Valproic Acid

Pregnancy

Major malformations: the EURAP prospective registry (Battino et al., 2024) included 1,549 valproate monotherapy pregnancies; 153 had major malformations (9.9%, 95% CI 8.5% to 11.5%). The study found increasing malformation prevalence with increasing valproate dose. Pregnancy loss: the Scottish matched population cohort (Moore et al., 2025) examined over 900,000 singleton pregnancies, including 365 with valproate monotherapy. Valproate was associated with a composite pregnancy-loss outcome (adjusted OR 1.92, 95% CI 1.50 to 2.47); the primary outcome combined spontaneous loss and termination. In supplementary analyses, valproate remained associated with termination, but not spontaneous loss. MotherToBaby says one recent study found increased miscarriage chance, but its fact sheet does not name that study, and its companion reference list does not identify a miscarriage study; that specific claim is not verified. Forbes et al. (2024) studied all antiseizure medicines, not valproate alone, and found no association after restricting the comparison to women with epilepsy (adjusted HR 0.98, 95% CI 0.89 to 1.08). Neurodevelopment: the prospective observational NEAD study included 305 mothers and 311 children; 224 children completed six-year follow-up. Mean IQ at age 6 was 97 after valproate exposure versus 105 to 108 after carbamazepine, lamotrigine, or phenytoin; there was no unexposed comparison group. A five-cohort study published in 2026 included 2,910,206 live-born children and reported an association between valproate fills overlapping the 60 days before birth and combined NDDs in Canadian data (adjusted HR 1.38, 95% CI 1.16 to 1.65; 134 exposed cases). A 2026 US epilepsy cohort found associations between second-half-pregnancy valproate exposure and several NDD outcomes, but the abstract does not report valproate-specific estimates. Fetal growth: the MHRA's June 2025 update says product information reflects studies reporting lower mean birth weight and a higher risk of low birth weight or small-for-gestational-age birth. The EURAP 2026 cohort found lower birthweight centiles with valproate monotherapy than lamotrigine (adjusted b -2.54, 95% CI -4.50 to -0.58; 1,829 valproate-exposed offspring). A 2024 Nordic cohort found no association between valproate and small-for-gestational-age birth or microcephaly. These studies measure different outcomes and use different comparison groups. Newborn: MotherToBaby reports temporary jitteriness, irritability, and low blood sugar in some newborns exposed before birth. A 2010 case report described one newborn exposed to both valproate and phenytoin who had episodic hypoglycemia; it cannot estimate frequency or isolate valproate's effect. Switching or withdrawal: a 2026 systematic review included nine studies, with three meta-analyzed; 277 women withdrew or switched and 2,206 remained on valproate. The odds of tonic-clonic seizure recurrence were higher after withdrawal or switching (OR 1.73, 95% CI 1.06 to 2.84). Evidence on fetal outcomes and maternal complications remained limited.

Breastfeeding

LactMed reports low levels in milk and infant serum levels from undetectable to low. Kacirova et al. (2019) retrospectively studied 30 nursing women, measuring maternal serum, mature milk, and infant serum between postpartum days 6 and 32. The mean milk concentration was 1.6 mg/L and the mean infant serum concentration was 4.2 mg/L; 67% of milk and 33% of infant concentrations were below quantification, and none reached the lower limit of the reference range for people treated for epilepsy. The study measured drug levels, not a rate of infant injury. LactMed reports no definite adverse reactions in breastfed infants, but notes a theoretical risk of liver injury, a questionable thrombocytopenia report, and a rare case of infant hair loss that might have been caused by valproate in milk. MotherToBaby also describes a theoretical liver concern. DailyMed advises monitoring for jaundice or unusual bruising or bleeding. The 2025 EudraVigilance report summarized 922 suspected infant adverse-reaction reports across medicines and did not provide a valproate-specific incidence, so it is not used to estimate valproate risk.

What was tested

The pregnancy evidence is mainly observational: prospective international registries, national population cohorts, and a prospective study of children's cognition. Studies measured major malformations, pregnancy loss, child development concerns, neurodevelopmental diagnoses, birthweight, and newborn symptoms. The studies differ in their exposure windows, comparison groups, outcome definitions, and follow-up. The breastfeeding evidence includes small milk and infant blood-level studies and case reports; it does not establish a reliable rate for infant harm.