Use of antiepileptic drugs (AEDs) during pregnancy is associated with an increased risk of congenital malformations. Spina bifida aperta has been linked specifically to valproic acid (VPA) (estimated risk, 1 to 2%). The actual risk, the exclusive association of VPA with spina bifida and not anencephaly, and the precise causative relation remain matters of discussion. A prospective cohort study of pregnant women with epilepsy receiving AEDs and referred for prenatal diagnosis before week 22 of gestation was conducted, with follow-up to 3 months after birth. Pregnancies (291 singleton and 6 twin) in 261 women were evaluated. The prevalence of anomalies after exposure to any AED was 6.9%. For fetuses exposed to VPA, the prevalence was 9.4%, including six cases of spina bifida, two of which were in monozygotic twins (giving a prevalence rate of 6.3%, or 5.4%, if twins counted as one). Spina bifida was associated with a significantly higher average daily dose of VPA as compared with pregnancies with normal outcome (1.640 +/- 136 mg/d vs 941 +/- 48 mg/d, p = 0.0001). No relation was observed between the occurrence of spina bifida and type of maternal seizure or epilepsy, family history of epilepsy or neural-tube defects, or medical history. From these results we suggest that when the use of VPA during pregnancy cannot be avoided, the teratogenic risk might be diminished by reduction of the daily dose.