The FDA reviewed the literature on the subject and issued a notification in December 2011, stating, “given the conflicting results from different studies, it is premature to reach any conclusion about a possible link between SSRI use in pregnancy and PPHN.”9 The agency advised health care professionals “not to alter their current clinical practice of treating depression during pregnancy.” Health care professionals were also advised to “weigh the small potential risk of PPHN that may be associated with SSRI use in pregnancy against the substantial risks associated with under-treatment or no treatment of depression during pregnancy.”
These studies offer an excellent opportunity to understand which study designs are best-suited to evaluate rare outcomes. If an outcome occurs rarely (eg, PPHN), an extremely large number of patients with and without the risk factor (eg, treatment of the pregnant woman with an antidepressant) would have to be followed to be able to evaluate whether patients who did or did not have the risk factor were equally at risk for the outcome.
Early in an investigation of a potential association, it is not viable to study very large numbers of subjects. Instead, a case-control study is often used. Case-control studies are well suited to initial investigation of rare outcomes and outcomes that take a long time to develop (eg, lung cancer with smoking). The researchers start with a sufficient number of subjects who have the outcome of interest, identify an appropriate control group without the outcome, and then assess the proportion of subjects with or without the outcome who were exposed to the risk factor being studied.
Because case-control studies can have certain problems, such as recall bias (subjects may be more likely to remember exposure to the risk factor than the controls), their findings usually need to be confirmed by large cohort studies. In cohort studies, subjects with or without the risk factor are identified and followed up over time to determine whether they develop the outcome of interest.
When studying potential risk factors, confounders need to be taken into account. The confounder is a third factor that is associated with both the putative risk factor and the outcome and, therefore, results in an apparent association between the putative risk factor and the outcome.
Factors other than the putative risk factor being evaluated that are associated with an outcome (ie, potential confounders) can be controlled for in several ways. First, inclusion criteria can limit the subjects to a more homogeneous subgroup (eg, the studies cited in this article usually only included infants born after 33 or 34 weeks of gestation). Second, in case-control studies, the controls can be matched to cases on specific factors. To illustrate, Chambers and colleagues1 matched the controls to the cases on the hospital of birth and the approximate date of birth. Third, the potential confounders can be controlled for in the statistical analysis.
