Prenatal antidepressant exposure does not appear to increase autism risk

Both psychiatric control and discordant-sibling designs do not corroborate the association of exposure to prenatal antidepressants with autism, results of a meta-analysis have shown. Moreover, discordant-sibling designs effectively address surveillance bias in pharmacovigilance reports derived from national registries and other large databases.

In this study, the utility of comparison group operationalization in reducing surveillance bias was examined. The investigators performed a systematic search of several databases through August 2017 and identified observational studies of the association between prenatal antidepressant exposure and autism. They used the Newcastle-Ottawa Scale to assess study quality.

Random-effects meta-analysis yielded summary measures with 95 percent confidence intervals (CIs) stratified by comparator group composition, antidepressant class and exposure trimester.

Of the 14 studies that met the eligibility criteria, 13 reported results using a population-based comparison group, five using a psychiatric control group and four using a discordant-sibling control group. Eight studies were regarded as poor due to inadequate control for prenatal depression and maternal ethnicity.

Risk estimates for autism following prenatal exposure to any antidepressant were different for population-based designs (hazard ratio [HR], 1.42, 95 percent CI, 1.18–1.70; odds ratio [OR], 1.58, 95 percent CI, 1.25–1.99) in comparison with psychiatric control (HR, 1.14, 95 percent CI, 0.84–1.53; OR, 1.24, 95 percent CI, 0.93–1.66) and discordant-sibling designs (HR, 0.97, 95 percent CI, 0.68–1.37; OR, 0.85, 95 percent CI, 0.54–1.35).

Similar findings were observed for prenatal exposure to selective serotonin reuptake inhibitors.

In meta-regression of population-based studies, ethnicity differences remained a significant source of study heterogeneity despite statistical adjustment.

“Observational studies of prenatal antidepressant safety are hindered by methodological concerns, including susceptibility to surveillance bias,” the investigators said. “Some studies address potential bias by using alternative strategies to operationalize study comparison groups.”