Pregnancy registries: strengths, weaknesses, and bias interpretation of pregnancy registry data.

The chronic nature of epilepsy treatment, as well as the peak in incidence in children and adolescents, makes it likely that women of childbearing age will be exposed to antiepileptic drugs (AEDs). This has led to the establishment of several pregnancy registries to monitor the safety of anticonvulsants in terms of the risk of major birth defects following in utero exposure. Many registries are now approaching their second decade of data collection. With the release of substantial information on a range of AEDs (both older and newer generation drugs), it becomes increasingly important to understand how the methodological variations between cohorts may affect the risks of major birth defects being reported. This chapter explores the key methodological elements of these registries in terms of patient enrolment, the source of exposure and outcome information, outcome definition and ascertainment and comparator groups and how these elements may influence the reported risk estimates. With multiple registries collecting and releasing substantial data on AED safety in pregnancy, the chapter also reflects on how an increasing number of "signals" around increased risk of specific birth defects is likely. Possible methods for interpreting such signals, that the registries were not originally designed to detect and interpret, are also discussed.