Balancing maternal seizure control against fetal drug-exposure risk across preconception, pregnancy, and delivery — an illustrative model of antiepileptic drug (AED) teratogenicity management
The single highest-leverage intervention in AED teratogenicity management happens before conception. Roughly half of pregnancies are unplanned, which means every clinician managing a person with epilepsy of childbearing potential should treat every visit as a preconception visit. The goal of preconception counseling is not to stop medication reflexively, but to systematically ask whether the current drug and dose represent the lowest-risk regimen that still reliably controls seizures.
A preconception AED review typically works through several linked questions:
• Diagnosis confirmation: is the epilepsy syndrome correctly classified? Misclassification (e.g. treating juvenile myoclonic epilepsy as focal epilepsy) can lead to an unnecessarily high-risk drug choice when a lower-risk alternative would control the actual syndrome.
• Monotherapy vs. polytherapy: combination AED regimens carry substantially higher cumulative malformation risk than monotherapy at equivalent seizure control. Simplifying to the fewest effective agents, ideally one, is a primary goal.
• Dose minimization: risk for most AEDs scales with dose/serum concentration, not just drug identity. Confirming the lowest dose that maintains seizure freedom — rather than a historical or reflexively escalated dose — reduces exposure without sacrificing control.
• Switching feasibility: if a person is seizure-free on a higher-risk agent (e.g. valproate) for a syndrome that responds equally well to a lower-risk agent (e.g. lamotrigine or levetiracetam), a supervised cross-titration well before conception is often appropriate. Switching during an established pregnancy is generally avoided because of destabilization risk during the taper.
• Timing: switches and dose optimization should ideally be completed and stabilized for several months before conception, allowing time to confirm continued seizure control on the new regimen.
Preconception counseling is most protective when it happens routinely, not only when a pregnancy is already planned — because a meaningful share of pregnancies in people with epilepsy are unplanned, and first-trimester organogenesis is complete before many patients realize they are pregnant.
The preconception dashboard visualization abstracts this review process: a set of drug options are compared side by side, with the currently selected agent and dose highlighted. Adjusting the sliders lets you see how switching drug class or trimming dose shifts the projected risk metrics in the panel — a simplified stand-in for the kind of individualized risk-benefit conversation that happens between a patient and their neurologist or epileptologist before conception.
One of the clearest findings from large pregnancy registries (the North American AED Pregnancy Registry, UK/Ireland Epilepsy and Pregnancy Register, EURAP) is that teratogenic risk varies substantially by drug and, within a drug, by dose. Valproate stands out consistently across registries as carrying both the highest major congenital malformation rate and the most pronounced dose-dependent neurodevelopmental risk, while lamotrigine and levetiracetam are consistently among the lower-risk options at typical doses.
Valproate's elevated risk is not a single-defect signal but a broad pattern: neural tube defects (spina bifida), craniofacial anomalies, cardiac defects, hypospadias, and limb defects are all reported at increased frequency. Beyond structural malformations, valproate exposure is also associated with reduced cognitive outcomes and increased risk of neurodevelopmental disorders including autism spectrum features, in a dose-dependent manner — an effect that is less clearly demonstrated, or demonstrated at much lower magnitude, for lamotrigine, levetiracetam, or carbamazepine.
Because of this pattern, most contemporary guidelines recommend avoiding valproate in people of childbearing potential whenever an equally effective alternative exists, reserving it for situations (such as certain generalized epilepsies poorly controlled by anything else) where its unique efficacy outweighs the risk — with explicit, informed, shared decision-making.
Registry data consistently place valproate's major malformation risk at roughly three to four times that of lamotrigine or levetiracetam, and the gap widens further at higher valproate doses — making valproate the clearest example of a dose- and drug-dependent teratogenicity signal in this drug class.
Carbamazepine occupies an intermediate position: its malformation risk is lower than valproate but somewhat higher than lamotrigine or levetiracetam, with a particular signal for neural tube defects and, in some analyses, cardiac and urinary tract anomalies. Risk again scales with dose. For many focal epilepsies, carbamazepine remains a reasonable option when lamotrigine or levetiracetam are not effective or not tolerated, particularly at the lowest effective dose.
The Stage 2 canvas renders each of the four drugs as a bar whose height reflects its illustrative malformation-risk estimate at the currently selected dose. Moving the dose slider re-scales all four bars simultaneously (since serum concentration effects apply across drugs, just with different steepness), while the drug selector highlights which bar corresponds to the agent currently reflected in the metrics panel. The intent is to make the relative — not just absolute — risk gap between agents visually obvious.
Independent of which AED is chosen, folic acid supplementation is one of the most consistently protective interventions available for reducing neural tube defect (NTD) risk in pregnancies exposed to antiepileptic drugs. Because several AEDs interfere with folate metabolism, and because NTD risk is elevated across the class (most sharply for valproate and carbamazepine), high-dose folic acid — well above the standard population-level recommendation — is commonly advised for people with epilepsy who may become pregnant.
