Professionally and non-professionally applied household insecticides during pregnancy and early life and their associations with autism spectrum disorder and developmental delay in the CHARGE case-control study

Abstract

Background:

Autism Spectrum Disorder (ASD) is an increasing concern globally, with risks attributed to both genetic and environmental factors, including pesticide exposures. The CHARGE case-control study collected data to examine the relationship between household insecticide use and ASD or developmental delay (DD).

Methods:

Participants (n=1,526) aged 2 to 5 years from the CHARGE study encompassed children with clinically confirmed ASD (n = 810), DD (n = 186), and typically developing controls (n = 530) frequency matched to ASD children by age, sex, and geographic region. Household insecticide use was determined from parent interviews, as were the timing, frequency, and type of applications (professionally or non-professionally applied; indoor, outdoor, and flea applications) from three months pre-conception to the child’s second birthday. Logistic regression models were adjusted for confounders to estimate odds ratios (OR) for ASD or DD associated with insecticide exposure.

Results:

Professionally applied indoor insecticides were associated with greater than two-fold increased odds of ASD for all time periods (OR: pre-pregnancy, 2.62 (1.26, 5.44); pregnancy, 2.52 (1.41, 4.48); year 1, 2.20 (1.33, 3.64); year 2, 2.13 (1.29, 3.49)). Odds were also elevated for any outdoor application during years 1 and 2 (OR: 1.37 (1.05, 1.79) and 1.34 (1.03, 1.74), respectively), however, significance did not hold after false discovery rate correction. Higher exposure frequency was associated with greater odds of ASD for nearly all application types and time periods. Non-professional indoor insecticide use was associated with increased odds of DD in trimester 3 and the pregnancy period (OR: 1.72 (1.03, 2.89) and 1.58 (1.03, 2.40), respectively).

Conclusion:

Professionally applied indoor insecticides were consistently associated with ASD. The strong trend of increased odds with more frequent exposures, additional evidence for DD, and widespread use of household insecticides support increased regulatory scrutiny and public health interventions to minimize exposure during critical developmental windows.

1. Introduction

Autism spectrum disorder (ASD), a complex neurodevelopmental condition, presents with a spectrum of symptoms including, but not limited to, impairments in social reciprocity and communication skills, repetitive behaviors, restricted interests, and sensory sensitivities.¹ For over three decades, epidemiological data have documented an increasing prevalence of ASD, which currently affects approximately 1% of the global population. The prevalence is notably higher in the United States, estimated at 1 in 36 children.² The etiology of ASD is multifaceted, with evidence pointing towards both a genetic and environmental component.^3–7

Household insecticide exposure represents a critical area of investigation within the context of neurodevelopmental disorders. Insecticides are a large and heterogeneous group of chemicals, which, by their very nature, are designed to disrupt the neurological systems of pests. Up until the early 2000s organophosphorus (OP) compounds were the more prominent active ingredients found in domestic insecticide formulations for home use. Notably, OP insecticides such as chlorpyrifos and diazinon were extensively employed in residential settings before their phased elimination from household-approved products by the United States Environmental Protection Agency (EPA) in 2001 for chlorpyrifos and in 2002 and 2004 for indoor and outdoor formulations of diazinon, respectively.^8,9 Due to the phase out of OPs, pyrethroids have emerged as the principal active compounds.¹⁰

Evidence of associations between both OP and pyrethroid insecticides and neurodevelopmental outcomes, including ASD, has been documented in numerous epidemiological studies.^11–21 However, the majority of these investigations have concentrated on agricultural insecticide applications or biomarkers measured at specific time points during pregnancy or early childhood.^14,22–24 While biomarker-based studies, including those measuring urinary metabolites of pyrethroids like 3-PBA, offer several strengths, including the avoidance of recall bias and more precise exposure quantification, they are not without limitations. Important research gaps remain due to the inherent limitations of biomarker measurements, particularly their inability to fully capture chronic or long-term exposure. The urinary elimination half-life of 3-PBA, for example, is as short as 5.7 hours,²⁵ reflecting only short-term exposure, often within 24 to 48 hours prior to sample collection. This brief detection window can obscure the effects of intermittent but repeated exposures over time, which are critical for understanding the cumulative impact of pesticide exposure during sensitive developmental windows.

Furthermore, reliance on biomarkers at single time points can result in exposure misclassification, as such measurements might not represent an individual’s overall exposure burden across pregnancy or early childhood. This is particularly pertinent in the context of neurodevelopmental disorders like ASD, where exposure to neurotoxic agents during specific developmental stages (e.g., critical periods of brain growth and differentiation) is of primary concern. In contrast, repeated and detailed assessments of exposure timing and frequency, as performed in this study, offer a more comprehensive view of potential exposure patterns.

While the link between OP and neurodevelopment is well established, the relationship between pyrethroid exposure and neurodevelopment remains less clearly defined. In this study, we utilize extensive household insecticide use self-report data from the CHARGE (Childhood Autism Risk from Genetics and the Environment) study, which includes timing and frequency of use of insecticides from three months before conception through the first two years of a child’s life to examine the role of these insecticides with subsequent ASD or developmental delay (DD) in the child.

2. Materials and Methods

2.1. Study Population

The CHARGE Study was launched in 2003 as a population-based case-control study designed to address environmental influences on autism and other neurodevelopmental impairments, along with gene-by-environment interactions.²⁶ In brief, CHARGE enrolled children who received services for ASD or developmental delay without ASD (DD) through Regional Centers (RC), entities that contract with the California Department of Developmental Services (DDS). General population (GP) controls were concurrently identified from California State Vital Statistics birth files and were frequency-matched to the ASD cases by age, sex assigned at birth, and broad residential RC catchment area distribution. Inclusion criteria for enrollment in the CHARGE study stipulated that children must have been born in the state of California, be within the age range of 24 to 60 months at the time of recruitment, and were currently residing with at least one biologically related parent who was proficient in either English or Spanish, within the defined study catchment area. This study was approved by institutional review boards for the University of California. Written informed consent was obtained by the parent or guardian before collection of any data.

