Prenatal exposure to certain antiseizure medications (ASMs) was associated with an increased risk for neurodevelopmental disorders (NDDs) in children, with risk varying by individual drug.

Results of a large US population-based cohort study showed that valproate exposure was associated with an increased risk for NDD, whereas levetiracetam and phenytoin were not; zonisamide showed a potential signal, though estimates were imprecise because of the relatively small number of children exposed to this drug in utero.

Lead investigator Loreen Straub, MD, MS, instructor of medicine at Harvard Medical School and investigator in the Division of Pharmacoepidemiology and Pharmacoeconomics at Brigham and Women’s Hospital, both in Boston, said the key clinical question is not whether ASMs are broadly “safe” or “unsafe,” but which specific drug offers the best balance of efficacy and safety for an individual patient during pregnancy taking into account epilepsy type, disease severity, and prior treatment response.

“Our study adds to that individualized evidence base for one especially important outcome category — neurodevelopmental disorders in their offspring,” she told Medscape Medical News.

The study was published online on March 11 in The BMJ.

Evaluating Individual Agents

Valproate use during pregnancy has consistently been associated with impaired neurodevelopment in children. Current guidelines advise against its use in this setting and recommend alternatives such as lamotrigine, levetiracetam, or oxcarbazepine when appropriate, while noting that robust data on neurodevelopmental outcomes remain limited for many ASMs.

Given the widespread use of ASMs in women of childbearing age and limited comparative data on their neurodevelopmental safety, the researchers conducted the study to evaluate the risks associated with individual agents.

For the study, the researchers analyzed healthcare utilization data from two large US databases — the Medicaid Analytic eXtract/Transformed Medicaid Statistical Information System (MAX/TAF) and the Merative MarketScan Commercial Claims and Encounters Database — for the years 2000 through 2021.

The study included pregnant patients aged 12-55 years with epilepsy and their liveborn children. The final cohort included 14,993 children (MAX/TAF, n = 10,345; MarketScan, n = 4648) exposed to at least one ASM during the second half of pregnancy and 8887 unexposed children whose mothers had no ASM dispensation from 3 months before pregnancy through delivery.

Exposure was defined as at least one prescription fill during the second half of pregnancy, a critical period for neurodevelopment when rapid formation of neural connections occurs in the fetal brain. The drugs evaluated included carbamazepine, lacosamide, lamotrigine, levetiracetam, oxcarbazepine, phenobarbital, phenytoin, topiramate, valproate, and zonisamide.

In the Medicaid cohort (MAX/TAF), participants were racially and ethnically diverse. The proportion of White participants ranged from 51% to 62%, Black or African American participants from 18% to 33%, and Hispanic or Latino participants from 12% to 19%, depending on the antiseizure drug. The mean maternal age ranged from 24.5 years in unexposed pregnancies to 26.9 years among those exposed to lacosamide. In the commercially insured cohort (MarketScan), data were available only for phenytoin, with a mean maternal age of 32.6 years; race and ethnicity were not reported.

The researchers adjusted for multiple potential confounders, including maternal age, race or ethnicity, mental health conditions, substance use, healthcare utilization, concomitant drugs, and other comorbidities.

Children were followed for a mean of 3.4 years; 5505 were followed for at least 5 years and 2516 for at least 8 years after birth.

The study’s primary outcomes included attention-deficit hyperactivity disorder (ADHD), autism, behavioral disorders, developmental coordination disorder, intellectual disability, learning disability, speech or language disorder, and any NDD.

Risk by Medication

By age 8 years, 34.3% (95% CI, 32.0%-36.7%) of children who were unexposed prenatally had an NDD. Among those exposed, the incidence ranged from 22.7% (95% CI, 12.2%-39.8%) for lacosamide to 42.6% (95% CI, 31.9%-55.3%) for zonisamide.

For any NDD, weighted hazard ratios (HRs) suggested a potential increased risk with valproate (HR, 1.29; 95% CI, 1.06-1.56) and zonisamide (HR, 1.41; 95% CI, 1.06-1.88), whereas no meaningful risk increase was seen for levetiracetam or phenytoin.

