Personal and In-Home Exposures to Airborne Organophosphate Pesticides among Rural Residents of California’s Central Valley

Abstract

Residents of agricultural communities may experience higher exposures to pesticides due to their proximity to agricultural operations. We applied a novel measurement approach, using Ultrasonic Personal Air Samplers (UPAS), to quantify particulate matter and organophosphate pesticides in air in California’s Central Valley. We collected 124 personal, 126 in-home, and 32 outdoor air samples with 66 adults from 37 rural households in 2023 and 2024. We detected chlorpyrifos, acephate, malathion, diazinon, and naled in air samples. We detected gas-phase chlorpyrifos in 63% of personal samples and 86% of homeseven though use of chlorpyrifos has been banned in California (with few exceptions) since January 2021at 24 h average concentrations ranging up to 13 ng m^–3 (personal) and 5.8 ng m^–3 (in-home). We did not detect chlorpyrifos in outdoor air samples. Using linear mixed models, we found that higher indoor air temperatures and having more carpets/rugs were associated with higher indoor chlorpyrifos concentrations. The concentrations we measured were well below the California Department of Pesticide Regulation’s health screening level of 510 ng m^–3 for chronic exposure to chlorpyrifos in air; nevertheless, our results suggest that persistent chlorpyrifos in home environments continues to contribute to nondietary exposure among California residents.

1. Introduction

Organophosphates (OPs) are a class of agricultural pesticides widely used to prevent insect damage to crops. In California’s Central Valley, which is one of the most agriculturally productive regions in the United States (US), large volumes of OPs are applied annually to nuts, fruits, vegetables, cotton, and alfalfa. These compounds are typically sprayed or distributed in granular form across large areas of croplands, often near rural residential communities. , Despite their effectiveness in pest management, OPs pose significant risks to human health and the environment due to their toxicity and persistence. Exposure to OP pesticides has been associated with a range of adverse health outcomes. Acute exposures can cause nausea, dizziness, and respiratory distress. Chronic OP exposure has been linked to neurological and developmental disorders (including Parkinson’s disease, Alzheimer’s disease, and cognitive impairments in children) − as well as respiratory disease. −

One OP that is particularly harmful to human healthchlorpyrifoswas banned from household use in the US in 2000. Most chlorpyrifos products were further banned from agricultural use in California as of 1 January 2021, following years of advocacy by environmental justice and labor organizations. − However, some products that included chlorpyrifos in granular form were exempt from this ban. The US Environmental Protection Agency also banned agricultural use of chlorpyrifos nationwide in 2022. Although the nationwide ban was overturned in 2023, the California ban remains in place.

Humans can be exposed to agricultural pesticides through: (a) contact with treated crops; (b) consumption of treated foods; (c) dermal contact with, ingestion of, or inhalation of household dust that is contaminated with pesticides; as well as (d) direct inhalation of pesticides present in gas or particulate phases in air. Higher exposures to OP pesticides have been documented among individuals who live near and/or work in agricultural fields. OPs such as chlorpyrifos and methyl parathion have been detected in home environments in agricultural communities. − These OPs might enter living spaces through infiltration of outdoor air after spraying, tracking-in of contaminated soil, and/or “take home” occupational exposures. Furthermore, results from prior studies indicated that OPs can persist in indoor environments, and thereby continue to contribute to human exposure, for months or years after application. ,, Because OPs are semivolatile organic compounds (SVOCs), they can be present in the gas phase and also sorbed onto airborne particulate matter (PM), settled dust, and surface materials (e.g., carpet/rug fibers and foam). −

Characterization of human exposure to OPs remains a challenge. Recent personal exposure to OPs can be quantified by collecting and analyzing urine samples for DAP and 3,4,6-trichloro-2-pyridinol (TCPy) metabolites; however, this approach does not allow researchers to determine the route(s) that contributed to exposure or to determine whether exposure is chronic. , To assess the potential for chronic human exposure to pesticides in air, the California Department of Pesticide Regulation (DPR) measures 24 h average concentrations of OPs once per week at each of four outdoor Air Monitoring Network (AMN) sites. − Personal exposures to OPs in air are often quantified by sampling air onto XAD-2 sorbent or polyurethane foam disks worn by individuals. , The California DPR has established health screening levels for OPs in air; for chlorpyrifos, these screening levels range from 510 ng m^–3 for chronic exposure to 1200 ng m^–3 for 24 h acute exposure. DPR advises that concentrations below these levels are not guaranteed to be safe, and could warrant further investigation, but are not considered to pose significant health concerns and therefore do not necessitate further investigation.