2.2. Diagnostic and outcome measurements

Upon enrollment, a comprehensive assessment of all participants was conducted at the UC Davis MIND Institute (Sacramento, CA) or the UCLA Neuropsychiatric Institute. Both the Autism Diagnostic Observation Schedules Second Edition (ADOS-2)^27,28 and the Autism Diagnostic Interview–Revised (ADI-R)^29–31 were used to confirm ASD diagnoses in children with a previous ASD diagnosis or clinician suspicion of ASD. We used the Social Communication Questionnaire (SCQ)³² (a shortened version of the ADI-R questionnaire), with a cutoff score of 15 or above,^33,34 in conjunction with clinical judgement, to screen all other enrolled children for evidence of ASD symptoms. Based on these criteria, the ADI-R and ADOS-2 were administered on a second visit, to confirm ASD diagnosis. Finally, all children underwent a thorough clinical evaluation of cognitive functioning using the Mullen Scales of Early Learning (MSEL)³⁵ and their adaptive functioning was assessed using the Vineland Adaptive Behavior Scales (VABS).³⁶ Children classified as DD came from either the DDS or GP recruitment pools, did not meet criteria for ASD, and scored below 70 on either the MSEL or VABS. Typically Developing (TD) classification was assigned to those children drawn from the GP that did not meet criteria for ASD or DD and scored at or above 70 on both MSEL and VABS. Children with DD were excluded from these analyses if they were considered syndromic, defined as having a known genetic or other cause for their condition.

2.3. Household Insecticide Exposure

During a comprehensive telephone interview, a parent (typically the mother) was asked about the use of specific insecticide containing products, by them, others in the household, or professional applicators, during the period spanning from 3 months prior to conception with the index child up to the interview date. An environmental exposure questionnaire asked about flea or tick soaps, shampoos, sprays, dusts, powders, skin applications, or collars on pets; professional pest control or extermination; insect control products. Further questions addressed product type (e.g., spray, bait, fogger); brand name; whether the application was indoors, outdoors, or on a pet; who applied the product, and use of professional pest control services. The wording of the insecticide use questions was changed in 2011, after n = 1,036 were interviewed in the original form; subsequently n = 490 were interviewed in the new form. Details on the changes in wording can be found in the Supplement. To assess if the wording change in the questionnaire may have influenced the findings, we stratified all models by the form that was used (Supplemental Figures S1–S3). For these analyses, in order to avoid any influence by pesticide ingredient changes (see sensitivity analyses discussed in Section 2.4), we excluded those that would have likely been exposed to OPs by restricting to those with a year of conception equal to or greater than 2003 for professionally applied indoor, 2005 for non-professionally applied indoor and professionally applied outdoor, and 2007 for non-professionally applied outdoor insecticides. We did find that there were differences in effect estimates by form, however, adjusting for form in models did not change effect estimates.

We created the following categories for type of exposure: insecticide professionally applied indoors, insecticide non-professionally applied indoors, any insecticide applied indoors, insecticide non-professionally applied outdoors, any insecticide applied outdoors, and pet treatments in the form of flea collar use, flea skin application, and a combined variable for flea soap, shampoo, and powder use. For each insecticide type, we also asked about timing and frequency of exposure. For pre-pregnancy and each trimester, we created three frequency categories for each use: no exposure, exposed for 1 or 2 months, exposed during all 3 months. Likewise, for the whole pregnancy period, the three frequencies were: no exposure, exposed for 1–5 months, exposed for 6–9 months.

Text variables for brands used were manually cleaned and cross-referenced with the EPA’s Insecticide Product and Label System (https://ordspub.epa.gov/ords/insecticides/f?p=PPLS:1) to ensure the product contained a known insecticide. Use of borax-containing poisoned bait containers was not included because they have a small surface area of insecticide, which would result in low volatilization and thus limited exposure.

2.4. Statistical Analyses

Exposure and covariate distributions were assessed using univariate analyses. Wilcoxon rank sum and Pearson’s chi-squared tests were used to compare distributions of continuous and categorical demographic variables, respectively, between diagnostic outcomes (ASD or DD vs TD). Data were collected from in-clinic forms, parent-reported questionnaires, and a telephone-administered exposure questionnaire or study interview. All variables listed in Table 1 were collected or derived from the study interview questions. Demographic and pregnancy-related data were cross-referenced with existing sources in order to identify discrepancies that led to deeper cleaning. Demographic data, including sex, race/ethnicity, maternal education level and maternal place of birth, was validated with child’s birth records, prioritizing data from the study interview over the child’s birth records. Pregnancy-related data such as gestational age, conception age, year of conception, and parity were validated with prenatal medical records, and secondarily with child’s birth records. Data on mental health conditions, financial hardship, and living in single-family home was derived from the study interview questions. The urban area of the child’s birth address was derived from parent-reported addresses at birth and Census data.

Table 1.

