The safety of antidepressants in pregnancy

BMJ 2017; 357 doi: https://doi.org/10.1136/bmj.j2544 (Published 31 May 2017) Cite this as: BMJ 2017;357:j2544
  1. Lars Henning Pedersen, associate professor
  1. Department of Obstetrics and Gynaecology, Aarhus University Hospital, Institute of Clinical Medicine, Aarhus University, 8200 Aarhus N, Denmark
  1. LHP{at}clin.au.dk

Maternal antidepressants are implicated in ADHD, but so is maternal depression

The safety of antidepressants in pregnancy is controversial, partly because of profound methodological difficulties in separating the fetal effects of antidepressants from those related to maternal depression (confounding by indication). One central concern is the potential impact of these drugs on fetal brain development. Such effects may be subtle and possibly only detectable years after exposure, such as an increased susceptibility to (multifactorial) neurodevelopmental conditions.

Attention-deficit/hyperactivity disorder (ADHD) is the most prevalent neurodevelopmental disorder linked to prenatal use of antidepressants. The study by Man and colleagues in this issue (doi:10.1136/bmj.j2350) suggests that the association is at least partly explained by shared family factors.1 This is potentially an important finding, but the results should be interpreted within the context of the existing literature.

An effect of antidepressants on the fetal brain is biologically plausible. Monoamines modulate neuronal proliferation and migration and axonal wiring.2 In animals, exposure to antidepressants during brain development equivalent to the second and third trimester in humans causes histological and behavioural changes that persist into adulthood.3 However, maternal stress also has profound effects on the developing brain of experimental animals.

These mechanisms are difficult to investigate in humans. All human studies rely on observational designs, in which it might be impossible to adjust for prenatal and postnatal environments and relevant genetic traits. Previous studies have compared exposed children with the offspring of women with untreated depression4 or the offspring of women who discontinued treatment before pregnancy,5 or unexposed siblings,6 but there are still risks of systematic error in these designs.

Man and colleagues used data from Hong Kong population based electronic medical records in which 1252 out of 190 618 children were exposed to antidepressants. The crude association between antidepressant exposure and ADHD was 2.39 (95% confidence interval 1.9 to 3.0), but in analyses adjusting for maternal depression the estimate was 1.39 (1.1 to 1.8). The authors used several advanced approaches to try and isolate the effect of antidepressants, including several negative control analyses. Overall they found consistent signs of confounding by indication, in accordance with a recent study from Sweden.7

As the authors point out, however, the importance of maternal depression or related factors in these studies does not rule out a causal effect of antidepressants. For example, potential errors in the data such as non-differential misclassification of the exposure can bias results towards the null. In studies using administrative registries, misclassification happens when prescribed drugs in the exposed group are not used, or are used long after the prescription is cashed. In a Monte Carlo simulation8 assuming a “true” twofold risk of ADHD associated with antidepressants, misclassification of 20% of “exposed” participants (which would follow 80% compliance with prescribed treatment) coupled with a sensitivity and specificity of ADHD diagnosis of 0.90 and 0.99, would reduce the risk estimate to a potentially misleading 1.45.

Antidepressant use in pregnancy has been associated with other outcomes, including malformations,9 neonatal problems,10 and autism.6 Some of the associations diminish in analyses adjusting for maternal depression (eg, autism).67 Neonatal “adaptation” problems, however, seem to be caused by treatment.10 Other associations are still disputed, such as a link with septal heart defects.9 This link is so consistent across populations, that it should be discussed with patients regardless of the true causality of the relation.

There is no easy way to determine causality. Randomised trials are often unsuitable because of the rarity of many of the outcomes and, most importantly, ethical considerations. Advanced epidemiological methods continue to be the most useful tools despite the caveats, especially in combination with biological markers or imaging modalities,11 and preferably with potential biases quantified.8

Setting aside the debate about causality, use of antidepressants in pregnancy is a marker for pregnancies with risks of various adverse outcomes and for children with an increased susceptibility to a number of conditions, including ADHD.

Future care of pregnant women with depression should ideally consider detailed clinical and background information, including type of treatment and pharmacogenetic traits.] Importantly, the nature and severity of maternal mental illness must be taken into account. Because the uncertainties around many of the risks statistically associated with antidepressants cannot be disregarded, pregnant women with mild depression may benefit from non-drug treatment. However, pregnant women with severe depression need effective treatment12 that will in most cases include antidepressants.

Decision models that incorporate the risks of maternal depression and drugs and also take into account individual factors, including genetic traits, would be helpful. Such models will require more research on the importance of these factors, and greater collaboration to reach sample sizes big enough to generate answers to these important questions.


  • Research, doi: 10.1136/bmj.j2350
  • Competing interests: I have read and understood the BMJ policy on declaration of interests and declare the following: none.

  • Provenance and peer review: Commissioned; not peer reviewed.


View Abstract

Sign in

Log in through your institution