Another factor to consider during exposure assessments is that exposure to OPs rarely occurs in isolation. Instead, humans are typically exposed to multiple harmful air pollutants simultaneously. For example, in addition to being a major hub of agricultural production, the Central Valley of California is also known for having some of the worst ambient PM_2.5 pollution in the US. In a prior study of children living in an agricultural community, simultaneous exposure to higher levels of OP pesticides and fine particulate matter (PM_2.5) was associated with a greater increase in a biomarker of pulmonary inflammation (leukotriene E4) than separately higher exposures to OPs or PM_2.5. Efforts to understand the health risks associated with human exposures to OPs and PM_2.5 would therefore benefit from a measurement approach that allows both pollutants to be measured simultaneously using a single wearable air sampler.

In the present study, we used the Ultrasonic Personal Air Sampler (UPAS v2.1 PLUS, Access Sensor Technologies, Fort Collins, CO, USA)a device that is quieter, smaller, lighter, and thus more ergonomic than conventional wearable air samplersto sample OPs and PM simultaneously outside homes, inside homes, and in the breathing zones of people living in rural Central Valley agricultural communities. We analyzed the resulting data to answer the following questions: (1) Can the UPAS be used to evaluate exposures to OP pesticides and PM simultaneously? (2) Which OPs are most prevalent in the air to which rural Central Valley residents are exposed, and at what concentrations are those OPs present? (3) Which factors are associated with the presence and concentrations of OP pesticides in home air?

2. Materials and Methods

We sampled personal exposures to, in-home levels of, and outdoor levels of PM and OP pesticides in air in rural Fresno and Tulare Counties in California, US. Personal and in-home air samples were collected during three campaigns that took place in November–December 2023, May–June 2024, and September 2024; however, only Tulare County households were included in the September 2024 campaign. Methods related to participant recruitment and air sample collection were described previously. Here, we provide a summary of those methods along with more in-depth explanations of sample collection and analysis methods related to OPs.

The communities where sampling took place were identified by the Central California Environmental Justice Network and were selected due to their proximity to agricultural operations. In these communities, many homes were adjacent to agricultural fields or facilities that processed raw agricultural products. Some homes were surrounded by fields on all sides. Many households that participated in this study also participated in prior studies on OP exposure and had consented to being contacted about similar studies conducted by this research team. Additional households were recruited through participant referrals and by networking with local organizations. We visited each household once per campaign to collect 24 h personal and in-home air samples. If a household was not available to participate in the second or third campaign, we recruited an additional household from the same county to keep the number of households and participants sampled during each campaign approximately constant. Within each participating household, adults (18 years of age or older) were recruited for personal sampling. All recruitment and study methods were approved by the Colorado State University (CSU) Institutional Review Board (protocol #2527). All participants provided written informed consent in either English or Spanish.

2.1. Sample Collection

During the 24 h sampling period, up to two adults from the household were monitored for personal exposure to PM and gas-phase OP pesticides. If two adults participated, one was monitored for exposure to PM₁₀ and OPs while the other was monitored for exposure to PM_2.5 and OPs. If only one adult participated, then that participant was monitored for exposure to either PM₁₀ and OPs or PM_2.5 and OPs. If an individual participated during multiple campaigns, that individual was monitored for the same PM size fraction each time. Along with each personal air sample, a matching in-home sample (i.e., a sample of the same PM size fraction plus gas-phase OP pesticides) was collected concurrently in a common area such as the living room or dining area (not in a bedroom, bathroom, adjacent to the kitchen stove, or next to a window). Participants were also surveyed about occupational exposures (e.g., “How many people in the home work in agriculture or have contact with pesticides at work?”), their homes (e.g., “In how many rooms is there carpet?”), use of consumer pesticide and herbicide products in and around their homes, as well as concerns they had about the environment where they lived. A complete list of survey questions is provided by Li et al.

All air samples were collected using the UPAS v2.1 PLUS. Each UPAS sampled air through a PM_2.5 or PM₁₀ inlet, followed by (1) a 37 mm diameter polytetrafluoroethylene (PTFE) membrane filter (PT37P-PF03, Measurement Technology Laboratories, Minneapolis, MN, USA) to collect PM and (2) a packet of XAD-2 sorbent (20279, Sigma-Aldrich, St. Louis, MO, USA) to collect gas-phase OPs (Figure ). Packets containing XAD-2 were manufactured at Access Sensor Technologies (AST). Each packet was constructed from two 37 mm diameter discs of 60 μm nylon mesh (U-CMN-60-C, Component Supply Company, Sparta, TN, USA) that were cut out on a laser cutter, soaked in ethanol, allowed to air-dry, and then adhered together around the outer circumference using adhesive transfer tape (467MP, 3M, St. Paul, MN, USA). Each packet was then filled with 0.3 ± 0.02 g of purified 20–60 mesh size XAD-2 before the full circumference of the packet was sealed. Finally, each XAD-2 sorbent packet was installed behind a PTFE membrane filter in a 37 mm UPAS sample cartridge as shown in Figure . Each personal and in-home air sample was collected at a flow rate of 2 L min^–1.

1.

Particulate matter and organophosphate pesticides were sampled with the UPAS v2.1 PLUS. Air was drawn through a PM_2.5 or PM₁₀ inlet, followed by a PTFE membrane filter and a XAD-2 sorbent packet.

