Article Text

Systematic review
Meta-analysis: antidepressant exposure during pregnancy is associated with poor neonatal adaptation
  1. Lars Henning Pedersen
  1. Department of Obstetrics and Gynecology, Institute of Clinical Medicine, Aarhus University, Aarhus, Denmark
  1. Correspondence to: Dr Lars Henning Pedersen, Department of Obstetrics and Gynecology, Institute of Clinical Medicine, Aarhus University, Brendstrupgaardsvej 100, Aarhus 8200, Denmark; lhp{at}

Statistics from

Commentary on: OpenUrlCrossRefPubMed


Antidepressants are used by 7.6% of pregnant women in the USA.1 Treatment of depression during pregnancy must balance the well-being of the mother with the potential effect on the fetus. Importantly, if prenatal antidepressant exposure causes neonatal symptoms, easy access to neonatal care may be warranted for women treated during pregnancy.


Grigoriadis and colleagues summarise existing knowledge of the association between prenatal antidepressant exposure and neonatal symptoms. After initial literature searches, they evaluated the detected studies using 19 criteria, calculated a quality rating score (‘high’, ‘moderate’, ‘low’ and ‘very low’) and excluded studies that were considered of ‘very low’ quality. Quality assessment was based on existing instruments and checklists, with a number of choices made specifically for this area. Exposure was classified as use of any antidepressants, although the majority had used serotonin reuptake inhibitors (SSRI). The outcome was defined as the presence of one or more of 15 signs of ‘poor neonatal adaptation syndrome’ (PNAS), including respiratory distress or seizures. Additionally, as PNAS was not uniformly defined in the studies, the authors specifically investigated respiratory distress (or tachypnoea if respiratory distress was not listed) and tremor (or ‘shaking’ if tremor was not listed). Potential publication bias was investigated by funnel plots and the Duval and Tweedie L estimator. The findings are presented as OR with 95% CI.


The meta-analysis found an association between exposure to any antidepressant and the risk of PNAS (OR=5.07, 95% CI 3.25 to 7.90) based on eight studies. For respiratory distress and tremor the estimates were OR=2.2 (95% CI 1.81 to 2.66) and OR=7.89 (95% CI 3.33 to 18.73), respectively. The authors found no evidence for the presence of publication bias for the main PNAS analysis but there were signs of publication bias for the individual symptoms analyses. With Duval and Tweedie's trim-and-fill procedure, the estimates changed to OR=2.00 (95% CI 1.59 to 2.52) for respiratory distress and OR=5.17 (95% CI 2.11 to 12.7) for tremor.


Based on the potential adverse effects of antidepressants in adults and the fact that the drugs cross the placenta, some infant symptoms after prenatal exposure seem likely. Toxicity and withdrawal are among the potential explanations for PNAS. Regardless of the causal mechanism, Grigoriadis and colleagues provide solid evidence for an association that may in itself influence obstetric and neonatal management.

Clinically, a number of questions remain. A higher risk associated with paroxetine and clomipramine compared with the other SSRIs has been suggested elsewhere,2 but this meta-analysis did not investigate the potential differences. Tapering of antidepressant treatment before expected labour is controversial: the potential consequences have not been studied and it does seem problematic to taper a treatment for depression just before a potential traumatic experience. Timing during pregnancy and dose-response effects were not covered by the meta-analysis. The authors thoroughly discuss the timing of the symptoms after birth, but more data is needed in order to determine if and for how long the exposed infants need to be observed in the postpartum period. The potential long-term effects of PNAS are uncertain and optimal treatment of infants with the condition is yet to be determined.

There are some limitations with the current paper. Importantly, the meta-analysis was unable to adjust for maternal depression. Secondly, the assessed studies differ substantially in the confounders included in the adjusted models, while the non-collapsibility of ORs might further hamper interpretation. Thirdly, the heterogeneity in the definitions of the outcomes makes it difficult to avoid potential information bias (eg, the term ‘shaking’ is potentially more susceptible to information bias than ‘seizures’). The estimates may consequently be quantitatively less valid and less precise than the CIs suggest. Yet, the direction of the associations found in the meta-analysis seems convincing and, crucially, the authors found no signs of publication bias in the main analysis. Despite the above limitations, Grigoriadis and colleagues provide important evidence for decision makers when discussing, for example, access to neonatal care after prenatal antidepressant exposure.


View Abstract


  • Competing interests None.

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.