				

UNITED STATES ENVIRONMENTAL PROTECTION AGENCY

WASHINGTON, D.C.  20460

OFFICE OF

PREVENTION, PESTICIDES AND

TOXIC SUBSTANCES



MEMORANDUM

October 18, 2007

SUBJECT:	A Review of “Maternal Residence near Agricultural Pesticide
Applications and Autism Spectrum Disorders among Children in the
California Central Valley” Roberts et al., 2007 [EHP, 115; 10: 
1482-1489]  (Endosulfan (PC code 079401) and Dicofol (PC code 010501))

DP Barcode D342660	

FROM:	Carol Christensen, MPH

		Chemistry and Exposure Branch

				Health Effects Division (7509P)

THRU:	David J. Miller, Chief	

		Chemistry and Exposure Branch

				Health Effects Division (7509P)

TO:		Tracey Perry, Chemical Review Manager

		Special Review and Reregistration Division (7508P)

		In the study “Maternal Residence near Agricultural Pesticide
Applications and Autism Spectrum Disorders among Children in the
California Central Valley” by Roberts et al. (2007) [Environmental
Health Perspectives, vol. 115, Issue 10, pp.1482], the authors
investigated the association between maternal pesticide exposure and
development of autism spectrum disorder (ASD) in offspring.  This
memorandum provides a review of the study for use in the endosulfan and
dicofol Reregistration Eligibility Documents (RED).  The pesticide
registrants Makhteshim-Agan of North America (MANA) also submitted a
review of this study (8/20/07).  The Agency acknowledges receipt of this
review and notes in this review many of the same points made by MANA. 
The study authors and HED acknowledge this is a hypothesis-generating
study and would not attempt causal inference based upon this study
alone. 

	Background

	At this time, little is known about the association between pesticides
and pediatric neurological development.  Therefore, the authors studied
the hypothesis that infants born to women living in close proximity to
pesticide treated agricultural fields experience greater risk of
developing autism spectrum disorder (ASD) than infants born to women who
live further away from agricultural fields.  Many agree that ASD is a
functional deficit caused by an alternation to specific brain structures
during certain temporal windows during in utero development. 
Researchers also agree that while there is a strong genetic component
suggested by previous epidemiologic investigations (studies in twins and
siblings), the environment also likely plays a role; gene by environment
interactions in the development of ASD are strongly implicated. 
Although the specific causal “pathway to autism” is unknown, the
authors posit that pesticides which produce neurologic effects in animal
studies may play a role in the development of ASD.  Many studies have
shown pesticides can “drift” from the site of application and are
detected in nearby residential homes.  The study was conducted as part
of a demonstration project of the utility of environmental public health
tracking, an initiative funded by the CDC to generate clues for further
etiologic study into environmental causes of disease.

Study Design and Results

Using publicly available datasets, the authors assessed both pesticide
exposure and an ASD outcome.  Pesticide exposure was determined through
use of California Department of Pesticide Regulation (DPR) pesticide use
records correlated with maternal residence recorded on the birth
certificate.  Geocoding these two types of variables allowed the authors
to ascertain the distance between maternal residence and agricultural
fields for analysis.  ASD outcome was ascertained through the state’s
Department of Developmental Services (DDS) which collects information
concerning ASD cases to provide social services to families with ASD
affected children.  Families are referred to the DDS through doctors,
educators, family or friends.  According to the authors, the DDS is used
widely across various racial and economic groups in the population.  To
be eligible for DDS services, children must meet the criteria for ASD or
other developmental disorders as administered by the state.  The authors
did not limit the age range for eligibility into the study.  Using DDS
records of ASD diagnosis and birth certificate data, the authors
identified 465 ASD cases from 19 counties in Central Valley of southern
California who met the study criteria.  Authors excluded infant deaths,
multiple births, premature infants (less than 37 weeks) and low birth
weight infants (less than 2500 grams).  Controls (n=6975) were frequency
matched to cases (15:1) by gestational age using incidence-density
sampling procedures.

