A comparison of temporal trends in United States autism prevalence to trends in suspected environmental factors

Abstract

Background: The prevalence of diagnosed autism has increased rapidly over the last several decades among U.S. children. Environmental factors are thought to be driving this increase and a list of the top ten suspected environmental toxins was published recently.

Methods: Temporal trends in autism for birth years 1970-2005 were derived from a combination of data from the California Department of Developmental Services (CDDS) and the United States Individuals with Disabilities Education Act (IDEA). Temporal trends in suspected toxins were derived from data compiled during an extensive literature survey. Toxin and autism trends were compared by visual inspection and computed correlation coefficients. Using IDEA data, autism prevalence vs. birth year trends were calculated independently from snapshots of data from the most recent annual report, and by tracking prevalence at a constant age over many years of reports. The ratio of the snapshot:tracking trend slopes was used to estimate the "real" fraction of the increase in autism.

Results: The CDDS and IDEA data sets are qualitatively consistent in suggesting a strong increase in autism prevalence over recent decades. The quantitative comparison of IDEA snapshot and constant-age tracking trend slopes suggests that ~75-80% of the tracked increase in autism since 1988 is due to an actual increase in the disorder rather than to changing diagnostic criteria. Most of the suspected environmental toxins examined have flat or decreasing temporal trends that correlate poorly to the rise in autism. Some, including lead, organochlorine pesticides and vehicular emissions, have strongly decreasing trends. Among the suspected toxins surveyed, polybrominated diphenyl ethers, aluminum adjuvants, and the herbicide glyphosate have increasing trends that correlate positively to the rise in autism.

Conclusions: Diagnosed autism prevalence has risen dramatically in the U.S over the last several decades and continued to trend upward as of birth year 2005. The increase is mainly real and has occurred mostly since the late 1980s. In contrast, children's exposure to most of the top ten toxic compounds has remained flat or decreased over this same time frame. Environmental factors with increasing temporal trends can help suggest hypotheses for drivers of autism that merit further investigation.

Figure 1

Autism prevalence vs. birth year for California IDEA data, derived using two independent methods: 1) Constant-age tracking of 9 year-olds over 20 years of annual reports from 1991–2010 (red) and 2) Age-resolved snapshot from the most recent report in 2010 (blue). The slope of each curve over the same birth year interval, 1993–2001, is estimated with a least squares linear fit. The snapshot fit (grey) spans ages 9–17 in the 2010 report. The constant-age tracking fit spans report years 2002–2010. The snapshot:tracking slope ratio over the 1993–2001 birth year interval is 0.80.

Figure 2

Temporal trends in CDDS autistic disorder (blue) and California IDEA autism data (red). For both CDDS and IDEA, age-resolved snapshots for 2002 (squares) and constant-age tracking data for 5 year-olds (triangles) are shown. Trends slopes (symbol b, in units of autism count per 10⁴ per year) are approximated using least squares linear regression over birth year intervals 1988–1995 (for 2002 snapshots) and 1995–2001 (for 5 year-old tracking).

Figure 3

Temporal trend in blood lead (Pb) concentration in U.S. children compared to the temporal trend in autism, constructed from a composite of CDDS 2002 snapshot data (covering birth years 1970–1997) and California IDEA 5 year-old tracking data for birth years 1995–2005 (see Section Method s for details).

Figure 4

Temporal trend in autism compared to trends in highway sector emissions of direct PM2.5 and indirect PM2.5 and ozone precursor species. Emissions are normalized to the 1970 values for CO (=163 Mtons), NOx (=13 Mtons), and VOCs (=17 Mtons) and to the 1990 values for SO2 (=0.5 Mtons), and direct PM2.5 (=0.3 Mtons).

Figure 5

Temporal trend in autism compared to temporal trend in cumulative amount of postnatal aluminum adjuvant administered to U.S. children by 18 months of age. Red circles are years with published immunization schedules. Red triangles reflect educated guesses (see Additional file 1 ) for details) in gap years without published schedules. The red curve is lagged 1 year because 18 month-olds born, e.g., in 1994 will likely follow the 1995 schedule.

Figure 6

Temporal trend in autism compared to temporal trend in U.S. application of glyphosate to genetically-modified corn and soy crops, as estimated from US Department of Agriculture data (see Additional file 1 ).