To assess the relative contribution of antiepileptic drugs (AEDS) to occurrence of congenital malformations, we compared two prospective studies. We analyzed data for 14 AEDs for total daily doses (drug score) and eight background factors. From the first study, the drug score and polytherapy--particularly the use of valproate plus carbamazepine--were suspected to be primary factors for increased incidence of congenital malformation. In the other study, the drug score for each case was decreased, and polytherapy--particularly valproate plus carbamazepine--was changed to monotherapy before conception. These changes significantly decreased the incidence of malformations. Among risk factors, only the doses of methylphenobarbital for mothers of infants with malformations were significantly higher than those for mothers of infants without malformations. Statistical differences were seen in drug score, number of AEDs, maternal age at delivery, seizure type, and etiology of epilepsy between the two groups. When data were corrected for seizure type, maternal age at delivery, or etiology of epilepsy, the difference in the incidence of malformations did not disappear, but it did disappear when data were corrected for drug score or number of AEDs. These results support our previous observations that AEDs are primary factors for the increased incidence of congenital malformation in infants of mothers with epilepsy.