Subject Characteristics by CHARGE Study Diagnostic Group

	TD (n=530)	ASD (n=810)	p-valueA	DD (n=186)	p-valueB
Male Sex (%)C	431 (81.3%)	671 (83.5%)	0.31	131 (70.4%)	<0.01
Child’s Race/Ethnicity
Non-Hispanic White	271 (51.1%)	380 (46.9%)	0.31	62 (33.3%)	<0.01
Hispanic	147 (27.7%)	249 (30.7%)		89 (47.8%)
Other	112 (21.1%)	181 (22.3%)		35 (18.8%)
Year of ConceptionD	2004 (2002, 2007)	2003 (2000, 2009)	0.02	2004 (2002, 2007)	0.25
Season of ConceptionD
Nov – Feb	165 (31.7%)	265 (33.5%)	0.75	60 (32.8%)	0.82
Mar – Jun	184 (35.4%)	266 (33.6%)		60 (32.8%)
Jul – Oct	171 (32.9%)	260 (32.9%)		63 (34.4%)
Mother’s Age at Conception (yrs)E	30.0 (26.5, 34.0)	30.0 (26.0, 34.0)	0.56	28.0 (24.0, 33.0)	<0.01
≥ 35 yrs (% yes)	104 (20.0%)	170 (21.5%)	0.52	35 (19.3%)	0.85
Father’s Age at Conception (yrs)F	32.0 (28.0, 36.0)	32.0 (28.0, 37.0)	0.82	32.0 (26.0, 37.0)	0.13
≥ 35 yrs (% yes)	178 (34.9%)	261 (34.3%)	0.84	60 (34.9%)	>0.99
Maternal Pre-pregnancy BMIG	24.2 (21.6, 28.2)	24.5 (21.8, 29.2)	0.08	25.4 (22.0, 29.7)	0.01
Underweight	15 (2.9%)	30 (3.8%)	0.048	4 (2.2%)	0.09
Normal	284 (54.1%)	404 (50.8%)		80 (44.2%)
Overweight	137 (26.1%)	181 (22.8%)		55 (30.4%)
Obese	89 (17.0%)	180 (22.6%)		42 (23.2%)
Gestational Age < 37 weeks (% yes)D	48 (9.2%)	83 (10.5%)	0.46	31 (16.9%)	<0.01
ParityH
1	210 (39.8%)	396 (49.4%)	<0.01	67 (36.0%)	0.09
2	203 (38.5%)	280 (35.0%)		64 (34.4%)
3+	114 (21.6%)	125 (15.6%)		55 (29.6%)
Gestational Diabetes (% yes)I	30 (5.7%)	76 (9.5%)	0.01	19 (10.3%)	0.03
Had any mental health condition before or during pregnancy (% yes)J	101 (22.3%)	244 (42.7%)	<0.01	52 (33.3%)	<0.01
Mother Born in US (% yes)K	440 (83.2%)	600 (74.6%)	<0.01	135 (72.6%)	<0.01
Private Health InsuranceL	442 (84.7%)	624 (78.8%)	<0.01	111 (60.3%)	<0.01
Home Ownership (% yes)M	395 (75.1%)	507 (63.5%)	<0.01	98 (53.6%)	<0.01
Financial Hardship (%yes)N	80 (15.2%)	167 (20.9%)	<0.01	51 (27.9%)	<0.01
Mother’s EducationC
Some College	69 (13.0%)	121 (15.0%)	0.01	61 (32.8%)	<0.01
AA or Technical Degree	176 (33.2%)	325 (40.4%)		71 (38.2%)
Bachelor’s Degree	197 (37.2%)	240 (29.9%)		42 (22.6%)
Graduate Degree or higher	88 (16.6%)	118 (14.7%)		12 (6.5%)
Lived in single family home
PrepregnancyO	381 (75.7%)	501 (66.7%)	<0.01	115 (68.0%)	0.049
Entire PregnancyP	379 (73.9%)	483 (63.5%)	<0.01	105 (60.3%)	<0.01
Year 1Q	395 (76.8%)	501 (65.7%)	<0.01	104 (59.8%)	<0.01
Year 2R	411 (80.1%)	525 (69.1%)	<0.01	109 (63.0%)	<0.01
Child’s birth address in urban areaS	468 (88.8%)	724 (91.9%)	0.06	166 (91.2%)	0.36

Values are N (%) or median (IQR). Percentages are calculated among those with non-missing data. P-values are from Pearson chi-square tests (categorical variables) or Wilcoxon Rank Sum test (numeric variables) comparing

^A.ASD to TD and

^B.DD to TD

^C.Missing n = 6 ASD;

^D.Missing n = 10 TD, 19 ASD, 3 DD;

^E.Missing n = 10 TD, 19 ASD, 5 DD;

^F.Missing n = 20 TD, 50 ASD, 14 DD;

^G.Missing n = 5 TD, 15 ASD, 5 DD;

^H.Missing n = 3 TD, 9 ASD;

^I.Missing n = 3 TD, 9 ASD, 2 DD;

^J.Missing n = 78 TD, 238 ASD, 30 DD;

^K.Missing n = 1 TD, 6 ASD, 0 DD;

^L.Missing n = 8 TD, 18 ASD, 2 DD;

^M.Missing n = 4 TD, 12 ASD, 3 DD;

^N.Missing n = 2 TD, 10 ASD, 3 DD;

^O.Missing n = 27 TD, 59 ASD, 17 DD;

^P.Missing n = 17 TD, 49 ASD, 12 DD;

^Q.Missing n = 16 TD, 47 ASD, 12 DD;

^R.Missing n = 17 TD, 49 ASD, 13 DD;

^S.Missing n = 3 TD, 22 ASD, 4 DD. Abbreviations: ASD, autism spectrum disorder; DD, developmental delay; TD, typically developing.

We employed Directed Acyclic Graphs (DAGs) to visually represent the presumed causal relationships among the study variables, based on existing literature (Supplemental Figure S4).³⁷ The Dagitty web application³⁸ was employed to identify minimally sufficient adjustment sets of variables to estimate the adjusted association between household insecticides and ASD. This led to a minimally sufficient adjustment set that included the following covariates: year of conception (modeled as a cubic polynomial), maternal education (bachelor’s degree and higher vs. some college or less), maternal mental health problems during pregnancy (yes/no), home ownership at time of enrollment (yes/no), urbanicity (mother’s address at time of delivery within an area classified as urban by the US Census closest to their birth year (2000, 2010, or 2020); yes/no), and season of conception. The addition of season of conception in the final models had no impact on the effect estimates, and thus, to increase power, was not included in the final models. Maternal mental health data was not collected in the first three years of CHARGE (i.e., before Oct 2005; n = 382). To avoid dropping those children from analyses, we conducted multiple imputation of maternal mental health with 50 imputations, based on a Markov Chain Monte Carlo imputation method that assumes a multivariate normal joint distribution of variables. We designated the pooled results from multiple imputation³⁹ as the primary analyses and compared them with results obtained from a “complete-case” analysis that restricted to participants with non-missing maternal mental health. Given that the ASD group was disproportionately larger during the years before mental health was collected as compared with the subsequent years (70% vs. 50%, respectively) and similarly the TD group was smaller (20% vs. 31%, respectively), we also performed sensitivity analysis to test the missingness assumptions of our primary analyses, that mental health was independent of insecticide exposures. For these analyses, we modified the imputed exposure values by increasing those for TDs by 20% and 50%, separately, and then compared the results with our primary analyses. There was no discernible difference in the results.

