Background: Most women with epilepsy need to take antiepileptic drugs (AEDs) in pregnancy to prevent the potentially harmful effects of seizures. Retrospective studies have demonstrated an increased chance of having a child with a birth defect (BD) in women with epilepsy taking AEDs. It is uncertain how much of this risk is directly caused by the AEDs and whether certain drugs or combinations are associated with a greater risk. Aims: To establish a register to evaluate prospectively the incidence of adverse pregnancy outcomes in women exposed to specific AEDs; to determine whether certain AEDs or combinations were associated with a greater risk; and to determine whether other factors influenced the risk. Methods: An Australia-wide, prospective, voluntary, telephone-interview based, observational register. Three groups of pregnant women were enrolled: those with epilepsy taking AEDs, those with epilepsy not taking AEDs, and those taking AEDs for a non-epileptic indication. The pregnancy outcomes were evaluated by follow-up interviews and by reference to hospital and treating doctors' records. Results: Over the first 30 months of the study (till December 2001) 334 eligible women were enrolled, with all states and territories being represented. Two hundred and ninety two pregnancies had been completed, of which 256 (88%) resulted in a healthy live birth, 19 (6.5%) a live birth with a birth defect, four an induced abortion because of a detected malformation on ultrasound, one premature labour with a stillbirth and 12 (4%) spontaneous abortions. Of the completed pregnancies, 269 were exposed to at least one AED during the first trimester. The incidence of birth defects in relation to specific AEDs was: valproate (16.7%), phenytoin (10.5%), lamotrigine (7.7%) and carbamazepine (3.3%), none of which was significantly different from that in women with epilepsy not taking an AED (4.3%, n.s.). The dose of valproate taken was higher in pregnancies with BD compared to those without (mean 2081 mg vs. 1149 mg, p < 0.0001). The incidence of folate supplementation being taken prior to conception did not differ for pregnancy outcomes with or without BD (70% vs. 66%, n.s.). Conclusions: The model for the Australian Pregnancy Register appears to be successful, with strong enrolment from all regions over the first 30 months. The study is prospective and includes reference to all new AEDs approved in Australia over the past decade. Analysis of the pregnancy outcomes to date may reveal early trends, but numbers are still to small for any definitive conclusions to be made regarding the relative risk in pregnancy of individual AEDs.