At least one field blank was collected each day personal and in-home samples were started. Each of these field blanks was collected by (a) installing a cartridge containing a PTFE membrane filter and XAD-2 sorbent packet in a UPAS (as shown in Figure ) at the same time when the cartridges used to collect that day’s samples were installed in UPAS, (b) transporting the UPAS containing the blank filter and sorbent packet to participating households along with the UPAS used to collect samples, (c) transporting the UPAS containing the blank cartridge back to the sample handling area at the end of the day, and (d) unloading the blank cartridge along with any sample cartridges that were returned to the handling area that dayall without sampling any air through the blank filter and sorbent packet.

PM and OPs in outdoor air were also sampled at a subset of participating homes. Fourteen-day integrated samples of PM₁₀ and OPs were collected outside of three homes in Fresno County and three homes in Tulare County between May 17 and July 7, 2024. Paired sets of seven-day integrated PM₁₀/OP and PM_2.5/OP samples were collected outside of the same three Tulare County homes between August 15 and September 10, 2024. Additionally, 24 h paired outdoor PM₁₀/OP and PM_2.5/OP samples were collected concurrently with the 24 h personal and in-home samples collected at these three Tulare county homes in September 2024. See SI Section S1.1 for additional information about outdoor sample collection and sample storage.

2.2. Sample Analysis

Each XAD-2 packet was extracted into acetonitrile and analyzed, using liquid chromatography-tandem mass spectrometry (LC-MS/MS), for 11 pesticides: chlorpyrifos, acephate, malathion, diazinon, naled, bensulide, dimethoate, oxydemeton-methyl, phorate, phosmet, and tribufos.

To extract the OP sample, each XAD-2 sorbent packet was placed in a 4 mL glass vial, and 2 mL of acetonitrile was added to the vial. Each vial stood for 30 min and was then sonicated for 30 min. After sonication, the extract was transferred to a separate vial and dried down to 500 μL with nitrogen. Next, 150 μL of the concentrated extract was transferred to an LC-MS/MS autosampler vial. Another 150 μL of a 0.1% formic acid/LC-MS-grade water solution was then added to the autosampler vial. The resulting solution was vortex mixed.

Calibration samples were prepared from standards for each of the 11 OP pesticides. Each standard consisted of 100 μg mL^–1 of the pesticide of interest in methanol (P-094S for chlorpyrifos, P-200S-A for acephate, P-060S for malathion, P-033S for diazinon, M-622-18 for naled, P-204S for bensulide, P-039S for dimethoate, P-290S for oxydemeton-methyl, APP-9-181 for phorate, M-8141A-1-08 for phosmet, and S-13194A1-1ML for tribufos, AccuStandard, New Haven, CT, USA).

All samples were analyzedincluding the calibration samples and a set of control samplesusing an Agilent 1290 UHPLC coupled to an Agilent 6460 triple quadrupole mass spectrometer equipped with an Agilent Jet Stream electrospray ionization source (Agilent, Santa Clara, CA, USA). Pesticides were first chromatographically separated on an Agilent Poroshell C18 column (2.1 × 100 mm, 2.7 μm) held at 40 °C. A sample volume of 10 μL was injected along with a mixture of: (A) water with 5 mM ammonium formate/0.05% formic acid and (B) 5 mM ammonium formate/0.05% formic acid in methanol at a flow rate of 0.4 mL min^–1. The gradient elution used was 10% B for 1 min, 15% B at 1.5 min, 70% B at 2.5 min, to 100% B at 10 min. The ionization source conditions were as follows: nebulizer at 45 psi, gas flow of 12 L min^–1 at 300 °C; sheath gas flow of 12 L min^–1 at 375 °C. The electrospray ionization polarity was set to positive for all analytes. Two ion transitions (m/z) were monitored for each analyte and deuterium-labeled internal standards. These ion transitions and the corresponding fragmentor and collision energy voltages are displayed in Table S1. Compound identifications were confirmed by retention time and the product ion ratios (±20%). Data were collected and processed using Agilent MassHunter Quantitative software (v.B.08.01). Quantitation was performed with linear regression using 6-point calibration curves for concentrations ranging from 0.5 to 500 ng mL^–1.

After the gas-phase OP samples collected on the XAD-2 packets were analyzed, 15 air samples containing the highest gas-phase concentrations of chlorpyrifos, acephate, and malathion were identified. Each PM sample collected upstream of the XAD-2 sorbent packet during one of these 15 air samples was then extracted from the PTFE filter (using the same procedures used to extract samples from XAD-2 but with 50 mL beakers instead of 4 mL vials) and analyzed for all 11 OPs.

