The well-known teratogenicity of several anticonvulsant medications is associated with an elevated level of oxidative metabolites that are normally eliminated by the enzyme epoxide hydrolase. In this study, we attempted to determine whether infants who are at risk for congenital malformations could be identified prenatally by the measurement of epoxide hydrolase activity. Before fetuses at risk could be identified, it was necessary to measure epoxide hydrolase activity in a randomly selected sample of amniocytes from 100 pregnant women. According to a thin-layer chromatographic assay, the randomly selected sample population had an apparently trimodal distribution, suggestive of an enzyme regulated by a single gene with two allelic forms. Fetuses homozygous for the recessive allele would have low epoxide hydrolase activity and would therefore be at risk if exposed to anticonvulsant drugs during gestation. In a prospective study of 19 pregnancies monitored by amniocentesis, an adverse outcome was predicted for four fetuses on the basis of low enzyme activity (less than 30 percent of the standard). In all four cases, the mother was receiving phenytoin monotherapy, and after birth the infants had clinical findings compatible with the fetal hydantoin syndrome. The 15 fetuses with enzyme activity above 30 percent of the standard were not considered to be at risk, and all 15 neonates lacked any characteristic features of the fetal hydantoin syndrome. These preliminary results suggest that this enzymatic biomarker may prove useful in determining which infants are at increased risk for congenital malformations induced by anticonvulsant drugs.