To determine the primary factors responsible for the increased incidence of malformation in the off-spring of antiepileptic drug (AED)-treated epileptic mothers, special attention was paid to drug combinations in a prospective study of 172 deliveries. Variables used for analysis were eight antiepileptic drugs (AEDs) and total daily dosages (drug score), and seven background factors consisting of maternal age at delivery, gravida, outcome of previous pregnancy, etiology and type of epilepsy, occurrence of seizures in the first trimester of pregnancy, and seizure frequency during pregnancy. The overall rate of malformation was 14.0%. Thirty-one patients were administered a single drug, and the rate of malformation was 6.5%. The remaining 141 patients were treated with multiple AEDs, and the rate of malformation was 15.6%. The drug score of the latter group was significantly higher than the former (p = 0.01). There was no definite dose-dependent increase in the incidence of malformations associated with any individual AEDs. There was no relationship between the type of defect and individual AEDs. Wilcoxon rank-sum test revealed significant association between the drug score, valproate (VPA), and congenital malformation. Carbamazepine (CBZ) also reached an almost significant level. Furthermore, VPA polypharmacy produced the highest incidence of malformation, higher than that produced by any other AED or drug combination. There was no significant association between the presence of malformations and the other putative risk factors. These results suggest that high dose of AEDs reflecting polypharmacy, VPA polypharmacy in particular, are primary factors responsible for the increased incidence of congenital malformation in the offspring of treated epileptic mothers.