2.3. Data Management and Statistical Analysis

The time-averaged mass concentration of each pesticide in each air sample (ng m^–3) was calculated from (a) the mass concentration of the pesticide in the sample as analyzed using LC-MS/MS (ng mL^–1) and (b) the volume of air sampled by the UPAS (m³). The limit of detection for the LC-MS/MS analysis was 0.5 ng mL^–1, which translated to a concentration of 0.17 ng m^–3 in a 24 h air sample collected at 2 L min^–1. If, during LC-MS/MS analysis, a pesticide was detected at a concentration <0.5 ng mL^–1, that pesticide was assigned a concentration of $0.5/\sqrt{2}$ ng mL^–1, which translated to 0.12 ng m^–3 in a 24 h air sample collected at 2 L min^–1. If a pesticide was not detected in a sample during LC-MS/MS analysis, then that pesticide was assigned a concentration of zero.

For each pesticide and sample type (personal, in-home, or outdoor), we calculated the detection rate as well as the minimum, 25th percentile, median, 75th percentile, and maximum gas-phase concentrations in air (for all of the samples collected on XAD-2 sorbent packets).

For all pairs of collocated measurements of gas-phase chlorpyrifos in home air (ng m^–3), we calculated the mean difference, mean percent difference, and Spearman correlation coefficient. The percent difference between each pair of samples was calculated as the difference between the two measurements divided by the mean of the two measurements. If both measurements were zero, the percent difference was set equal to zero. Additionally, we fit a linear model to compare paired measurements using Deming regression with assumed constant error.

For each pesticide, we also calculated the Spearman correlation between 24 h average gas-phase in-home concentrations and personal exposures. The concentrations measured in duplicate, concurrent in-home samples were averaged before calculating these correlation coefficients.

Factors associated with the 24 h average concentration of chlorpyrifos in the air inside household i on visit j (c _ij ; ng m^–3; n = 126) were investigated by fitting the linear mixed model shown in eq :

$$c_{ij}=\alpha +{\alpha }_i+\sum _{k=1}^K{\beta }_k{x}_{kij}+{\epsilon }_{ij}$$ 1

where α was a fixed intercept, α _i was a random intercept for household i, β _k was the slope associated with one of K fixed effects, and ε _ij was the random error.

Additionally, factors associated with the presence (detected vs not detected) of chlorpyrifos, acephate, and malathion, respectively, in home air (n = 126 for each pesticide) were investigated by fitting logistic mixed models of the form shown in eq to the data:

$$\log \mathrm{it}(p_{ij})=\alpha +{\alpha }_i+\sum _{k=1}^K{\beta }_k{x}_{kij}$$ 2

where p _ij was the probability of the pesticide (chlorpyrifos, acephate, or malathion) being detected in household i on visit j.

The models in eqs and were fit using Bayesian methods in the R programming environment using the ‘rstanarm’ package. For all parameters, we used the default weakly-informative priors centered on zero. These priors act in the same way as ridge regression with a small penalty applied to the parameters to reduce variance inflation, add stability to the estimates, and account for the effects of multicollinearity; therefore, the models shown in eqs and were fit once, to the full data set of n = 126 indoor chlorpyrifos concentrations, with K = 15 fixed effects included. One of these fixed effects was the sampling campaign (reference = Nov–Dec 2023 vs May–Jun 2024 or Sep 2024). The other 14 fixed effects, which are listed in Table , were selected because we expected them to relate to (a) primary or secondary occupational exposure to pesticides (e.g., number of people living in the home who were employed in agriculture; see SI Section S1.3.1 for additional information), (b) other potential sources of pesticide exposure (e.g., a metric related to the cumulative mass of the pesticide applied for agricultural use at various distances from the home in 2018–2023, as indicated in the California DPR Pesticide Use Reporting data; see SI Section S1.3.2 for additional information), (c) dust and SVOC reservoirs inside the home (e.g., the sum of the number of rooms with carpet and the number of rooms with area rugs), (d) surface wear and resuspension of dust inside the home (e.g., number of people living in the home), (e) home ventilation rates and outdoor air infiltration rates (e.g., binary indicators of whether the home was cooled using air conditioning, a swamp cooler, by opening windows, and/or using fans), or (f) vaporization of SVOCs (24 h average indoor air temperature estimated as described in SI Section S1.3.3). As a sensitivity analysis, the model in eq was also fit to a subset of data (n = 120) that excluded six data points from one household where the highest chlorpyrifos concentrations were measured.

1. Rural Households Visited, Participants Sampled, and Samples Collected Successfully in Each County during Each Campaign .

