In medicine, there are often situations that require patients and their providers to make difficult management decisions. We turn to systematic evidence to guide our approach and counsel our patients, but many times the information is limited or inadequate. The treatment of women with depression who are either pregnant or breastfeeding presents a number of issues for which we have insufficient data. These include questions such as, What is the relative likelihood of becoming depressed and requiring treatment during pregnancy or the several months after delivery? What are the short- and long-term consequences for children exposed to maternal psychiatric illness or to pharmacological treatment? What are the short- and long-term consequences for a neonate exposed to an antidepressant as a result of breastfeeding? The absence of sufficient data is not the result of a lack of interest among researchers but derives largely from the ethical and practical issues that make research in this area difficult. For example, it is neither ethical nor practical to randomly assign depressed, pregnant women to antidepressant agents versus placebo in order to evaluate their clinical outcomes and the sequelae for their offspring. Rather, we must turn to observational, cohort studies and reports that rely on administrative databases. An issue with the latter source of information is that it can be difficult to disentangle confounding effects from main effects. In the case of exploring maternal and fetal outcomes among women who are depressed, pregnant, and undergoing pharmacological treatment, it is not always possible to determine the role of the underlying illness (depressive symptoms or the biology of depression), unhealthy behaviors (e.g., hazardous substance use), and antidepressant exposure. Bad news seems to make a better story than good news, and thus we tend to hear more about the problems associated with antidepressants than about the reassuring news that a compound is not a major teratogen or is not likely to compromise behavioral outcomes over the long term.
The sensationalism that often accompanies stories about negative outcomes reinforces the importance of continuing to conduct rigorous new research and publish thoughtful syntheses of the literature. The August (1) and September (2) issues of the Journal contained reviews and treatment recommendations regarding clinical management of either pregnant or lactating women with depressive disorders. In the current issue, Dietz and colleagues provide information regarding how frequently a woman is likely to be diagnosed and treated for a depressive disorder before, during, and after pregnancy.
The article by Freeman from the August issue (1) reviews several studies that explored whether specific fetal anomalies are associated with in utero exposure to selective serotonin reuptake inhibitors (SSRIs), either as a class or as individual agents. Generally, there has been less support for a “class effect” for the SSRIs, while several administrative databases (3, 4) and teratogen information services (5) suggest an association between paroxetine and cardiac malformations. However, the weight of evidence is a moving target since two large case-cohort studies published in June (6, 7) did not replicate this finding but, instead, found additional possible associations between other anomalies and paroxetine as well as other antidepressants. The mixed findings are likely a result of attempts to identify associations between relatively infrequent antidepressant exposure and rare defects, which require large databases that may have residual confounders. While there may be risks when antidepressants are used in pregnancy, these effects are not large effects, as would be expected for a major teratogen (8).
When one turns from the effects of in utero exposure to the consequences of neonatal exposure through breast milk, there is far less information and the most data are derived from case reports and small cohort studies, as outlined by Payne in the September issue of the Journal(2). The take-home messages are that breastfeeding has many benefits, the amount of medication in breast milk varies according to when the drug is taken and what part of breast milk is assayed, but usually maternal use does not lead to substantial levels in the neonate. However, it is best that neonates be monitored for difficulty feeding, weight gain, sleep or state changes, etc., if the mother is undergoing antidepressant treatment while breastfeeding.
While it is important to have guidelines for managing depression in pregnancy (9), we also need to know if pregnancy decreases and the postpartum period increases the risk of depression. It is commonly thought that depression increases during the postpartum time, but many systematic studies do not support this (10). However, most cohort studies have had rather modest sample sizes, and there is other research showing that treatment visits and psychiatric hospitalizations increase after, compared to before, delivery (11). This finding was replicated by Dietz et al. in the current issue of the Journal. The authors used the database of a large health maintenance organization and assessed rates of any depressive disorder 39 weeks before, during, and after pregnancy. The rates for recorded depressive diagnoses were 8.7%, 6.9%, and 10.4%, respectively, for those time intervals. Diagnoses were made significantly more frequently during the 39 weeks postdelivery than during pregnancy, although over 50% of the women who were diagnosed with a depressive disorder after delivery were also diagnosed with a depressive disorder before delivery. Whether the higher rate of depressive disorders postpartum is a result of patients’ willingness to disclose depressive symptoms after delivery, clinicians’ increased vigilance in diagnosing depressive disorders postnatally, or a true difference in the incidence of depression is not clear.
The challenge for answering a number of questions regarding the incidence of pregnancy-related mood disorders and the risks and benefits associated with their treatment is a classic trade-off between power and precision. Given the ethical and practical constraints, large cohorts are needed to address some of these questions. Administrative databases provide the largest sample sizes and the most power. But the cost of this is a lack of precision regarding information about possible confounders, such as health habits and illness characteristics. Such is the issue in the study by Dietz et al. and the previous study by Munk-Olsen et al. (11). But we must use the information that we have and hope for increased opportunities to enhance that database with future, rigorously controlled studies.