select search filters
roundups & rapid reactions
before the headlines
Fiona fox's blog

antidepressant use in pregnancy and psychiatric disorders

Scientists publishing in the BMJ report an association between antidepressant use in pregnancy and diagnosis of psychiatric disorders in children.

A roundup accompanied this analysis.


Title, Date of Publication & Journal

Antidepressant use during pregnancy and psychiatric disorders in offspring: Danish nationwide register based cohort study by Xiaoqin Liu et al.

Published: September 7 2017



Study’s main claims – and are they supported by the data

The study does not claim that antidepressants taken during pregnancy cause psychiatric disorders in childhood. The authors acknowledge there is no evidence in the paper to support such a claim.

The study rather claims there is an association between antidepressant use in pregnancy and diagnosis of psychiatric disorders in children.

There is evidence from the analysis conducted that exposure to antidepressants is potentially related to an increased risk of psychiatric disorders. Compared to an unexposed group who had no maternal antidepressant use before or during pregnancy, increased risks of psychiatric disorders were identified in three antidepressant exposure groups: (1) the discontinuation group – those whose mothers had used antidepressants before but not during pregnancy, (2) the continuation group – those whose mothers used before and during pregnancy and (3) the new user group – those whose mothers had only used antidepressants during pregnancy. However the authors also note that this association could instead be due to the mothers’ underlying illness, or a combination of the antidepressants and mothers’ underlying illness. The methods used in this study do not enable a direct relationship between antidepressant use and diagnosis of psychiatric disorders in children to be inferred.

The association is unsurprising since children with mothers with a history of psychiatric disorders are more likely to have psychiatric disorders themselves, given that we know psychiatric disease has a genetic component (in addition to upbringing styles). If we take antidepressant use in pregnancy as a marker for psychiatric disease, these results do not seem unexpected.

The authors attempt to control for the underlying illness by focussing in on the risk for psychiatric disorders among offspring in the continuation group (2) in comparison to the discontinuation group (1). The risk for psychiatric disorders among offspring in the continuation group was higher than that in the discontinuation group with a hazard ratio 1.27, (95% confidence interval 1.17 to 1.38), which represents an increased relative risk of psychiatric disorders of 27% for children of mothers who continued relative to discontinued. This effect is relative – it is important to consider the absolute effect to establish the number of individuals this actually impacts. The authors use the relative risk to estimate the proportion of psychiatric cases in the population that could have been prevented if mothers from the continuation group had discontinued antidepressant treatment as 0.5%. However this estimated value assumes the association to be causal, which the authors have already indicated cannot be inferred.  This figure should therefore be interpreted with some caution.




The study used a large, population based data set (from Danish national registers). Data on antidepressant use was collected prospectively, which reduces the risk from recall bias. Some results have been reported in absolute terms, however see note above on assumptions underlying some of these. A number of sensitivity analyses were conducted to explore the robustness of the main conclusions and the results of these were supportive of the primary results. The authors have carefully thought through potential confounders and adjusted for a number in the analysis, including demographic and psychiatric characteristics of the mothers, including inpatient and outpatient psychiatric treatment and comorbid psychotropic drug use. However this does not eliminate the potential for other unmeasured confounders to be having an impact on results and causing spurious associations (see limitations below).


Cause and effect is unclear. It is not possible to tell whether the antidepressant use in pregnant women is causing any effect on psychiatric disorders in children. The methodology doesn’t completely eliminate the risk of an additional unmeasured confounding factor (or factors) affecting all results e.g. underlying genetic factors. It is noted that reported effect sizes were reduced after adjustment for the confounders considered which indicates that some of the observed association is explained by factors other than the antidepressants themselves. Not all absolute effect sizes are clear; some rely on strong, untestable assumptions.

 The main results presented in the abstract and press release are the adjusted 15 year cumulative incidence of psychiatric disorders by antidepressant user group. Whilst it is indicated that children were followed for a maximum of 16.5 years, many would have been followed for much less time than this (births during 1998-2012 were included). It is not clear exactly how many of the study population reached 15 years of follow-up, therefore how reliable these figures really are. 



Confounder = a variable that influences both the outcome (psychiatric disorder) and exposure variable of interest (depressant use) causing a spurious association.

 The population attributable fraction = the proportion of psychiatric cases in the population that could have been prevented if mothers from the continuation group had discontinued antidepressant, assuming the association to be causal and no unmeasured confounding.

Hazard ratio  = relative change the in risk of an outcome (psychiatric disorder) experienced by people exposed compared to those unexposed.


Any specific expertise relevant to studied paper (beyond statistical)?



* ‘Antidepressant use during pregnancy and psychiatric disorders in offspring: Danish nationwide register based cohort study’ by Xiaoqin Liu et al. published in the BMJ on Thursday 7 September 2017.


The SMC also produced a Factsheet on genome editing which is attached and available here: 


Before The Headlines is a service provided to the SMC by volunteer statisticians: members of the Royal Statistical Society (RSS), Statisticians in the Pharmaceutical Industry (PSI) and experienced statisticians in academia and research. A list of contributors, including affiliations, is available here.

in this section

filter Headlines by year

search by tag