Abstract Introduction Preterm birth is a major contributor to neonatal morbidity and mortality and its rate has been increasing over the past two decades. Antidepressant medication use during pregnancy has also been rising, with rates up to 7.5% in the US. The objective was to systematically review the literature to determine the strength of the available evidence relating to a possible association between antidepressant use during pregnancy and preterm birth. Methods We conducted a computerized search in PUBMED, MEDLINE and PsycINFO through September 2012, supplemented with a manual search of reference lists, to identify original published research on preterm birth rates in women taking antidepressants during pregnancy. Data were independently extracted by two reviewers, and absolute and relative risks abstracted or calculated. Our a priori design was to group studies by level of confounding adjustment and by timing of antidepressant use during pregnancy; we used random-effects models to calculate summary measures of effect. Results Forty-one studies met inclusion criteria. Pooled adjusted odds ratios (95% CI) were 1.53 (1.40–1.66) for antidepressant use at any time and 1.96 (1.62–2.38) for 3rd trimester use. Controlling for a diagnosis of depression did not eliminate the effect. There was no increased risk [1.16 (0.92–1.45)] in studies that identified patients based on 1st trimester exposure. Sensitivity analyses demonstrated unmeasured confounding would have to be strong to account for the observed association. Discussion Published evidence is consistent with an increased risk of preterm birth in women taking antidepressants during the 2nd and 3rd trimesters, although the possibility of residual confounding cannot be completely ruled out.

Citation: Huybrechts KF, Sanghani RS, Avorn J, Urato AC (2014) Preterm Birth and Antidepressant Medication Use during Pregnancy: A Systematic Review and Meta-Analysis. PLoS ONE 9(3): e92778. https://doi.org/10.1371/journal.pone.0092778 Editor: Shannon M. Hawkins, Baylor College of Medicine, United States of America Received: October 22, 2013; Accepted: February 25, 2014; Published: March 26, 2014 Copyright: © 2014 Huybrechts et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: There are no current funding sources for this study. Krista Huybrechts is supported by a career development grant K01MH099141 from the National Institute of Mental Health. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing interests: The authors have declared that no competing interests exist.

Introduction Preterm birth is a major clinical problem throughout the world. It is the leading cause of infant mortality: approximately 75% of all perinatal deaths occur among preterm infants [1]. It is also a major contributor to both short- and long-term morbidity: surviving infants are at increased risk of health problems ranging from neurodevelopmental disabilities such as cerebral palsy and mental retardation to other chronic health problems such as asthma [2]. Although the risk is highest in very preterm infants (<32 gestational weeks), it has been well documented that moderate (32–33 gestational weeks) and mild (34 to 36 gestational weeks) preterm birth infants are also at increased risk for neonatal and post-neonatal mortality and morbidity [3]–[6]. Rates of preterm birth have been increasing over the past two decades and it is a major public health concern [7], with costs to society that have been estimated to be as high as $26.2 billion per year in the US [7], and £939 million per year in the UK [8]. It has been reported that two thirds of these costs are incurred for the care of babies born moderately prematurely [8]. In many developed countries, the use of antidepressant medications has increased sharply between 1996 and 2005, and now surpasses antihypertensives as the most commonly prescribed drug class in ambulatory care [9]. During this same time period, rates of antidepressant use during pregnancy have increased approximately 4-fold, with reported rates of up to 3–6% in Europe [10]–[12] and up to 8% in the US [13], [14]. Numerous studies, of varying size and quality, have examined the effects of antidepressant medication use on pregnancy outcomes, including preterm birth. They differ in terms of the timing of the antidepressant exposure during pregnancy and adjustment for potential confounding variables, including lifestyle factors, co-morbidities, and the severity of the underlying psychiatric illness. The extent to which such differences contribute to variability in findings remains to be elucidated. The objective of this review was to determine the strength of the available evidence relating to a possible association between antidepressant use during pregnancy and preterm birth, and to assess this relationship in terms of (1) the timing of the antidepressant use studied, and (2) attempts to control for the possible confounding effects of depression itself.

