Antidepressants during Pregnancy and Autism Risk: Evidence of Confounding

Background: A number of studies have reported associations between maternal antidepressant use during pregnancy and childhood autism. The causal meaning of these associations is still unclear. The possibility of genetic confounding and confounding by severity of underlying depression (the indication for the medications) has been hypothesised but has not been directly investigated to date.

Objectives: In a large population-based cohort we aimed to assess whether mothers with greater genetic risk for autism are more likely to take antidepressants during pregnancy and whether severity of depression is associated with autism risk

Methods: We used data from the Avon Longitudinal Study of Parents and Children (ALSPAC), a birth cohort study in England which recruited 14,541 pregnant women in the Bristol area in 1991-1992, resulting in 14062 live births. Medication use was ascertained prospectively via questionnaires and coded into WHO Anatomical Therapeutic Classification codes by a pharmacologist. Children with an ASD diagnosis were identified by record linkage with health and education records and parental reports. Autism related traits were ascertained by 4 scales that independently predict ASD in ALSPAC: the Social and Communication Disorders Checklist, the Children's Communication Checklist (coherence subscale), a repetitive behaviour measure, and the Emotionality, Activity and Sociability temperament scale (sociability subscale). These measures were dichotomised, with approximately 15% defined on each scale as having highest autistic traits. Maternal polygenic risk scores (PRS) for autism were constructed using summary data from the Psychiatric Genomics Consortium GWAS. Maternal depression was assessed prospectively at 18 and 32 weeks gestation using the Edinburgh Postnatal Depression Scale.

Results: Mothers in the top decile of an optimal polygenic risk score for autism (p-value threshold 0.05) had a 2-fold odds of taking antidepressants during pregnancy [Odds Ratio 1.97 (95%CI 1.05-3.70)] and for having a child with autism [2.19 (1.30-3.68)] as compared to those in the lower 90 percentiles. There was no evidence of an association between the mother's autism PRS and severity of depression (assessed both by continuous EPDS scores and in quintiles). A dose response association was observed between higher severity quintiles of maternal depression during pregnancy and social communication, speech coherence and repetitive behaviours but not with a diagnosis of autism.

Conclusions: The results demonstrate the potential for a significant contribution of genetic confounding, and confounding by severity of indication in the associations between antidepressant use during pregnancy and offspring autism. The implications of these findings to the wider literature will be discussed.