Report on Autism Environmental Epidemiology Studies


Leading epidemiologists from around the world met at the International Society for Environmental Epidemiology on August 28, 2012, in Columbia, S.C., to present a symposium titled Challenges and Pitfalls of Multi-Site Collaboration in International Autism Environmental Epidemiology Studies. Epidemiological studies of environmental risk factors and their interaction with genetic susceptibilities have led to a deeper understanding of the causes of autism spectrum disorders (ASDs). The goals of this symposium were to:

  • Inform epidemiologists outside the field of autism on new findings.
  • Share study approaches and new technology.
  • Engage environmental epidemiologists interested in studying autism in groups within and outside the United States.
  • Discuss solutions and ideas to address challenges in studying environmental epidemiology of ASD internationally.

Because not everyone who would benefit from these presentations could attend the meeting, a summary of the main points and key take home messages are posted here on this website. Autism Speaks Senior Director of Environmental and Clinical Sciences Alycia Halladay, Ph.D., moderated the symposium. For more information, contact Dr. Halladay at

Here is a summary of the five presentations.

New Findings in the Environmental Epidemiology of ASD

Heather Volk, Ph.D., of the University of Southern California


ASDs are characterized by poor social skills and communication, and are often accompanied by repetitive or restricted behaviors. However, behaviors and symptoms of autism can vary greatly from one individual with ASD to the next. One factor shown to be associated with autism is preterm birth. Some studies have shown that as high as 8 percent of preterm infants are diagnosed with ASD. In the U.S., studies examining the prevalence of ASD (the number of cases in the population) seem to indicate that more children are being diagnosed with an ASD. Additionally, studies examining the incidence (number of new cases among those at risk) of ASD in California suggest that this increase cannot be fully explained by the fact that children are being diagnosed at an earlier age or because of a change in diagnostic criteria. It is likely that many factors, both genetic and environmental, contribute to ASD. However, despite this apparent increase in the number of children with ASD, we should not restrict our focus of research to just those risk factors that we think have become more prevalent in the environment in the last couple of decades. Moreover, we can also seek to understand risk factors that may explain a subset of individuals with autism. There may be subgroups of autism that might be more susceptible to certain environmental exposures. Future research that combines the efforts of epidemiologists, neurobiologists, geneticists and clinicians can help us address these factors and learn about factors, not previously identified, which increase risk for ASD. For more information, contact Dr. Volk at

Use of Existing Databases to Study Environmental Factors in Relation to ASD

Gayle Windham, Ph.D., of the California Department of Public Health


ASDs are developmental disorders thought to result from the disruption of normal neurobiological mechanisms during the prenatal and early postnatal period. Several lines of evidence suggest environmental factors should be examined for possible contributing factors. But several methodological difficulties hamper studies of environmental exposures and ASDs. Traditionally ASDs are not diagnosed until school age, yet the critical exposure period occurs during development in the womb. ASD is too infrequent to justify cohort studies, so case-control studies are typical, but have limitations. These include possible recall bias, a mother’s lack of knowledge about exposures or not being able to collect bio-specimens during the time period of most interest. 

An important way to examine the effects of exposure is by linking to databases with measurements of environmental exposures or access to archived bio specimens. Thus far, studies of autism have been conducted in relation to hazardous air pollutants, mercury releases, pesticide applications and proximity to freeways. As existing environmental data is assembled for other reasons, it often requires several layers of manipulation for use in health studies, or may not cover the ideal time or geographic area. Such studies have other limitations that must be considered, including the importance of having appropriate address history, exposures based on estimates or surrogates and a lack of information on personal habits that effect exposure pathways or on possible confounders. For more information, contact Dr. Windham at

Key points from the symposium:

1. Existing studies of autism, including from registry or surveillance systems, could be used to examine associations with environmental exposures by linking to existing sources of exposure data. This requires collection of address information, or a proxy, for geo-linking, during appropriate time period, with pregnancy (or birth) of most concern.  Examples of toxic exposures of interest with already collected measurements include:

  • Hazardous air pollutants (HAPs)
  • Criteria Air Pollutants
  • Traffic Density data
  • Water monitoring, disinfection by-products
  • Pesticide use or applications
  • Toxic release inventories

2. There are a number of issues that should be considered when setting up a study. They are:

  • Temporal variability – what is the unit of time of exposure measurement (day, month, year) and can it be used to assign specific times during development of each subject?
  • Work across geographic boundaries or multiple organizations’ data
  • Spatial variability - what is the unit of geographic measurement (block, census tract, county) and does it provide enough difference from person to person in the study to see patterns?
  • Availability of information on potential confounders

3.  There are existing birth cohort studies that have extensive follow-up data during susceptible periods, or biospecimens with measured exposures, that could be used to study exposures and autism by linking to existing registries or administrative data bases, or adding evaluation for ASD or autism traits.

