral density and osteomalacia [25, 47]. The median lethal dose in adults is estimated at between 2 and 5 g, with nystagmus, ataxia and dysarthria as the initial symptoms of overdose [25].

7.3. Interactions with reproductive consequences

Enzyme induction by phenytoin has two direct consequences in women of childbearing age. The first is reduced efficacy of hormonal contraceptives, which may lead to an unplanned pregnancy under exposure; oral contraceptives and oestrogens are among the drugs whose effect is reduced by phenytoin, and the European summary states that, through enzyme induction, phenytoin may result in loss of therapeutic effect of hormonal contraceptives [4, 25]. The British guideline includes carbamazepine, oxcarbazepine, phenytoin and topiramate among the drugs that may compromise the efficacy of hormonal contraceptives [47]. The Scottish guideline completes the list of hepatic inducers with eslicarbazepine, perampanel above 12 mg per day, phenobarbital, primidone, rufinamide and topiramate above 200 mg per day, and notes that enzyme induction persists 14 to 28 days after discontinuation of the antiepileptic [24]. The review of lifelong counselling classifies phenytoin as an inducer with a strong interaction with hormonal contraceptives and recommends preferring the intrauterine device, adding a barrier method if an oral contraceptive is continued [114]. The second consequence is reduction of folic acid concentrations, relevant to periconceptional supplementation [4, 25].

Among interactions between antiepileptics, phenytoin reduces the concentrations of carbamazepine, lamotrigine, topiramate, oxcarbazepine, lacosamide and felbamate, has an unpredictable effect on phenobarbital and valproate, and increases the risk of valproate-associated hyperammonaemia when used in combination [4, 67]. The 2021 European review also added to the product information interactions with the direct oral anticoagulants, namely dabigatran and rivaroxaban, with reduction of their plasma concentrations, with lacosamide and with ticagrelor, through CYP3A4 induction [67].

7.4. Pharmacokinetics and monitoring in pregnancy

Pregnancy measurably alters phenytoin kinetics, in two directions that may cancel each other out in the interpretation of the total concentration. Plasma clearance increases: in a study of 86 pregnant women with epilepsy, with free and total concentrations by trimester compared with values obtained at least ten weeks after delivery, plasma clearance of phenytoin increased from the first trimester, while the increase in clearance of the free fraction was less pronounced and only significant in the third trimester [106]. At the same time, the free fraction rises: binding to serum proteins decreases in pregnancy, especially in the last eight weeks, partly through a fall in albumin, which raises the free fraction of diazepam, phenytoin and valproate [115]. The fosphenytoin label describes that plasma clearance of phenytoin generally increases during pregnancy, with a peak in the third trimester, and returns to the pre-pregnancy level some weeks or months after delivery [5].

The practical consequence is explicit in the product information: decreased serum concentrations and increased seizure frequency may occur during pregnancy because of altered pharmacokinetics, so the serum concentration should be measured periodically and the dose adjusted, and, because of changes in protein binding, monitoring should be based on the unbound fraction; after delivery, restoration of the original dose is likely to be indicated [4]. The British review notes that phenytoin is, with lamotrigine, one of the drugs whose concentrations are most affected in pregnancy and postpartum [22]. The North American guideline recommends monitoring concentrations throughout pregnancy according to the pharmacokinetics of the drug and the clinical situation, and adjusting the dose if concentrations fall or seizure control worsens (both level B) [21]. The British guideline sets out the pathway in detail: obtain a baseline concentration before conception for carbamazepine, lamotrigine, levetiracetam, oxcarbazepine, phenobarbital and phenytoin, checking adherence at the same time; monitor and adjust according to the regulator's guidance; discuss the results with the woman; and establish an antenatal plan for return to preconception doses, to begin in the first days after delivery [47]. The Oxford Desk Reference adds particular monitoring of lamotrigine and, possibly, oxcarbazepine and levetiracetam concentrations, with adjustment according to seizure control and consideration of a dose increase if there is a pronounced fall relative to the optimal pre-pregnancy value [8]. The magnitude of the fall has been quantified: in pregnancy, the phenytoin concentration falls to about 78% of baseline, with comparable reductions for ethosuximide (70%), lacosamide (83%), valproate (84%) and phenobarbital (89%), and variation below 10% for carbamazepine, clonazepam and topiramate [114].

The guidelines diverge on routine monitoring. The British obstetric guideline recommends against routine serum measurement of antiepileptics in pregnancy, reserving it for individual circumstances such as suspected non-adherence, toxicity or intractable seizures, and notes that lamotrigine may fall by up to 70% [23]. The Scottish guideline also considers routine monitoring not indicated, but expressly identifies three situations in which it is useful, adjustment of the phenytoin dose, assessment of adherence and suspected toxicity, adding that circulating phenytoin concentrations tend to correlate with seizure control and that total concentrations may mislead because of variation in the free fraction [24]. The British guideline and regulator, by contrast, recommend obtaining a preconception baseline concentration and monitoring throughout pregnancy for a set of drugs that includes phenytoin [22, 47]. In practice, the intermediate position is that of the Scottish guideline, which separates indiscriminate measurement from measurement targeted at the drug with non-linear kinetics.

7.5. Comparative regulatory table

Access and reimbursement in Portugal and Brazil do not appear in the sources consulted, so this article does not document the national situation of these medicines or any differences in labelling approved by the Portuguese and Brazilian authorities [4, 25].

7.6. Evidence base: study designs and their limits

None of the risk data come from randomised trials, which is inevitable: it is not possible to randomise exposure to a teratogen, and the EURAP analysis of fetal growth itself did not include a control group without exposure, because stopping the antiepileptic in a woman with active epilepsy is generally not an option and the clinical question is not whether treatment can be avoided, but which treatment is safest [45]. The available studies are observational of necessity.

Several limitations condition the reading of these figures and should accompany any communication of risk. The studies are observational and subject to confounding by indication, given that the drug is chosen according to the type and severity of the epilepsy [17, 45]. The phenytoin estimates rest on small denominators in contemporary cohorts, between 142 and 1604 exposures, against more than 10,000 for lamotrigine, which produces wide confidence intervals and reduces the confidence assigned by the guideline to the low classification [18, 21]. The temporal reduction in the prevalence of major malformations observed in the European registry, from 6.1% to 3.7% between the periods 1998–2004 and 2015–2022, is an effect of prescribing composition and not a reduction in the risk of each drug, and should not be read as an improvement in the safety of phenytoin [18]. It is not legitimate to compare percentages directly between studies with different designs, periods and definitions of malformation: the difference between 2.8% in the North American registry and 5.4% in the cohort meta-analysis does not measure a temporal change in risk, but a difference in population and in detection method [17, 19]. Relative risks should be read together with the absolute rate: a relative risk of 3.81 corresponds to a risk difference of 0.03, that is, three additional cases per 100 exposed [17]. Null and negative results are included and should not be omitted: most comparisons of phenytoin with other traditional antiepileptics did not reach significance, no specific malformation reached significance in the comparisons with controls, and the dose analysis of fetal growth in EURAP had intervals that include the null value for all outcomes [17, 45]. In a comparison in which heterogeneity was substantial, with an I² of 58% between levetiracetam and phenytoin, significance disappeared in the random-effects model [17]. The risk ratios published in the 2025 summary of the North American registry against lamotrigine are numerically identical to those of the 2012 summary, so the absolute proportions are preferred here [19, 117]. There are additional limitations specific to the growth analysis: the cohort was not population-based, had limited representation of ethnicities other than white and Asian, did not include exposures for indications other than epilepsy, did not allow control for maternal weight or nutritional status, and used a standard reference instead of country-specific centiles, with an overall prevalence of small for gestational age of 20%, similar to the 17 to 22% already described but higher than the 11% estimated by the North American registry [45].

7.7. Choice of drug before conception

The central recommendation of the joint North American guideline is to choose the optimal drug and dose for seizures and for the fetus as early as possible before conception, for example at the time of starting the drug after menarche, within a shared decision with the woman (level B for both) [21]. For reducing malformations, the guideline recommends considering lamotrigine, levetiracetam or oxcarbazepine (level A), avoiding valproate (level A), avoiding phenobarbital to reduce cardiac malformations (level A) and avoiding phenobarbital and topiramate to reduce oral clefts (level B), as well as informing the woman that the baseline prevalence of major malformations is 2.4 to 2.9% (level A) [21]. For fetal growth, it recommends avoiding valproate or topiramate (level B) and monitoring growth when these are used (level B) [21]. For neurodevelopment, it recommends avoiding valproate (level A) [21]. Phenytoin does not appear in any of these specific recommendations: it is neither among the drugs of lowest risk nor among those recommended to be avoided for a given outcome, occupying in the quantitative data an intermediate position, above lamotrigine, levetiracetam and oxcarbazepine and below valproate [21].

The European guidance is more directive with regard to phenytoin in particular. The harmonised text requires that, if a pregnancy is planned, every effort be made to switch to a suitable alternative treatment before conception and before stopping contraception, and that, if no alternative exists after weighing risks and benefits and treatment is continued, the lowest effective dose be used; if the woman becomes pregnant while on phenytoin, she should be referred to a specialist to reassess the treatment and consider alternatives [25, 67]. The British committee's review recommends discussing the risks at the start of treatment and at the annual review, referring urgently women who are planning pregnancy, using monotherapy at the lowest effective dose and reporting cases through the national system [22]. The Oxford Desk Reference summarises the preconception pathway: refer to a neurologist to assess withdrawal, if the woman has been seizure-free for several years, simplification to monotherapy instead of polytherapy, or dose reduction if seizures are controlled [8]. Optimisation should be carried out and efficacy assessed before conception, with reassessment of the indication, choice of a drug suited to the type of epilepsy and of lowest teratogenic potential, monotherapy at the minimum effective dose, in particular for tonic-clonic seizures, and measurement of the plasma concentration once the dose is optimised [8].

The British obstetric guideline, endorsed by five societies and colleges, adds the reassuring side of counselling, which is part of the same consultation: the risk of malformation is low without exposure to antiepileptics in the periconceptional period, with 2.8% (26/939) in children of women with epilepsy without antiepileptics in the first trimester, and the risk depends on the type, number and dose of drugs; the same guideline identifies the association of phenobarbital and phenytoin with cardiac malformations and of phenytoin and carbamazepine with cleft palate, and records the evidence as insufficient for eslicarbazepine, gabapentin, lacosamide, oxcarbazepine, perampanel, pregabalin, topiramate and zonisamide [23]. As for neurodevelopment, the same source states that there is very little evidence for levetiracetam and for phenytoin, and cites the Cochrane review that found no difference in developmental quotient between carbamazepine, lamotrigine and phenytoin and the children of women without epilepsy or with untreated epilepsy [23]. No change should be abrupt: faced with an unplanned pregnancy, the guidance is urgent discussion with the specialist, never immediate discontinuation or substitution, and an invitation to enrol in the national pregnancy registry [23].

The report of the International League Against Epilepsy task force on women and pregnancy, published in 2019 with its own executive summary, covers the maternal and fetal risks of seizures, structural and behavioural teratogenicity, the effect of pregnancy on seizure control and on pharmacokinetics, preconception optimisation and management of pregnancy, and formulates expert-opinion recommendations [118, 119]. The same task force's survey of the organisation's 118 national chapters, with 77 responses, showed that 68% had guidelines, but that 34% of these dated from 2014 or earlier, that at least 20% contained no information on cognitive risk, risks by drug or choice of antiepileptic, that 34% did not address concentration measurement and that 19% gave no guidance on breastfeeding [120]. The full text of the numbered 2019 recommendations is not open access and is not reproduced here; in particular, the recommendation of monthly concentration monitoring with phenytoin and phenobarbital, attributed to that source by one of the published cases, was not confirmed in the original document [54, 118].

7.8. Folic acid: the divergence between recommendations

The two main sources diverge on the dose. The joint North American guideline recommends folic acid at a dose of at least 0.4 mg per day, before conception and during pregnancy, to reduce the risk of neural tube defects (level B) and possibly to improve neurodevelopment (level A), with counselling on adherence (level B), expressly stating that the optimal dose and timing are unknown [21]. The British review recommends that all women taking antiepileptics who are planning pregnancy be offered folic acid 5 mg per day before any possibility of pregnancy [22], and the British guideline refers to this guidance [47]. The British obstetric guideline recommends the same dose of 5 mg daily to all women with epilepsy, from at least three months before conception until the end of the first trimester, justifying it by the possible cognitive benefit and by the safety of the supplement, and declaring the studies on the effect of folate on malformations to be discordant, with two in favour and two showing no benefit [23]. The Scottish guideline grades the recommendation in a revealing way: 400 µg daily in women with epilepsy not taking antiepileptics has grade A, as do 5 mg in women not taking antiepileptics with a family history or a previous child with a neural tube defect or with a body mass index above 30, whereas 5 mg daily in women on antiepileptics has only grade D, because the data are insufficient to guide the dose; the same source adds that phenytoin and phenobarbital may exert their teratogenic effect through mechanisms other than folate depletion [24]. The international survey shows that practice reflects this uncertainty: 91% of national chapters recommended folate, with doses between 0.4 mg and 4 mg or more per day [120]. The Oxford Desk Reference likewise adopts 5 mg daily in the preconception period in women on antiepileptics [8], with the justification that the high dose is appropriate when the mother needs to continue anticonvulsant therapy, this being the dose used in the prevention of recurrence of neural tube defects, from at least one month before conception until 12 weeks, against 400 µg daily in primary prevention [108]. The source notes that there are no known adverse effects or contraindications with 5 mg daily, adding that in women with epilepsy the anticonvulsant treatment should be reviewed [108].

