On 2018 Jan 06, Rima Obeid commented:

Oversampling a high risk group and sampling from a hospital setting where kids were born and later treated for autism (a single Center study) strongly suggest a selection bias. Selection bias limits the external validity and generalizability of results from the Boston cohort. Recent metaanalysis and other large studies do not show an association or show even inverse association (example of links below). Moreover, sensitivity analysis is missing in the study by Raghavan et al.<br> https://www.ncbi.nlm.nih.gov/pubmed/29026508 https://www.ncbi.nlm.nih.gov/pubmed/29299606

On 2017 Dec 12, M Daniele Fallin commented:

True! This is an "enriched risk" ASD cohort in that it was over-sampled for pre-term birth, a consistent risk factor for ASD. In a low-risk or general US population, you would expect somewhere between 1 and 2% based on current CDC estimates of 8 year olds, although the prevalence at younger ages, and in diverse ethnic populations (such as this one) is less clear. This enriched risk design makes the study of ASD possible in this birth cohort (enough ASD events), and is a common design in ASD prospective studies - although typically the risk enrichment is due to family history (baby sibling studies).

On 2017 Dec 06, Aiguo Ren commented:

86 out of 1257 infants (6.8%) had ASD in this cohort! Isn't this incidence too high?

On 2017 Dec 06, Aiguo Ren commented:

86 out of 1257 infants (6.8%) had ASD in this cohort! Isn't this incidence too high?