Valproate's pregnancy risks confirmed again; levetiracetam and lamotrigine look safer

BMJ Group

Before the data, the dilemma. A pregnant woman with epilepsy needs her medication. Uncontrolled seizures endanger both her and her child. But some antiseizure drugs carry their own risks to fetal development. The question is not whether to treat, but which drug carries the least harm.

After the data: the answer is clearer for some drugs, murkier for others, and urgent for those still too new to have a solid evidence base.

A large observational study published in The BMJ in March 2026 analyzed insurance claims data from two major U.S. databases spanning 2000 to 2021, comparing 14,993 children exposed to at least one antiseizure medication during the second half of pregnancy with 8,887 unexposed children born to mothers with diagnosed epilepsy.

Valproate: the known danger, confirmed

Valproate showed associations with several neurodevelopmental disorders, including ADHD, autism spectrum disorder, and intellectual disability. This is not new information, but the study reinforces it with a larger and more diverse population than many previous analyses. The consistency of the signal across two separate insurance databases and multiple outcome measures strengthens confidence in the finding.

For clinicians, the message is the same one that regulatory agencies have been delivering for years: valproate should be avoided during pregnancy whenever an alternative exists. For patients currently taking valproate, the decision to switch must be weighed against the risk of seizure breakthrough, a conversation that requires individualized clinical judgment.

Levetiracetam and lamotrigine: no red flags

Levetiracetam and phenytoin were not associated with an increased risk of any of the studied neurodevelopmental outcomes. Lamotrigine showed no meaningful associations across most outcomes, though there was a potential signal for intellectual disability that the authors describe as based on small numbers and requiring confirmation.

These two drugs are already the most commonly prescribed alternatives to valproate for women of childbearing age. The new data provide additional reassurance for their use during pregnancy, though the authors note that absence of evidence is not evidence of absence, particularly for rarer outcomes.

Zonisamide: a new flag

Zonisamide showed associations with several neurodevelopmental disorders in this analysis. The drug is increasingly prescribed, and unlike valproate, it does not yet carry strong warnings about pregnancy risks. The authors call for further evaluation of zonisamide's safety during pregnancy, noting that this signal needs to be confirmed or refuted by additional studies.

Carbamazepine and oxcarbazepine showed modest risk increases for ADHD and behavioral disorders. Several other drugs showed associations with intellectual disability, but these estimates are based on small numbers and should be interpreted cautiously.

Observational data, inherent limits

This is an observational study using insurance claims data, not a randomized controlled trial. Claims data capture diagnoses coded for billing purposes, which may not perfectly reflect clinical reality. Some conditions may be overcoded, others undercoded. The data cannot capture disease severity, seizure frequency, or the specific clinical reasoning behind drug choices.

The study adjusted for numerous potential confounders including maternal age, ethnicity, mental health, substance use, other medications, and underlying conditions. But unmeasured confounders inevitably remain. The type and severity of epilepsy, which directly influences drug choice, could not be fully accounted for. Sicker patients may receive different drugs and may also have children at higher baseline risk of neurodevelopmental disorders.

Follow-up periods varied. Of the nearly 24,000 children studied, only 5,505 were followed for at least five years and 2,516 for at least eight years. Some neurodevelopmental disorders may not be diagnosed until later childhood, meaning the study could undercount outcomes for children with shorter follow-up.

Despite these limitations, the use of two large nationwide databases of insured pregnant women linked to their children provides strong generalizability, and results were consistent across multiple sensitivity analyses.

The monitoring imperative

Antiseizure drugs are increasingly prescribed to women of childbearing age, not only for epilepsy but also for bipolar disorder and migraine prevention. The expanding patient population makes ongoing safety monitoring essential. The authors emphasize that newer drugs with limited pregnancy data need particular attention, and that the few signals that emerged in this study, even those based on small numbers, warrant continued surveillance rather than dismissal.