The divergence is not only about dose, but about target. The systematic review of the North American guideline concluded that there is probably no demonstrated benefit of folic acid at a dose of at least 0.4 mg daily in preventing major malformations in children of women with epilepsy, and that the same dose is possibly associated with fewer autistic traits at 3 years and probably with a full-scale intelligence quotient about 6 points higher at 6 years [21]. In the NEAD study, periconceptional folate was associated with a mean intelligence quotient of 108 (95% CI 106–111) against 101 (98–104) in its absence (p=0.0009) [23, 82].

The timing of supplementation is today the most consequential piece of data, and it comes from a target trial emulation using registries from Denmark, Iceland, Norway and Sweden between 1997 and 2020, in women on antiepileptics, with 18,255, 13,619 and 12,365 eligible pregnancies in the three windows analysed: starting high-dose folate, of 4 or 5 mg daily, between one and twelve weeks before pregnancy was associated with major congenital anomalies in 2.6% against 4.8% in non-initiators, an absolute reduction of 2.2 percentage points (95% CI 0.5–3.5) and a relative reduction of 45% (relative risk 0.55; 0.25–0.91), with no reduction when initiation occurred between one and twelve weeks after conception (1.27; 0.94–1.64) or between thirteen and fifty-two weeks before (0.99; 0.65–1.36); in the subgroup exposed to valproate, topiramate, carbamazepine, phenobarbital or phenytoin, initiation between one and twelve weeks before was associated with a reduction of 86% (0.14; 0.03–0.34) [121]. The gap between recommendation and practice has been measured in the same region: among 8695 pregnancies with treated epilepsy from 2006 to 2017, 54.3% had a prescription for high-dose folate, with 74.3% in Sweden, 41.4% in Norway and 34.3% in Denmark, and 42% of users did not start before conception [122]. In the Indian Kerala registry, the use of folate of 5 mg or more per day in the month before conception rose from 43.9% to 81% between 1998 and 2017 and in the first trimester from 52.7% to 86.6%, with a fall in seizures during pregnancy from 57.2% to 32.9% and replacement of phenobarbital, phenytoin and clonazepam by newer drugs, without any change in the rate of major malformations, from 7.5% to 7.3% [123].

Against the high dose weighs an oncological signal. In the registry cohort from Denmark, Norway and Sweden with 3,379,171 children born between 1997 and 2017, exposure to prescriptions of folic acid at a dose equal to or greater than 1 mg daily between 90 days before the start of pregnancy and delivery was associated, in the 27,784 children of mothers with epilepsy, of whom 5934 were exposed to a mean dose of 4.3 mg, with an adjusted risk of 2.7 (95% CI 1.2–6.3) of cancer before the age of 20, with an absolute risk of 1.4% (0.5–3.6) in the exposed against 0.6% (0.3–1.1) in the unexposed, with 18 exposed cases and 29 unexposed; in children of mothers without epilepsy the adjusted risk was 1.1 (0.9–1.4), and antiepileptics without high-dose folate were not associated with cancer [124]. A later analysis in the women who gave birth themselves, with 64,485 exposed among 1,465,785, gave an adjusted risk of 1.2 (1.1–1.2), attenuated to 1.1 (1.04–1.2) with a six-month lag, with 2.0 (1.3–2.9) for non-Hodgkin lymphoma and 1.3 (1.0–1.8) in women with epilepsy [125]. The results were contested and the authors maintained them in a published response, so the controversy is documented and unresolved [126]. The practical reading is that the high dose has a demonstrated benefit when started before conception and a risk signal that justifies not prolonging it indefinitely or using it without indication, and that the divergence between the recommendations of at least 0.4 mg and of 5 mg daily is not resolved by the available evidence [21, 23, 24, 121, 124]. In the 6-year assessment of the MONEAD programme, folic acid in the first 12 weeks was positively associated with cognition and behaviour, with no risk signal at higher doses [127]. In the EURAP fetal growth analysis, folate intake after conception reduced the risk of low birthweight relative to no supplementation or underdosing [45]. It should be noted that phenytoin reduces folate concentrations and that folic acid reduces phenytoin concentrations, so introducing a high dose into an established regimen justifies monitoring of the serum concentration of the antiepileptic [4, 25]. The recommendation of 4 mg daily attributed to a public health authority by one of the published cases was not verified in the primary source and is not adopted here [54].

7.9. Perinatal vitamin K

The two product information documents recommend administration of vitamin K to the mother and to the newborn to prevent or correct haemorrhagic disease due to reduced vitamin K-dependent clotting factors: the US label indicates administration to the mother before delivery and to the neonate after birth, and the British summary indicates administration to the mother during the last month of gestation and to the newborn after delivery [4, 25]. Neither product information specifies the dose, route or duration, and the joint North American guideline does not address this intervention [4, 21, 25]. The two British guidelines fill that gap as regards the newborn and diverge as regards the mother. Both recommend 1 mg of vitamin K intramuscularly at birth: the obstetric guideline frames it as a point of good practice for all children of women with epilepsy taking enzyme-inducing antiepileptics, listing among these carbamazepine, phenytoin, phenobarbital, primidone, oxcarbazepine, topiramate and eslicarbazepine, and the Scottish guideline frames it as administration to all children of women with epilepsy, on the same terms on which any newborn receives it [23, 24]. As for antenatal maternal administration, the obstetric guideline considers the evidence insufficient to recommend it routinely, both for preventing haemorrhagic disease of the newborn and for preventing postpartum haemorrhage, basing itself on a systematic review without randomised trials and on two observational studies showing no benefit, in which the newborns also received 1 mg intramuscularly at birth; the Scottish guideline, differing, allows consideration of maternal oral phytomenadione 10 mg per day in the third trimester when there are additional risk factors for haemorrhagic disease, such as maternal liver disease or anticipated preterm delivery, with grade D [23, 24]. The approach the two share, and which is supported, is neonatal prophylaxis; maternal administration is optional, directed at cases of increased risk, and its dose is sourced only from the Scottish guideline.

A head-on divergence between the same two guidelines concerns antenatal corticosteroids in women on enzyme inducers. The Scottish guideline recommends a double dose of betamethasone or dexamethasone, 48 mg over 12 to 24 hours, with grade D; the obstetric guideline expressly recommends against doubling the dose, also with grade D, acknowledging that phenytoin, carbamazepine and phenobarbital increase corticosteroid metabolism but that there are no studies demonstrating benefit from doubling [23, 24]. Without comparative evidence, the decision depends on the local protocol, and the divergence should be made explicit to the woman and to the neonatal team.

7.10. Management in established pregnancy and at delivery

Once pregnancy is established, the rule is therapeutic stability with increased surveillance. Withdrawal or change of antiepileptic should be avoided, monitoring should be closer, the dose should be adjusted according to seizure control, a dose increase should be considered if there is a pronounced fall in concentrations relative to the optimal pre-pregnancy value, and detailed fetal morphological assessment should be offered [8, 21]. The British guideline recommends referring to the epilepsy team those who are planning pregnancy or are pregnant, ensuring information sharing between epileptology, obstetrics and primary care, reinforcing adherence and not stopping without supervision, discussing risks and benefits of any adjustment, and increasing the frequency of reviews in those with intellectual disability, under 16 years of age, active epilepsy with a seizure in the last 12 months, bilateral tonic-clonic seizures or modifiable risk factors for sudden death [47]. Delivery and the three days around it are the period of greatest seizure exacerbation, with an odds ratio of 3.7 (95% CI 2.4–5.9) in women who had seizures in the month before pregnancy; tonic-clonic seizures occur in 1 to 2% of deliveries and in another 1 to 2% in the following 24 hours, with 3.5% (60/1956) of seizures at delivery in the European registry and 3.5 to 5% in the Scottish estimate [23, 24]. The recommended approach is to continue the antiepileptic during labour, parenterally if necessary, with phenytoin among the options available by that route, and to consider prophylactic clobazam when the risk is very high [23]. A seizure during labour is treated with benzodiazepines, namely intravenous lorazepam 0.1 mg/kg in adults, usually as a 4 mg bolus, with an additional dose after 10 to 20 minutes (no maximum dose defined in the source), or diazepam 5 to 10 mg by slow intravenous injection, and, without intravenous access, rectal diazepam 10 to 20 mg repeatable once after 15 minutes or buccal midazolam 10 mg; if the seizure persists, phenytoin or fosphenytoin is used [23]. The loading dose indicated by the sources is not uniform: the British obstetric guideline indicates 10 to 15 mg/kg by intravenous infusion, with about 1000 mg in adults, the Scottish guideline indicates 18 mg/kg at 50 mg per minute with electrocardiographic monitoring and reduction of the rate in case of hypotension or arrhythmia, and the fosphenytoin label indicates 15 to 20 mg of phenytoin equivalents per kg; the difference is one of emergency regimen and not of teratogenicity, but should be resolved by the institution's protocol [5, 23, 24]. When in doubt between eclampsia and an epileptic seizure, the Scottish guideline indicates magnesium sulfate 4 g intravenously over 5 minutes followed by 1 g per hour for 24 hours, in addition to the benzodiazepine [24]. Analgesia and anaesthesia require caution with pethidine, because of its epileptogenic metabolite, with a preference for diamorphine, and avoidance of pethidine, ketamine and sevoflurane in general anaesthesia [23]. In the first ten days after delivery, any dose increase made during pregnancy should be reviewed, because of the risk of toxicity [23].

In the perioperative setting and in emergencies, fosphenytoin retains its indication in status epilepticus, and its use in a woman of childbearing age in a context of clinical need is expressly envisaged in the European text, which acknowledges that the magnitude of the fetal risk with short-term use in emergency situations is unknown, the risk of fetal harm having to be weighed against the risk of the indication itself for the fetus and the pregnant woman [5, 67].

7.11. Breastfeeding

The positions diverge. Phenytoin is excreted in human milk, according to the US label, which adopts the neutral wording of weighing the benefits of breastfeeding for development and health against the mother's clinical need and the possible adverse effects in the infant [4]. The British summary states that phenytoin appears to be excreted in low concentrations in human milk and, on that basis, does not recommend breastfeeding in women receiving the drug [25]. The British guideline, in the opposite direction, considers breastfeeding generally safe in women on antiepileptics and recommends encouraging it [47]. The most recent empirical data come from the MONEAD programme: in 273 children of women with epilepsy and 84 controls, at 6 years there was no difference in language (107.93 against 107.18) or in adaptive behaviour (102.91 against 102.14), and breastfeeding for six months or more was associated with better adaptive behaviour than its absence (estimate 4.68; 95% CI 1.34–8.03), with 5.88 (2.21–9.55) across all monotherapies [128]. The regimens in this cohort were contemporary, mainly lamotrigine and levetiracetam, and phenytoin was not analysed separately [128].

For phenytoin there are, however, specific data. Transfer into milk is minimal, with a ratio between the milk concentration and the maternal serum concentration of 0.06 to 0.19, the lowest in the group after valproate, against 0.36 to 0.41 for carbamazepine, 0.6 for lamotrigine, 0.66 to 1.1 for topiramate and 1 to 3.09 for levetiracetam [23]. The drugs and lactation database gives absolute values of 0.26 to 4.5 mg/L in milk, depending on the dose and the time of collection, with 1.0 to 2.6 mg/L in the first four days on 300 mg daily and 1.4 mg/L before the first dose and 4.2 mg/L two hours after the last on 100 mg three times a day, corresponding to an estimated infant dose of 0.5 to 8% of the weight-adjusted maternal dose; in six infants of mothers on 200 to 400 mg daily, only two had detectable serum phenytoin between the first and third month, at 126 and 197 µg/L [129]. The same source concludes that breastfeeding on phenytoin monotherapy does not appear to affect growth or development, that in one study breastfed children had higher intelligence quotients and verbal abilities at 6 years, and that the mother's need for the drug is not a reason to stop breastfeeding; the adverse effects described are few, with 28 infants of mothers on 100 to 200 mg three times a day without reactions, one old report of methaemoglobinaemia, two hyperexcitable infants when phenytoin ceased to be detectable in serum at three to six weeks, and one fatal case of vitamin B12 deficiency with macrocytic anaemia and tremors in a 4-month-old infant exclusively breastfed by a mother who had been on phenytoin for thirteen years without supplementation [129]. Combination with sedating anticonvulsants or with psychotropics is the risk scenario for sedation or withdrawal symptoms in the infant, and the Scottish guideline adds enhanced surveillance in the preterm infant, in jaundice, in polytherapy and at high doses [24, 129]. The British obstetric guideline encourages breastfeeding with grade C, noting that in the NEAD programme, which included phenytoin, breastfeeding had no effect on the intelligence quotient at 3 years and that psychomotor development at 6 and 18 months was better in breastfed children [23, 129].