	Fresno County		Tulare County
	Nov–Dec 2023	May–Jun 2024	Nov–Dec 2023	May–Jun 2024	Sep 2024
Households, Participants, and Sample Sizes
households	19	18	13	14	12
participants	31	30	22	22	22
personal samples	29	30	22	21	22
in-home samples	31	30	22	22	21
outdoor samples		7		9	16
Household Characteristics Used as Fixed-effect Predictors in Linear Mixed Models
number of people living in home	4 (2, 10)	4 (2, 11)	5 (1, 8)	5 (1, 7)	4.5 (1, 8)
number of people living in the home who were employed in agriculture	1 (0, 4)	1 (0, 3)	1 (0, 3)	0 (0, 3)	1 (0, 5)
warm-blooded pets inside the home	16%	17%	46%	7%	33%
chemical products used in or outside the home to control pests or flea/tick treatment applied to a pet within the past month	58%	39%	62%	86%	100%
weighted mass of chlorpyrifos applied for agricultural use upwind of home in 2018–2023 (kg)	5300 (4850, 5470)		3710 (3630, 3790)
weighted mass of acephate applied for agricultural use upwind of home in 2018–2023 (kg)	4840 (4100, 5070)		2990 (2880, 3040)
weighted mass of malathion applied for agricultural use upwind of home in 2018–2023 (kg)	7790 (7170, 7850)		6430 (6390, 6540)
participants owned their home	37%	44%	46%	43%	67%
number of rooms with carpet + number of rooms with area rugs	1 (0, 4)	1 (0, 6)	1 (0, 7)	1 (0, 4)	2.5 (0, 4)
mobile home	32%	33%	38%	50%	42%
home cooled using air conditioning (central, window unit, or mini split)	79%	72%	62%	64%	92%
home cooled using a swamp cooler	26%	39%	54%	50%	42%
home cooled by opening windows	58%	78%	77%	79%	50%
home cooled using fans	89%	78%	85%	64%	83%
estimated 24-h-average indoor air temperature (°C)	21 (18, 26)	26 (21, 30)	22 (17, 26)	25 (21, 28)	26 (22, 29)
swept or vacuumed during sample	74%	78%	46%	64%	75%

^aEach household characteristic used as a predictor in linear mixed models is also summarized on a per-household basis (except for the sampled air temperature, which is summarized on a per-in-home-sample basis). For binary characteristics, the percentage of households is listed. For numeric values, the median and (range) is listed.

^bOutdoor samples were collected at three homes in each county between May 17 and July 7, 2024 and at three homes in Tulare County between August 15 and September 10, 2024.

^cSee SI Section S1.3.2 for detailed information on how these values were calculated.

^dThese questions were not asked in a way that guaranteed rooms with area rugs were exclusive of rooms with carpet (i.e., some rooms could have had carpet and area rugs). Any rug larger than a door mat or bath mat was considered an area rug.

^eMost homes were houses or mobile homes, but two households in Tulare County lived in apartments.

^fHouseholds could use multiple cooling methods; see SI Section S1.3.1 for more information.

^gSee SI Section S1.3.3 for detailed information on how these values were estimated.

We fit four Markov chain Monte Carlo (MCMC) runs for each model with 8000 iterations in each chain. We discarded the first 4000 iterationsas burn-in. We assessed convergence using effective sample size and R̂ statistics. We assessed the fit of the model shown in eq using fitted vs residual plots, quantile-quantile plots, and posterior predictive diagnostics (the latter using the ‘bayesplot’ package in R). After examining these plots, we deemed the residuals to be acceptably homoscedastic and approximately normally distributed, thus confirming our decisions to not log-transform c _ij and to include all zero observations (i.e., nondetects) when fitting the model.

For each fixed effect in each model, the median, 50% credible interval (CI), and 95% CI were calculated. For logistic models, these values were converted to odds ratios.

3. Results and Discussion

Personal and in-home samples were collected with 31 individuals from 19 households in Fresno County and 35 individuals from 18 households in Tulare County (Table ). All participants identified as Hispanic or Latino/a. Exposure to pesticides is not equitably distributed in the US, and a prior analysis of California Pesticide Use Reporting data demonstrated that Central Valley communities with higher proportions of Hispanic residents tended to experience higher rates of chlorpyrifos application in 2011–2015. When asked, “What are the main concerns you and your family have about your environment?” as part of the visit surveys, 62% (41/66) of participants mentioned pesticides at least once. Exposure to pesticides is a pervasive and well-documented concern in predominantly Latino/a Central Valley agricultural communities. ,

Across all sampling campaigns, we made 76 unique visits to homes and collected 302 air samples. Twenty samples were excluded from data analyses because of quality assurance issues: no XAD-2 sorbent packet was installed in the sample cartridge for one personal sample, the sample log file was missing for one in-home sample and four outdoor samples, and 14 outdoor samples ran for <2 h because the internal temperature of the UPAS was too high. Overall, 124 personal samples, 126 in-home samples, and 32 outdoor samples were deemed successful. All “successful” samples ran for >14 h, and 97% (273/282) of successful samples ran for >20 h.

