Maternal Residence Near Agricultural Pesticide Applications and Autism
Maternal Residence Near Agricultural Pesticide Applications and Autism
The objective of this study was to systematically explore the general hypothesis that residential proximity to agricultural pesticide applications during pregnancy could be associated with ASD in offspring. By separately considering the parameters identifying a) compounds and compound groupings, b) spatial proximity, and c) temporal windows, we tested 249 hypotheses that met our predetermined criteria. Application of a priori analytic standards to reduce the statistical problems associated with testing and interpreting this large number of hypotheses led us to dismiss nearly all of these hypotheses. Statistical approaches (Aickin and Gensler 1996; Nichols and Hayasaka 2003) and visual inspection concurred that the association between organochlorine pesticide applications immediately before and during the period of CNS embryogenesis and ASD risk merited further investigation. This association was strongest for residences closest to pesticide applications and was attenuated with increasing distance. A posteriori analysis indicated that the association was strongest, in these data, among those residing near the highest nonzero quartile of pesticide applied during the 8 weeks immediately following cranial neural tube closure. The magnitude of this association was substantially larger than any we could generate through additional testing using alternative time periods and/or quartiles of pesticide applications. We adopted as an a posteriori hypothesis that these 8 weeks reflect the period of actual maximum embryonic vulnerability to the organochlorine pesticides. Findings were insensitive to choices of covariates available for our exploratory model, although the inclusion of sex as a covariate attenuated the association slightly. Within the limits of our data, ASD risk increased monotonically with the amount of organochlorine applications during this a posteriori time period.
Organochlorines include a chemically diverse group of halobenzene-derivative compounds used mostly as insecticides; in the study area, nearly all of the pesticide applied in this class was dicofol or endosulfan, both of which were used on cotton, fruits, vegetables, beans, and nuts.
In general, halobenzene derivatives are metabolized through the cytochrome P450 system in humans (Rietjens et al. 1997). Dicofol is chemically similar to dichlorodiphenyltrichloroethane (DDT), the difference being that dicofol possesses a hydroxy moiety on one of its two aliphatic carbon atoms. Dicofol is not metabolized to dichlorodiphenyldichloroethylene (DDE), is cleared from the body more quickly, and bioaccumulates less than DDT (U.S. Environmental Protection Agency 1998). Following oral dosage in studies with rats and mice, peak serum concentrations are reached within 24–48 hr, with most of the compound cleared from the body within 8 days (EXTOXNET 1996a). Environmentally, dicofol's geographic and temporal distribution follows the patterns of its application because of its relative solubility, generally being detectable in field runoff only during seasons of field applications (Domagalski 1996).
In rats, endosulfan is converted by the liver after oral administration to endosulfan sulfate and endosulfan diol; peak serum concentrations are reached within hours and elimination achieved within days (Agency for Toxic Substances and Disease Registry 2000; Chan and Mohd 2005; Chan et al. 2005). In humans, the diol compound in particular has been detected in both placenta and neonatal cord blood (Cerillo et al. 2005). Less soluble than dicofol, endosulfan breaks down in soil and water over periods of weeks to months (EXTOXNET 1996b).
Generally speaking, the brain has not been highlighted as the primary target organ for the toxicity of either dicofol or endosulfan. The latter compound has been noted to have estrogenic effects as well as some effects on the thyroid gland (Schantz and Widholm 2001; Soto et al. 1994), which may be relevant to concerns about the role of the fetal hormonal milieu in ASD pathogenesis (Baron-Cohen et al. 2005). Sexual differentiation of brain structures in higher mammals occurs during the fetal period (weeks 9–38) (Tsuruo 2005), and aromatase, the enzyme that converts androgens into estrogen, is expressed by nerve cells localized in specific brain structures during this period.
Both dicofol and endosulfan noncompetitively bind gamma amino-butyric acid (GABA) receptor–mediated chloride ion channels in nerve cells (Sunol et al. 1998). GABA-mediated neurotransmission is known to play important roles in gestational brain development, and the theory that altered GABA metabolism could play a role in ASD has been advanced (Cohen 2001). GABA is a neurotransmitter largely unique to interneurons, and GABA-mediated activity regulates cell migration, proliferation, synaptogenesis, and, by extension, the overall patterning of neural networks (Herlenius and Lagercrantz 2004). Mice with abnormal genes for glutamic acid decarboxylase, which is essential for GABA synthesis, develop epilepsy, abnormal neural activity, and increased anxiety-like behavior. Different forms of this enzyme appear sequentially throughout development, which is thought to imply the existence of multiple distinct functions for GABA as a neurotransmitter during different periods (Varju et al. 2001).