	The authors’ goal was to assess whether in utero exposure to
pesticides (measured as proximity of maternal residence to agricultural
fields) is associated with the probability (or odds) of developing ASD
in childhood.  Numerous specific hypothesis were generated by combining
information concerning potential developmental or reproductive toxicity
or neurotoxicity of specific pesticides (n=54); maternal residential
distance from agricultural fields (250, 500, 750 and 1,000 meters); and,
three critical windows of gestational development (central nervous
system development (CNS), neural tube closure and entire gestational
period) during which exposure to pesticides may be a putative risk
factor for ASD.  Applying the decision criteria of at least 5 cases or
control subjects per quartile of exposure (distance from field) to the
total number of hypotheses generated by combining the three data sources
listed above, the authors analyzed a total of 249 separate hypotheses. 
The authors employed published statistical methods to adjust the
significance level (i.e., p-value) to control for multiple comparisons. 
Conditional logistic regression was utilized to measure the association
for each separate hypothesis, adjusting for race/ethnicity, maternal
education, and DDS receiving center.  All variables were modeled as
categorical; no linearity assumption was applied.  Upon analysis of
these 249 a priori hypotheses, eight hypotheses were identified as
significant using both the unadjusted and multiple-comparison adjusted
p-values.  Four of the eight hypotheses relate to organochlorine
pesticides at each exposure level (250m, 500m, 750m, and 1,000 m) during
both the gestational period of CNS development and the entire
gestational period.  Only the four organochlorine pesticide hypotheses
noted above met the criteria of at least one case for each quartile of
exposure at the selected radius of 500m from the treated field for a
posteriori analysis of varying exposure time-windows.  To refine the
critical period of gestational development, the a posteriori analysis
employed an 8-week moving time-window for the period 300 days prior to
and 300 days post-conception to identify the optimal model (the
gestational time period during which pesticide exposure was most closely
associated with ASD).  As a result, the time period of 26-81 days
post-conception was identified as the most informative exposure window. 
In the study area (19 counties in CA Central Valley), the
organochlorines endosulfan and dicofol accounted for nearly all of the
OC use (>98%).  It should be noted there were only 88 subjects exposed
to dicofol and 27 exposed to endosulfan at the distance 500m from the
treated field during this gestational time window.  Odds ratios are only
presented for the OC class and not these individual chemicals due to low
numbers of exposed cases and controls.  Additional sensitivity analyses
and varying statistical models did not substantially change the
conclusion.  The results of this analysis reveals that in utero exposure
to OCs (as measured by maternal residence distance from the treated
field of 500m) is associated with a 6-fold increase in risk of ASD
[OR=6.1 (95% C.I. 2.4-15.3)] in childhood when compared to those
children who did not experience in utero exposure to OC (as measured by
maternal residence distance from treated field of greater than 1,000m). 
The most commonly applied organochlorines in this study population are
endosulfan and dicofol.   The authors state that this is the first study
to explore whether the risk of ASD is associated with residential
exposure at drift concentrations and the “results require replication
in further studies and should be treated with caution.” Overall, this
hypothesis-generating study suggests, according to the authors, “the
possibility of a connection between gestational exposure to
organochlorine pesticides and ASD and requires further study.”

	

	Review

	Strengths of this study include novel use of publicly available
datasets to investigate a new hypothesis; the large size of the study
(465 cases and 6975 controls) allowed for investigation of multiple
hypotheses inexpensively; the applicability of the case-control study
design to a study of a rare disease such as ASD; the utility of the
study design for evaluating numerous hypotheses within one study; and,
the sophisticated use of multiple-comparison control methods to avoid
Type I errors.  This is a hypothesis generating study.  This work
provides leads which should be further investigated in subsequent
etiologic studies of OC’s and endosulfan and dicofol.  The study is a
necessary first step to investigating the role of environmental
chemicals, pesticides, and ASD.  