We estimated odds ratios (OR) and 95% confidence intervals (CI) for associations between the various application types of household insecticides and ASD or DD, adjusted for confounders, using multiple logistic regression. To account for multiple hypothesis testing, the false discovery rate (FDR) method was employed for the main analyses (i.e., each pesticide application type (yes/no) vs ASD). Specifically, the Benjamini-Hochberg procedure was applied to control the expected proportion of false positives among the set of significant test statistics, ensuring robust results across independent comparisons. Associations with P after FDR correction (P_FDR) <0.05 were considered as statistically significant, and those with P_FDR <0.10 as borderline significant.

In sensitivity analyses, indoor and outdoor insecticide models were stratified by year of conception to assess differences in insecticide ingredients (OPs vs pyrethroids), dichotomized at 2003 for professionally applied indoor, 2005 for non-professionally applied indoor and professionally applied outdoor, and 2007 for non-professionally applied outdoor insecticides, to account for people using old insecticides they still had in their homes, as well as different ban dates for indoor and outdoor OPs. In another sensitivity analysis, outdoor insecticide models were restricted to those that lived in a single-family house during the period of interest, as we suspected these subjects would be more aware of the amount/type of insecticides being applied around the outside of their homes. Lastly, given the discrepancy in diagnosis rates by child’s sex, we also stratified all models by child’s sex as a sensitivity analysis. Stata (SE version 18.0; StataCorp, College Station, TX) was used for all analyses.

3. Results

3.1. Participant Characteristics

At the time of this study, n = 1,734 participants had been enrolled and classified as ASD, DD or TD (n = 859 ASD, n = 332 DD, and n = 543 TD). Of these, n = 81 were excluded for not completing the questionnaire on insecticide exposures (n = 49 ASD, n = 19 DD, and n = 13 TD) and n = 127 DD were excluded for having a known genetic cause for their condition. Of the n = 1,526 remaining, n = 810 were classified as ASD, n = 186 as DD, and n = 530 as TD. Enrolled children were conceived between 1997 and 2018, with enrollment in the first few years for TD children delayed due to a lengthy process for obtaining access to the State’s birth files (n = 26 TD vs n = 144 ASD enrolled in first two years).

Table 1 compares ASD children and their mothers with TD children and their mothers. ASD children were more likely to be first born. Mothers of children with ASD were more likely to have gestational diabetes or any mental health condition before or during pregnancy, and less likely to own their home, have private insurance, be born in the US, or live in a single-family home during any of the pregnancy periods. Mothers of children with ASD had less education and were also more likely to report experiencing financial hardship.

Children with DD were less likely than TD children to be male, and more likely to be Hispanic or born prematurely (Table 1). Compared to mothers of TD children, mothers of children with DD tended to be younger and have a higher pre-pregnancy BMI. They were also more likely to have gestational diabetes, or a mental health condition before or during pregnancy. In general, compared to mothers of TD children, mothers of children with DD tended to have lower socioeconomic status (SES) based on numerous indicators e.g.: education, home ownership, experiences of financial hardship, and source of payment for their delivery (Table 1).

3.2. Distribution of Insecticide Exposure

More participants in our study reported using any indoor insecticide (44.1%) during the index period than any other application, followed by any outdoor insecticide (40.1%; Supplemental Table S1). Flea collar use was the least commonly reported product type (6.3%), followed by indoor professionally applied insecticides (11.8%). Apart from flea skin applications, more participants reported use of insecticides during trimester 2 than during trimesters 1 or 3. Reported use of all insecticides, other than flea collars, was lower during the pregnancy period than during the child’s first and second years of life.

3.3. Indoor Household Insecticide Exposure and ASD

Indoor professional insecticide application reported during each time period was associated with more than two-fold increased odds of ASD compared with no such application, after adjustment for covariates (e.g., pregnancy OR (95% CI): 2.52 (1.41, 4.48); year 1 OR: 2.20 (1.33, 3.64); Figure 1). After FDR correction, these associations remained significant. Similarly, reports of any indoor insecticide use, professionally applied or not, for each period analyzed, was associated with approximately 40–50% increased odds of ASD compared with no indoor insecticide use (e.g., pregnancy OR: 1.47 (1.11, 1.96); year 1 OR: 1.46 (1.12, 1.91)). However, after FDR correction, only the whole pregnancy period, the pre-pregnancy or pregnancy period, and years 1 and 2 remained statistically significant (P_FDR <0.05). While odds of ASD were increased by around 30% among those that reported any non-professionally applied indoor insecticide use during any period, the confidence intervals for each period contained the null.

Figure 1.

Associations Between Indoor Insecticide Exposures and Autism Spectrum Disorder. Values are adjusted odds ratios and 95% confidence intervals (bars) from unconditional logistic regression, modeled separately for each period, and adjusted for child’s year of conception (cubic), mother’s education (bachelor’s degree or higher vs less than bachelor’s degree), home ownership at time of enrollment (yes/no), mother’s address in urban area at time of delivery (yes/no), and self-report of any maternal mental health condition during pre-pregnancy or pregnancy (yes/no). Numbers exposed for each diagnosis group are listed to the right of the figure. Abbreviations: ASD, autism spectrum disorder; TD, typically developing.

When stratified by child’s sex, associations between indoor insecticides and ASD were generally higher among males than females (Supplemental Figure S5). Compared to models including all children, associations among males were similar, though slightly attenuated. Associations between professionally applied indoor insecticides and ASD could not be evaluated for female children due to small sample sizes. Non-professionally applied indoor insecticides or any insecticides showed no association with ASD among females.

