Background: Sibling sex ratios have been applied as an indirect test of a hypothesized association between prenatal testosterone levels and risk for autism, a developmental disorder disproportionately affecting males. Differences in sibling sex ratios between those with and without autism would provide evidence of a shared risk factor for autism and offspring sex. Conclusions related to prenatal testosterone, however, require additional assumptions. Here, we used directed acyclic graphs (DAGs) to clarify the elements required for a valid test of the hypothesis that sibling sex ratios differ between children with and without autism. We then conducted such a test using a large, population-based sample of children.
Methods: Over 1.1 million subjects, born in California from 1992-2007, and identified through birth records, were included. The association between autism diagnosis, determined using the administrative database of the California Department of Developmental Services, and the sex of the subsequent sibling was examined using generalized estimating equations. Sources of potential bias identified using DAGs were addressed.
Results: Among male children with autism, 52.2% of next-born siblings were brothers, versus 51.0% for unaffected males. For females with autism, 50.2% of following siblings were brothers versus 51.2% among control females. The relative risk of a subsequent male sibling associated with autism diagnosis was 1.02 (95% confidence interval: 0.99, 1.04).
Conclusions: In a large, population-based sample we failed to find evidence suggesting an excess of brothers among children with autism while controlling for several threats to validity. This test cannot rule out a role of any given exposure, including prenatal testosterone, in either risk of autism or offspring sex ratio, but suggests against a common cause of both.