	Limitations of this study include possibility of selection bias and
misclassification of both the exposure and outcome; the authors
acknowledge these limitations.  If participation in the DDS varies by
exposure-status (proximity to treated field) possibly due to differences
in race, income or other SES factor, selection bias may be induced.  If
present, selection bias may affect the validity of the study results. 
While there is no way to directly assess this potential, the authors
state the DDS participants are diverse in race-ethnicity and income
level.  In addition, while there are standard measures now available for
diagnosing ASD, application of those measures may vary by state DDS
investigator, referring MD or educational personnel.  The authors note
that more mild cases of ASD may be missed in this study.  ASD is a
complex, multi-faceted condition with no clear indicator of disease
status such as a genetic marker, therefore, outcome misclassification
may occur.  Misclassification of exposure level (pesticide exposure) may
occur based upon the indirect measure of exposure utilized, proximity of
maternal residence to treated field.  Pesticide exposure as a result of
use of products in the home environment or in occupational settings is
not directly measured in this study.  Errors in maternal residence
records or migration of participants between time of conception, time of
birth and time of diagnosis may also adversely affect exposure
classification.  The misclassification should be non-differential
leading to an attenuation of the risk estimate.  The study authors did
not collect all potentially confounding variables such as prenatal
vitamin intake or level of prenatal care which may affect ASD outcome.  

	However, this study indicates additional etiologic study of the
association between pesticide exposure and the organochlorines
endosulfan and dicofol specifically and development of ASD warrants
further investigation.  Indeed, the CHARGE (Childhood Autism Risks from
Genetics and Environment) study was initiated to investigate such
hypotheses.  Pesticides are among the environmental exposures to be
evaluated in this population-based cohort of children in California
(Hertz-Picciotto, 2006).  

	Recommendation/Next Steps

	As a hypothesis-generating study, causal inference is not possible
based upon these results.  To determine whether an exposure is causally
related to a disease, it is generally agreed that numerous criteria must
be met.  For this hypothesis, some of the criteria are now met, but not
all.  Evidence suggestive of a potential causal association includes the
biological plausibility that organochlorines may affect the developing
fetal brain given prior evidence of the developmental toxicity of OCs
and their close structural similarity to other development toxicants
such as DDT.  In this study, a dose-response trend was observed and
temporality is established.  Another factor suggestive of a potential
causal association includes the strong magnitude of the association
measured, although it is based on only a small number of cases.  Factors
which must still be elucidated to make a causal inference concerning the
association between OC exposure and ASD include the need to repeat these
results in other etiologic studies, further evaluation of the mechanisms
of ASD development and how OC chemicals may be involved, and refutation
of other alternative explanations for the association between pesticide
exposure and ASD such as other occupational exposures, quality of
prenatal care or other high-penetrance genetic factors.  More work is
needed in this area before causal inference can be attempted.

	HED will follow the work of the CHARGE study and other etiologic
investigations of the association between OC exposure and autism
spectrum disorder.  

	

cc:  	Dr. Elissa Reaves, Reregistration Branch II

	Health Effects Division (7509P)

References

Robert, E.M., English, P.B., Grether, J.K., Windham, G.C., Somberg, L.,
and C. Wolff. (2007)  “Maternal Residence Near Agricultural Pesticide
Applications and Autism Spectrum Disorders among Children in the
California Central Valley.”  Environmental Health Perspectives 115;
10:1482-1489.

Hertz-Picciotta, I., Croen, L.A., Hansen, R., Jones, C.R., van de Water,
J., and I.N. Pessah.  (2006)  “The CHARGE Study:  An Epidemiologic
Investigation of Genetic and Environmental Factors Contributing to
Autism.”  Environmental Health Perspectives, 114;7:1119-1125.

	

	

 PAGE   

  PAGE   1 

	

  PAGE   

	

  PAGE   6 