7.12. Treatment of the affected child

Major malformations follow the management proper to each, surgical or medical: repair of cleft lip and palate, surgical correction of heart disease, correction of hypospadias, orthopaedic treatment of talipes and of positional deformities [1, 130]. Intervention for neurodevelopment consists of rehabilitation and educational support, guided by formal assessment, and the North American guideline recommends age-appropriate developmental screening in all children exposed to any antiepileptic (level B), with specific counselling about valproate relative to the other drugs, including phenytoin (level A) [21]. The European summary requires that, after birth, children be closely monitored for neurodevelopmental disorders [25]. There is no drug with an approved indication for the embryopathy and no nutritional or neuromodulation intervention of demonstrated efficacy in this condition [21, 112].

7.13. Developments of the last twelve months

The relevant developments in the window between October 2025 and October 2026 are all at observational or documentary level, without any new regulatory decision on the teratogenicity of phenytoin. On 11 March 2026, the North American administrative-data population cohort on exposure in the second half of pregnancy concluded that phenytoin and levetiracetam were not associated with an increased risk of any neurodevelopmental disorder, unlike valproate and zonisamide [90]. In June 2026, the target trial emulation using Nordic registries showed that the benefit of high-dose folic acid depends on its being started between one and twelve weeks before conception [121]. Also in 2026, the review of lifelong counselling quantified the fall in phenytoin concentration in pregnancy to 78% of baseline and the fracture risk associated with chronic exposure [114], and the analysis of maternal and cord concentrations from the MONEAD programme confirmed the almost complete placental passage of antiepileptics, without an estimate of its own for phenytoin because only a single pair was available [96]. In 2024, two works entered the evidence base of this article: the EURAP analysis of 9840 monotherapies, which gave 6.3% major malformations with phenytoin and an odds ratio of 2.39 against levetiracetam [18], and the systematic examination of the digits that documented distinct effects of phenytoin, phenobarbital and carbamazepine [72]. On 16 July 2025, the update of the North American registry covering 1997 to 2023 was published, with 7311 participants on monotherapy and 2.8% major malformations with phenytoin (12/423), and with topiramate rising to 5.1% (26/510) [19]. In September 2026, the EURAP fetal growth analysis associated phenytoin with a higher, not lower, birthweight centile than lamotrigine, reversing the expectation created by the historical description of growth deficiency [45]. In the same month, the breastfeeding analysis of the MONEAD programme found no adverse effect at 6 years and associated prolonged breastfeeding with better adaptive behaviour [128]. The British guideline was updated on 5 August 2026, in a change relating to sulthiame unrelated to pregnancy [47]. On the regulatory front, the current versions of the US product information date from 12 May 2026 for the chewable tablets and 28 August 2025 for fosphenytoin, both retaining the description of fetal hydantoin syndrome, and no new authorisations, opinions, suspensions or withdrawals relating to phenytoin were identified in this window [4, 5]. This absence results from the sources consulted and does not constitute proof of non-existence.

8. Reproductive Options and Family Counselling

8.1. Nature of the recurrence risk

The condition is not hereditary and the risk in a subsequent pregnancy depends on the exposure, not on transmission of a genotype [2]. The practical consequence is that the risk is modifiable: keeping the same drug at the same dose, the risk of an affected sibling is high; by changing the exposure before conception, the risk approaches that of the general population treated with the drugs of lowest teratogenic potential [15, 21].

Quantification of this risk requires a correction to the most cited source. The Oxford Desk Reference estimates the sibling recurrence risk at between 39 and 55%, if the mother continues the same drug at the same dose, attributing the estimate to the population-based retrospective cohort from the Grampian region of Scotland [8]. Reading the full text of that study shows that the lower limit corresponds to one of its results and that the upper limit does not appear in the article. The study assessed mothers on antiepileptics in pregnancy between 1976 and 2000, identified in hospital obstetric records, with 149 families and 293 children assessed, and described neonatal withdrawal in 20% of exposed children, congenital malformations in 14% of exposed pregnancies against 5% in unexposed siblings, developmental delay in 24% against 11%, or 19% against 3% excluding a family history of delay, facial dysmorphism in 52% against 25%, medical problems in childhood in 31% against 13%, behavioural disorders in 20% against 5%, and major malformation or developmental delay in 31% of the exposed [92]. The recurrence calculation itself started from the 68 families, out of 149, with one child affected by major malformation or developmental delay: counting the first affected child of each family as the proband, the probands had 65 siblings, of whom 19 were also affected, giving a crude recurrence rate of 29%, reduced to 26% (14/54) excluding families with a family history of developmental delay; applying Weinberg's proband method, to correct for incomplete ascertainment in single-child families, recurrence rises to 43%, or 39% excluding a positive family history [92]. The figure of 39% is therefore the corrected estimate with exclusion of family history, and the limit of 55% is not reproducible in the source cited by the manual; the nearest value in the article is the proportion of 52 to 70% of children with fetal anticonvulsant syndrome facies among those exposed to carbamazepine, valproate, phenytoin or polytherapy, a finding that, for phenytoin alone, did not reach significance (p=0.12) [92]. The outcome to which the recurrence applies is major malformation or developmental delay, and not major malformation alone, and the cohort is not specific to phenytoin: it brings together antiepileptics in general, in an era of polytherapy, with an overall relative risk of 2.38 (95% CI 1.03–5.51) for the regimens observed [92]. The authors themselves noted that some of the cases counted as malformations were diagnosed late, for example a soft palate cleft identified at one year and eight months, and that their exclusion would reduce the apparent malformation rate to 8% [92].

The guidelines give lower recurrence figures for malformation and on another basis. The British obstetric guideline reports a major malformation recurrence of 16.8 per 100 in women with a previously affected child [23], and the Scottish guideline grades it as 9.8% without a previous history, 16.8 to 35.7% with one previously affected child and 50% with two affected children [24]. The difference between these figures and those of the Scottish study results from the outcome measured, major malformation in one case and major malformation or developmental delay in the other, and should be made explicit in counselling.

The severity of the effect in a subsequent pregnancy is, moreover, unpredictable, because it depends on drug metabolism by the fetoplacental unit, partly genetically determined and variable between siblings; the classic example is a woman exposed to the same dose in three pregnancies with only one affected child [15]. The same principle applies in reverse: the phenotypic discordance described in heteropaternal dizygotic twins shows that fetuses with the same exposure can have different outcomes [48].

8.2. Content of preconception counselling

Counselling should begin long before any intention to become pregnant. The joint North American guideline recommends choosing the optimal drug and dose for seizures and for the fetus as early as possible, for example when starting treatment after menarche, within a shared decision (level B) [21]. The British review recommends discussing the risks at the start of treatment and at the annual review, and referring urgently women who are planning pregnancy [22]. The European text requires that a woman of childbearing age be fully informed of the risk of fetal harm, that she understands the risks and the need to plan pregnancy, and that she be advised to consult her doctor as soon as she plans to become pregnant, to discuss switching to another treatment before conception and before stopping contraception [25, 67].

The figures to communicate are the baseline prevalence of major malformations of 2.4 to 2.9% in the population without epilepsy [21], the risk with phenytoin of 2 to 3 times that value according to the European regulatory text [67], or about 6% in absolute terms according to the British SmPC [25], with the values from prospective registries between 2.8 and 6.3% [18, 19], and the lower-risk alternatives, which are lamotrigine, levetiracetam and oxcarbazepine (level A) [21]. Information on neurodevelopment should make explicit the divergence between the data specific to phenytoin, which show no deficit against carbamazepine or lamotrigine, and the regulatory position, which holds that a risk cannot be excluded because the studies are contradictory [21, 22, 67, 82]. Communication should equally include the maternal risk of uncontrolled seizures, which is an integral part of the decision [8, 21].

8.3. Contraception

Contraception is part of counselling and not an adjacent topic, because enzyme induction compromises the efficacy of hormonal contraceptives [47, 67]. The European text requires effective contraception during treatment and for up to one month after its discontinuation, with at least one effective method, such as an intrauterine device, or two complementary methods including a barrier method, chosen according to individual circumstances and with the woman's participation in the decision [25, 67]. A pregnancy test should be considered before starting treatment in a woman of childbearing age [67]. The Australian guidance reproduces the same content, adding that counselling should be given before starting treatment and repeated regularly during treatment, and recalling that these drugs are also used in indications other than epilepsy, such as migraine, bipolar disorder, trigeminal neuralgia, anxiety and neuropathic pain [116].

The magnitude of contraceptive failure has been quantified and justifies preference for methods independent of hepatic metabolism. The British obstetric guideline gives an oral contraceptive failure rate of 3.1 per 100 woman-years with enzyme inducers, about three times that expected, and the Scottish guideline indicates failure of up to 7 per 100 woman-years [23, 24]. In the eligibility classification reproduced by the obstetric guideline, combined hormonal methods and the progestogen-only pill fall into category 3 with inducers, the implant into category 2, and injectable medroxyprogesterone acetate, the levonorgestrel intrauterine system and the copper intrauterine device into category 1, with a recommendation for consistent condom use in the categories indicated [23]. The Scottish guideline is more restrictive regarding progestogens, in not recommending the progestogen-only pill or levonorgestrel or etonogestrel implants with inducers, while leaving unrestricted the progestogen injection, with the caveat of reduced bone mineral density, and the levonorgestrel intrauterine system [24]. If the woman continues combined oral contraception, both indicate an oestrogen of at least 50 µg of ethinylestradiol, up to 70 µg in case of breakthrough bleeding, with tricycling and a four-day free interval, without efficacy data to support it [23, 24]. For emergency contraception, the obstetric guideline recommends only the copper intrauterine device, pragmatically allowing double-dose levonorgestrel, 3 mg as a single dose up to 120 hours, and excludes ulipristal; the Scottish guideline indicates levonorgestrel 3 mg as a single dose, preferably within the first 72 hours and outside the conditions approved in the summary of product characteristics, with the non-hormonal intrauterine device up to five days as an alternative, likewise excluding ulipristal [23, 24]. Enzyme induction persists 14 to 28 days after discontinuation of the antiepileptic, so these rules continue to apply in that period, and postpartum the Scottish guideline contraindicates the progestogen-only pill in women on inducers [24]. The contraception recommendations described here date from 2015 to 2018 and may have been changed by later updates from the reproductive health societies, which were not consulted [23, 24].

8.4. Cascade testing, carriers and mosaicism

There is no place for cascade testing, carrier studies or assessment of germline mosaicism, because there is no causal variant to screen for [2]. Genotyping of CYP2C9 and HLA-B is maternal pharmacogenetics, intended for the dose and safety of the woman herself, and should not be presented to the family as a test of fetal risk [14, 21]. Measurement of epoxide hydrolase activity in amniocytes, proposed in 1990, is not validated and should not be offered [11, 51].

Genetic investigation nonetheless has a place in counselling for another reason. Maternal epilepsy may have a heritable genetic basis: the empirical risk of epilepsy in the offspring of a person with isolated or idiopathic epilepsy, without clear familial inheritance, lies between 1.5 and 7.5%, with a central value of 4% for the cumulative risk up to 20 years of age when one parent is affected, through multifactorial inheritance [107]. The table of empirical risks for idiopathic epilepsy, as cumulative risk of clinical epilepsy up to 20 years of age and excluding febrile convulsions, gives about 60% for a monozygotic twin, about 10% for a dizygotic twin, 6% for a sibling with onset before 10 years, 1 to 2% for a sibling with onset after 25 years, 2.5% for the overall sibling risk, 4% for the offspring of an affected parent, about 10% when a parent and a sibling are affected, about 15% when both parents are affected, and about 1% in the general population; in multifactorial idiopathic forms, the risk for second-degree relatives barely exceeds that of the population [107]. The Scottish guideline gives concordant values with a distinction by parent: 2.8 to 8.7% when the mother has epilepsy, 1.0 to 3.6% when it is the father, 2.3 to 6% when onset in the parent occurred before 20 years of age and 1.0 to 3.6% when it occurred afterwards, with no increase when onset was after 35 years [24]. About 15% of cases of epilepsy with associated manifestations have a pathogenic or possibly pathogenic copy number variation, with a similar proportion of de novo variants in candidate genes in epilepsy with developmental delay or epileptic encephalopathy [107]. When the exposed child has dysmorphism or delay, identification of a microdeletion inherited from the mother, for example at 15q11.2, 15q13.3 or 16p13.11, substantially alters counselling, because it introduces a Mendelian or near-Mendelian recurrence risk to be added to the teratogenic risk [8].