3.1. Detection Rates and Concentrations

Of the 11 pesticides for which samples were analyzed, five were detected in at least one sample: chlorpyrifos, acephate, malathion, diazinon, and naled (Table ). Although most uses of chlorpyrifos were banned in California starting on January 1, 2021, chlorpyrifos was detected more frequently than the other four pesticidesin 63% and 71% of personal and in-home air samples, respectively. Chlorpyrifos was detected in the indoor air during every one of 2+ visits to 18/37 (49%) study homes and at least once in 14 additional homes. There were only five homes in which chlorpyrifos was never detected in indoor air. Our ability to detect chlorpyrifos in air samples collected 3 to 4 years after the statewide ban on most applications was consistent with Shin et al.’s estimated half-life of ∼5 years for chlorpyrifos in homes. Chlorpyrifos was not detected in (a) any of the outdoor air samples (Table S3), (b) any of the 29 field-blank XAD-2 sorbent packets, or (c) any of the 15 particle-phase samples that were extracted from filters and analyzed. The latter result was also consistent with Shin et al.’s estimate that, given the octanol–water partition coefficient and Henry’s law constant for chlorpyrifos, > 99% of chlorpyrifos in indoor air should be present in the gas phase rather than in the particle phase. Additional results for chlorpyrifos are discussed below; results for acephate, malathion, diazinon, and naled are discussed in the Supporting Information.

2. Detection Rates as well as Median, 75th Percentile, and Maximum 24 h Average Gas-Phase Concentrations (ng m^–3) of Each Organophosphate Pesticide in Home and Personal Air Samples .

	Nov–Dec 2023				May–Jun 2024				Sep 2024
	in-home		personal		in-home		personal		in-home		Personal
Chlorpyrifos
detection rate	40/53	75%	30/51	59%	35/52	67%	35/51	69%	14/21	67%	13/22	59%
median	0.66		0.45		0.63		0.66		0.69		0.55
75th percentile	0.90		0.67		1.48		1.33		0.96		0.86
maximum	5.82		8.69		4.27		13.28		3.53		5.05
Acephate
detection rate	18/53	34%	12/51	24%	10/52	19%	13/51	25%	9/21	43%	7/22	32%
median	0		0		0		0		0		0
75th percentile	0.68		0		0		0.17		1.35		1.14
maximum	53.32		30.93		324.37		107.86		20.90		14.86
Malathion
detection rate	7/53	13%	6/51	12%	16/52	31%	17/51	33%	7/21	33%	7/22	32%
median	0		0		0		0		0		0
75th percentile	0		0		0.21		0.30		0.20		0.19
maximum	0.67		1.24		1.34		1.51		1.14		0.83
Diazinon
detection rate	5/53	9%	7/51	14%	18/52	35%	14/51	27%	2/21	10%	3/22	14%
median	0		0		0		0		0		0
75th percentile	0		0		0.12		0.12		0		0
maximum	2.18		1.50		1.73		2.49		0.21		0.31
Naled
detection rate	0/53	0%	0/51	0%	0/52	0%	0/51	0%	3/21	14%	5/22	23%
median									0		0
75th percentile									0		0
maximum									3.06		4.61

^aAll minimum values and all but one 25th percentile were zero.

The median (maximum) 24 h average concentrations of gas-phase chlorpyrifos in home air were 0.66 (5.82), 0.63 (4.27), and 0.69 (3.53) ng m^–3 in Nov–Dec 2023, May–Jun 2024, and Sep 2024, respectively. The median (maximum) 24 h average personal exposures to gas-phase chlorpyrifos were 0.45 (8.69), 0.66 (13.3), and 0.55 (5.05) ng m^–3 in Nov–Dec 2023, May–Jun 2024, and Sep 2024, respectively (Table and Figure ).

2.

Twenty-four hour average concentrations of gas-phase chlorpyrifos in home and personal air samples. Each circular marker represents a 24 h sample in which chlorpyrifos was detected. Boxes illustrate median (thick center line), 75th percentile (thin top line), and 25th percentile (thin bottom line) concentrations (Table ). Median, 75th percentile, and 25th percentile values were calculated with concentrations of zero (i.e., samples in which chlorpyrifos was not detected) included. Boxes that extend below the x-axis indicate 25th percentile concentrations of zero, which cannot be shown on the log scale. For each sample type and campaign, the number of samples in which chlorpyrifos was detected, as a fraction of the total number of successful samples, is shown at the top of the graph.

In-home levels and personal exposures to chlorpyrifos were strongly correlated (Spearman coefficient of 0.79), suggesting that personal exposures occurred primarily in the home (Figure ). In a prior analysis of time- and location-resolved PM exposure data from this study, most participants were found to spend the majority of their time at home. The median time spent at home was 96% during the 85/124 personal samples for which participant location data were available for ≥80% of the sample duration. The high proportion of time spent at home was likely attributable to the fact that 75% of samples were collected with participants who reported being homemakers, retired, unemployed, disabled, or employed seasonally. There were five personal samples with 24 h average chlorpyrifos exposures >1 ng m^–3 higher than the level measured in the participant’s home (Figure ). We hypothesize that these discrepancies occurred because the participant experienced higher exposures: (a) in a different location or (b) during an activity (e.g., cleaning, home maintenance) in which pollutant concentrations were higher in the participant’s breathing zone compared to the ambient concentration in the home common area.