Methods To identify all available studies on the topic of antidepressant medication use during pregnancy and preterm delivery, we performed a computerized search in PUBMED, MEDLINE, and PsycINFO using the key words: (“antidepressant*” or “tricyclic antidepressant*” or “selective serotonin reuptake inhibitor*” or “serotonin-norepinephrine reuptake inhibitor*”) and (“preterm birth*” or “preterm deliver*” or “pregnanc*” or “pregnancy complication*”). The databases were searched from their inception through September 12, 2012. Reference lists of selected articles were also searched to identify additional studies that reported on preterm births and antidepressant exposure. Studies were included if they identified a group of pregnant women exposed to antidepressants at some point during their pregnancy as well as a comparison group, and reported on preterm birth rates, irrespective of whether preterm birth was a pre-specified study endpoint or one of several pregnancy characteristics reported. No restrictions were imposed on study size or design. Once relevant studies were identified, for each study two investigators with clinical and epidemiologic expertise (ACU, KFH) independently abstracted preterm birth rates in the group(s) exposed to antidepressants and in the comparator group(s). When these were not expressed as percentages in the manuscript, they were calculated. Relative risks (expressed as odds ratios in all studies) were also taken directly from the manuscript. Whenever available, preference was given to relative risks adjusted for potential confounding variables. When relative risks were not reported in the manuscript, we estimated the unadjusted odds ratio and its corresponding 95% confidence limits (Wald method) based on the available information. When relative risks were only presented graphically, we contacted the authors to obtain the corresponding numerical estimates. Types of antidepressants used, numbers of users and other pertinent exposure information (e.g., time and duration), and potential confounders accounted for were also retrieved. Any discrepancies in data abstraction were resolved by consensus between the reviewers. A funnel plot was examined for evidence of publication bias [15]. Between-study heterogeneity was examined using the Cochran Q and I2 tests [16]. We used a random-effects meta-analysis model to calculate summary measures of effect while accounting for heterogeneity across studies [17]. Because of the critical importance of (1) the timing of antidepressant use during pregnancy, and (2) the potential for confounding by the presence of depression itself (confounding by indication), we had determined a priori to group studies by timing of antidepressant exposure and level of confounding adjustment. We chose this approach to address the inherent issues of clinical and methodological diversity [18]. Finally, we conducted a sensitivity analysis to identify the strength of the residual confounding that would be necessary to fully explain the estimated association between antidepressant medication use and preterm birth [19].