Because this is the first study to explore whether risk of ASD is associated with residential exposure to organochlorine pesticides at drift concentrations, our results require replication in further studies and should be treated with caution. In particular, we want to draw the reader's attention to the small numbers of subjects classified as "exposed" under our model that generated the largest magnitude of ASD risk (a 500-m distance between field and residence and the a posteriori temporal window). Using this model, a total of 113 case and control subjects were connected with pesticide applications within the spatial–temporal window, with 29 subjects (8 cases) in the fourth nonzero quartile of exposure.
Among control subjects, the prevalence of our a posteriori–defined exposure was 14.3/1,000 births in the Central Valley region. Assuming a baseline risk for ASD of 6.5/1,000 births (Bertrand et al. 2001; Chakrabarti and Fombonne 2001), the OR of 6.1 suggests a putative population attributable risk on the order of 7% for births to Central Valley residents. This calculation assumes that the relationship between exposure and outcome is causal and considers only exposure to drift from agricultural applications. Associations of ASD with exposure from other sources could not be considered using the present study design.
To our knowledge, neither dicofol nor endosulfan are used in household products or are used in any quantities outside of the commercial agricultural setting. Residues of both compounds are commonly detected in a wide variety of foods, however (Groth et al. 2000), as are those of persistent organochlorine compounds no longer in use as pesticides (Schafer and Kegley 2002). Both compounds have structural similarities with relatively common toxic contaminants such as hexachlorobenzene and polychlorinated biphenyls. Most chlorinated aromatic ring compounds are substantially more persistent in the environment and in human tissues than dicofol and endosulfan. The ability to detect associations between putative pesticide exposure due to agricultural drift and ASD risk in the present analysis may have been augmented by the presence of the compounds in relatively defined spatial and temporal windows.
The availability of pesticide application data in California has provided an opportunity to detect a possible link, but replication of our results and further evaluation in laboratory studies are essential to determine whether these compounds could be etiologically related to the occurrence of ASDs in some children. In this context, it may be relevant to note that the total applied poundage of endosulfan and dicofol decreased in California by approximately one-half between 1998 and 1999 but appears to have remained steady since that time (Pesticide Action Network of North America 2006).
One strength of this study was the ability to locate pesticide applications with relatively high resolution in both space and time, which allowed us to operationalize hypotheses referring to specific temporal periods of vulnerability. We were able to use this and the large number of compounds for which we had data to characterize many associations, identifying those likely to arise through multiple testing and contrasting those with associations that appeared more compelling. We were able to meet many standards set for the epidemiologic study of neurodevelopmental effects of in utero chemical exposure (Amler et al. 2006), particularly with regard to the methodical definition and testing of plausible hypotheses a priori. Another strength is that prior studies have demonstrated good diagnostic validity for children reported by DDS to have autism when electronic statewide records are compared with data in RC records (Grether J, personal communication) or results of standardized evaluations conducted for specific studies (e.g., using the Autism Diagnostic Interview–Revised and the Autism Diagnostic Observation Scale) (Hertz-Picciotto et al. 2006).
Misclassification of exposure is the primary limitation of this study. It has been estimated that one in three women change addresses during pregnancy (Canfield et al. 2006). Furthermore, we were unable to assess time spent at home during the time periods in question or the influence of wind speed and direction on drift. Although it is impossible to assess, it is likely that this misclassification is nonsystematic in nature; further, the specificity of our exposure metric is likely to decrease as larger buffer distances between fields and residences are employed.
Little information was available to us describing the mothers and children in the sample other than basic demographic characteristics, so we were unable to adjust for confounders potentially important to gestational neurodevelopment, such as the use of prenatal vitamins (Shaw et al. 1995). Anecdotal evidence suggests that mothers from a wide variety of socioeconomic backgrounds and occupations were represented in the "exposed" categories, but we cannot dismiss the possibility that these women may be disproportionately employed in agriculture and therefore subject to occupational exposures to pesticides beyond drift concentrations.
ASD is relatively rare, and more mildly affected children may be less represented in our case group. The proportion of mothers in the sample living in proximity to pesticide applications during our specific time periods of interest was small. This limited the numbers of people classified as exposed at any particular level of pesticide compounds. As mentioned above, for example, only 29 subjects were classified in the fourth nonzero quartile of exposure to organochlorines using our a posteriori parameters. Of these, eight subjects had ASD; although this is significantly greater than the expected number (1.8), the need for replication of these findings in other, larger populations is clear.