The neural tube closes very early in embryogenesis — by roughly the fourth week after conception, often before a person knows they are pregnant. This means folic acid supplementation is most protective when it is already at steady state at the time of conception, not started after a positive pregnancy test. This is a second major reason preconception counseling (Stage 1) matters: it creates the window needed to establish adequate folate status before the neural tube is even forming.
Folate is a cofactor in one-carbon metabolism, supporting DNA synthesis and methylation reactions essential for rapid cell division during neurulation. Several AEDs (valproate and carbamazepine most notably) are thought to interfere with folate metabolism through enzyme induction or direct antifolate effects, compounding baseline NTD risk from the drug itself.
Because the neural tube closes so early, folic acid protects most effectively when supplementation predates conception — reinforcing that preconception planning and folate optimization are two parts of the same intervention, not separate steps.
The Stage 3 canvas animates small folate molecules converging on and stabilizing a developing neural tube structure as it closes along its length. This is a simplified, illustrative metaphor for a biochemical process (one-carbon metabolism supporting neural crest and neural plate cell division) rather than a literal depiction — folate does not physically "coat" the tube — but it communicates the protective, cofactor role folic acid plays during this critical developmental window.
Pregnancy is not pharmacokinetically neutral: plasma volume expands, protein binding shifts, renal blood flow and glomerular filtration increase, and hepatic metabolic enzyme activity changes — sometimes substantially. For several AEDs, this means a dose that maintained seizure freedom before pregnancy may no longer produce an adequate serum level as pregnancy progresses, making regular level monitoring and dose adjustment a core part of safe management.
Lamotrigine is cleared substantially through hepatic glucuronidation (UGT enzymes), and pregnancy — driven largely by rising estrogen — markedly increases UGT activity. The result can be a two- to three-fold increase in lamotrigine clearance by the third trimester in some patients, meaning serum levels can fall well below the preconception baseline even without any dose change. Falling levels can translate directly into breakthrough seizures if not anticipated.
Because of this, many clinicians establish an individual target level before conception, then monitor levels each trimester (and around any change in seizure frequency), increasing the dose as needed to maintain that target — then reducing the dose again postpartum as clearance normalizes over roughly the first few weeks after delivery, to avoid postpartum toxicity from a dose that is now "too high" for the non-pregnant state.
A dose increase made to compensate for pregnancy-related clearance is not a signal that the drug has become "more dangerous" — it is a pharmacokinetic correction to maintain the same effective exposure and the same seizure control that existed before conception.
Levetiracetam clearance also rises during pregnancy, driven by increased renal elimination as glomerular filtration rate climbs — again supporting trimester-based level checks. Valproate and carbamazepine show smaller but still clinically relevant pharmacokinetic shifts, including changes in protein binding that affect how "total" versus "free" drug levels should be interpreted.
The general principle across all agents: an individualized target level established when the patient is stable and seizure-free (ideally preconception) is the reference point; levels are rechecked periodically through pregnancy and postpartum, and dose is adjusted to keep the patient near that reference — not simply held constant regardless of what pregnancy is doing to clearance.
It is tempting to think of AED teratogenicity management as simply "minimize drug exposure." But uncontrolled maternal seizures — particularly generalized tonic-clonic seizures — carry their own well-documented fetal risks: transient maternal and fetal hypoxia, physical trauma from falls, placental abruption, and in the most severe cases, status epilepticus, which carries meaningful maternal and fetal mortality risk. The clinical goal is therefore not zero drug exposure, but the lowest effective dose of the safest appropriate drug that reliably prevents seizures.
A common misconception is that discontinuing AEDs during pregnancy removes fetal risk. In reality, for most people with a clinically significant epilepsy diagnosis, discontinuation substantially raises the risk of breakthrough seizures — and generalized convulsive seizures during pregnancy are themselves associated with adverse fetal outcomes, including reduced fetal heart rate during and after the seizure, increased miscarriage risk, and, in severe or prolonged cases (status epilepticus), significant maternal and fetal mortality risk.
This reframes the clinical question away from "drug vs. no drug" and toward "which drug, at which dose, achieves seizure freedom with the least fetal risk" — the same question addressed in Stages 1 and 2, now viewed through the lens of the alternative (uncontrolled seizures) rather than in isolation.
Because untreated seizures carry their own fetal risk, the clinical target is not the absence of medication but the lowest effective dose of the lowest-risk drug that reliably controls seizures — a genuine risk-risk balance rather than a simple minimization problem.
The Stage 5 canvas renders a scale (seesaw) with "seizure risk" on one side and "drug exposure risk" on the other. As you increase the dose slider, drug-exposure risk rises on its side of the scale; conversely, an implicit assumption in the model is that higher doses of a given drug improve seizure control, so very low doses tip the scale toward seizure risk instead. The selected drug's risk slope (from Stage 2) determines how steeply the drug-exposure side responds to dose changes — visually reinforcing why the same dose increase means very different things for valproate versus lamotrigine or levetiracetam.