4. These projects require collaborations among researchers and clinicians with different backgrounds and culture, committed funding and some methods development.


Measuring Early Courses of Autism:  A Challenge of the Hamamatsu Birth Cohort Study

Kenji J. Tsuchiya, M.D., Ph.D., of the Hamamatsu University School of Medicine, Japan.


The causes and early developmental trajectory of ASD have not been fully understood. One comprehensive way to tackle this big question is to track and monitor behavioral traits, biological markers and environmental exposures of children as often as possible in their development. To follow this research strategy, Hamamatsu Birth Cohort (HBC) Study was established in November, 2007, in Hamamatsu, Japan. There are 1,200 mother child dyads that are seen 11 different times for four years. There is high follow up rate and a variety of data collected from biological to behavioral. This is one example of a birth cohort study mentioned by Dr. Windham in her presentation that could be used:

Key points from the symposium:

1. The HBC Study has a number of strengths in depicting developmental trajectories.

2. Major limitations include a moderate sample size; an international collaboration with birth cohorts of a similar sort should be sought.

3. Parental age effects, i.e., an increase in fathering and mothering age, appear to be associated with various aspects of delay in child development, including fetal development and emergence of ASD. Investigations of the underlying mechanisms with this regard are under way.    

4. Japan is exceptional among the OECD (Organisation for Economic Co-operation and Development) countries in that the percentage of children with low birth weight has been consistently increasing for the last three decades. It reached 10 percent in 2010. Among the participants of the HBC Study, the proportion of low birth weight children was 11 percent. Careful follow-up of these children should be performed since birth weight is deemed an intermediate marker that reflects a range of biological and environmental challenges for fetal development. In some studies, low birth weight children have been suggested to be at heightened risk for ASD.

For more information, contact Dr. Tsuchiya at


Seek and Ye Shall Find:  Epidemiological Approaches to Developmental Psychopathology

Young-Shin Kim, M.P.H., Ph.D., of Yale University

Our total population-based ASD prevalence study of South Korean children demonstrated that 2.6 percent of all children have an ASD. When looking at already identified cases of autism – those who are recorded in special education records – we found a prevalence of 0.8 percent. This is similar to the 1 percent prevalence seen in the U.S. using a similar approach of records review. But when we went into schools to directly screen for ASD among all school children, we found an additional 1.8 percent of kids who were affected. When combined with the 0.8 percent, it amounted to a prevalence estimate of 2.6 percent. Furthermore, this 1.8 percent of the ASD population had characteristics that were distinct from the already-diagnosed population with regard to ASD severity, IQ and sex ratio. These data highlight the importance of the use of an epidemiologically ascertained, population-based, representative sample in ASD studies, especially since previous ASD studies have missed a major portion of ASD individuals who have distinct characteristics. These data will likely have important implications for the analyses of environmental factors that contribute to ASD.   

Key points from the symposium:

1. Two-thirds of children with autism in South Korea who attended mainstreamed schools were undiagnosed and not receiving autism services including special education, despite their significant functional impairment.  

2. Those who were undiagnosed had differences in characteristics than those with a diagnosis.

3. Therefore, using a total population approach is useful to identifying risk factors associated with ASD.


For more information, contact Dr. Kim at

Addressing Challenges in Multinational Registry Analysis: iCARE

Diana Schendel, Ph.D., of the Centers for Disease Control and Prevention


The International Collaboration for Autism Registry Epidemiology (iCARE) was established as the first multinational consortium to promote and facilitate research on ASD by integrating existing national or statewide, population-based, individual-level data systems. The iCARE approach includes rigorous data harmonization processes and data federation that yields unprecedented autism study sample size and flexible research application capabilities. The resource occupies a unique niche for analyses that require very large sample sizes or comparison of findings across different populations with the aim of enhancing the ability to detect environmental and genetic contributions to the causes and life course of autism.

Key points from the symposium:

Establishing iCARE required solutions to three main challenges:

  • Differences existed among partners in language and culture; different research environments and expectations; administrative differences; work calendars and time zone differences. This was solved by developing a Memorandum of Agreement (available upon request) comprised of guiding principles and processes, such as membership rights and responsibilities, communication, data sharing and authorship.
  • There was diverse data availability and formats over time and by site. The solution to this problem was to harmonize to a common format prior to analyses via a multi-step process.
    • Define minimum dataset of variables of interest and variables to assess data quality.
    • Conduct data surveys by site to gather information on local data sources and variables.
    • Implement data harmonization protocol comprised of specific instructions for variable creation.
    • Quality control measures to ensure accuracy and uniformity in local implementation of the harmonization protocol and data upload.
  • In the countries involved in this project, export of local data to a remote data archive is not permitted. The project overcame this barrier by creating an infrastructure in which each site’s data set is maintained and stored locally; data are federated with other sites’ data electronically during individual analytic sessions.