Fig 1.
Possible causal structures underlying a hypothesized association between autism and sibling sex.
In the common cause scenario (A), an association arises because risk of autism and sex of each child share a common antecedent (for example, maternal testosterone levels). In the sex independence scenarios (B, C), risk of autism in siblings shares a common cause (B), or observation of autism in one sibling directly influences observation of autism in the other (C), but sex is independent. In the separate causes scenario (D), both sex of siblings within a given family and risk of autism in those children are correlated, but the association arises through separate antecedent factors. In the autism independence scenario (E), sex of siblings shares a common cause but risk of autism does not. Depending on method of analysis, an association between whether autism is diagnosed in one child (A1) and sex of the sibling (S2) could potentially be observed under any of these scenarios.
Fig 2.
Possible causal structures underlying a selection bias induced association between autism and sibling sex.
Preferences regarding sex composition may affect future fertility decisions, and hence observation (O) of additional siblings. This is represented here by arrows from sexes of child 1 and 2, S1 and S2, to O3, the observation (or not) of a third child. If either autism in previous children (A1 and A2), or shared factors associated with autism (U; ex. parental age at the start of childbearing) also affect future fertility, selection bias may result. This is due to de facto conditioning on O3 (only children who are conceived and born may be observed), opening a blocked path between autism and child sex. Conditioning on the sex of prior children will block these paths. A box drawn around a variable represents conditioning on it, which, unless it is a collider, blocks any paths through it.
Table 1.
Summary of criteria for a valid sibling sex ratio test.
Table 2.
Identification of eligible subjects from total population of California births, 1992–2007 and percentage male among included and excluded births.
Table 3.
Relative risks that next sibling is male among subjects with versus without Department of Developmental Services (DDS) autism diagnoses among California births, 1992–2003.