8.5. Prenatal and preimplantation diagnosis

There is no prenatal diagnosis of the embryopathy because there is no molecular target, and genetic prenatal diagnosis is only available for conditions with a known mutation or cytogenetic abnormality [107]. What is offered is morphological surveillance: detailed anatomy ultrasound in all pregnant women treated with any antiepileptic (level B) and specialist antenatal monitoring directed at the described malformations when phenytoin is continued [21, 25]. The Oxford Desk Reference includes the offer of prenatal diagnosis among the recommendations for the pregnant woman on an antiepileptic, in the sense of this detailed assessment [8]. Preimplantation genetic diagnosis does not apply, and effective prevention is modification of the exposure before conception [21, 67].

8.6. Counselling of adolescents and minors

The age for counselling is that of menarche, not that of a pregnancy plan. The joint North American guideline expressly places the choice of the optimal drug at the moment of starting it after menarche [21], and the European text requires counselling before starting treatment in any woman of childbearing age [67]. The British guideline recommends that information on contraception, folic acid, conception, pregnancy, breastfeeding and the menopause be given to girls and women with epilepsy, and that reviews be more frequent below 16 years of age [47]. The content should be adapted to the adolescent's capacity to understand, with involvement of legal guardians where appropriate, and repeated at each review, given that reproductive decisions are dynamic and that the drug chosen in adolescence conditions the exposure of a pregnancy that may not be planned [22, 47].

8.7. Follow-up of an already exposed pregnancy and registration

When exposure has already occurred, counselling changes its object: it becomes individual risk assessment and surveillance, rather than prevention. The woman should be referred to a specialist to reassess treatment and consider alternatives, without stopping the drug on her own initiative, and the lowest effective dose should be maintained if treatment is continued [8, 67]. Individualised risk assessment can be obtained from teratology information services, which also collect the exposure and the outcome for surveillance purposes [15]. Enrolment in pregnancy registries is recommended by the US label and is done by the patient herself [4, 29]. Registries are the instrument that generated virtually all the available evidence and their value depends on prospective enrolment: the European registry had 31,000 pregnancies enrolled as of 16 October 2026, with clinicians from 47 countries [28].

9. Special Considerations: Population Diversity and Epidemiology

9.1. Prevalence of the entity

There is no validated estimate of the prevalence of hydantoin embryopathy: Orphanet classifies the worldwide point prevalence as unknown, and the same applies to fetal carbamazepine syndrome and phenobarbital embryopathy [2, 9, 10]. This absence is a consequence of the design of the available studies, which measure proportions of outcomes among the exposed and not the frequency of the entity in the population.

The three figures frequently cited as prevalence are not prevalences. The risk of about 10% that an exposed fetus will develop the syndrome is a proportion among the exposed, derived from a prospective cohort of 35 infants in which 11% had sufficient features and another 31% had some compatible features [1, 40]. The 5.4% major malformations of the cohort meta-analysis and the 2.8% of the North American registry are proportions of malformation among the exposed, not of the complete syndrome [17, 19]. The 2- to 3-fold increase referred to in the regulatory texts is a relative risk against a baseline of 2 to 3% [67].

The population denominator can nonetheless be estimated indirectly. About 0.5% of pregnant women take anticonvulsants for epilepsy during pregnancy, according to the Oxford Desk Reference, which does not state the denominator [8], and antiepileptics are among the teratogenic drugs most prescribed to women of childbearing age [8]. Epilepsy affects about 3 to 5 per 1000 pregnant women, according to the framing cited in one of the published cases [54]. The British guidelines give convergent estimates with an explicit denominator: the prevalence of epilepsy in pregnancy is 0.5 to 1%, with about 2500 newborns per year of women with epilepsy in the United Kingdom, and in Scotland 0.7% of pregnancies occur in women with epilepsy, about 400 births per year in the 2011 data [23, 24]. Fetal exposure to antiepileptics occurs in about one in every 250 pregnancies [24]. The fraction of these pregnancies exposed to phenytoin is today small: 1.4% of the monotherapies in the European registry [45].

Maternal risk frames these figures. The risk of death is ten times higher in pregnant women with epilepsy, with 14 maternal deaths attributed to epilepsy in the United Kingdom between 2009 and 2012, 12 of them from sudden unexpected death in epilepsy, which corresponds to 0.40 per 10,000 pregnancies (95% CI 0.22–0.68), to which were added another 12 deaths between six weeks and one year after delivery; status epilepticus occurs in 0.55 to 1.8% of pregnancies [23, 24]. By contrast, two thirds of women have no worsening of seizures in pregnancy, 74 to 92% of those who had been seizure-free for nine months to a year remain seizure-free, and in the European registry the proportion without seizures was 74% in idiopathic generalised epilepsies and 60% in focal epilepsies [23, 24]. Obstetric morbidity has been quantified in a systematic review of 38 studies and 2,837,325 women, which gave, in women with epilepsy against those without, odds ratios of 1.54 (1.02–2.32) for spontaneous abortion, 1.49 (1.01–2.20) for antepartum haemorrhage, 1.37 (1.21–1.55) for hypertensive disease, 1.67 (1.31–2.11) for induction of labour, 1.40 (1.23–1.58) for caesarean section, 1.16 (1.01–1.34) for preterm birth, 1.26 (1.20–1.33) for fetal growth restriction and 1.29 (1.13–1.49) for postpartum haemorrhage, with no difference in gestational diabetes or perinatal death; comparing women with and without antiepileptics, induction (1.40; 1.05–1.85), growth restriction (3.51; 1.23–10.01), postpartum haemorrhage (1.33; 1.16–1.54) and neonatal intensive care admission (1.42; 1.13–1.78) remained increased [23]. Postpartum depression affects 29% of these women against 11% of the others, with multiparity and polytherapy as associated factors [23].

9.2. Data by region

The available evidence is unevenly distributed across regions, and each data set measures something different.

The comparison between the North American and the European registry is the most informative contrast and also the one most subject to misreading. The values of 2.8% and 6.3% do not measure different risks of the same drug in different populations: they differ in the method of recruitment, which is voluntary enrolment on the pregnant woman's initiative in one case and enrolment by participating clinicians in the other, in the timing and mode of outcome assessment, and in the number of exposures [18, 19]. The overall prevalence of small for gestational age illustrates the same problem: it was 20% in the European cohort, similar to the 17 to 22% already described in children of women with treated epilepsy, but higher than the 11% estimated by the North American registry [45].

9.3. Founder populations and ethnic diversity

No founder population is described as having a particular teratogenic risk for phenytoin in the sources consulted, which is consistent with the nature of the condition: exposure does not follow lineages. Where ancestry has a documented consequence is in maternal safety. The HLA-B*15:02 allele is the recognised risk factor for severe cutaneous reactions in patients of Asian ancestry on phenytoin, and the British guideline translates the datum as an alert for people of Han Chinese or Thai origin [4, 47]. The prevalences of the CYP2C9 poor metaboliser phenotype vary by population, with about 2 to 3% in white populations, 0.5 to 4% in Asian populations and less than 1% in African Americans, and those of the intermediate phenotype about 35%, 15 to 36% and 24%, respectively [4]; the pharmacogenetics consortium's estimates give 2.6% poor metabolisers in Europeans, 3.0% in Near Eastern populations and 0.6% in East Asians [12]. The frequency of HLA-B*15:02 carriers is above 15% in Hong Kong, Thailand, Malaysia and parts of the Philippines, about 10% in Taiwan, 4% in northern China, 2 to 4% in South Asia and below 1% in populations of European ancestry, which defines the regions where genotyping before starting phenytoin has the greatest yield [65, 66]. There are also individual reports suggesting an increased, although rare, incidence of hypersensitivity reactions, including rash and hepatotoxicity, in Black patients [25].

The representativeness of the cohorts is a declared limitation. The fetal growth analysis of the European registry notes limited representation of ethnicities other than white and Asian, the absence of a population base and the inclusion only of exposures for epilepsy [45]. The North American neurodevelopment programme has the complementary limitation that 232 of 298 children assessed at 6 years (78%) were exposed to lamotrigine or levetiracetam, which restricts assessment of the other drugs [127].

9.4. Special populations

Three groups deserve note. The newborn has unpredictable oral absorption of phenytoin and possibly depressed metabolism, so serum monitoring is especially important when the drug is administered to them [25]. Children have gingival hyperplasia more often than adults, especially with poor oral hygiene, and the adverse effect profile is otherwise similar [25]. The woman with intellectual disability and epilepsy is expressly identified by the British guideline as needing more frequent reviews, alongside those under 16 years of age, with active epilepsy with a seizure in the last 12 months, with bilateral tonic-clonic seizures or with modifiable risk factors for sudden death, which matters because it is a group in which the capacity to plan pregnancy and to adhere to contraception may be compromised [47].

Non-epileptic indications constitute another special group, highlighted by the Australian regulator: antiepileptics are also approved or used in migraine, bipolar disorder, trigeminal neuralgia, anxiety and neuropathic pain, and counselling about contraception and preconception reassessment applies equally in these cases [8, 116]. Phenytoin has a second-line indication in trigeminal neuralgia [25]. The available cohorts include only exposures for epilepsy, so extrapolation to these indications, although plausible, is not documented [45].

9.5. Declared gaps

The sources consulted contain no data from Portugal or Brazil on the frequency of the embryopathy, on the prescribing of phenytoin in pregnancy or on access to and reimbursement of these medicines, and no targeted geographical search was carried out. The European registry includes Portuguese centres, namely a reference centre for refractory epilepsies integrated into the European network, whose collaborators are among the authors of the fetal growth analysis, so Portuguese pregnancies are represented in the aggregated data without the possibility of separate national reading [45]. The absence of data in the sources consulted does not demonstrate the absence of published data.

Other material gaps are the lack of within-group dose analysis for phenytoin in the European registry, with contradictory evidence between studies about the existence of a dose-response relationship [17, 76]; the absence of data specific to phenytoin in recent neurodevelopment studies, which are dominated by lamotrigine and levetiracetam [127, 133]; the absence of a phenytoin row in the autism spectrum disorder results of the North American guideline and in the Nordic registries [21, 94]; and the absence of a vitamin K dose in the product information, supplied only by the British guidelines for the newborn and by the Scottish guideline for the mother [4, 23–25]. A gap of another nature is the unevenness of national guidance: in the International League Against Epilepsy's survey of its 118 chapters, with 77 responses, 68% had guidelines, but 34% of these dated from 2014 or earlier, and at least 20% contained no information on cognitive risk, risks by drug or choice of antiepileptic [120]. The full 2019 report of the same task force is not open access, so its numbered recommendations are not reproduced in this article [118]. The neurocognitive extension of the European registry, with 162 children assessed at 6 and 7 years, did not include a phenytoin-exposed group [133].

9.6. Networks, registries and associations

The relevant European reference network for maternal epilepsy is EpiCARE, dedicated to rare and complex epilepsies and launched in 2017 [26]. For dysmorphological assessment and developmental follow-up of the exposed child, the relevant network is ERN ITHACA, dedicated to rare malformation syndromes and neurodevelopmental disorders [27].

Pregnancy registries are the infrastructure that underpins the evidence. The European registry is a prospective observational study launched in Europe in 1999 by a consortium of independent research groups and later extended to other continents, with clinicians from 47 countries in Europe, Oceania, Asia, Latin America and Africa, headquarters in Milan and 31,000 pregnancies enrolled as of 16 October 2026 [28]. The North American registry is the non-profit philanthropic arm of a medical school, with enrolment by the patient herself through a dedicated telephone line and recommended in the phenytoin and fosphenytoin labels [4, 5, 29]. The United Kingdom and Ireland registers, begun in 1996, provided data on 5206 informative outcomes up to 2012 and 3607 cases in the 2006 report [30, 131], and the British obstetric guideline recommends inviting all pregnant women with epilepsy to enrol in that register [23]. The Indian Kerala registry, with enrolment since 1998, is the main source of data from South Asia, with 2328 pregnancies analysed up to 2019 [77]. The neurodevelopment research programmes, now completed, are registered as observational studies [112].

For patients and families, there is in the United Kingdom an association dedicated to fetal anticonvulsant syndrome [8] and specific information on epilepsy and pregnancy made available by the British epilepsy association [31]. Teratology information services are the resource for immediate counselling: the British national service, free for health professionals, with a dedicated line, online reviews and patient leaflets; the European network of teratology information services; and the North American service with its own telephone line [15]. The current activity status of the Canadian service cited in the older sources was not confirmed [15].

