How can population cohort studies contribute to our understanding of low prevalence clinical disorders? The case of autism spectrum disorders