April 24, 2014
The last decade has brought much attention to the risks of fetal exposure to antidepressant medications, with numerous studies describing the risks of adverse outcomes that range from congenital malformations, persistent pulmonary hypertension of the newborn (PPHN), and neonatal adaptation syndrome, to preterm birth and spontaneous abortion. It has been challenging for clinicians to make clinical sense out of these studies, which vary in their methodologic rigor, and to determine how to fit these data into the treatment algorithm for managing depression during pregnancy.
To date, rigorous studies evaluating whether treatment with antidepressants during pregnancy increases the risk of spontaneous abortion have been relatively sparse. The available data, largely from small cohort studies, have suggested that selective serotonin reuptake inhibitor (SSRI) exposure during pregnancy is associated with an increased risk of spontaneous abortion relative to unexposed pregnancies – but the increases have been relatively modest and rates typically have not exceeded estimated rates in the general population. Cohort studies are often underpowered with a small sample of women, hindering the ability to show a statistical difference for an uncommon outcome like spontaneous abortion. And even if there is a statistical difference, a critical issue to consider is whether that difference is clinically meaningful.
Studies using large administrative databases both in the United States and internationally have provided the capacity to look at exposure to psychiatric and other medications during pregnancy in large numbers of women. But many have not adequately addressed the problem of “confounding by indication,” because they compare women taking the antidepressant to untreated women who do not have the underlying disorder for which the medication is used. This is critical because depression may be independently associated with some of the adverse obstetrical and neonatal outcomes that have been ascribed to exposure to antidepressants (Arch. Gen. Psychiatr. 2010;67:1012-24).
But a population-based study published last year using data from the Danish birth, hospital, and psychiatric registries from February 1997 to December 2008 addressed the confounding issue in more than 1 million pregnancies and represents an incremental contribution to the literature on this topic (PLoS One 2013;8:e72095 [doi: 10.1371/journal.pone.0072095]).
Of about 1,005,000 pregnancies, the rate of spontaneous abortion (before 22 weeks’ gestation) was 11%. The risk of spontaneous abortion among pregnancies in women with depression treated with an antidepressant in early pregnancy (22,061 pregnancies) was compared with the risk among pregnancies in women with depression who did not report use of antidepressants (1,843 pregnancies) and in women with no diagnosis of depression and no antidepressant use (981,415). The antidepressants included SSRIs, serotonin-norepinephrine reuptake inhibitors, and tricyclics, and the risk ratios were adjusted for maternal age, education, and other factors.
The risk of spontaneous abortion associated with the use of antidepressants was slightly higher compared with no antidepressant use, with a relative risk of 1.14. However, when the analysis was restricted to women diagnosed with depressive disorder, as documented in the psychiatric registry (which includes information on all admissions to psychiatric hospitals and outpatient treatments in Denmark), the risk associated with antidepressant use dropped to 1.0.
Essentially, after adjustment for the depression diagnosis, the risk moves toward the null, and spontaneous abortion rates are comparable between unexposed and exposed pregnancies. This has never been shown in such a large study, and is a critical finding because this represents an effort in a very large dataset to parse out the relative effects of exposure to the disease versus exposure to the medicine.
When considered in the context of previously available data, where do these results leave the clinician and their patients? Women with depression often ask about the risks of preterm birth and spontaneous abortion associated with antidepressants. This concern is particularly relevant for women with depression undergoing assisted reproductive technology procedures for infertility.
Patients and clinicians can be somewhat reassured by the findings of this study, which indicate that even if there is a risk of increased spontaneous abortion, it is a very small one, and it appears to be even smaller when factoring in the depression diagnosis.
There is no perfect decision with respect to the challenging question of whether to use an antidepressant in women with significant psychological distress, including major affective disorder. Women with the same illness history presented with the same information may make different decisions about psychiatric medication use during pregnancy, and some women may choose not to use an antidepressant.
Finally, this study highlights the importance of factoring in the effect of the underlying illness for which the patient is being treated in clinical studies that examine the fetal effects of exposure to psychiatric medications, as well as for other classes of medications for other nonpsychiatric illnesses. In large datasets, it is challenging to adjust for variables that are relevant to the outcome of interest, and this could explain some of the disparate findings of studies regarding spontaneous abortion and other obstetrical and neonatal outcomes.
When we see large studies with large numbers of patients, while exciting, we need to be more critical in interpreting the results, paying attention to the comparison groups and whether the study adjusts for the variables that can be very relevant to the outcome of interest.
Dr. Cohen is the director of the Center for Women’s Mental Health at Massachusetts General Hospital in Boston, which provides information about reproductive mental health at www.womensmentalhealth.org. He has been a consultant to manufacturers of antidepressant medications, and the Danish study was funded with one of the author’s post doc grants from the Danish Council for Independent Research Companies. To comment, e-mail him at email@example.com.