In addition to studies in animals, a variety of other approaches are used to identify possible drug teratogenicity and to assess the relation between drug exposure and fetal outcome. The first accounts of adverse fetal outcomes after exposure to a marketed drug are usually published in the form of case reports. These reports can be either very useful or useless in establishing teratogenic risk on the basis of relatively simple statistical considerations. If the drug in question is taken by relatively small numbers of women (e.g., isotretinoin) or causes a rare malformation (e.g., ear agenesis), then a small number of cases can establish a strong association. Warfarin, diethylstilbestrol, and isotretinoin were originally identified as human teratogens on the basis of case reports. If, on the other hand, the drug is taken by many pregnant women (e.g., Bendectin), a small number of case reports of abnormalities may simply reflect the spontaneous occurrence of malformations in the general population, which ranges from 1 to 5 percent, unless there is a characteristic pattern of malformations (as, for example, with alcohol or thalidomide). To date, prenatal exposure to many of the known human teratogens has been associated with characteristic patterns of malformations, and this has become an important tenet in establishing teratogenicity.
Epidemiologic studies are typically designed to determine whether mothers who took a specific drug during pregnancy have a larger number of malformed children than mothers who did not (cohort studies) or whether mothers of children with a specific malformation took the drug more often than mothers of children without the malformation (case–control studies).
With the international development of teratology-information services, a new source of data for prospective observational research has emerged. Pregnant women taking prescription or over-the-counter drugs voluntarily call these centers for risk-assessment counseling, usually during the first trimester. Since the exposure data are recorded prospectively, the probability of recall bias is reduced, and follow-up of exposed pregnancies can extend well beyond parturition. Collaboration among these services can yield the large samples needed to study rare events more effectively.
Drug manufacturers may perform postmarketing cohort studies of prospectively reported exposures. Such studies were useful in establishing the safety and risk of Bendectin, isotretinoin, fluoxetine, and acyclovir.
Because most studies of teratogenic risk are limited in size, meta-analyses of studies of similar design are becoming more frequent. A detailed, stepwise methodologic approach to meta-analysis of teratologic studies has been described. The appropriate use of this approach depends to a large extent on establishing sound a priori criteria for methodologic quality and ensuring the inclusion of data from all available studies, in order to obviate any publication bias against negative results.
Long-term studies are increasingly important, because it is becoming clear that the long-term effects of teratogenic drugs on neurobehavioral development can have a more devastating effect on children and their families than structural anomalies. To date, several drugs have been shown to affect brain development, including carbamazepine, isotretinoin, phenytoin, valproic acid, and warfarin. Carbamazepine and valproic acid may cause cognitive brain dysfunction as part of the neural-tube defects they induce. Originally, isotretinoin was found to cause structural abnormalities that affected brain development, but recent studies have suggested that even phenotypically normal children may have abnormal neurodevelopment. Warfarin was initially associated with chondrodysplasia punctata and mental retardation and has subsequently been found to cause the Dandy–Walker brain malformation in an estimated 1 to 2 percent of exposed fetuses.