Earlier studies indicated that the prevalence of congenital anomalies is greater in infants of epileptic mothers treated with anticonvulsants than in infants of mothers without epilepsy. We carried out a study of women in the General Practice Research Database who delivered liveborn infants between January 1988 and March 1993 and who were exposed to an anticonvulsant drug during the first trimester of pregnancy, and women with epilepsy not treated with anticonvulsants during pregnancy. We matched two nonexposed women without epilepsy to each exposed woman for age at delivery, date of baby's birth, and general practice. Two hundred ninety-seven women treated for epilepsy had 10 liveborn infants with major anomalies (3.4%) compared with 6 of the 594 nonexposed women (1.0%, RR = 3.3, 95% CI 1.2-9.2). We conclude that the infants of women with epilepsy who are treated with an anticonvulsant during the first trimester of pregnancy have an increased risk of major congenital anomalies.