The objective of this study was to systematically explore the general hypothesis that residential proximity to agricultural pesticide applications during pregnancy could be associated with ASD in offspring. By separately considering the parameters identifying a) compounds and compound groupings, b) spatial proximity, and c) temporal windows, we tested 249 hypotheses that met our predetermined criteria. Application of a priori analytic standards to reduce the statistical problems associated with testing and interpreting this large number of hypotheses led us to dismiss nearly all of these hypotheses. Statistical approaches (Aickin and Gensler 1996; Nichols and Hayasaka 2003) and visual inspection concurred that the association between organochlorine pesticide applications immediately before and during the period of CNS embryogenesis and ASD risk merited further investigation. This association was strongest for residences closest to pesticide applications and was attenuated with increasing distance. A posteriori analysis indicated that the association was strongest, in these data, among those residing near the highest nonzero quartile of pesticide applied during the 8 weeks immediately following cranial neural tube closure. The magnitude of this association was substantially larger than any we could generate through additional testing using alternative time periods and/or quartiles of pesticide applications. We adopted as an a posteriori hypothesis that these 8 weeks reflect the period of actual maximum embryonic vulnerability to the organochlorine pesticides. Findings were insensitive to choices of covariates available for our exploratory model, although the inclusion of sex as a covariate attenuated the association slightly. Within the limits of our data, ASD risk increased monotonically with the amount of organochlorine applications during this a posteriori time period.
Organochlorines include a chemically diverse group of halobenzene-derivative compounds used mostly as insecticides; in the study area, nearly all of the pesticide applied in this class was dicofol or endosulfan, both of which were used on cotton, fruits, vegetables, beans, and nuts.
In general, halobenzene derivatives are metabolized through the cytochrome P450 system in humans (Rietjens et al. 1997). Dicofol is chemically similar to dichlorodiphenyltrichloroethane (DDT), the difference being that dicofol possesses a hydroxy moiety on one of its two aliphatic carbon atoms. Dicofol is not metabolized to dichlorodiphenyldichloroethylene (DDE), is cleared from the body more quickly, and bioaccumulates less than DDT (U.S. Environmental Protection Agency 1998). Following oral dosage in studies with rats and mice, peak serum concentrations are reached within 24–48 hr, with most of the compound cleared from the body within 8 days (EXTOXNET 1996a). Environmentally, dicofol's geographic and temporal distribution follows the patterns of its application because of its relative solubility, generally being detectable in field runoff only during seasons of field applications (Domagalski 1996).
In rats, endosulfan is converted by the liver after oral administration to endosulfan sulfate and endosulfan diol; peak serum concentrations are reached within hours and elimination achieved within days (Agency for Toxic Substances and Disease Registry 2000; Chan and Mohd 2005; Chan et al. 2005). In humans, the diol compound in particular has been detected in both placenta and neonatal cord blood (Cerillo et al. 2005). Less soluble than dicofol, endosulfan breaks down in soil and water over periods of weeks to months (EXTOXNET 1996b).
Generally speaking, the brain has not been highlighted as the primary target organ for the toxicity of either dicofol or endosulfan. The latter compound has been noted to have estrogenic effects as well as some effects on the thyroid gland (Schantz and Widholm 2001; Soto et al. 1994), which may be relevant to concerns about the role of the fetal hormonal milieu in ASD pathogenesis (Baron-Cohen et al. 2005). Sexual differentiation of brain structures in higher mammals occurs during the fetal period (weeks 9–38) (Tsuruo 2005), and aromatase, the enzyme that converts androgens into estrogen, is expressed by nerve cells localized in specific brain structures during this period.
Both dicofol and endosulfan noncompetitively bind gamma amino-butyric acid (GABA) receptor–mediated chloride ion channels in nerve cells (Sunol et al. 1998). GABA-mediated neurotransmission is known to play important roles in gestational brain development, and the theory that altered GABA metabolism could play a role in ASD has been advanced (Cohen 2001). GABA is a neurotransmitter largely unique to interneurons, and GABA-mediated activity regulates cell migration, proliferation, synaptogenesis, and, by extension, the overall patterning of neural networks (Herlenius and Lagercrantz 2004). Mice with abnormal genes for glutamic acid decarboxylase, which is essential for GABA synthesis, develop epilepsy, abnormal neural activity, and increased anxiety-like behavior. Different forms of this enzyme appear sequentially throughout development, which is thought to imply the existence of multiple distinct functions for GABA as a neurotransmitter during different periods (Varju et al. 2001).