In sensitivity analyses assessing potential differences in indoor insecticide ingredients (OPs vs pyrethroids), associations appeared stronger for insecticides that likely contained OPs compared to those that likely contained pyrethroids (Supplemental Figure S6). While sample sizes for professionally applied indoor insecticides that likely contained OPs in shorter time periods (e.g., pre-pregnancy, each trimester) were too small to assess, associations during longer time periods, such as pregnancy and each year of life, were similar between the two ingredients.

3.4. Outdoor Household Insecticide Exposure and ASD

Reports of any outdoor insecticide use during trimester 1 and the first and second years of life were associated with approximately 35% increased odds of ASD compared to no outdoor insecticide use, after adjustment for covariates (OR: 1.38 (1.00, 1.91), 1.37 (1.05, 1.79), and 1.34 (1.03, 1.74), respectively; Figure 2). However, the significance did not hold for trimester 1 and year 2 after FDR correction (P_FDR = 0.167 and 0.124, respectively), and the association in year 1 was borderline significant (P_FDR = 0.09).

Figure 2.

Associations Between Outdoor Insecticide Exposures and Autism Spectrum Disorder. Values are adjusted odds ratios and 95% confidence intervals (bars) from unconditional logistic regression, modeled separately for each period, and adjusted for child’s year of conception (cubic), mother’s education (bachelor’s degree or higher vs less than bachelor’s degree), home ownership at time of enrollment (yes/no), mother’s address in urban area at time of delivery (yes/no), and any maternal mental health condition during pre-pregnancy or pregnancy (yes/no). Numbers exposed for each diagnosis group are listed to the right of the figure. Abbreviations: ASD, autism spectrum disorder; TD, typically developing.

In sensitivity analyses restricting to those that lived in a single-family home (and hence may have more knowledge about insecticide applications than those in apartment or multi-plex residences), observed ORs for ASD were stronger than those observed among all home types from both outdoor non-professionally applied insecticides and any outdoor insecticide use, during every period (Supplemental Figure S7). In this group, reports of any outdoor insecticide application during pre-pregnancy, trimesters 1, 2, and 3, and the first two years of life were associated with 50% increased odds of ASD compared to those with no outdoor insecticide applications, after adjustment for covariates.

Associations between outdoor insecticides and ASD were higher among females than males in most time periods for both non-professionally applied outdoor insecticides and any outdoor insecticides, with any outdoor insecticides in trimester 2 reaching statistical significance (OR: 2.26 (1.00, 5.09)) and trimesters 1 and 3 borderline significant (2.18 (0.97, 4.91) and 2.19 (0.92, 5.21), respectively), albeit with wide CIs (Supplemental Figure S8). Models restricted to males yielded similar results as observed in models including all children.

In sensitivity analyses assessing potential differences in ingredients for outdoor insecticides applied by a non-professional, associations appeared stronger for insecticides that likely contained OPs compared to those that likely contained pyrethroids (Supplemental Figure S9), with some time periods showing a statistically significant protective effect of pyrethroid exposure.

3.5. Pet Flea Insecticide Exposure and ASD

Of the flea applications, reported use of flea soaps, shampoos, or powders during the second year of life was borderline associated with 45% higher ASD odds compared to no exposure (OR: 1.45 (0.97, 2.15); Figure 3). This borderline association did not hold after FDR correction. Associations among all other flea applications and time points were all close to the null.

Figure 3.

Associations Between Pet Flea Control Product Exposures and Autism Spectrum Disorder. Values are adjusted odds ratios and 95% confidence intervals (bars) from unconditional logistic regression, modeled separately for each period, and adjusted for child’s year of conception (cubic), mother’s education (bachelor’s degree or higher vs less than bachelor’s degree), home ownership at time of enrollment (yes/no), mother’s address in urban area at time of delivery (yes/no), and any maternal mental health condition during pre-pregnancy or pregnancy (yes/no). Numbers exposed for each diagnosis group are listed to the right of the figure. Abbreviations: ASD, autism spectrum disorder; TD, typically developing.

Female children in households with flea skin applications on pets were at greater risk for ASD than males in similar households, however the reverse was true for pet flea soaps, shampoos, and powders, where males were at higher risk. (Supplemental Figure S10). Flea collar use could not be evaluated in relation to ASD risk among female children, due to small sample sizes. Among males, the associations between each of the flea applications and ASD were, again, similar to results from all children.

3.6. Frequency of Household Insecticide Applications and ASD

High frequency (every month) of non-professionally applied insecticides in all time periods were associated with elevated odds of ASD, but the CIs in all but one instance (trimester 1), included the null value (Figure 4). Less frequent use showed no associations. In contrast, high frequency use (monthly) of any indoor insecticides, for every period examined, was associated with nearly two-fold increased odds of ASD (range: 1.65 to 2.09) compared with no reported indoor insecticide use, after adjusting for covariates. Trimesters 1 and 2, and the whole pregnancy had the strongest associations (Figure 4). Again, for all periods, results for lower frequency of reported indoor insecticide applications were null. There were too few participants in the group with the highest frequency of exposure for professionally applied indoor insecticides (e.g., exposure in every month of a trimester) to examine the association by frequency of use.

Figure 4.

Associations Between Frequency of Indoor Insecticide Exposures and Autism Spectrum Disorder. Values are adjusted odds ratios and 95% confidence intervals (bars) from unconditional logistic regression, modeled separately for each period, and adjusted for child’s year of conception (cubic), mother’s education (bachelor’s degree or higher vs less than bachelor’s degree), home ownership at time of enrollment (yes/no), mother’s address in urban area at time of delivery (yes/no), and any maternal mental health condition during pre-pregnancy or pregnancy (yes/no). Associations between frequency of professionally applied indoor insecticides and ASD could not be evaluated due to small sample sizes. Numbers exposed for each diagnosis group are listed to the right of the figure. Abbreviations: ASD, autism spectrum disorder; TD, typically developing.

