Should we treat women with depression during pregnancy with antidepressants? This ostensibly simple question is presently without answer for a variety of methodological, conceptual, and clinical reasons, many of which are explicitly stated and implicitly demonstrated in the article by Misri and colleagues in this issue of the Journal and in several recent articles published in other journals (1, 3).
Misri and colleagues report the effects of prenatal exposure to selective serotonin reuptake inhibitors (SSRIs) on internalizing behaviors (emotionally reactive, anxious/depressed, withdrawn, somatic complaints) in 4- to 5-year-old children who have been followed since birth. These authors’ earlier studies of the children exposed to SSRIs in utero found evidence of transient neonatal maladaptation and decreased affect expressivity, which raised the possibility that the in utero exposure to drugs had a longer-term toxic developmental effect on these children. However, by ages 4 to 5, at the time of the study in this issue of the Journal, no differences in internalizing behaviors were detected between children with in utero exposure to SSRIs and comparison children. Notably, maternal anxiety at the time of testing was correlated with the total internalizing behavior score.
The study by Misri and colleagues has many strong points. Its longitudinal design permits the assessment of the same individuals over time and demonstrates, as a cross-sectional design cannot, that behavioral abnormalities detected in the first year of life do not necessarily translate into affective internalizing symptoms later in childhood. The longitudinal assessment of maternal symptoms, an obvious but often missing covariate, is also worth applauding and generated the one predictor of internalizing behavior in this study: maternal anxiety.
Unfortunately, limitations of this study preclude any premature celebration of the findings. The sample size (22 exposed dyads and 14 comparison dyads) is very small. While the design is longitudinal, the reported component of the larger study is essentially cross-sectional, and the very low sample sizes are not powered to detect anything except very large differences between the two groups. This is particularly troubling given the unexplained dropout of 24 of the exposed mother-infant dyads and nine of the comparison dyads. The association between maternal anxiety and internalizing behavior, while conceptually appealing, was observed with parental ratings only, suggesting that the ratings may have reflected the maternal affective state rather than behavioral characteristics of the children. Additionally, had the correlations been corrected for multiple comparisons, none would have achieved statistical significance. Finally, a 3-point behavioral rating scale does not allow for more subtle gradations of behavior that may, nonetheless, be clinically relevant. Limitations notwithstanding, the Misri and colleagues study could be good news, perhaps reassuring us that neonatal irritability reported following prenatal SSRI exposure does not necessarily translate into persistent adverse behavioral effects.
More worrisome conclusions are found in an article by Chambers and colleagues, published in the New England Journal of Medicine(1), which describes a sixfold greater likelihood of the occurrence of persistent pulmonary hypertension of the newborn in infants exposed to an SSRI after the 20th week of gestation. Although most earlier reports (2) suggest that SSRIs are not associated with greater risk of congenital malformation, the report by Chambers and colleagues raises anew the concern that fetal development is adversely affected by the presence of SSRI antidepressants. Further, although clinical practice recommends the avoidance of medication during the first trimester in order to prevent teratogenesis, the Chambers and colleagues study suggests that it is late, not early, exposure to SSRIs that increases the risk of pulmonary hypertension of the newborn.
Avoidance of SSRIs during pregnancy is not without risks. In an article published in the Journal of the American Medical Association, Cohen and colleagues (3) point out that there is an adjusted fivefold increase in the risk of depressive relapse in women with recurrent major depression who discontinue antidepressant medication proximate to pregnancy, compared with those who maintain their medications. Given reports ascribing both postnatal complications and infant behavioral abnormalities to maternal depression (4, 5), it may be said that exposure to depression is as potentially toxic as exposure to medications. Indeed, the finding by Misri and colleagues—that maternal anxiety is related to a child’s internalizing behavior—is consistent with other articles in this issue of the Journal, reviewed in the editorial by Wamboldt and Reiss, which point to the long-term impact of maternal illness on parenting behavior and its devastating effect on the child. What should we conclude from these studies, and how should we treat our patients?
Our therapeutic confusion derives in part from the requirement to calculate risk profiles for two individuals (mother and infant), involving multiple predictors and outcomes. Even when drug exposure is unequivocally associated with risk of an adverse outcome (e.g., pulmonary hypertension of the newborn), the majority of exposed infants do not develop such a condition, and we do not understand the source of this differential susceptibility. Further, there is no calculus for evaluating and balancing risks, a problem compounded by the confusion of relative risk with absolute risk. Is a fivefold increase of maternal depressive relapse worth the price of avoiding a sixfold risk of development of neonatal pulmonary hypertension of the newborn? Does the answer to this question change if we consider the absolute risks: a 68% likelihood of depressive relapse in untreated women with recurrent depression versus a sixfold decrease in the risk of a condition that occurs in 1/1000 infants?
Several factors contribute to the absence of clear guidelines for the treatment of depression (or a history of depression) during pregnancy. First, the risks of antidepressant treatment during pregnancy cannot be meaningfully determined without a comparison group of depressed women who are untreated during pregnancy. Although, as mentioned above, several studies suggest that untreated depression during pregnancy is associated with obstetrical complications and infant behavioral abnormalities, studies comparing neonatal outcomes in treated and untreated depressed women during pregnancy are virtually absent. Second, clinicians often experience the therapeutic algorithm for treating pregnant women with an affective disorder history as a thousand Ns of one. Prenatally or during pregnancy, the clinician must determine whether the patient is depressed or euthymic, on or off medication, unipolar or bipolar, first episode or recurrence, nulliparous or multiparous, with or without a past history of postpartum depression, intending to breast-feed or not, etc. No two patients are exactly alike, and the absence of predictive data from large sample studies leaves the clinician and the mother uncertain about the best strategy. Third, women who experience depression during pregnancy and postpartum often suffer a double dose of cultural neglect. Despite overwhelming data to the contrary, depression is still regarded as a personal or moral failing rather than a serious, potentially fatal illness. Depressive disorders that are confined to women are disparaged all the more, demonstrated in the response by some to Brooke Shields’ experience with postpartum depression. How can one otherwise explain the conspicuously small number of studies on postpartum depression despite its occurrence in 10% of women and its recent identification in the United Kingdom as the leading cause of maternal death—more than hemorrhage or sepsis (6)?
Notwithstanding the absence of predictors, several therapeutic principles can guide the clinician treating women during pregnancy. 1) As demonstrated by Cohen and colleagues, pregnancy does not protect against the occurrence of depression, and the likelihood of relapse is very high in untreated women with recurrent illness. 2) As Misri and colleagues—as well as others in this issue—report, maternal depression adversely affects child development, and prenatal depression may adversely affect the offspring. 3) When attempting to balance risks, transient postnatal behavioral abnormalities in the offspring of treated mothers must not be assumed to portend long-term compromise, particularly if the cost of attempting to avoid the abnormalities is worsening of maternal depression. 4) SSRIs, the most commonly used and best-tolerated treatment for depression, carry a small but significant risk for a serious medical consequence. The decision to use SSRIs must take the severity of the mother’s illness and the risk to the fetus into account. Practical help for one such complication, decreased maternal weight gain, is discussed in this issue of the Journal in the clinical case conference by Bodnar and colleagues. The stark reality is that pregnancy in the context of a history of affective illness often leaves the mother, the baby, and their doctors between Scylla and Charybdis.