A behavioral health outpatient program recently decided that psychiatrists would no longer be permitted to provide new prescriptions for antidepressants to pregnant women because the sponsoring institution had determined that the risk was too high. As a psychiatrist specializing in the treatment of women in the perinatal period, I am one of many practitioners who must frequently decide whether or not to use medication to treat depressed pregnant women.

Currently, the medical literature contains studies of more than 20,000 pregnancy outcomes for women exposed to antidepressants (1). This literature has been helpful in identifying possible effects of this exposure on the outcome of pregnancy. The results of studies that demonstrate increased risk for adverse pregnancy or birth events are frequently publicized in the media, which has resulted in many pregnant women refusing medication. Some law firms have seized on this information to seek plaintiffs to bring lawsuits. Many clinicians hear of positive studies through the media or see advertisements from law firms, scan the medical literature for abstracts, and may conclude that the risks of medication to the mother and fetus outweigh possible benefits. However, negative studies receive minimal media attention.

Clinicians who prescribe a treatment have a duty to be aware of all possible side effects of the treatment. Clinicians are also responsible for understanding and informing their patients of the risks and benefits of all treatment decisions, including the decision not to treat. There is considerable evidence that the decision not to prescribe antidepressants to a woman who is depressed or likely to have a recurrence during pregnancy may generate greater risks to the woman and her fetus than the risks of exposure to the medication (2–10). Research on both the effects of antidepressant exposure and the effects of treated and untreated depression on the outcome of pregnancy must continue. However, it is incumbent on those of us who conduct this research or specialize in these treatments, as well on the medical journals that publish our work, to make clear the meaning and limitations of our evidence-based research for other clinicians and the women they treat. Ultimately, our clinical judgment about the overall risks and benefits, informed by a critical and detailed reading of the medical literature, needs to be clearly communicated to everyone involved-patients, their families, and other physicians involved in the care of pregnant women.

Because of ethical concerns, gold standard prospective controlled clinical treatment trials do not include pregnant women. Most studies of the use of medications during pregnancy are retrospective and may rely on prescription databases, teratology services, birth registries, or population records of congenital defects. Little is known in these studies about the actual dosage and gestational timing of exposure, the use of other prescribed or over-the-counter drugs, or the medical or psychiatric condition of the mother.

Selected birth defects often differ from study to study, and findings are seldom replicated. A fundamental principle is that teratogenic exposures induce specific patterns of malformation and do not increase the incidences of all defects. Lack of consistency across these studies with respect to specific drugs and specific malformations makes it difficult for physicians to translate the findings into clinical practice. If antidepressants are teratogenic, then we would expect to see similar findings for similar drug exposures and specific defects in all studies (11). But such patterns are rare in the current literature.

While many studies have contradictory outcomes, emphasis is typically placed on those with positive findings while negative studies may be disregarded. Occhiogrosso et al. (12), in a review in this issue of the Journal, describe the strengths and weaknesses of six studies on the association between prenatal exposure to selective serotonin reuptake inhibitors (SSRIs) and persistent pulmonary hypertension of the newborn (PPHN). Three studies found no association between PPHN and SSRIs. Of the three studies that found associations, two of the studies examined the same population database, incorporating additional recent births. Examined together, the six studies identified a total of 50 SSRI-exposed infants with PPHN among approximately 25,000 who had SSRI exposure, which was fewer than the estimated number in the general population.

Findings must be interpreted in the context of the existing 3% baseline rate of birth defects (13). With reference to risk estimates, specific defects are rare and absolute risks are usually small (14). For example, the baseline prevalence for right ventricular outflow obstruction defects is estimated to be 5.5 cases per 10,000 births; thus, if a specific SSRI increases the risk by a factor of 2, the risk of having an affected child would be only 0.1%. Studies should report absolute risks and interpret them in some meaningful way that clinicians can understand and communicate to their patients.

While data sets may provide information on prescriptions, it is not clear whether the medication was actually taken by the mother. Petersen et al. (15) observed that pregnancy was a major determinant in the cessation of antidepressant medication, and most women do not receive further antidepressant prescriptions beyond 6 weeks of gestation.

Additional problems include insufficient power and inability to separate SSRI use from the effects of antenatal depression. In a recent population-based case-controlled trial that received widespread media attention (16), the authors suggested a higher rate of autism spectrum disorder in the children of 20 mothers who were treated with SSRIs in the year before and during pregnancy. In addition to having a small number of case subjects, the study did not control for psychiatric illness in the mother at the time of the study, although the association between psychiatric disorders and autism spectrum disorder is well documented (17). Autism spectrum disorder is a heritable disease, but family genetic loading was not reported. In addition, no data were reported on underlying medical illnesses, use of illicit substances, use of over-the-counter or prescribed medications, smoking history, or alcohol use.

Studies are often limited by analyses that are unadjusted for confounders such as alcoholism or smoking. For example, another recent article (18) suggested that taking fluoxetine and paroxetine during pregnancy increases the risks for cardiac defects and that taking citalopram during pregnancy increases the risk for spina bifida in exposed infants relative to unexposed referent offspring. The authors also found that mothers who took SSRIs during pregnancy had a 10-fold increase in the rate of fetal alcohol syndrome, which indicates substantial maternal alcohol ingestion, a significant confounder. Such confounders are an important methodological concern in birth defect research and must be taken into account for the proper interpretation of findings (19).

Depression and anxiety also have adverse effects on the maternal-fetal environment and on infant outcome. The depressed pregnant woman is more likely to use nicotine, illicit drugs, and alcohol, and she is less likely to attend to her prenatal care (2). Antepartum depression is associated with preeclampsia, low birth weight and gestational age, and premature birth (3–5). Adverse neurobehavioral effects include toddler developmental delay and behavioral problems (6, 7). Antepartum depression is also the strongest predictor of postpartum depression (8), which in turn can have additive adverse effects on the infant's temperament, mother-infant interaction, and later infant cognitive ability (9, 10). A recent analysis from the Centers for Disease Control and Prevention's surveillance system from 17 states over 4 years found 94 pregnancy-related suicides and 139 homicides from 2003 to 2007 (20). One-half of these violent deaths were associated with intimate partner violence and psychiatric illness, which are common in pregnancy-associated suicide (21).

We saw an inappropriate decline in the prescription of antidepressants for adolescents and young adults because of a Food and Drug Administration black box warning about suicidality. However, untreated depression itself has a far greater impact on suicidal behavior than the adverse effects of the antidepressants. Psychiatrists have generally used appropriate judgment in prescribing antidepressants and psychotherapies for this patient population, but other physicians did not (22). We now need to communicate clearly about the risks and benefits of treatment for depressed pregnant women if we wish to avoid similar adverse consequences to their well-being and the well-being of their infants.