In covariate adjusted models, high frequency non-professionally applied outdoor insecticides versus no such applications during trimesters 1 and 2, and the entire pregnancy period were associated with ORs (95% CIs) of 2.68 (1.35, 5.31), 1.91 (1.06, 3.44), and 2.08 (1.01, 4.32), respectively (Figure 5). For any outdoor applications, high frequency use (monthly in 3-month periods, and 6–9 months for the whole pregnancy) was consistently associated with similarly elevated but more precise ORs, ranging from 1.83 (1.18, 2.84) (pre-pregnancy) to 2.31 (1.49, 3.60) (trimester 1), in all time periods. The ORs were consistently higher for those households living in single-family residences with “any outdoor” applications (compared with all households; Supplemental Figure S11).

Figure 5.

Associations Between Frequency of Outdoor Insecticide Exposures and Autism Spectrum Disorder. Values are adjusted odds ratios and 95% confidence intervals (bars) from unconditional logistic regression, modeled separately for each period, and adjusted for child’s year of conception (cubic), mother’s education (bachelor’s degree or higher vs less than bachelor’s degree), home ownership at time of enrollment (yes/no), mother’s address in urban area at time of delivery (yes/no), and any maternal mental health condition during pre-pregnancy or pregnancy (yes/no). Numbers exposed for each diagnosis group are listed to the right of the figure. Abbreviations: ASD, autism spectrum disorder; TD, typically developing.

High frequency use (compared with no use) of flea soaps, shampoos and powders on pets were generally associated with some elevation in the odds of ASD, but confidence intervals included null values with the exception of trimester 3 (OR: 2.05 (1.06, 3.94)). Frequent applications of flea soaps, shampoos and powders during trimester 1 or the full pregnancy period had ORs with borderline associations (Figure 6). Most individuals with exposure to flea collars used them continuously so, there were very few participants in the low frequency of exposure group, and thus we were unable to examine the association by frequency of use.

Figure 6.

Associations Between Frequency of Pet Flea Control Product Exposures and Autism Spectrum Disorder. Values are adjusted odds ratios and 95% confidence intervals (bars) from unconditional logistic regression, modeled separately for each period, and adjusted for child’s year of conception (cubic), mother’s education (bachelor’s degree or higher vs less than bachelor’s degree), home ownership at time of enrollment (yes/no), mother’s address in urban area at time of delivery (yes/no), and any maternal mental health condition during pre-pregnancy or pregnancy (yes/no). Associations between frequency of flea collar exposure and ASD could not be evaluated due to small sample sizes. Numbers exposed for each diagnosis group are listed to the right of the figure. Abbreviations: ASD, autism spectrum disorder; TD, typically developing.

Associations between frequency of each of the insecticide applications (indoor, outdoor, and flea) and ASD among males were similar to models including all children (Supplemental Figure S12). Associations among female children could not be evaluated due to small sample sizes.

3.6. Household Insecticide Exposure and DD

Reports of indoor non-professionally applied insecticides (vs. no non-professional applications) during trimester 3 or the full pregnancy period were associated with an elevated odds of DD: OR = 1.72 (1.03, 2.89) and 1.58 (1.03, 2.40), respectively, in models adjusted for covariates (Supplemental Figure S13). There were too few DD participants with exposure to professionally applied indoor insecticides to examine that association.

While odds of DD were increased by nearly 70% among those that reported any non-professionally applied outdoor insecticide use during trimester 3, the confidence intervals contained the null (OR: 1.67 (0.93, 3.03); Supplemental Figure S14). Associations among all other outdoor applications and time points were essentially null. When restricting to those that lived in a single-family home, associations observed between outdoor insecticide use (Supplemental Figure S15) were similar to those observed among all home types.

Associations among flea applications and DD could not be evaluated due to small sample sizes.

3.7. Frequency of Household Insecticide Exposure and DD

The strongest results for development of DD were from non-professionally applied high frequency indoor insecticide use during the full pregnancy period and trimester 3, however, the confidence intervals contained the null (OR: 1.83 (0.93, 3.58) and 1.83 (0.97, 3.46), respectively; Supplemental Figure S16). Associations between reported high frequency applications of any indoor insecticides and DD were similar to those of non-professionally applied indoor insecticides. No associations with DD were observed for less frequent applications, neither for any nor for non-professionally applied indoor insecticides.

Associations between frequent use of outdoor and flea insecticides and DD could not be evaluated due to small sample sizes.

4. Discussion

The findings from the present study contribute to an evolving body of literature underscoring the potential neurodevelopmental hazards posed by household insecticide exposure during critical periods of prenatal and early postnatal development. Our study, leveraging the comprehensive data from the CHARGE study, elucidates the associations between various modalities of household insecticide use and the risk of ASD in offspring. The differential risk patterns observed across the spectrum of insecticide types, professional and non-professional applications, and exposure periods offer insights into the complex relationship between environmental exposures and neurodevelopmental outcomes.

The observed association between professional indoor insecticide application and ASD is particularly striking, with a greater than twofold increase in odds during all time points. This relationship is underscored by the potency of these chemicals, designed to penetrate the neurological systems of pests, coupled with their persistence when applied indoors. This persistence arises from limited direct sunlight indoors, resulting in minimal degradation by photolysis.⁴⁰ Additionally, the increased risk associated with any outdoor applications during the child’s first two years of life suggests that the hazard is not confined to professional applications alone but extends to consumer-grade products as well, which is particularly concerning given the pervasive nature of these products in residential settings. This relationship was further potentiated in the context of high-frequency exposure scenarios, highlighting a dose-response relationship that warrants attention. The stratified analysis revealing stronger associations for male children with indoor insecticide exposure parallels the higher baseline risk of ASD in males and may reflect underlying sex-specific neurobiological vulnerabilities. Our results are consistent with previous studies that have reported associations of ASD with prenatal and early life exposures to both pyrethroids and OPs and ASD.^{13,19,41–43}

The potential mechanisms through which household insecticide exposure may contribute to the development of ASD are multifaceted, reflecting the complex etiology of ASD itself, which encompasses both genetic and environmental components. The neurotoxic properties of the primary classes of insecticides involved in this study – organophosphates and pyrethroids – provide a biological basis for the observed associations.