10. Illustrative Clinical Case

10.1. Severe, lethal form in an unsupervised pregnancy (India, 2023)

A 26-year-old woman, in her second pregnancy with one previous living child, with an unsupervised pregnancy and spontaneous conception, presented at 29 weeks and 2 days with severe, persistent antepartum haemorrhage due to grade 4 placenta praevia [54]. She had been taking phenytoin 300 mg per day, in three divided doses, for three years, medication started because of persistent seizures in the postpartum period of the previous pregnancy [54]. She had had no first-trimester appointments and a single second-trimester appointment, without supplements, without the recommended vaccinations, without ultrasound and without antenatal testing; there was no fever, rash or other documented exposure to teratogens, and neither antenatal corticosteroids nor magnesium sulfate were given [54].

An emergency caesarean section was performed. The newborn did not cry and was intubated for absent respiratory effort, with an Apgar score of 3 at one minute and 4 at five minutes, mechanical ventilation in assist-control mode with an inspired oxygen fraction of 100%, a peak inspiratory pressure of 20 cmH₂O and a positive end-expiratory pressure of 5 cmH₂O; birthweight was 1091 g [54]. Examination revealed microcephaly, hypertelorism, complete cleft lip and palate, low-set ears, bilateral finger contractures, bilateral oligosyndactyly of the lower limbs and a two-vessel umbilical cord [54]. There was a pansystolic murmur at the apex, and echocardiography showed a large ventricular septal defect with double outlet right ventricle [54]. Informed of the prognosis, the parents decided not to escalate treatment or to resuscitate; bradycardia occurred and death followed at three hours of life [54].

The published clinical photograph documents the features of the case and was consulted directly, being accessible at https://pmc.ncbi.nlm.nih.gov/articles/PMC10756398/figure/FIG1/; the legend identifies, in the six panels, the hypertelorism with broad nasal root, the cleft lip and the low-set ears, the hypoplastic distal phalanges of the right hand with contractures, the oligosyndactyly, the absence of distal phalanges in both feet and the single umbilical artery [54].

The authors attributed the severity to the absence of medical supervision, of serum concentration monitoring and of folate supplementation, through lack of antenatal follow-up [54]. Two recommendations cited by them in the background, monthly therapeutic monitoring with phenytoin or phenobarbital attributed to an international epilepsy society and a preconception folate dose of 4 mg daily attributed to a public health authority, were not verified in the primary sources and are not adopted in this article [54]. The authors themselves noted that a global survey by that society found guidelines in about two thirds of its national chapters, many revised before 2014 and with wide variation in the recommended folate dose [54]. The case illustrates the severe form in a situation of total absence of antenatal care and does not represent the typical outcome of exposure, which in most exposed fetuses does not produce embryopathy [53, 54].

10.2. Survival with multiple malformations and anaesthetic implications (India, 2012)

A 4-year-old boy, weighing 20 kg, was scheduled for bilateral tibialis anterior lengthening for congenital talipes equinovarus [130]. His mother had epilepsy and had been taking phenytoin 100 mg twice daily for seven years, including throughout the pregnancy, having been followed in an antenatal clinic, where she was advised to continue the drug because of persistent seizures; a previous pregnancy, also ending in caesarean section, had produced a healthy child without defects [130]. Birth was at term, by caesarean section, for failure to progress in labour [130]. The child had seizures on the first day of life and further episodes on two later occasions, without prescribed medication, as well as recurrent upper respiratory infections and delayed developmental milestones [130].

Examination showed abnormal facies, cutaneous syndactyly of the fingers, low-set ears, hypertelorism, incomplete cleft palate and spina bifida occulta of L5, without neurological deficit, with normal routine tests [130]. Chest radiography revealed dextrocardia; the electroencephalogram showed frontoparietal spikes; brain magnetic resonance imaging showed bilateral peritrigonal and periventricular cystic encephalomalacia of the deep white matter, suggestive of prenatal hypoxic-ischaemic encephalopathy; spinal magnetic resonance imaging showed bony deformity of L5 without cord or root involvement; the electrocardiogram was in sinus rhythm; and echocardiography revealed a 4.8 mm ostium secundum atrial septal defect, without left-to-right shunt, with an ejection fraction of 62% [130].

Anaesthetic management consisted of oral midazolam 0.5 mg/kg, corresponding to 10 mg, thirty minutes beforehand, topical anaesthetic cream, propofol 75 µg/kg/min for sedation, and subarachnoid block at L3-L4 with 1.2 mL of 0.5% hyperbaric bupivacaine and 10 µg of fentanyl, through a 25 G, 5 cm paediatric Quincke needle, with oxygen by mask; there was haemodynamic stability for an hour and a half, an hour's stay in post-anaesthesia care and no postoperative sensory or motor deficit [130]. Regional anaesthesia was chosen because of the reactive and potentially difficult airway, the type of procedure and postoperative analgesia [130].

The approach described corresponds to one team's decision in an individual case and does not constitute a routine recommendation for children with this embryopathy; the choice between regional and general technique depends on assessment of the airway, the heart defect and the procedure in each case [130]. The article also states that phenytoin had been classified by the US authority in pregnancy risk category D, a nomenclature abandoned with the entry into force of the pregnancy and lactation labelling rule, so that classification is not reproduced as current practice [4, 130].

10.3. What the two cases document together

The two reports cover the two extremes of the clinical spectrum and the two situations in which the diagnosis arises. In the first, the complete embryopathy with lethal malformations occurs in a pregnancy without any surveillance, with continuous exposure to 300 mg daily and without folate [54]. In the second, the child survives with multiple correctable malformations, developmental delay and seizures, in a supervised pregnancy in which continuation of the drug was a clinical decision justified by persistent maternal seizures, and the practical question shifts to surgical and anaesthetic management and to developmental follow-up [130]. The second case also documents the variability between siblings under the same maternal exposure, given that the previous pregnancy had produced a child without defects [15, 130].

11. Prognosis and Long-Term Follow-Up

11.1. Overall prognosis

Prognosis depends on three independent components: the major malformations present, neurodevelopment and, in the first days of life, haemorrhagic disease. Most exposed fetuses do not develop embryopathy: about 90% of exposed infants under good management showed no evidence of teratogenesis in the historical series [53], and the proportions of major malformation among the exposed range between 2.8% and 6.3% in contemporary prospective registries [18, 19]. When embryopathy is present, major malformations determine the organ prognosis and follow the course proper to each lesion, from successful surgical repair of cleft lip and palate and of an atrial septal defect to the lethal outcome of complex heart defects in the setting of extreme prematurity [54, 130].

As to growth, the classic description indicates mild to moderate deficiency of prenatal onset, which may become more marked in the first months, with a frequent relative growth delay in the first months of unknown cause [1]. Contemporary data do not confirm growth restriction as an effect attributable to phenytoin, which in the European registry was associated with a weight centile 8.04 points higher than that of lamotrigine (95% CI 2.81–13.27), with 15.1% of newborns small for gestational age against 14.6% with lamotrigine, and with an estimated frequency of 14.4 per 1000 in the systematic review of the North American guideline [21, 45]. The growth prognosis should therefore be communicated with this distinction: the deficiency is part of the phenotype of the complete syndrome, not of the average effect of the exposure.

As to neurodevelopment, intellectual impairment is borderline to mild and occasional, and childhood performance may be better than that predicted by progress in early infancy, with verbal abilities lower than non-verbal abilities and no dose effect observed [1]. The mean intelligence quotient values in those exposed to phenytoin are 99 at 3 years in a series of 48 children, with a 95% confidence interval between 94 and 104, and 108 at 6 years [1, 81], values confirmed by the 6-year assessment of the respective research programme, with 108 (95% CI 104–112), higher than the 97 with valproate and similar to the 105 with carbamazepine and the 108 with lamotrigine [82], and by the guideline's systematic review, with a full-scale quotient of 103.2 (95% CI 93.0–113.4; n=76) [21]. The regulators' position is more cautious, holding that, despite inconsistent results, the possibility of adverse neurodevelopmental effects cannot be excluded [22, 67].

Two individual markers substantially modify the expected cognitive prognosis. Microcephaly was associated with a deficit of 23.7 points of full-scale intelligence quotient relative to exposed children with normal head circumference; and, with microcephaly excluded, the presence of midface or digital hypoplasia was associated with deficits of 12.7 points in the verbal quotient, 10.0 in the performance quotient and 12.8 in the full-scale quotient (p=0.0061), with major malformations in themselves having no effect on the intelligence quotient [87]. The practical consequence is that the dysmorphic phenotype, and not the presence of an organ malformation, is the indicator that mandates formal developmental assessment.

As to neurodevelopmental disorders defined by diagnosis, the roughly fourfold higher risk attributed by the Oxford Desk Reference to antiepileptics in general rests on two studies exclusively about valproate and without a phenytoin-exposed group, so it does not transfer to this drug [8, 88, 89]. Phenytoin does not appear in the specific results of the network meta-analysis, of the North American guideline or of the large Nordic registry study [21, 84, 94], and in the largest recent population cohort, with exposure in the second half of pregnancy and pregnant women with untreated epilepsy as reference, it was not associated with an increased risk of any of the disorders assessed [90].

11.2. Surveillance table

Fetal growth surveillance and fetal cardiac screening appear in the table because they are formal recommendations for exposures to valproate, topiramate and phenobarbital, and they signal by contrast that phenytoin has, in current guidelines, no specific fetal surveillance recommendation beyond the detailed anatomy ultrasound common to all antiepileptics [21].

11.3. Surveillance of the mother postpartum

Phenytoin clearance returns to the pre-pregnancy level some weeks or months after delivery, which makes the postpartum period a time of risk of toxicity if the dose increased during pregnancy is continued [5]. The label states that restoration of the original dose will probably be necessary after delivery, and the British guideline recommends an antenatal plan for the return to preconception doses, to begin in the first days after delivery [4, 47]. The decision about breastfeeding should be taken before delivery, given the divergence between the summary of product characteristics, which does not recommend it, and the guidelines and the drugs and lactation database, which consider it compatible with phenytoin, with minimal transfer into milk, and recommend encouraging it [23, 25, 47, 129]. The newborn breastfed by a mother on phenytoin is monitored for sedation, feeding difficulty and, in the first weeks, signs of hyperexcitability when their own serum concentration disappears [129]. The British obstetric guideline adds review of the maternal dose in the first ten days after delivery and identifies the three peripartum days as the period of greatest risk of seizure exacerbation [23].

11.4. Neoplasms and oncological surveillance

There is no recommendation for oncological surveillance. The association with neoplasms rests on case reports of neural crest tumours, mainly neuroblastoma, mentioned in the labels of both jurisdictions, and the role of the drug relative to other factors has not been determined [4, 25, 99, 100]. The historical report recommended examination at birth and surveillance for bleeding problems, noting that a neural crest tumour may arise at birth or later in childhood [99]. Experimentally, the carcinoma-promoting effect is unequivocal in rodents for phenobarbital and for phenytoin, but the causal relationship in humans remains under debate and there is no quantified risk to justify a screening protocol [101]. The only quantified oncological signal in this population does not concern the antiepileptic, but the supplement: in children of mothers with epilepsy exposed to folic acid at a dose equal to or greater than 1 mg daily, the risk of cancer before the age of 20 was 1.4% against 0.6% in the unexposed (adjusted risk 2.7; 95% CI 1.2–6.3), whereas antiepileptics without high-dose folate were not associated with cancer; the finding is contested and has not given rise to any surveillance recommendation [124, 126].

11.5. Transition to adulthood

The consequences that persist into adulthood are the sequelae of the corrected malformations, cognitive impairment when present and the educational and support needs arising from it [8, 87]. The adult woman who was exposed in utero and who herself has epilepsy, a possible scenario given the genetic component of maternal epilepsy, should receive the preconception counselling proper to her condition, with the choice of the drug of lowest teratogenic potential from the start of treatment [21, 107]. Follow-up of the affected adult is not protocolised in any of the guidelines consulted, which are limited to paediatric developmental surveillance [21, 47].

12. Summary and Final Considerations

Hydantoin embryopathy is a drug-exposure embryofetopathy, without a causal gene, coded as ORPHA:1912 and without an entry in the catalogue of Mendelian disorders, whose cardinal phenotype combines hypoplasia of the distal phalanges and nails, a facies with a broad, depressed nasal root, short nose and bowed upper lip, orofacial clefts, congenital heart disease, growth abnormalities and borderline to mild cognitive impairment. The designation was proposed by Hanson and Smith in 1975, after Meadow had raised the association in 1968 and the epidemiological studies of 1973 had demonstrated it, and systematic screening of 128,049 deliveries established in 2001 that the pattern is due to the drugs and not to maternal epilepsy.

