Discordant Sibling Pair Comparisons in Observational Studies: A Research Design Simply Explained.

Abstract

When studying how (eg) gestational exposure to antidepressant drugs influences the risk of (eg) autism spectrum disorder (ASD) in offspring, conventional observational studies adjust analyses for available covariates and confounds. In such analyses, a significant association between antidepressant exposure and ASD outcome can never be asserted to be causal because of the possibility of residual confounding arising from confounding by indication (or severity thereof), confounding by genetic risk factors, and confounding by environmental risk factors. Confounding by indication and severity thereof can sometimes be addressed through propensity score matching, but the adjustment can never be perfect. Additionally, adjustment for genetic and environmental risk factors is hard or impossible to do because these are inadequately measured, unmeasured, and/or unknown variables. Sibling comparison studies have recently emerged as an option to address the genetic and environmental risk factors. In such studies, sibs discordant for exposure are compared for risk of outcome (cohort design) or sibs discordant for outcome are compared for odds of exposure (case-control design). The assumption is that sibs share similar genetic and environmental risk factors and so, when sibs are compared, these risk factors cancel out whether they are measured or not, known or unknown. If antidepressant exposure remains significantly associated with ASD in sibling comparisons, a possible conclusion is that antidepressants and not genetic or environmental factors drive the ASD risk. If antidepressant exposure loses significance in the sibling comparisons, it suggests that shared genetic and/or environmental factors, rather than antidepressant exposure, explain the ASD risk. The interpretation, however, is nuanced. Strengths, limitations, and interpretations of sibling comparison studies are explained. To illustrate the usefulness of sibling comparisons, results are presented from a recent study of ASD risk after gestational or early infancy exposure to antibiotics.