IMFAR Analysis in Support of NDAR Strategic Requirements

Background: The National Database for Autism Research (NDAR) is a collaborative biomedical informatics system sponsored by the National Institutes of Health.  Primary objectives of NDAR are to provide broad access to high quality, detailed human subjects data that underlie research findings relevant to autism spectrum disorders (ASD) and to facilitate data sharing. Sharing of phenotypic, genetic and imaging data provides the opportunity for researchers to (a) validate research results, (b) pool standardized information, (c) use data collected by others to explore hypotheses not considered by the original investigator.  NDAR also uses a global unique identifier (GUID) that allows researchers to identify data from a particular subject across multiple studies using a secure method.
Objectives:

To characterize the type of research data and typical sample sizes of recent ASD research studies using information available from abstracts of the 2008 International Meeting for Autism Research (IMFAR).  This analysis was carried out to better understand how NDAR can contribute to advancing and accelerating ASD research.

Methods:

The 779 poster and oral presentation research abstracts were reviewed for the following information: 

·         Abstract topic - by IMFAR category

·         Whether the research included human subjects or biomaterials

·         Sample size of enrolled human subjects with ASD as well as controls

·         Specific clinical assessments/measures disclosed

·         Whether the study focused on NDAR relevant categories of neuroimaging, genetics, treatment/intervention, or phenotyping

A priori exclusion criteria were established to confine subsequent analyses to studies most appropriate for inclusion in NDAR. The exclusion criteria were:

·         The study did not involve human subjects or the number of human subjects was not reported (123 studies)

·         The research focused on the development or validation of  measures/assessments (40)

·         The study involved cell/animal models (25)

·         The research involved a literature review, bioinformatics system review, or survey of treatment providers (19)

·         The study reported on a large self-selected survey population or online questionnaire (19)

·         The study was an epidemiology study (38)

Results:

Of the 779 abstracts, 510 (65%) involved human subjects or human biomaterials.  The average sample size of IMFAR studies involving human subjects is 67 with a standard deviation of 118.  The median sample size is 24.

Analysis of sample sizes for specific types of studies: Mean (SD)

Treatment/Intervention (70 studies) – 31 (73)

Phenotyping (323 studies) – 60 (99)

Genetics (47 studies) - 218 (219)

Neuroimaging (57 studies) – 21 (15)

Conclusions:

This analysis provides evidence that relatively small sample sizes are the norm in ASD research and that a large portion of autism studies focuses on phenotype data.  The mean sample sizes differ by study type with genetics studies having larger sample sizes than imaging studies. Of particular note are the small samples sizes for treatment/intervention studies, which generally require larger numbers of participants to have sufficient power to demonstrate efficacy. It is our belief that community adoption and support of a bioinformatics platform such as NDAR could help increase scientific collaboration/corroboration, thereby increasing study power, and data sharing, and other efficiencies.