OPs are able to cross the blood-brain barrier and the placental barrier,⁴⁴ making the developing fetus particularly susceptible to the effects of OPs. However, critical to understanding how insecticides influence early child neurodevelopment is the distinction between acute effects, which are well-known to occur after high-level exposures, versus impacts after lower but chronic exposures. Although high dose exposure to OPs inhibits acetylcholinesterase (AChE) activity, resulting in acute neurologic toxicity, this contrasts with the neurodevelopmental impacts in children following prenatal organophosphate exposures at doses that are insufficient to induce those acute symptoms, as recognized in 2016 by the U.S. EPA.⁴⁵ In fact, across dozens of studies demonstrating adverse neurodevelopmental impairments after prenatal OP exposures, no evidence emerged of acute toxicity in the mother. Furthermore, the developmental deficits only become manifest months or years later. Household exposures, such as those of CHARGE Study children in the pre- and postnatal periods, were unlikely to have reached concentrations inducing pathogenic cholinesterase inhibition. Hence, OP interference with brain development most likely involves other mechanisms. The literature indicates that these may include effects on one or more of the following: dopaminergic and glutamatergic neurotransmission; endocannabinoid and serotonergic systems; neuroinflammatory pathways, particularly in the hippocampus; effects on protein-kinase C receptor signaling; insulin resistance; and disruption of functioning in nuclear transcription.^46–49 Studies in multiple animal species confirm the neurodevelopmental impacts from low doses of chlorpyrifos in the absence of AChE inhibition.⁴⁹

Pyrethroids, while often considered safer than OPs, act by altering the function of sodium channels in neuronal membranes.⁵⁰ This action can lead to prolonged sodium influx, which can excite nerve cells, potentially leading to neurotoxicity if exposure occurs at critical developmental stages. As with OPs, pyrethroids are able to cross the blood-brain barrier and the placental barrier.⁴⁴ The susceptibility of the developing brain to these disruptions may explain the link between prenatal and early-life exposure to pyrethroids and the increased risk of neurodevelopmental disorders such as ASD. Aberrations in synaptic formation and maintenance have been postulated as underlying mechanisms in ASD pathophysiology,⁵¹ offering a potential pathway through which pyrethroid exposure may contribute to ASD risk.

Furthermore, both classes of insecticides have been implicated in oxidative stress and inflammatory responses,^52–58 which have been associated with ASD in multiple studies.^59–61 Chronic exposure may exacerbate these biological processes, potentially contributing to the pathophysiology of ASD. The precise mechanisms by which these exposures interact with genetic susceptibilities to increase ASD risk remain under investigation. Epigenetic modifications induced by environmental toxins such as insecticides could play a role in altering gene expression without changing the underlying DNA sequence, thereby affecting neurodevelopment.⁶²

4.1. Limitations and Strengths

This study has limitations that should be considered in interpreting the findings. One limitation is the lack of specific insecticides used. While we did ask participants for specific brand names of insecticides used, there were substantial recall issues with the exact name of each insecticide. For example, many participants would simply report “Raid” instead of “Raid Max Ant and Roach Killer” or “Raid Wasp & Hornet Killer.” This made it difficult to identify the specific insecticide(s) each participant was exposed to. Additionally, even though organophosphates were banned in the US for indoor residential use in the early 2000s, many people still had these insecticides in their homes, making it difficult to determine which active ingredient was applied. In sensitivity analyses, we stratified by year of conception to attempt to tease out the effects of OPs from that of pyrethroids in indoor and outdoor insecticides. While we found similar effect estimates for OPs and pyrethroids in professionally applied indoor insecticides, sample sizes were too small to examine these effects in shorter time periods (e.g., by trimester). For outdoor insecticides applied by a non-professional, the effect of pyrethroids was smaller, and sometimes protective, compared to the effect of OPs. Given that the survey had changed around the same time as the pesticide change, we can’t be certain if these effects were due to the different ingredients or the change in the way the question was asked.

Another key limitation of our study is the reliance on self-reported data for household insecticide use. Self-report methods are inherently susceptible to errors of recall, particularly when asking participants to remember and accurately report past events or exposures over extended periods. These errors can be random or systematic. Systematic error is a particular concern in studies involving children with serious health or developmental conditions, such as cancer, malformations, or ASD, where parents may differentially report exposures they believe to be related to their child’s condition, leading to reporting bias. For example, parents whose children are not affected may give less thought to their exposures and hence be less likely to report them as compared with parents of an affected child. Few studies have examined the extent of recall bias during pregnancy.^63,64 One study⁶³ compared self-report of multiple exposures during pregnancy among parents of children with and without malformations in comparison with their obstetric records and found random errors with 75% of examined exposures and evidence of systematic differences by case-control status in the remaining 25% of exposures, with the authors concluding that recall bias is exposure-specific.

Additionally, very little research has evaluated validity or reliability of self-reported home pesticide applications, especially during pregnancy. Although household pesticides are marketed as being safe, concerns about certain classes of these chemicals have been publicized in the mass media; as a result, parents whose children have been diagnosed with ASD may be sensitized about such exposures, leading to differential reporting of home use insecticides in our study. Such biases could potentially exaggerate the strength of the associations between household insecticide use and ASD outcomes we observed. However, self-reported household pesticide use has demonstrated reliability in a case-control study of cutaneous melanoma among individuals in Rome, re-administered the same pesticide questionnaire approximately one year apart.⁶⁵ Additionally, validity was observed in a cohort of older male orchardists (i.e., farmers) in Washington state, who provided consistent recall of pesticide use 20–25 years later.⁶⁶ Information on timing, frequency, dose, types of pesticide applications could also be less accurately recalled.

