Table 1.

Reproductive-outcome information sources

Study typea	Description
Single-case report	Women with adverse pregnancy outcomes may become subjects of case reports. Positive case reports are signals that require further study. The total number of exposures associated with the adverse outcome is unavailable; therefore, the strength of the association is unknown. Publication of case reports creates concern and avoidance of the implicated drug, which make it difficult to collect data to define the true risk.
Case series	Case series reports provide information about several women with similar patterns of adverse pregnancy outcomes. The signal is stronger, because rare exposures and rare defects are not likely to occur by chance in a small sample. This mechanism was used to identify the association of birth defects, especially limb-reduction anomalies, with thalidomide in the 1960s.
Registries	Registries require large numbers of patients, and often years, to yield information. They must include women whose exposures were documented before knowledge of pregnancy outcomes in order to provide accurate information about the ratio of adverse events to the number of exposures. Registries allow generation of a meaningful rate of adverse outcomes that can be compared with background adverse-event rates. Registries are useful for monitoring—for example, to rapidly identify or exclude a drug as a major teratogen.
Retrospective, case-control	Sample populations of patients with and without an outcome of interest (e.g., congenital cardiovascular defects) are compared for exposure to an agent suspected of being associated with the outcome (e.g., paroxetine). The exposure occurrence is derived from interviews or records. Data from cases and controls may be derived from existing population data, such as registries. These studies yield an odds ratio.
Prospective, cohort	Prospective observational studies are focused on the study of samples of exposed and nonexposed women followed prospectively to compare the rate of the outcome(s). Ideally, women not exposed (to antidepressants) are assessed for the level of activity of the underlying disease condition (depression). These studies generally require large samples and lengthy observational periods, and they are difficult to implement for rare events, such as birth defects. They yield relative risk.
Meta-analysis	A meta-analysis combines data from studies of similar design based on predetermined inclusion/exclusion rules to create a summary effect size from the resulting larger sample.

^aThese are observational studies, rather than experimental studies, because the investigator cannot specify exposure conditions, i.e., assign pregnant women to be exposed vs. not exposed to antidepressant medication.