To compare the incidence of foetal malformations (FMs) in pregnant women with epilepsy treated with different anti-epileptic drugs (AED) and doses, and the influence of seizures, family and personal history, and environmental factors. A prospective, observational, community-based cohort study. Methods. A voluntary, Australia-wide, telephone-interview-based register prospectively enrolling three groups of pregnant women: taking AEDs for epilepsy; with epilepsy not taking AEDs; taking AEDs for a non-epileptic indication. Four hundred and fifty eligible women were enrolled over 40 months. Three hundred and ninety six pregnancies had been completed, with 7 sets of twins, for a total of 403 pregnancy outcomes. Results. 354 (87.8%) pregnancy outcomes resulted in a healthy live birth, 26 (6.5%) had a FM, 4 (1%) a death in utero, 1 (0.2%) a premature labour with stillbirth, 14 (3.5%) a spontaneous abortion and 4 lost to follow-up. The FM rate was greater in pregnancies exposed to sodium valproate (VPA) in the first trimester (16.0%) compared with those exposed to all other AEDs (16.0% vs. 2.4%, P<0.01) or no AEDs (16.0% vs. 3.1%, P<0.01). The mean daily dose of VPA taken in pregnancy with FMs was significantly greater than in those without (1975 vs. 1128 mg, P<0.01). The incidence of FM with VPA doses ≥1100 mg was 30.2% vs. 3.2% with doses <1100 mg (P<0.01). Conclusions. There is a dose-effect relationship for FM and exposure to VPA during the first trimester of pregnancy, with higher doses of VPA associated with a significantly greater risk than with lower doses or with other AEDs. These results highlight the need to limit, where possible, the dose of VPA in pregnancy.