New research published in the BMJ studies the association between antidepressant use during pregnancy and autism.
A roundup accompanied this analysis.
Title, Date of Publication & Journal
Antidepressants during pregnancy and autism in offspring: population based cohort study
Published: Wednesday 19 July
Study’s main claims – and are they supported by the data
The study claims there is a small association between antidepressant use in pregnancy and autism without intellectual disability. This claim is plausible given the evidence presented in the paper. The study finds no link between antidepressant use and autism with intellectual disability.
The study does not claim that antidepressant use causes autism, and there is no evidence in the paper to support such a claim.
Although the association is statistically robust, the number of children this affects is very small – around 1 in every 100 babies born to a mother on antidepressants. Additionally none of the ‘excess’ cases of autism will be autism with associated intellectual disability. The link is mostly of interest because it tells us something new and interesting about the origins of autism, rather than because it uncovers a significant risk to new babies.
The study uses five different statistical methods to investigate the relationship between antidepressant use in pregnancy and autism. The methods all give similar results which make the conclusions more robust than if any one statistical method was used in isolation.
However this methodology doesn’t completely eliminate the risk of a chance confounding factor affecting all five tests e.g. there may be underlying genetic factors that causes both antidepressant use in mothers and autism in babies.
It’s worth noting that the association was only seen for autism without intellectual disability and there was no statistically significant effect for autism with associated intellectual disability.
It is important to emphasise the significant risks of stopping antidepressants during pregnancy before discussion with a doctor. These can range from serious maternal depression meaning the baby cannot be cared for properly, to clinical effects on the baby such as withdrawal symptoms.
There are two main strengths to this study:
The first is a very large sample size, which was the entire population of Stockholm County in Sweden from 1996-2007. This helped the researchers find subtle effects in their data, which is important when looking at a rare outcome such as children with autism being born to mothers taking antidepressants.
The second is a robust selection of methods to assess the claim, including rare but powerful statistical tests such as a negative control design. This is good because it makes it less likely that the authors are ‘cherry picking’ the one statistical analysis that they think is the most interesting.
The main weakness of the study is an inability to prove whether antidepressant use actually causes autism, for example by investigating whether there is a dose-response relationship between severity of depression / amount of antidepressants and a diagnosis of autism. This study can only show an association.
The authors were unable to analyse the severity of the psychological disorders suffered by the mothers. Although they attempted to adjust for confounders like this, the methods used mean the authors cannot control for all residual issues. A potential confounding factor could be that women who continue to take antidepressants during pregnancy have more severe mental illnesses.
Although the sample size is large, the number of mothers who met the criteria was quite small, which makes subgroup analysis challenging. In particular the findings on the effect of individual antidepressant drugs had a very small sample size – probably much smaller than is appropriate given the number of comparisons the authors are making.
Intellectual disability – A condition characterised by significant limitations in general mental capacity (IQ) and adaptive behaviour (social and practical skills such as using money or telling the time). Usually an IQ of less than 70 indicates a possible intellectual disability
Negative control trial – A trial in which no response is expected, because an experimental condition is removed. If a response is detected, that proves there must be a confounder.