3.

Comparison of 24 h average gas-phase concentrations of chlorpyrifos measured in home air and occupants’ personal air samples. If two concurrent in-home air samples were collected, the value shown on the x-axis represents the mean of the two measurements. Values of zero, which cannot be shown on the log scale, are represented by the points that are partially visible along the x- and y-axes.

3.2. Strengths and Limitations Associated with Our Measurement Approach

A key strength of this study is that we measured personal exposures to OP pesticides in air directly. These measurements allowed us to quantify exposures associated with the specific route of inhalation. We also collected these measurements using a novel air-sampling approach. Specifically, we used UPAS to sample OPs onto an XAD-2 sorbent that was installed behind a PTFE PM sampling filter. This approach had two key advantages: (1) it allowed us to measure exposures to PM and gas-phase OP pesticides simultaneously; (2) the UPAS was easier to deploy, smaller, lighter, and quieter during operation compared with conventional personal sampling equipment. As a result, we were able to assess exposures while imposing a smaller ergonomic burden on the study participants.

One potential limitation is that, using our measurement approach, the residence time of sampled air in the XAD-2 sorbent bed was shorter than it would have been if we had used OSHA Method 62 or NIOSH Method 5600 to measure exposure to OPs. , We used the UPAS to sample 2 L min^–1 air through 300 mg of XAD-2, which we estimated resulted in 0.56× the residence time that would have been achieved if we had sampled 1 L min^–1 air (the maximum flow rate specified in OSHA Method 62 and NIOSH Method 5600) through 270 mg of XAD-2 in the OSHA Versatile Sampler (OVS)assuming the same XAD-2 bed density and porosity in our sorbent packets and in the OVS (see SI Section S1.4 for additional information). It is possible that, due to this shorter residence time, not all of the OPs were captured on the XAD-2 sorbent and that our measurements are underestimates of the true OP concentrations in the sampled air.

We detected no chlorpyrifos, acephate, malathion, nor naled on any of our 29 field blanks. As a result, we are confident that our measurements were not affected by background contamination and do not believe that our measurements could be overestimates of OP concentrations in homes and breathing zones of rural Fresno County and Tulare County residents.

The agreement we observed between duplicate measurements of 24 h average gas-phase chlorpyrifos concentrations in home air provided confidence in the precision of our measurements (Figure ). We collected 50 pairs of successful collocated in-home air samples. For each pair, one sample consisted of PM_2.5 + gas-phase OPs and the other sample consisted of PM₁₀ + gas-phase OPs. Duplicate measurements were strongly correlated with a Spearman coefficient of 0.83 and a mean absolute difference of 0.2 ng m^–3. A linear model fit to the duplicate in-home measurements using Deming regression had a slope of 0.953 (95% confidence interval: 0.914, 0.993) and a y-intercept of −0.032 (95% confidence interval: −0.146, 0.082).

4.

Twenty-four hour average gas-phase concentrations of chlorpyrifos measured in 50 pairs of collocated in-home air samples. Each circular marker represents a pair of samples. Each sample was collected on an XAD-2 sorbent packet installed downstream of a PTFE filter. Values on the x- and y-axes represent samples collected with PM₁₀ and PM_2.5 inlets upstream of the filter samples, respectively. The gray line is y = x. The orange dash-dotted line represents the linear model fit to the data using Deming regression: y = 0.953x – 0.032.

Among the 37 pairs of collocated indoor samples for which the mean of the two 24 h average concentrations was below 1 ng m^–3, there were 10 pairs (27%) in which chlorpyrifos was not detected in either sample, 19 pairs (51%) in which chlorpyrifos was detected in both samples (with a minimum mean concentration of 0.38 ng m^–3 and a mean absolute percent difference between the duplicate measurements of 15%), as well as 8 pairs (22%) in which chlorpyrifos was detected in one sample but not in the other (Figure ). Similarly, among the 99 paired indoor-personal samples for which the mean of the personal and indoor concentrations was below 1 ng m^–3, there were 23 pairs (23%) in which chlorpyrifos was not detected in either sample, 47 pairs (47%) in which chlorpyrifos was detected in both samples, and 29 pairs (29%) in which chlorpyrifos was detected in one sample but not in the other (Figure ). These results suggested that our measurements were less precise at 24 h average concentrations <1 ng m^–3; however, a methodology that can quantify airborne OP concentrations as low as 1 ng m^–3 precisely remains a valuable tool for personal exposure assessment in epidemiological research. For the 13 pairs of collocated indoor samples for which the mean of the two 24 h average concentrations was between 1 and 6 ng m^–3, the mean absolute percent difference between duplicate measurements was 9.4%.