The pathophysiology has four proposed pathways, none validated as a clinical test or therapeutic target: accumulation of arene oxides through low activity of the epoxide hydrolase encoded by EPHX1, blockade of the hERG channel IKr current with embryonic arrhythmia and injury from hypoxia and reoxygenation, bioactivation by embryonic peroxidases with oxidative stress, and interference with placental folate transport. Fetal susceptibility explains the central observation of the condition, which is discordance between equally exposed fetuses, including between dizygotic twins and between pregnancies of the same woman. The pharmacogenetics of CYP2C9 and HLA-B conditions maternal dose and safety, with recommendations for a 25% or 50% reduction in maintenance according to the activity score and with an indication not to start the drug in carriers of HLA-B*15:02 without previous exposure, but does not predict the embryopathy.

Quantification of the risk depends on study design and requires stating what each figure measures. In contemporary prospective registries, major malformations occur in 2.8% of the exposed in the North American registry and in 6.3% in the European registry, which in 2024 came to have 9840 monotherapies and assigned phenytoin an odds ratio of 2.39 against levetiracetam; in the cohorts pooled by the Cochrane review, in 5.4%, with a relative risk of 3.81 against children of women without epilepsy and of 2.01 against untreated epilepsy, corresponding to a risk difference of three cases per 100 exposed; and the regulatory texts translate the whole as a 2- to 3-fold increase over a baseline of 2 to 3%, or about 6% in absolute terms. Phenytoin occupies an intermediate position: above lamotrigine, levetiracetam and oxcarbazepine, equivalent to carbamazepine, phenobarbital, primidone and topiramate in the direct comparisons that did not reach significance, and below valproate. It has no demonstrated dose-response relationship, unlike carbamazepine, phenobarbital, lamotrigine and valproate, and polytherapy consistently increases the risk, according to the drugs combined and not their number.

Two recent findings correct the historical description. As to fetal growth, the analysis of the European registry published in September 2026 associated phenytoin with a birthweight centile 8.04 points higher than that of lamotrigine, reversing the expectation of growth restriction, which is confirmed instead for topiramate, phenobarbital and oxcarbazepine. As to neurodevelopment, the intelligence quotients measured in those exposed to phenytoin, between 103 and 108, do not differ from those with carbamazepine and lamotrigine and are higher than those with valproate, and the largest recent population cohort, with exposure in the second half of pregnancy, found no association of phenytoin with any neurodevelopmental disorder, although the regulators maintain that, the studies being contradictory, a risk cannot be excluded, a position supported by the small size of the samples specific to phenytoin, between 40 and 76 children.

Diagnosis is clinical and has no confirmatory test: it requires documentation of the exposure with type and dose, targeted dysmorphological examination and exclusion of genetic and metabolic causes, with particular attention to microdeletions predisposing to epilepsy that may have been inherited from the mother. The facies is not enough, as shown by the study in which 47% of the exposed were correctly identified but 45% of the unexposed had some of the facial features. Measurement of epoxide hydrolase activity in amniocytes, proposed in 1990 in a series of 19 pregnancies, was never validated and should not be offered. The differential requires distinguishing the antiepileptic embryopathies from one another, supported by the available orienting markers, and separating them from warfarin embryopathy, maternal vitamin K malabsorption, fetal alcohol spectrum disorders, maternal diabetic embryofetopathy and the genetic causes of nail and distal phalanx hypoplasia, above all Coffin–Siris syndrome.

There is no treatment of the embryopathy: there is management of the exposure. Prevention is played out before conception, with the choice of drug and dose at the very start of treatment after menarche, monotherapy at the minimum effective dose, effective contraception, which enzyme induction may compromise, and folic acid supplementation, whose dose divides the sources between at least 0.4 mg daily and 5 mg daily, with benefit documented mainly for neurodevelopment and not for the prevention of malformations, and whose protective effect on congenital anomalies, demonstrated by target trial emulation in Nordic registries, depends on the supplement being started between one and twelve weeks before conception, counterbalanced by an oncological risk signal in children exposed to doses equal to or greater than 1 mg daily. Once pregnancy is established, the rule reverses to therapeutic stability with increased surveillance, including measurement of the free fraction, because pregnancy increases clearance and reduces protein binding. In the perinatal period, vitamin K prevents haemorrhagic disease: the product information does not state a dose, but the British guidelines recommend 1 mg intramuscularly to the newborn, and only the Scottish guideline provides for maternal oral phytomenadione of 10 mg per day in the third trimester, and only with additional risk factors, diverging from the obstetric guideline, which considers that maternal administration to lack sufficient evidence; the same two guidelines diverge head-on on doubling the dose of antenatal corticosteroids in women on enzyme inducers. The European regime is more restrictive than the North American, in requiring that phenytoin not be used in women of childbearing age unless the benefit outweighs the risks after consideration of the alternatives, with text harmonised in all Member States since August 2021, content replicated by the Australian regulator in 2022.

Follow-up of the child rests on dysmorphological examination with measurement of head circumference, on age-appropriate developmental screening recommended for all children exposed to any antiepileptic, on formal developmental assessment when there is microcephaly or midface or digital hypoplasia, because these are associated with deficits of 10 to 24 intelligence quotient points, and on vision and hearing screening. Oncological surveillance is not recommended, despite the reports of neural crest tumours.

The material limitations of this body of evidence are recognisable and should accompany clinical communication. All the studies are observational and subject to confounding by indication; the phenytoin denominators in contemporary cohorts are small, which led the joint North American guideline to classify confidence in its estimates as low; direct comparison of percentages between registries with different designs is not legitimate; there is no published within-group dose analysis for phenytoin; recent neurodevelopment studies are dominated by lamotrigine and levetiracetam; the maternal vitamin K dose is sourced from a single guideline and is contested by another; and the sources consulted contain no data from Portugal or Brazil on frequency, prescribing or access, even though Portuguese pregnancies are represented in the aggregated data of the European registry. The clinical relevance of the entity has changed in nature: phenytoin today represents 1.4% of the monotherapies in the European registry and is no longer a usual choice in women of childbearing age, so knowledge of the embryopathy serves mainly to recognise the child already born of an exposed pregnancy, to counsel the woman who becomes pregnant while on the drug because alternatives have failed, and to frame acute exposure in the setting of status epilepticus or neurosurgery.