Because this is the first study to explore whether risk of ASD is associated with residential exposure to organochlorine pesticides at drift concentrations, our results require replication in further studies and should be treated with caution. In particular, we want to draw the reader's attention to the small numbers of subjects classified as "exposed" under our model that generated the largest magnitude of ASD risk (a 500-m distance between field and residence and the a posteriori temporal window). Using this model, a total of 113 case and control subjects were connected with pesticide applications within the spatial–temporal window, with 29 subjects (8 cases) in the fourth nonzero quartile of exposure.
Among control subjects, the prevalence of our a posteriori–defined exposure was 14.3/1,000 births in the Central Valley region. Assuming a baseline risk for ASD of 6.5/1,000 births (Bertrand et al. 2001; Chakrabarti and Fombonne 2001), the OR of 6.1 suggests a putative population attributable risk on the order of 7% for births to Central Valley residents. This calculation assumes that the relationship between exposure and outcome is causal and considers only exposure to drift from agricultural applications. Associations of ASD with exposure from other sources could not be considered using the present study design.
To our knowledge, neither dicofol nor endosulfan are used in household products or are used in any quantities outside of the commercial agricultural setting. Residues of both compounds are commonly detected in a wide variety of foods, however (Groth et al. 2000), as are those of persistent organochlorine compounds no longer in use as pesticides (Schafer and Kegley 2002). Both compounds have structural similarities with relatively common toxic contaminants such as hexachlorobenzene and polychlorinated biphenyls. Most chlorinated aromatic ring compounds are substantially more persistent in the environment and in human tissues than dicofol and endosulfan. The ability to detect associations between putative pesticide exposure due to agricultural drift and ASD risk in the present analysis may have been augmented by the presence of the compounds in relatively defined spatial and temporal windows.
The availability of pesticide application data in California has provided an opportunity to detect a possible link, but replication of our results and further evaluation in laboratory studies are essential to determine whether these compounds could be etiologically related to the occurrence of ASDs in some children. In this context, it may be relevant to note that the total applied poundage of endosulfan and dicofol decreased in California by approximately one-half between 1998 and 1999 but appears to have remained steady since that time (Pesticide Action Network of North America 2006).
One strength of this study was the ability to locate pesticide applications with relatively high resolution in both space and time, which allowed us to operationalize hypotheses referring to specific temporal periods of vulnerability. We were able to use this and the large number of compounds for which we had data to characterize many associations, identifying those likely to arise through multiple testing and contrasting those with associations that appeared more compelling. We were able to meet many standards set for the epidemiologic study of neurodevelopmental effects of in utero chemical exposure (Amler et al. 2006), particularly with regard to the methodical definition and testing of plausible hypotheses a priori. Another strength is that prior studies have demonstrated good diagnostic validity for children reported by DDS to have autism when electronic statewide records are compared with data in RC records (Grether J, personal communication) or results of standardized evaluations conducted for specific studies (e.g., using the Autism Diagnostic Interview–Revised and the Autism Diagnostic Observation Scale) (Hertz-Picciotto et al. 2006).
Misclassification of exposure is the primary limitation of this study. It has been estimated that one in three women change addresses during pregnancy (Canfield et al. 2006). Furthermore, we were unable to assess time spent at home during the time periods in question or the influence of wind speed and direction on drift. Although it is impossible to assess, it is likely that this misclassification is nonsystematic in nature; further, the specificity of our exposure metric is likely to decrease as larger buffer distances between fields and residences are employed.
Little information was available to us describing the mothers and children in the sample other than basic demographic characteristics, so we were unable to adjust for confounders potentially important to gestational neurodevelopment, such as the use of prenatal vitamins (Shaw et al. 1995). Anecdotal evidence suggests that mothers from a wide variety of socioeconomic backgrounds and occupations were represented in the "exposed" categories, but we cannot dismiss the possibility that these women may be disproportionately employed in agriculture and therefore subject to occupational exposures to pesticides beyond drift concentrations.
ASD is relatively rare, and more mildly affected children may be less represented in our case group. The proportion of mothers in the sample living in proximity to pesticide applications during our specific time periods of interest was small. This limited the numbers of people classified as exposed at any particular level of pesticide compounds. As mentioned above, for example, only 29 subjects were classified in the fourth nonzero quartile of exposure to organochlorines using our a posteriori parameters. Of these, eight subjects had ASD; although this is significantly greater than the expected number (1.8), the need for replication of these findings in other, larger populations is clear.
Source...