Taking antidepressants during pregnancy does not appear to increase the child’s risk of autism, according to a new meta-analysis.

The review[1] examines 14 studies, many of which identified a connection between antidepressant use during pregnancy and autism. However, that research failed to account for ascertainment bias, which occurs when one group of patients or subjects undergoes testing more frequently than others, says study author Jeffrey Newport, director of the Women’s Reproductive Mental Health program at UT Health Austin’s Mulva Clinic for the Neurosciences and a professor of psychiatry at Dell Medical School at The University of Texas at Austin.

“In these studies, immigrant and Latina mothers consistently had both lower rates of antidepressant treatment and lower rates of autism diagnosis in their children. This is not surprising, as these minority groups are known to have poorer access to health care, including treatment for depression and careful diagnostic assessment of concerning behaviors in a child,”

Newport says.

Ethnic Bias

Newport discovered that family-based studies eliminated the bias problem by comparing children with antidepressant exposure or autism diagnosis with their siblings who did not have antidepressant exposure or autism. With the ethnic bias eliminated, the family-based studies revealed no association between prenatal antidepressant use and autism.

“This should remind us that although insurance databases and national registries have the advantage of huge numbers of participants, their data is not collected to answer research questions, but to manage business and clinical concerns. Thankfully, the results of this meta-analysis show that with thoughtful study designs, researchers can overcome the biases often encountered when using such databases,”

Newport says.

Ascertainment bias (also called detection bias) is caused when the actions of an investigator affect the results of a trial. Ascertainment bias can be unintentional and could arise from the investigator’s hopes or expectations of the trial. It is most likely to occur when:

  • the investigator’s personal judgement is used to measure subjective trial outcomes, and

  • the investigator knows which group each person is allocated to (for example treatment or placebo).

The most important design technique for avoiding detection bias in clinical trials is blinding. The potential effect of bias should also be taken into account during statistical analysis of trial data.

[1] Monica L. Vega, Graham C. Newport, Durim Bozhdaraj, Samantha B. Saltz, Charles B. Nemeroff, and D. Jeffrey Newport. Implementation of Advanced Methods for Reproductive Pharmacovigilance in Autism: A Meta-Analysis of the Effects of Prenatal Antidepressant Exposure. American Journal of Psychiatry, May 2020 https://doi.org/10.1176/appi.ajp.2020.18070766

For future updates, subscribe via Newsletter here or Twitter