Furthermore, a notable limitation in our methodology is the exclusive focus on household insecticide use without accounting for exposure in places of business, childcare facilities, or other environments outside the home. We did not ask about insecticide use at the mother’s workplace or whether or not the child attended daycare, because it was expected that they would not know the extent of use, if any. It is well-documented that many such locations routinely employ insecticide applications as part of their maintenance routines, likely exposing some mothers and children to additional sources of these chemicals beyond the home environment. This would lead to a non-differential underestimation of the total insecticide exposure experienced by participants, thereby diluting the observed associations between household insecticide use and ASD risk. Another possible source of insecticide exposure not accounted for in this study is that from food, however, a recent study found no association between pesticide residues on food and ASD.⁶⁷

Moreover, our analysis encountered constraints regarding sex stratification. While we were able to limit our analyses to male children, revealing no significant difference in effect estimates, the inability to examine females across all analyses precluded a detailed examination of potential sex-specific effects. This limitation stems from the large sex ratio, in which ASD cases are more than 4 times as likely to be male; to maximize efficiency of control for confounding by sex, we also frequency matched the population controls to the cases by sex. Expanding the number of female controls would have slightly improved power, but since the low rate of ASD in females was not within our capacity to change, the statistical power would still be strongly driven by the small number of female ASD cases (and the even smaller number who are exposed). Replication of findings in future studies with more females is needed.

Lastly, residual confounding remains a concern when interpreting the association between household insecticide use and ASD. While our findings suggest a potential causal link, the comparable associations observed across different developmental windows suggest the possibility of unmeasured confounding, and alternative explanations involving confounding factors must be considered. One possible source of residual confounding is SES. Households with SES may face higher pest burdens due to living conditions, such as crowded or poorly maintained housing, which may lead to increased pesticide use. At the same time, lower SES is associated with increased risk of ASD.^2,68,69 While SES variables, such as maternal education, home ownership, and urbanicity, were included in the analysis, residual confounding from more nuanced SES factors, such as neighborhood characteristics or long-term economic instability, may remain. Another potential source of residual confounding is parental ASD traits, which have been shown to increase the risk of ASD in offspring,⁷⁰ and which may also influence household behaviors, including insecticide use. Parents with ASD traits might be more likely to adopt household routines that involve frequent insecticide application, potentially confounding the observed relationship between insecticide exposure and ASD. While maternal mental health was included in the analysis to partially account for this, it does not fully capture the broader spectrum of parental ASD traits, and thus residual confounding from this source remains a concern. Another potential confounder is the need for insecticide use, which may reflect underlying environmental exposures to the pests targeted by these chemicals. Insects could carry biological agents or trigger environmental stressors, like asthma, that impact neurodevelopment, independently increasing ASD risk. To mitigate these concerns, future studies could employ methodologies such as negative control exposures or sibling-comparison designs. Additionally, a more granular examination of different insecticide types would help clarify whether specific agents are driving the association, strengthening the causal interpretation of the findings.

The strengths of this study are considerable and contribute valuable insights into the association between household insecticide use and ASD risk. Notably, the use of detailed data on the frequency and timing of insecticide exposure, and whether applied by a member of the household or professionally applied, represents a significant advantage over past studies. This granularity allows for a more nuanced analysis of exposure windows that are potentially critical for ASD development and distinguishes this study from others which rely on very small windows of exposure with urine/serum biomarkers, or nonspecific timing and/or lack of frequency of exposure with self-report. Furthermore, the distinction as to whether a member of the household or a professional service applied the insecticides is relatively novel, and illuminated a strong association with the professional applications, which, if replicated, could be an area of close scrutiny with respect to public health and/or policy change.

The diagnostic classification of ASD in this study deserves particular mention. All diagnoses were clinically confirmed using gold-standard standardized assessments, including the Autism Diagnostic Observation Schedules and the Autism Diagnostic Interview-Revised. This rigorous approach ensures that our diagnoses are both accurate and reliable, minimizing the potential for misclassification bias and lending credence to the observed associations.

The diversity of the study population is another noteworthy strength. Successful recruitment of participants from a variety of urban and rural environments, with a broad range of socioeconomic levels and ethnic backgrounds, this study enhances the generalizability of our findings. Lastly, the large sample size of this study, comprising over a thousand participating children and their households, provides respectable statistical power and precision of the findings. This robust sample enabled us to detect even modest associations between insecticide exposure and ASD, thereby contributing valuable evidence to the ongoing discourse on environmental risk factors for neurodevelopmental disorders.

4.2. Public Health Implications

These findings add to growing body of evidence that exposure to household insecticides during pregnancy and early life may be associated with ASD, underscoring the need for heightened awareness and precautionary measures regarding household insecticide use, particularly in environments inhabited by pregnant women and young children. Secondly, these results highlight the potential need for regulatory oversight in the approval and marketing of these chemicals, advocating for stricter safety standards and the promotion of non-chemical pest control alternatives.

Given the relatively low number of subjects exposed to professionally applied indoor insecticides (3.7 – 7.6%, in a given time period, and 12.1 % using it during any time period), replication of this study is needed, particularly in relation to the professional applications. In addition, future research should aim to elucidate the biological mechanisms underlying these observed associations, exploring potential gene-environment interactions and the role of neuroinflammation and oxidative stress pathways. Longitudinal studies with detailed exposure assessments, including repeated biomarker measures, would further refine our understanding of these relationships. Additionally, public health interventions and policies aimed at reducing environmental exposures to neurotoxic chemicals are crucial for mitigating ASD risk.

4.3. Conclusion

This study adds to the growing body of evidence linking prenatal and early life exposure to insecticides to increased ASD risk, and it is the first study to implicate professionally applied indoor insecticide applications in ASD risk. The findings call for a reevaluation of the use of these chemicals within residential settings, including monitoring of toxic insecticides in the households and in the bodies of the residents, as well as advocacy for safer practices and regulations to protect the most vulnerable populations. Through a multidisciplinary approach encompassing epidemiology, neuroscience, and public health policy, we can advance our understanding of ASD etiology and contribute to the development of effective intervention strategies for protecting those at risk for debilitating neurodevelopmental conditions.