Several antiseizure drugs were associated with an increased risk for intellectual disability in children with prenatal exposure:

• Lamotrigine: HR, 2.38 (95% CI, 1.20-4.73)
• Topiramate: HR, 1.84 (95% CI, 0.75-4.51)
• Oxcarbazepine: HR, 3.30 (95% CI, 1.29-8.41)
• Valproate: HR, 1.72 (95% CI, 0.77-3.83)
• Zonisamide: HR, 4.50 (95% CI, 1.43-14.18)
• Phenobarbital: HR, 3.70 (95% CI, 0.85-16.11)

The investigators emphasized that these findings were based on small numbers and should be interpreted cautiously.

Valproate was associated with an increased risk across all specific NDDs, whereas lamotrigine was linked only to intellectual disability. Levetiracetam and phenytoin were not associated with any of the outcomes examined.

Topiramate showed an association with intellectual disability and a potential signal for learning difficulty, with a HR of 1.23 (95% CI, 0.54-2.79).

Carbamazepine and oxcarbazepine showed a modest increased risk for ADHD and behavioral disorders, with HRs ranging from 1.23 to 1.40.

Straub noted that these findings were surprising, especially for oxcarbazepine, because it is generally considered relatively safe in terms of congenital malformation risk and commonly recommended in current epilepsy guidelines.

“Our study does not overturn that, but it reminds us that a reassuring malformation profile does not mean the neurodevelopmental profile is also benign or fully understood,” she said.

Zonisamide was associated with an increased risk for most NDDs except autism, although the small number of exposed pregnancies limits the accuracy of these estimates. For phenobarbital and lacosamide, the data were insufficient to draw reliable conclusions.

Finding a Balance

When treating pregnant patients with epilepsy, neurologists must balance the need for maternal seizure control with potential risks to the developing fetus.

Straub emphasized that seizure control should remain the primary goal, but when multiple effective options are available, clinicians should favor medications with the most reassuring safety data in pregnancy. She noted that the current findings add to the evidence base, particularly with respect to neurodevelopmental outcomes.

Overall, the results were largely reassuring for commonly used ASMs. Straub added that the findings further support lamotrigine and levetiracetam as relatively safe options during pregnancy, consistent with current guideline recommendations.

She added that the interpretation for phenytoin is more limited, noting that clinicians must weigh its broader safety profile, including established risks for congenital malformations — rather than focusing solely on neurodevelopmental outcomes. This underscores the importance of evaluating overall pregnancy safety, not just neurodevelopment.

Straub also noted emerging concerns with less-studied drugs, particularly potential signals for zonisamide. She emphasized that these findings underscore the need for continued monitoring and should be replicated in other studies before drawing conclusions or informing clinical recommendations.

Continued evaluation using additional data sources with longer follow-up will be especially important for drugs such as zonisamide and lacosamide, as well as for uncommon outcomes such as intellectual disability.

“Replication in other healthcare systems and registries is critical because consistency across independent datasets is an important way to distinguish a true signal from random variation,” she said.

Reassuring but Incomplete Data

Commenting for Medscape Medical News, Kimford J. Meador, MD, professor, Department of Neurology & Neurological Sciences, Stanford University School of Medicine, Stanford, California, and Stanford Neuroscience Health Center, Palo Alto, California, said the study reinforces the known neurodevelopmental risks associated with fetal valproate exposure and raises questions about zonisamide.

Meador noted that the study is important because randomized controlled trials of ASMs in pregnancy are often either unethical — such as when they would involve assigning patients to higher-risk medications — or impractical to conduct. As a result, observational studies play a critical role in this area.

However, he cautioned that such studies remain vulnerable to residual confounding, as treatment is not randomly assigned.

“We need to find a reliable signal across repeated observational studies to determine if observed relationships to ASM fetal exposure are to be trusted,” he added, and underscored the importance of consistent findings across multiple datasets.

Although the study includes a large overall cohort, Meador noted that sample sizes for some individual drugs and specific neurodevelopmental outcomes remain small, limiting the certainty of specific findings.

He said that the new data on oxcarbazepine and zonisamide are particularly valuable because prior neurodevelopmental safety information for these drugs has been limited.

Meador said the findings confirm known risks, including those associated with valproate, and raise potential concerns for less-studied agents such as zonisamide but emphasized the need for additional research.

The study was funded by the National Institute of Mental Health. Straub and Meador reported having no relevant financial disclosures.