The Autism Internet Research Survey was created by the parent of a child with autism hoping to identify possible causes for the rise in autism. The survey did not state whether the rise in autism was due to a rise in incidence or in the number of individuals registered for special education programs; however, parents who believe there is an increase in autism incidence may have been more inclined to take the survey. In order to quickly obtain the number of cases required for this analysis, the internet was used to solicit participants. Subsequently, this developed into a New Jersey-based nonprofit organization, Autism Internet Research Survey. Neither the organization nor the survey is related to any commercial entity.
The Autism Internet Research Survey invited parents to complete surveys for their children with or without AD. Whether a child had AD was self-reported by clicking on one of two links: "For those with autistic children who want to take the survey click here." or "For those who want to take the control survey (you have children, but not with any autism spectrum disorder) click here."
Ads for the surveys were placed online using Google and restricted to the United States. Individuals who performed online searches containing keywords (autistic, autism research, autism, MMR, autism education, etc.) were shown an ad requesting their participation in a research survey. The total number of keywords used was 306, and they were grouped into the following categories: autism and autistic features 262, treatment for autism 24, prominent people involved in autism 13, and possible causes of autism 7. Participants completed the surveys from February to April 2005. The surveys included 91 questions on breastfeeding, infant formula use, date of birth, and the nature of their child's development. Limiting the age range to children two to 18 years and the respondents to parents yielded 861 case and 123 control children.
Breastfeeding data was recorded from a drop-down menu with nine choices of duration of breastfeeding. This variable was recoded into five categories: none, less than 2 months, 2–6 months, more than 6 months, and unknown. These breastfeeding categories were tested for association with autism using logistic regression.
Infant formula use data was recorded from a drop-down menu with 39 brand-name choices as well as "Other", "None", and "I don't know". This variable was recoded into three categories: None, Formula without DHA/ARA, and Formula with DHA/ARA. Information regarding DHA/ARA supplementation was ascertained from the manufacturers websites. If parents chose the category "Other" or "I don't know", no determination could be made regarding DHA/ARA supplementation, and the data was excluded from further analysis (n = 38). The remaining three infant formula categories were tested for association with autism using logistic regression.
Children under two years old were excluded since AD is rarely diagnosed before age two. For analysis of breastfeeding, the age range was limited to 2–18 years. Eighteen years was chosen as the upper age for the range in an attempt to minimize recall bias from the parents.
For analysis of infant formula, 2–4 years was chosen as the age range. Four years was chosen as the upper age for this portion of the study since supplementation with DHA/ARA has only been available in the US since 2002. Children older than four would not have had the opportunity to use DHA/ARA supplemented formulas during the first year of life.
Parents of autistic children were also questioned about the nature of their child's development. Three choices were given in a drop-down menu: 1) My child developed normally, then regressed (lost skills). 2) My child developed normally, then stopped. 3) My child never developed in a normal way. For the purposes of this study, if response number 1 was chosen, the child was assumed to have a regression in development, i.e. lost skills that had previously been acquired.
In order to remove the effects of congenital conditions associated with autism from the odds ratios seen in this study, breastfeeding and infant formula use were also tested for association with autism for the subset of children with reported regression in development.
All analyses, including characterization of the population and logistic regression, were performed using SAS version 9.1 for Windows (SAS Institute Inc., Cary, North Carolina). This study was approved by the University of California, San Diego Human Research Protections Program and the Institutional Review Board at San Diego State University.