13. References

1. Jones KL, Jones MC, del Campo M. Smith's Recognizable Patterns of Human Malformation. 8th ed. Philadelphia: Elsevier; 2022. Fetal hydantoin syndrome (fetal Dilantin syndrome); p. 792–3.
2. Orphanet. Fetal hydantoin syndrome (ORPHA:1912) [Internet]. Paris: INSERM; 2026 [cited 2026 Oct 17]. Available from: https://www.orpha.net/en/disease/detail/1912
3. Jones KL, Jones MC, del Campo M. Smith's Recognizable Patterns of Human Malformation. 8th ed. Philadelphia: Elsevier; 2022. Genetic counseling: multifactorial inheritance; p. 949.
4. US Food and Drug Administration. DILANTIN INFATABS (phenytoin) chewable tablets — prescribing information. Parke-Davis Div of Pfizer Inc; SPL version effective 12 May 2026. ANDA 084427; SPL set id ca119a89-2394-4d34-8078-cd1fa4e8e2f1. Available from: https://api.fda.gov/drug/label.json?search=openfda.brand_name:%22Dilantin+Infatabs%22
5. US Food and Drug Administration. CEREBYX (fosphenytoin sodium) injection — prescribing information. Pfizer Laboratories Div Pfizer Inc; SPL version effective 28 August 2025. NDA 020450 (initial approval 5 August 1996).
6. Orphanet. Fetal anticonvulsant syndrome (ORPHA:370068) [Internet]. Paris: INSERM; 2026 [cited 2026 Oct 17]. Available from: https://www.orpha.net/en/disease/detail/370068
7. Orphanet. Fetal trimethadione syndrome (ORPHA:1913) [Internet]. Paris: INSERM; 2026 [cited 2026 Oct 17]. Available from: https://www.orpha.net/en/disease/detail/1913
8. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Fetal anticonvulsant syndrome (FACS); p. 732–5.
9. Orphanet. Fetal carbamazepine syndrome (ORPHA:370076) [Internet]. Paris: INSERM; 2026 [cited 2026 Oct 17]. Available from: https://www.orpha.net/en/disease/detail/370076
10. Orphanet. Phenobarbital embryopathy (ORPHA:1919) [Internet]. Paris: INSERM; 2026 [cited 2026 Oct 17]. Available from: https://www.orpha.net/en/disease/detail/1919
11. GeneGuidance (base local do projeto). EPHX1 — epoxide hydrolase 1 [registo curado]. Consultado 2026-10-17.
12. GeneGuidance (base local do projeto). CYP2C9 [registo curado]. Consultado 2026-10-17.
13. VariantValidator. Validation of NM_000120.4:c.337T>C and NM_000120.4:c.416A>G in GRCh38 [Internet]. Manchester: VariantValidator; 2026 [cited 2026 Oct 17]. Available from: https://rest.variantvalidator.org/
14. Karnes JH, Rettie AE, Somogyi AA, Huddart R, Fohner AE, Formea CM, et al. Clinical Pharmacogenetics Implementation Consortium (CPIC) guideline for CYP2C9 and HLA-B genotypes and phenytoin dosing: 2020 update. Clin Pharmacol Ther. 2021;109(2):302-9. doi:10.1002/cpt.2008
15. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Drugs in pregnancy; p. 718–21.
16. Vergano SS, Santen G, Wieczorek D, Matsumoto N. Coffin-Siris syndrome. 2013 Apr 4 [updated 2025 May 15]. In: Adam MP, Feldman J, Mirzaa GM, et al., editors. GeneReviews® [Internet]. Seattle (WA): University of Washington, Seattle; 1993–2026. Available from: https://www.ncbi.nlm.nih.gov/books/NBK131811/
17. Bromley R, Adab N, Bluett-Duncan M, Clayton-Smith J, Christensen J, Edwards K, et al. Monotherapy treatment of epilepsy in pregnancy: congenital malformation outcomes in the child. Cochrane Database Syst Rev. 2023;8(8):CD010224. doi:10.1002/14651858.CD010224.pub3
18. Battino D, Tomson T, Bonizzoni E, Craig J, Perucca E, Sabers A, et al.; EURAP Collaborators. Risk of major congenital malformations and exposure to antiseizure medication monotherapy. JAMA Neurol. 2024;81(5):481-9. doi:10.1001/jamaneurol.2024.0258
19. Hernandez-Diaz S, Quinn M, Conant S, Lyons A, Paik H, Ward J, et al. Use of antiseizure medications early in pregnancy and the risk of major malformations in the newborn. Neurology. 2025;105(3):e213786. doi:10.1212/WNL.0000000000213786
20. Co-ordination Group for Mutual Recognition and Decentralised Procedures — Human (CMDh). Annex I: scientific conclusions and grounds for the variation to the terms of the marketing authorisation(s) — phenytoin. PSUSA/00002392/202008. Adopted at the CMDh meeting of April 2021. Available from: https://www.ema.europa.eu/documents/psusa/phenytoin-cmdh-scientific-conclusions-grounds-variation-amendments-product-information-timetable/00002392/202008_en.pdf
21. Pack AM, Oskoui M, Williams Roberson S, Donley DK, French J, Gerard EE, et al. Teratogenesis, perinatal, and neurodevelopmental outcomes after in utero exposure to antiseizure medication: practice guideline from the AAN, AES, and SMFM. Neurology. 2024;102(11):e209279. doi:10.1212/WNL.0000000000209279
22. Medicines and Healthcare products Regulatory Agency. Antiepileptic drugs in pregnancy: updated advice following comprehensive safety review. Drug Safety Update. 7 January 2021 (post-publication notes to June 2024). Available from: https://www.gov.uk/drug-safety-update/antiepileptic-drugs-in-pregnancy-updated-advice-following-comprehensive-safety-review
23. Royal College of Obstetricians and Gynaecologists. Epilepsy in Pregnancy. Green-top Guideline No. 68. London: RCOG; June 2016. Available from: https://www.rcog.org.uk/media/rzldnacf/gtg68_epilepsy.pdf
24. Scottish Intercollegiate Guidelines Network. Diagnosis and management of epilepsy in adults. SIGN 143. Edinburgh: SIGN; May 2015, revised 2018. Available from: https://www.sign.ac.uk/media/1079/sign143_2018.pdf
25. Viatris Products Limited. Epanutin 30mg/5ml oral suspension — summary of product characteristics. PL 46302/0285; text revised 12/2025. Available from: https://www.medicines.org.uk/emc/product/2257/smpc
26. EpiCARE — European Reference Network for Rare and Complex Epilepsies. Available from: https://epi-care.eu/
27. ERN ITHACA — European Reference Network for Rare Malformation Syndromes, Intellectual and Other Neurodevelopmental Disorders. Available from: https://ern-ithaca.eu/
28. EURAP — International Registry of Antiepileptic Drugs and Pregnancy. Available from: https://eurapinternational.org/
29. North American Antiepileptic Drug (AED) Pregnancy Registry. Available from: https://www.aedpregnancyregistry.org/
30. Campbell E, Kennedy F, Russell A, Smithson WH, Parsons L, Morrison PJ, et al. Malformation risks of antiepileptic drug monotherapies in pregnancy: updated results from the UK and Ireland Epilepsy and Pregnancy Registers. J Neurol Neurosurg Psychiatry. 2014;85(9):1029-34. doi:10.1136/jnnp-2013-306318
31. Epilepsy Action. Starting a family. Available from: https://www.epilepsy.org.uk/living/starting-a-family
32. Merritt HH, Putnam TJ. Sodium diphenyl hydantoinate in the treatment of convulsive disorders. JAMA. 1938;111:1068-73. Reimpresso: JAMA. 1984;251(8):1062-7. doi:10.1001/jama.251.8.1062
33. Meadow SR. Anticonvulsant drugs and congenital abnormalities. Lancet. 1968;2(7581):1296. doi:10.1016/s0140-6736(68)91781-9
34. Fedrick J. Epilepsy and pregnancy: a report from the Oxford Record Linkage Study. Br Med J. 1973;2(5864):442-8. doi:10.1136/bmj.2.5864.442
35. Monson RR, Rosenberg L, Hartz SC, Shapiro S, Heinonen OP, Slone D. Diphenylhydantoin and selected congenital malformations. N Engl J Med. 1973;289(20):1049-52. doi:10.1056/NEJM197311152892001
36. Speidel BD, Meadow SR. Maternal epilepsy and abnormalities of the fetus and newborn. Lancet. 1972;2(7782):839-43.
37. Hill RM, Verniaud WM, Horning MG, McCulley LB, Morgan NF. Infants exposed in utero to antiepileptic drugs. A prospective study. Am J Dis Child. 1974;127(5):645-53.
38. Aase JM. Letter: anticonvulsant drugs and congenital abnormalities. Am J Dis Child. 1974;127(5):758. doi:10.1001/archpedi.1974.02110240144024
39. Hanson JW, Smith DW. The fetal hydantoin syndrome. J Pediatr. 1975;87(2):285-90. doi:10.1016/s0022-3476(75)80604-4
40. Hanson JW, Myrianthopoulos NC, Harvey MA, Smith DW. Risks to the offspring of women treated with hydantoin anticonvulsants, with emphasis on the fetal hydantoin syndrome. J Pediatr. 1976;89(4):662-8. doi:10.1016/s0022-3476(76)80414-3
41. Hanson JW, Buehler BA. Fetal hydantoin syndrome: current status. J Pediatr. 1982;101(5):816-8. doi:10.1016/s0022-3476(82)80339-9
42. Hanson JW. Teratogen update: fetal hydantoin effects. Teratology. 1986;33(3):349-53. doi:10.1002/tera.1420330314
43. Hanson JW, Smith DW. Fetal hydantoin syndrome. Lancet. 1976;1(7961):692. doi:10.1016/s0140-6736(76)92805-1
44. Holmes LB, Harvey EA, Coull BA, Huntington KB, Khoshbin S, Hayes AM, et al. The teratogenicity of anticonvulsant drugs. N Engl J Med. 2001;344(15):1132-8. doi:10.1056/NEJM200104123441504
45. Perucca P, Malpas CB, Berry-Noronha A, Walker SP, Kalincik T, Moche Kamga B, et al. Antiseizure medication use in pregnancy and the risk of poor fetal growth in the EURAP international registry: a prospective cohort study. Lancet Neurol. 2026;25(9):840-51. doi:10.1016/S1474-4422(26)00216-4
46. Lindhout D, Meinardi H, Meijer JW, Nau H. Antiepileptic drugs and teratogenesis in two consecutive cohorts: changes in prescription policy paralleled by changes in pattern of malformations. Neurology. 1992;42(4 Suppl 5):94-110.
47. National Institute for Health and Care Excellence. Epilepsies in children, young people and adults. NICE guideline NG217. Published 27 April 2022; last updated 5 August 2026. Available from: https://www.nice.org.uk/guidance/ng217
48. Phelan MC, Pellock JM, Nance WE. Discordant expression of fetal hydantoin syndrome in heteropaternal dizygotic twins. N Engl J Med. 1982;307(2):99-101. doi:10.1056/NEJM198207083070206
49. Finnell RH, Chernoff GF. Variable patterns of malformation in the mouse fetal hydantoin syndrome. Am J Med Genet. 1984;19(3):463-71. doi:10.1002/ajmg.1320190307
50. Finnell RH. Phenytoin-induced teratogenesis: a mouse model. Science. 1981;211(4481):483-4. doi:10.1126/science.7455686
51. Buehler BA, Delimont D, van Waes M, Finnell RH. Prenatal prediction of risk of the fetal hydantoin syndrome. N Engl J Med. 1990;322(22):1567-72. doi:10.1056/NEJM199005313222204
52. Jones KL, Jones MC, del Campo M. Smith's Recognizable Patterns of Human Malformation. 8th ed. Philadelphia: Elsevier; 2022. Prenatal diagnosis; p. 950–4.
53. Buehler BA, Rao V, Finnell RH. Biochemical and molecular teratology of fetal hydantoin syndrome. Neurol Clin. 1994;12(4):741-8.
54. Aggarwal S, Barman M, Poudel B, Joshi K, Devi R, Singh P, et al. Fetal hydantoin syndrome: a case report. Cureus. 2023;15(11):e49663. doi:10.7759/cureus.49663
55. Danielsson BR, Lansdell K, Patmore L, Tomson T. Phenytoin and phenobarbital inhibit human HERG potassium channels. Epilepsy Res. 2003;55(1-2):147-57. doi:10.1016/s0920-1211(03)00119-0
56. Azarbayjani F, Danielsson BR. Embryonic arrhythmia by inhibition of HERG channels: a common hypoxia-related teratogenic mechanism for antiepileptic drugs? Epilepsia. 2002;43(5):457-68. doi:10.1046/j.1528-1157.2002.28999.x
57. Danielsson BR, Danielsson C, Nilsson MF. Embryonic cardiac arrhythmia and generation of reactive oxygen species: common teratogenic mechanism for IKr blocking drugs. Reprod Toxicol. 2007;24(1):42-56. doi:10.1016/j.reprotox.2007.04.005
58. Danielsson C, Azarbayjani F, Sköld AC, Sjögren N, Danielsson BR. Polytherapy with hERG-blocking antiepileptic drugs: increased risk for embryonic cardiac arrhythmia and teratogenicity. Birth Defects Res A Clin Mol Teratol. 2007;79(8):595-603. doi:10.1002/bdra.20378
59. Azarbayjani F, Borg LA, Danielsson BR. Increased susceptibility to phenytoin teratogenicity: excessive generation of reactive oxygen species or impaired antioxidant defense? Basic Clin Pharmacol Toxicol. 2006;99(4):305-11. doi:10.1111/j.1742-7843.2006.pto_416.x
60. Wells PG, Bhuller Y, Chen CS, Jeng W, Kasapinovic S, Kennedy JC, et al. Molecular and biochemical mechanisms in teratogenesis involving reactive oxygen species. Toxicol Appl Pharmacol. 2005;207(2 Suppl):354-66. doi:10.1016/j.taap.2005.01.061
61. Winn LM, Wells PG. Maternal administration of superoxide dismutase and catalase in phenytoin teratogenicity. Free Radic Biol Med. 1999;26(3-4):266-74. doi:10.1016/s0891-5849(98)00193-2
62. Will M, Barnard JA, Said HM, Ghishan FK. Fetal hydantoin syndrome: inhibition of placental folic acid transport as a potential mechanism for fetal growth retardation in the rat. Res Commun Chem Pathol Pharmacol. 1985;48(1):91-8.
63. Hernández-Díaz S, Werler MM, Walker AM, Mitchell AA. Folic acid antagonists during pregnancy and the risk of birth defects. N Engl J Med. 2000;343(22):1608-14. doi:10.1056/NEJM200011303432204
64. Kowalski TW, Gomes JDA, Feira MF, Dupont ÁV, Recamonde-Mendoza M, Vianna FSL. Anticonvulsants and chromatin-genes expression: a systems biology investigation. Front Neurosci. 2020;14:591196. doi:10.3389/fnins.2020.591196
65. GeneGuidance (base local do projeto). HLA-B [registo curado]. Consultado 2026-10-17.
66. US Food and Drug Administration. TEGRETOL (carbamazepine) prescribing information. East Hanover (NJ): Novartis Pharmaceuticals Corporation; SPL version effective 20 August 2026. NDA 016608/018927/020234.
67. Co-ordination Group for Mutual Recognition and Decentralised Procedures — Human (CMDh). Position of the CMDh on Periodic Safety Update Reports for phenytoin, with annexes on product information amendments. EMA/CMDh/209316/2021, PSUSA/00002392/202008. Amsterdam, 22 April 2021 (PRAC recommendation adopted on 9 April 2021). Reproduced in: Arzneimittelkommission der deutschen Ärzteschaft, Archiv 2021. Available from: https://www.akdae.de/Arzneimittelsicherheit/Weitere/Archiv/2021/20210927.pdf
68. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Coarse facial features; p. 106–9.
69. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Seizures with developmental delay; p. 308–11.
70. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Limb reduction defects; p. 212–5.
71. Spranger JW, Brill PW, Hall C, Nishimura G, Superti-Furga A, Unger S. Bone Dysplasias: An Atlas of Genetic Disorders of Skeletal Development. 4th ed. Oxford University Press; 2018. Brachydactyly B (MIM 113000); p. 455–6.
72. Holmes LB, Hunt AT. Different fetal effects on fingers from exposure to phenytoin, phenobarbital, and carbamazepine. Am J Med Genet A. 2024;194(9):e63620. doi:10.1002/ajmg.a.63620
73. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Chondrodysplasia punctata; p. 98–101.
74. Spranger JW, Brill PW, Hall C, Nishimura G, Superti-Furga A, Unger S. Bone Dysplasias: An Atlas of Genetic Disorders of Skeletal Development. 4th ed. Oxford University Press; 2018. Chondrodysplasia punctata, tibia-metacarpal type (MIM 118651); p. 371–4.
75. Spranger JW, Brill PW, Hall C, Nishimura G, Superti-Furga A, Unger S. Bone Dysplasias: An Atlas of Genetic Disorders of Skeletal Development. 4th ed. Oxford University Press; 2018. Chondrodysplasia punctata, Conradi-Hünermann type (MIM 302960); p. 353–6.
76. Tomson T, Battino D, Bonizzoni E, Craig J, Lindhout D, Perucca E, et al. Comparative risk of major congenital malformations with eight different antiepileptic drugs: a prospective cohort study of the EURAP registry. Lancet Neurol. 2018;17(6):530-8. doi:10.1016/S1474-4422(18)30107-8
77. Thomas SV, Jeemon P, Pillai R, Jose M, Lalithakumari AM, Murali S, et al. Malformation risk of new anti-epileptic drugs in women with epilepsy; observational data from the Kerala registry of epilepsy and pregnancy (KREP). Seizure. 2021;93:127-32. doi:10.1016/j.seizure.2021.10.015
78. Weston J, Bromley R, Jackson CF, Adab N, Clayton-Smith J, Greenhalgh J, et al. Monotherapy treatment of epilepsy in pregnancy: congenital malformation outcomes in the child. Cochrane Database Syst Rev. 2016;11(11):CD010224. doi:10.1002/14651858.CD010224.pub2
79. Veroniki AA, Cogo E, Rios P, Straus SE, Finkelstein Y, Kealey R, et al. Comparative safety of anti-epileptic drugs during pregnancy: a systematic review and network meta-analysis of congenital malformations and prenatal outcomes. BMC Med. 2017;15(1):95. doi:10.1186/s12916-017-0845-1
80. Tomson T, Battino D, Bonizzoni E, Craig J, Lindhout D, Sabers A, et al.; EURAP study group. Dose-dependent risk of malformations with antiepileptic drugs: an analysis of data from the EURAP epilepsy and pregnancy registry. Lancet Neurol. 2011;10(7):609-17. doi:10.1016/S1474-4422(11)70107-7
81. Meador KJ, Baker GA, Browning N, Clayton-Smith J, Combs-Cantrell DT, Cohen M, et al. Cognitive function at 3 years of age after fetal exposure to antiepileptic drugs. N Engl J Med. 2009;360(16):1597-605. doi:10.1056/NEJMoa0803531
82. Meador KJ, Baker GA, Browning N, Cohen MJ, Bromley RL, Clayton-Smith J, et al. Fetal antiepileptic drug exposure and cognitive outcomes at age 6 years (NEAD study): a prospective observational study. Lancet Neurol. 2013;12(3):244-52. doi:10.1016/S1474-4422(12)70323-X
83. Bromley R, Weston J, Adab N, Greenhalgh J, Sanniti A, McKay AJ, et al. Treatment for epilepsy in pregnancy: neurodevelopmental outcomes in the child. Cochrane Database Syst Rev. 2014;2014(10):CD010236. doi:10.1002/14651858.CD010236.pub2
84. Veroniki AA, Rios P, Cogo E, Straus SE, Finkelstein Y, Kealey R, et al. Comparative safety of antiepileptic drugs for neurological development in children exposed during pregnancy and breast feeding: a systematic review and network meta-analysis. BMJ Open. 2017;7(7):e017248. doi:10.1136/bmjopen-2017-017248
85. Cohen MJ, Meador KJ, Loring DW, Matthews AG, Brown C, Robalino CP, et al. Behavioral outcomes and neurodevelopmental disorders among children of women with epilepsy. JAMA Neurol. 2024;81(1):19-29. doi:10.1001/jamaneurol.2023.4315
86. Meador KJ, Cohen MJ, Loring DW, Matthews AG, Brown CA, Robalino C, et al. Association of prenatal exposure to antiseizure medications with creative and executive function at age 4.5 years. Neurology. 2024;102(12):e209448. doi:10.1212/WNL.0000000000209448
87. Holmes LB, Coull BA, Dorfman J, Rosenberger PB. The correlation of deficits in IQ with midface and digit hypoplasia in children exposed in utero to anticonvulsant drugs. J Pediatr. 2005;146(1):118-22. doi:10.1016/j.jpeds.2004.08.048
88. Bromley RL, Mawer GE, Briggs M, Cheyne C, Clayton-Smith J, García-Fiñana M, et al. The prevalence of neurodevelopmental disorders in children prenatally exposed to antiepileptic drugs. J Neurol Neurosurg Psychiatry. 2013;84(6):637-43. doi:10.1136/jnnp-2012-304270
89. Christensen J, Grønborg TK, Sørensen MJ, Schendel D, Parner ET, Pedersen LH, et al. Prenatal valproate exposure and risk of autism spectrum disorders and childhood autism. JAMA. 2013;309(16):1696-703. doi:10.1001/jama.2013.2270
90. Straub L, Hernández-Díaz S, Bateman BT, et al. Prenatal antiseizure drug exposure and risk of neurodevelopmental disorders in children: population based cohort study. BMJ. 2026;392:e085725. doi:10.1136/bmj-2025-085725
91. Adams J, Janulewicz PA, Macklin EA, Dhillon R, Phillips C, Schomer DL, et al. Neuropsychological effects in children exposed to anticonvulsant monotherapy during gestation: phenobarbital, carbamazepine, and phenytoin. Epilepsy Behav. 2022;127:108533. doi:10.1016/j.yebeh.2021.108533
92. Dean JC, Hailey H, Moore SJ, Lloyd DJ, Turnpenny PD, Little J. Long term health and neurodevelopment in children exposed to antiepileptic drugs before birth. J Med Genet. 2002;39(4):251-9. doi:10.1136/jmg.39.4.251
93. Thomas SV, Salim S, Jacob NS, Jose M, Salini RA, Selvaraj S, et al. Language, intelligence, and educational outcomes of adolescents with antenatal exposure to antiseizure medications: prospective data from the Kerala Registry of Epilepsy and Pregnancy. Seizure. 2022;100:76-81. doi:10.1016/j.seizure.2022.06.005
94. Dreier JW, Bjørk MH, Alvestad S, Gissler M, Igland J, Leinonen MK, et al. Prenatal exposure to antiseizure medication and incidence of childhood- and adolescence-onset psychiatric disorders. JAMA Neurol. 2023;80(6):568-77. doi:10.1001/jamaneurol.2023.0674
95. US Food and Drug Administration. MYSOLINE (primidone) prescribing information. Bridgewater (NJ): Bausch Health US; SPL version effective 10 September 2026. NDA 009170.
96. Avachat C, Meador KJ, Pennell PB, Birnbaum AK, et al. Maternal and umbilical cord plasma concentrations of antiseizure medications: results from the observational MONEAD study. Epilepsia. 2026;67(7):3401-9. doi:10.1002/epi.70129
97. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Hypertrichosis; p. 178–81.
98. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Optic nerve hypoplasia; p. 270–1.
99. Allen RW Jr, Ogden B, Bentley FL, Jung AL. Fetal hydantoin syndrome, neuroblastoma, and hemorrhagic disease in a neonate. JAMA. 1980;244(13):1464-5.
100. Sherman S, Roizen N. Fetal hydantoin syndrome and neuroblastoma. Lancet. 1976;1(7984):517. doi:10.1016/s0140-6736(76)90808-4
101. Singh G, Driever PH, Sander JW. Cancer risk in people with epilepsy: the role of antiepileptic drugs. Brain. 2005;128(Pt 1):7-17. doi:10.1093/brain/awh363
102. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Cleft lip and palate; p. 102–5.
103. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Hypospadias; p. 186–7.
104. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Approach to the consultation: dysmorphism and developmental delay; p. 4–7.
105. Kini U, Adab N, Vinten J, Fryer A, Clayton-Smith J; Liverpool and Manchester Neurodevelopmental Study Group. Dysmorphic features: an important clue to the diagnosis and severity of fetal anticonvulsant syndromes. Arch Dis Child Fetal Neonatal Ed. 2006;91(2):F90-5. doi:10.1136/adc.2004.067421
106. Tomson T, Lindbom U, Ekqvist B, Sundqvist A. Disposition of carbamazepine and phenytoin in pregnancy. Epilepsia. 1994;35(1):131-5. doi:10.1111/j.1528-1157.1994.tb02922.x
107. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Epilepsy in adults; p. 416–9.
108. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Neural tube defects; p. 500–5.
109. Firth HV, Hurst JA. Oxford Desk Reference: Clinical Genetics and Genomics. 2nd ed. Oxford: Oxford University Press; 2017. Maternal phenylketonuria (PKU); p. 758-9.
110. Jones KL, Jones MC, del Campo M. Smith's Recognizable Patterns of Human Malformation. 8th ed. Philadelphia: Elsevier; 2022. Fetal warfarin syndrome; p. 796–7.
111. Jones KL, Jones MC, del Campo M. Smith's Recognizable Patterns of Human Malformation. 8th ed. Philadelphia: Elsevier; 2022. Coffin-Siris syndrome; p. 816–9.
112. ClinicalTrials.gov. Query of API v2 (interventional studies; “phenytoin AND pregnancy”, “antiseizure medication AND pregnancy”) and records NCT01730170 (MONEAD) e NCT00021866 (NEAD). Available from: https://clinicaltrials.gov/
113. ClinicalTrials.gov. Queries of API v2 without restriction on study type: “epilepsy pregnancy” with antiseizure, antiepileptic and phenytoin; “phenytoin OR fosphenytoin” as intervention; “fetal hydantoin” and “fetal anticonvulsant syndrome”. Available from: https://clinicaltrials.gov/api/v2/studies [accessed 2026-10-17]
114. Tettenborn B, Ramantani G, Flügel D, Gaus V, et al. Women with epilepsy: evidence-based counseling across the lifespan. Epilepsia. 2026;67(6):2689-706. doi:10.1002/epi.70202
115. Perucca E, Ruprah M, Richens A. Altered drug binding to serum proteins in pregnant women: therapeutic relevance. J R Soc Med. 1981;74(6):422-6. doi:10.1177/014107688107400606
116. Therapeutic Goods Administration (Australia). Antiepileptic medicines — updated precautions for pregnancy and women of childbearing potential. Medicines Safety Update, 22 June 2022. Available from: https://www.tga.gov.au/news/safety-updates/antiepileptic-medicines-updated-precautions-pregnancy-and-women-childbearing-potential
117. Hernández-Díaz S, Smith CR, Shen A, Mittendorf R, Hauser WA, Yerby M, et al. Comparative safety of antiepileptic drugs during pregnancy. Neurology. 2012;78(21):1692-9. doi:10.1212/WNL.0b013e3182574f39
118. Tomson T, Battino D, Bromley R, Kochen S, Meador K, Pennell P, et al. Management of epilepsy in pregnancy: a report from the International League Against Epilepsy Task Force on Women and Pregnancy. Epileptic Disord. 2019;21(6):497-517. doi:10.1684/epd.2019.1105
119. Tomson T, Battino D, Bromley R, Kochen S, Meador K, Pennell P, et al. Executive summary: management of epilepsy in pregnancy: a report from the International League Against Epilepsy Task Force on Women and Pregnancy. Epilepsia. 2019;60(12):2343-5. doi:10.1111/epi.16395
120. Tomson T, Battino D, Bromley R, Kochen S, Meador KJ, Pennell PB, et al. Global survey of guidelines for the management of epilepsy in pregnancy: a report from the International League Against Epilepsy Task Force on Women and Pregnancy. Epilepsia Open. 2020;5(3):366-70. doi:10.1002/epi4.12420
121. Sun Y, Alvestad S, Cohen JM, Igland J, Gilhus NE, Vegrim HM, et al. Timing of high-dose folic acid supplementation in the periconceptional period among women taking antiseizure medications and risk of major congenital anomalies: a target trial emulation. J Neurol Neurosurg Psychiatry. 2026;97(6):542-52. doi:10.1136/jnnp-2025-337395
122. Vegrim HM, Dreier JW, Igland J, Gilhus NE, Alvestad S, Gissler M, et al. Prescribing patterns for higher dose folic acid in pregnant women with epilepsy treated with antiseizure medication. Epilepsia. 2024;65(6):1698-708. doi:10.1111/epi.17969
123. Keni RR, Jose M, Sarma PS, Thomas SV. Anti-epileptic drug and folic acid usage during pregnancy, seizure and malformation outcomes: changes over two decades in the Kerala Registry of Epilepsy and Pregnancy. Epilepsy Res. 2020;159:106250. doi:10.1016/j.eplepsyres.2019.106250
124. Vegrim HM, Dreier JW, Alvestad S, Gilhus NE, Gissler M, Igland J, et al. Cancer risk in children of mothers with epilepsy and high-dose folic acid use during pregnancy. JAMA Neurol. 2022;79(11):1130-8. doi:10.1001/jamaneurol.2022.2977
125. Vegrim HM, Dreier JW, Igland J, Alvestad S, Gilhus NE, Gissler M, et al. High-dose folic acid use and cancer risk in women who have given birth: a register-based cohort study. Epilepsia. 2025;66(1):75-88. doi:10.1111/epi.18146
126. Bjørk MH, Tomson T, Dreier JW, Alvestad S, Gilhus NE, Gissler M, et al. High-dose folic acid and cancer risk; unjustified concerns by von Wrede and colleagues regarding our paper. Epilepsia. 2023;64(9):2244-8. doi:10.1111/epi.17717
127. Meador KJ, Cohen MJ, Loring DW, Matthews AG, Brown C, Robalino CP, et al. Neuropsychological outcomes in 6-year-old children of women with epilepsy: a prospective nonrandomized clinical trial. JAMA Neurol. 2025;82(1):30-9. doi:10.1001/jamaneurol.2024.3982
128. Li Y, Loring DW, Cohen MJ, Gerard EE, Birnbaum AK, Matthews AG, et al. Breastfeeding and neurodevelopmental outcomes in children of women taking antiseizure medications. JAMA Netw Open. 2026;9(9):e2637040. doi:10.1001/jamanetworkopen.2026.37040
129. Phenytoin. In: Drugs and Lactation Database (LactMed) [Internet]. Bethesda (MD): National Institute of Child Health and Human Development; 2006–. Record NBK501273, last revision 15 November 2024. Available from: https://www.ncbi.nlm.nih.gov/books/NBK501273/
130. Singh R, Kumar N, Arora S, Bhandari R, Jain A. Fetal hydantoin syndrome and its anaesthetic implications: a case report. Case Rep Anesthesiol. 2012;2012:370412. doi:10.1155/2012/370412
131. Morrow J, Russell A, Guthrie E, Parsons L, Robertson I, Waddell R, et al. Malformation risks of antiepileptic drugs in pregnancy: a prospective study from the UK Epilepsy and Pregnancy Register. J Neurol Neurosurg Psychiatry. 2006;77(2):193-8. doi:10.1136/jnnp.2005.074203
132. Lausten-Thomsen U, Olsen RH, Christiansen M, Hedley PL, Heerfordt IM, Andersen JT, et al. Spontaneous adverse drug reaction reporting of congenital malformations: a Danish national register study. Pharmaceuticals (Basel). 2025;18(6):917. doi:10.3390/ph18060917
133. Stjerna S, Huber-Mollema Y, Tomson T, Perucca E, Battino D, Craig J, et al. Cognitive outcomes after fetal exposure to carbamazepine, lamotrigine, valproate or levetiracetam monotherapy: data from the EURAP neurocognitive extension protocol. Epilepsy Behav. 2024;159:110024. doi:10.1016/j.yebeh.2024.110024
134. Glover SJ, Quinn AG, Barter P, Hart J, Moore SJ, Dean JC, et al. Ophthalmic findings in fetal anticonvulsant syndrome(s). Ophthalmology. 2002;109(5):942-7. doi:10.1016/s0161-6420(02)00959-4