Discussion This systematic review of the literature concerning the relationship between antidepressant use and preterm birth identified 41 observational studies. Findings suggest the risk of preterm birth is increased in women treated with antidepressant medications during pregnancy, with pooled odds ratio estimates ranging between 1.16 (95% CI 0.92–1.45) and 1.96 (95% CI 1.62–2.38). The associations were stronger for antidepressant use later in pregnancy. Adjusting for a diagnosis of depression in most cases did not eliminate the effect, although the strength of the observed associations was somewhat attenuated. Sensitivity analyses demonstrated that very strong risk factors of preterm birth that are fairly imbalanced among exposure groups and independent of the adjusted confounders must be unmeasured and uncontrolled to explain the observed associations. Although it would be unlikely to miss such a strong single confounder, it is conceivable that several weaker confounders could have acted together to account for the apparent effect. Our findings are consistent with those from an earlier study which evaluated the association between prenatal antidepressant exposure and a range of adverse pregnancy and delivery outcome [73]. Preterm birth is a major problem worldwide [74]. The rate of antidepressant use during pregnancy has steadily increased over time in many industrialized nations, from less than 1% of women exposed in the early 90s to 3–6% in 2006 in Europe [10], [12], and to 7.5% in 2008 in the US [13]. It is therefore essential to determine if antidepressant use increases the risk of preterm birth. Studies in this area, however, are complicated due to several issues. First, antidepressant exposure in many pregnancies is not a “yes or no” phenomenon. Many women stop antidepressants when they discover they are pregnant, resulting in first trimester exposure only. Others stop, but may restart the medications later in the pregnancy. Still other patients do not take the medications for much of the pregnancy but start in the third trimester. The issue of classifying exposure is therefore complex. Several studies, for example, have been performed by Teratogen Information Services (TIS). Women who called the TIS and reported that they were taking antidepressants were classified as “exposed” in those studies. The advantages of this design include the fact that exposure is determined prior to outcome, eliminating recall bias, and a control group is readily identifiable–women who contact the TIS with exposure to another medication. Yet, there are also major drawbacks to this approach. Most women who contact TIS are concerned about the use of antidepressants during pregnancy; therefore many of them may not continue with the medication throughout the pregnancy. If exposure to antidepressants in the second and third trimesters is more likely to be associated with preterm birth (and available evidence suggests that this might be the case) then TIS studies that classify women as exposed solely on the basis of first trimester exposure would likely underestimate the association between antidepressant use and preterm delivery. Studies that rely on electronic healthcare databases, on the other hand, contain detailed information on filled prescriptions for antidepressant medications during the entire pregnancy. Automated pharmacy dispensing information is usually seen as the gold standard of drug exposure compared to self-reported information [75] or prescribing records in outpatient medical records [76]. Patient recall bias is not an issue in healthcare utilization databases since all data recording is independent of a patient’s memory or agreement to participate in a research study [77]–[80]. However, filling a prescription does not necessarily guarantee that it was ingested, which could result in some misclassification. Such misclassification of exposure which is independent of the outcome status is likely to bias results towards the null (i.e. attenuate the association between antidepressant use and preterm birth.). A major concern is the potential for confounding by the underlying depression and its severity, and associated poor health behaviors (e.g., nutrition, smoking, drug and alcohol use). More severely depressed women may be more likely to take antidepressants during pregnancy, and it has been suggested that it may be the depression itself that is causing the preterm birth and not the medication. Several of the studies in this systematic review made efforts to control for maternal depression and these studies continued to show increased rates of preterm birth in the antidepressant exposed pregnancies. The majority of studies we reviewed did not find increased preterm birth rates in depressed women unexposed to medication. However, despite these studies’ attempts to control for depression, it is likely that women with a diagnosis of depression who opt to continue treatment during pregnancy are inherently different from women with a depression diagnosis who discontinue treatment during pregnancy. It is therefore questionable whether these studies completely addressed confounding by indication severity. Nevertheless, the available data on whether depression itself is associated with preterm birth is inconsistent [81], and expert review panels have concluded that there is no clear association between depression and preterm birth. The Institute of Medicine reviewed this question and concluded that “Overall, recent prospective studies on depression do not suggest a strong pattern for depression as a general risk for preterm delivery, consistent with the results of earlier studies” [7]. A 2009 AAP/ACOG review similarly concluded: “Available data neither support nor refute a link between MDD [major depressive disorder] and these outcomes [preterm delivery and gestational age]” [82]. Despite the weak evidence to support the independent association between depression and adverse pregnancy outcomes, such as preterm birth, the prior belief is strong among some investigators [83]. Some limitations of this review, resulting from limitations in the source data, should be noted. The studies included were heterogeneous in terms of design, size, exposure assessment, timing and nature of the exposure, and confounding adjustment. For example, in some studies, women classified as exposed were only those taking medication throughout the pregnancy, and controls were those who stopped antidepressant use before the second trimester, while in others the “exposed” were patients taking an antidepressant in the first trimester, who may have stopped by the second trimester. We tried to address this by presenting the results separately by level of adjustment and timing of the exposure, but assignment of studies to these categories is somewhat subjective, and heterogeneity within categories remains. Although we would have liked to simultaneously investigate the effects of some of these factors through meta-regression, this was not feasible due to an insufficient number of studies. Given that all but one study defined preterm birth as a dichotomous outcome (<37 gestational weeks), we were not able to examine the strength of the association between antidepressant medication use and very-, moderate- and mild-preterm birth respectively. Nevertheless, even late preterm birth is a significant contributor to poor neonatal outcomes [3]–[6]. Likewise, the available evidence did not permit evaluation of the association with specific types of prematurity [83]. The most rigorous method for determining an association between antidepressant medication and preterm birth would be a randomized controlled trial, and some have argued for this [84]. Yet, numerous studies are now available that suggest antidepressant use during pregnancy may be associated with spontaneous abortion [85], birth defects [86], persistent pulmonary hypertension of the newborn [87], and newborn behavioral syndrome [88]. Whether it would be ethical to randomize depressed women to a medication arm with these possible effects is open for debate. Since it is unlikely a randomized trial will ever be conducted and since causality can never be ‘proven’ in the absence of trial data, we need to decide at what point the evidence is sufficiently strong to warrant informing patients, providers, and the public about the potential risks, so that they can be weighed against any expected benefits in a given patient. In conclusion, the findings from our review of the literature are consistent with an association between antidepressant use during pregnancy and preterm birth, although the possibility of residual confounding by depression severity cannot be completely ruled out based on the available evidence. Counseling of pregnant women must take into consideration the clinical circumstances of a given patient, the strength of the available evidence on the risks and benefits (i.e., avoidance of risks associated with untreated depression), and alternatives to medication use during pregnancy. While our study findings cannot prove causality, they reinforce the notion that antidepressants should not be used by pregnant women in the absence of a clear need that cannot be met through alternative approaches.

Acknowledgments We would like to acknowledge the contributions of Jessica L. Bauer MD, MS (Beth Israel Medical Center, Albert Einstein College of Medicine, New York, NY) and Wesley Eddings, PhD (Division of Pharmacoepidemiology, Brigham and Women’s Hospital, Boston, MA) in the preparation of this manuscript. This study was presented at the 29th International Conference on PharmacoEpidemiology and Therapeutic Risk Management, Montreal, August 25–28, 2013.

Author Contributions Conceived and designed the experiments: KH RS JA AU. Performed the experiments: KH AU. Analyzed the data: KH. Contributed reagents/materials/analysis tools: KH JA AU. Wrote the paper: KH RS JA AU.