Association Between Antenatal Exposure to Selective Serotonin Reuptake Inhibitors and Autism

Abstract

To the Editor: The recent systematic review and meta-analysis by Brown et al 1 that aimed to assess risk of autism spectrum disorder (ASD) in children whose mothers were exposed to selective serotonin reuptake inhibitors (SSRIs) during pregnancy was interesting. The authors reported an unclear association between antenatal exposure to SSRI and risk of ASD in offspring that warranted future studies. We applaud Dr Brown and colleagues' meaningful work. However, several important methodological issues regarding study selection should be noted.Of studies based on the same data sources and overlapping study periods and population, the authors retained the study with the highest rating score in quality assessment for metaanalysis. However, the authors failed to provide a clear theoretical background for that inclusion and exclusion. In the quality assessment section, a study by Hviid et al 2 was marked as moderate and was included for quantitative synthesis, but a study by Sørensen et al 3 was marked as low and was excluded because Hviid et al controlled for other psychotropic drugs. This reason for excluding the study by Sørensen et al is not adequate. First, no other included studies controlled for other psychotropic drugs and met the criteria for adequate consideration of distorting influences. Second, those 2 studies 2,3 were both rated as high quality using the NewcastleOttawa scale. 4 More importantly, the cohort study by Sørensen et al had an extra data source, the Danish National Hospital Register, and thus had a larger sample size with more ASD cases (91 cases in the SSRI-exposed group) than the study by Hviid et al (52 cases in the SSRI-exposed group), so the study by Sørensen et al might be more capable of assessing the causal association considering that autism cases are relatively rare. In this case, sensitivity analyses should be performed by excluding the study by Hviid et al and including the study by Sørensen et al.The rationale for excluding the case-control study by Gidaya et al 5 is also unclear. The authors stated that of the 2 studies that had the same final quality rating, the retrospective study by Hviid et al 2 was retained because of better control for distorting influences. That reason is not convincing, as the case-control study by Gidaya et al had a better exposure measure. Besides, data extracted from case-control studies and cohort studies were separately pooled in the quantitative analysis, so the interpretation of the results is based on the 2 pooled odds ratios. In consideration of the limited number of included studies, the results of a meta-analysis of case-control studies would be more robust with inclusion of the study by Gidaya et al.Overall, even though several studies used the same data sources, those studies were inconsistent in some methodological respects such as study design, sample size, or statistical analysis, which may potentially influence the association of interest. For studies with the same design, sensitivity analyses could be performed to assess the robustness of ...