In one line
The dangerous instinct in epilepsy and pregnancy is to stop the drug. A convulsive seizure carries an immediate risk of maternal death, fetal hypoxia and abruption, whereas the teratogenic risk of most modern antiseizure medications is a few percentage points above a background of about 2.4 to 2.9%. The work is done before conception, by getting her onto the right drug at the lowest effective dose; once she is pregnant, the priority shifts to keeping her free of convulsions.
Where this sits. The consultant layer above this chapter, covering the wider neurological differential and the woman whose seizures are not epilepsy at all, is the Finals chapter on neurological disorders in pregnancy.
The two risks, and their true sizes
Every decision here is a trade between two harms, so both need honest numbers.
Seizures. A generalised tonic-clonic seizure in pregnancy risks maternal trauma, aspiration, status epilepticus, and sudden unexpected death in epilepsy, which is a recognised cause of maternal death and is minimised by optimising seizure control. For the fetus it means abrupt hypoxia and acidosis, and it is associated with miscarriage, placental abruption and preterm birth. Focal seizures without loss of awareness carry far less of this, which is why the guidance is written specifically about convulsive seizures.
The 2024 AAN, AES and SMFM guideline puts this at its highest level of obligation: clinicians must minimise the occurrence of convulsive seizures in pregnancy to reduce risk to mother and fetus (Level A).
The medication. The birth prevalence of any major congenital malformation in the general population is about 2.4 to 2.9%, and every conversation should start there, because a woman told "3.1% risk" without that anchor hears something catastrophic.
| Antiseizure medication in monotherapy | Any major congenital malformation |
|---|---|
| Lamotrigine | ~3.1% |
| Oxcarbazepine | ~3.1% |
| Levetiracetam | ~3.5% |
| Valproate | ~9.7% |
Read those two columns together and the strategy is obvious. Lamotrigine, levetiracetam and oxcarbazepine sit essentially at background risk. Valproate is roughly three to four times background. The guideline reflects this exactly: clinicians must consider lamotrigine, levetiracetam or oxcarbazepine in women of childbearing potential where clinically appropriate (Level A), and must avoid valproate where clinically feasible (Level A).
Risk is dose-dependent. Prescribe the lowest appropriate dose of valproate, carbamazepine and lamotrigine (Level B). A woman on 500 mg of valproate is not in the same position as a woman on 2000 mg.
Monotherapy beats polytherapy. Malformation prevalence is higher with polytherapy, and every additional drug adds risk without necessarily adding control.
Which malformation, with which drug
This is worth holding as a pattern rather than a list, because it drives what you scan for.
- Valproate: the highest overall risk, and specifically neural tube defects, above all spina bifida, plus urogenital and renal malformations (Level B to avoid for this reason). The neural tube closes by 28 days after conception, which is often before the woman knows she is pregnant. This is the entire argument for acting preconceptually.
- Phenobarbital: cardiac malformations (Level A to avoid where feasible) and oral clefts (Level B). A woman on phenobarbital in pregnancy should have screening cardiac investigation of the fetus (Level B).
- Topiramate: oral clefts (Level B to avoid), and separately small for gestational age infants (Level B to avoid). A woman on topiramate needs fetal growth surveillance through the pregnancy (Level B). See Intrauterine growth restriction.
- Carbamazepine and phenytoin: intermediate risk, dose-dependent for carbamazepine.
- All women on any antiseizure medication: offer counselling on the benefits and harms of detailed fetal anatomical ultrasound (Level B).
Neurodevelopment, which is the harm that does not show on a scan
Malformation is visible and countable. The neurodevelopmental effect of valproate is neither, and it is arguably the larger harm.
The guideline states at Level A that clinicians must avoid valproate throughout pregnancy where clinically feasible to reduce poor neurodevelopmental outcomes including lower IQ and autism spectrum disorder, and must counsel any woman treated with or considering valproate that in utero exposure is associated with a decrease in full-scale, verbal and non-verbal IQ compared with carbamazepine, gabapentin, lamotrigine, levetiracetam, phenytoin and topiramate, and with an increased risk of autism spectrum disorder compared with carbamazepine, clonazepam, oxcarbazepine and lamotrigine.
Children exposed to any antiseizure medication in utero should have age-appropriate developmental screening (Level B). That is a paediatric follow-up arrangement made by the obstetric team before discharge, not a vague intention.
Folic acid, and a real disagreement
Every woman with epilepsy on an antiseizure medication takes folic acid preconceptually and through pregnancy. The guidelines disagree on the dose, and it is better to know that than to recite one number.
- AAN, AES and SMFM (2024): at least 0.4 mg daily, preconceptually and during pregnancy, to decrease neural tube defects (Level B) and, at Level A, to possibly improve neurodevelopmental outcomes including autism spectrum disorder and global IQ.
- UK practice, including NICE: 5 mg daily, on the basis that women with epilepsy are a high-risk group, the same dose used after a previous neural tube defect.
The AAN panel explicitly notes that there is considerable practice variation and that the optimal dose remains to be clarified. In South African practice the 5 mg dose is what is used, it is inexpensive, it aligns with the dose already given for previous neural tube defect and for pregestational diabetes, and no guideline argues it is harmful. Prescribe 5 mg, and be able to say that the American guidance sets a lower floor.
The timing is the part that gets missed. Folic acid must be started before conception, which means it belongs on the prescription of every woman with epilepsy of childbearing potential, not on the antenatal booking form.
Why the drug levels fall, and what to do about it
Pregnancy changes antiseizure medication pharmacokinetics substantially, in the same direction for almost every drug: levels fall.
The general mechanisms. Plasma volume expands, so the volume of distribution rises. Glomerular filtration rate rises by around 50%, so renally cleared drugs (levetiracetam is the obvious one) are eliminated faster. Serum albumin falls, which changes the free fraction of highly protein-bound drugs such as phenytoin, so a total phenytoin level can look low while the free level is adequate. Hepatic enzyme activity changes. Nausea and vomiting reduce absorption and adherence.
And one mechanism specific to lamotrigine, which is the drug most affected. Lamotrigine is cleared by glucuronidation, and UGT1A4 catalyses about 90% of it. Oestradiol up-regulates UGT1A4. So as oestrogen rises through pregnancy, lamotrigine clearance rises with it. Reported figures are striking: clearance increases by roughly 65 to 230% during gestation, and serum concentrations can fall by more than 60%. A woman who was perfectly controlled at 200 mg twice daily before conception can be subtherapeutic by the second trimester on the same prescription.