To the Editor: Dr. Urato is concerned that the group of pregnant women treated continuously with SSRIs was very different from the group of continuously depressed pregnant women and that we may not have adequately controlled for those group differences. Dr. Urato also states that many of these differences are risk factors for preterm birth and must be controlled for. The strategy for distributing confounding variables equally between groups is randomization. In an observational study, controlling for variables known to be associated with the outcome is an approach to address (but not resolve) this problem. We tested the variables identified by Dr. Urato in a multivariable model. Only maternal age significantly predicted shorter gestational length, and we controlled for race. We did not adjust for prior preterm birth in our model for SSRIs and current preterm birth because doing so may induce bias into the association between SSRI exposure and preterm birth in the index pregnancy (1). Conducting randomized clinical drug trials during pregnancy, which is generally considered unethical, is presently being weighed against the difficulty in interpreting and clinically applying data from observational studies as a result of this fundamental problem (2).
Dr. Urato’s comment that “these results are dramatic and join the accumulating evidence that now links SSRI use to preterm birth” does not acknowledge the ongoing debate about the effect of major depression on reproductive outcomes and is not a balanced interpretation of our data or that of other investigators. For example, Oberlander et al. (3) examined the effect of nonrandom maternal characteristics associated with SSRI treatment in pregnant depressed women. With a population health database, they used propensity score matching to control for characteristics of depression that could confound the examination of reproductive outcomes between SSRI-treated and unmedicated depressed women (such as maternal illness severity). Offspring exposed to SSRI in utero (compared with nonexposed offspring) had an increased risk for jaundice, feeding problems, lower birthweight, lower gestational age, higher rate of preterm birth, and respiratory distress. After propensity score matching, only an increased incidence of birth weight below the tenth percentile (not significant in the original comparison) and respiratory distress were significant (not preterm birth). Characteristics of the underlying disorder that are associated with being treated with antidepressants versus not being treated with antidepressants are important contributors to reproductive outcome.
In clinical practice, data from populations must be individualized for each pregnant woman in a risk-benefit discussion that includes her unique clinical history and personal values in addition to information about both gestational drug exposure and disease exposures.
Weinberg C: Toward a clearer definition of confounding. Am J Epidemiol 1993; 137:1–8
Lyerly AD, Little MO, Faden R: The second wave: toward responsible inclusion of pregnant women in research. Internat J Fem Appr Bioethics 2008; 1:6–22
Oberlander TF, Warburton W, Misri S, Aghajanian J, Hertzman C: Neonatal outcomes after prenatal exposure to selective serotonin reuptake inhibitor antidepressants and maternal depression using population-based linked health data. Arch Gen Psychiatry 2006; 63:898–906
The author’s disclosures accompany the original article.
This letter (doi: 10.1176/appi.ajp.2009.09050712r) was accepted for publication in June 2009.