3.3. Comparisons to Prior Studies

The range of chlorpyrifos concentrations that we measured in 24 h in-home air samples was almost identical to the range measured by Bradman et al. in air samples collected in California farmworker homes in 2002 (before agricultural use of chlorpyrifos was banned). , Our results are also consistent with another recent study that detected chlorpyrifos exposure in personal air samples collected in California after 1 January 2021. , Bennett et al. detected personal exposure to chlorpyrifos in air among two individuals living in rural Tulare County in 2021 and 2022 at 8 to 14 h average concentrations (28 and 13 ng m^–3) similar to the maximum 24 h average personal exposure to gas-phase chlorpyrifos in air that we measured among adults living in rural Fresno and Tulare counties in 2023 and 2024 (13.3 ng m^–3). Taken together, their results and ours indicate that nondietary exposure to chlorpyrifos among rural California residents persists post ban.

A prior study based on data collected in 1999 and 2000 suggested that exposure to chlorpyrifos among people living in the US was dominated by consumption of treated foodsbut that nondietary ingestion, dermal exposure, and inhalation exposure contributed to measurably higher exposures among individuals who lived in a region where frequent agricultural application of chlorpyrifos occurred. Authors of prior studies have also hypothesized that chlorpyrifos might persist in indoor environments (e.g., homes, schools) in California post ban. Chlorpyrifos was known to be present in rural Fresno and Tulare County homes prior to the ban. Having ≥1 household members that worked in agriculture was associated with higher levels of chlorpyrifos in household dust collected in 2019, suggesting that the presence of chlorpyrifos in the home might have resulted, at least partially, from agricultural workers taking the pesticide home on their skin, clothes, and/or shoes. Authors of prior studies hypothesized that chlorpyrifos might persist in indoor environments for months or years due to low light levels (compared to outdoors) and consequent low rates of photodegradation. ,,, Shin et al. estimated the half-life of chlorpyrifos in home environments to be ∼5 years and demonstrated that most chlorpyrifos would be sorbed onto surface materials (e.g., carpet fibers and pads). Results from the present study lend support, in three ways, to the hypothesis that the persistence of chlorpyrifos in home environments is one factor contributing to continued exposure among Central Valley residents:

• 1.Detection rates and gas-phase chlorpyrifos concentrations in home air remained relatively consistent across all three of our sampling campaigns, which took place between November 2023 and September 2024 (Table and Figure ).
• 2.Chlorpyrifos concentrations in personal air samples were similar to and strongly correlated with chlorpyrifos concentrations in home air samples, which suggested that most personal exposure to chlorpyrifos in air occurred in the home (Figure ).
• 3.Although chlorpyrifos was detected in the majority of personal and in-home air samples, chlorpyrifos was not detected in any outdoor air samples.

3.4. Factors Associated with Pesticide Concentrations in Home Air Samples

As shown in Figure , factors associated with higher chlorpyrifos concentrations in home air were: (1) having carpet and/or area rugs in a larger number of rooms (β = 0.18; 95% CI = 0.06–0.30) and (2) higher indoor air temperatures (β = 0.096; 95% CI = 0.042–0.150). Carpets and rugs serve as reservoirs for chlorpyrifos (sorbed onto fibers, pads, and embedded dust). Authors of prior studies estimated that >98% of chlorpyrifos in homes was sorbed onto carpet fibers, carpet pads, or deeply embedded dust. When six data points from the household with the highest indoor chlorpyrifos concentrations were removed from the data set for the sensitivity analysis, the 95% CI for the sum of the number of rooms with carpets + the number of rooms with area rugs included zero (β = 0.086; 95% CI = −0.016 to 0.19; Figure S4); however, the higher concentrations in the home that was excluded from the sensitivity analysis could have been (at least partially) related to the higher-than-median number of such rooms in that home.

5.

Association between each fixed effect and the outcome in the linear mixed model for the full data set of chlorpyrifos concentrations in home air (eq ). Each estimate represents the change in concentration (ng m^–3) associated with either (a) a change from 0 to 1 for each binary fixed effect or (b) a unit increase in each numeric fixed effect. For each fixed effect, the median estimate (central line), 50% credible interval (range of the box), and 95% credible interval (range of the line segment) are shown. Fixed effects for which there was a 95% chance the association was positive (negative) are shaded in red (blue).

The factor most consistently associated with a higher probability of detecting chlorpyrifos in a homeand higher airborne chlorpyrifos concentrationswas the indoor air temperature (see Figures , S4, and S6). We hypothesized that higher air temperatures were associated with higher airborne chlorpyrifos concentrations because higher temperatures caused more chlorpyrifos to volatilize from solid materials (e.g., rug fibers and pads). ,

The median estimated associations between different cooling methods and indoor chlorpyrifos concentrations were in the expected directions (air conditioning = positive vs swamp cooler and opening windows = negative). Participants using air conditioners might have maintained more airtight conditions in their homes to conserve energy and preserve thermal comfort, and as a result, higher gas-phase concentrations might have built up as chlorpyrifos volatilized from dust and/or surfaces. Additionally, lower ventilation rates would have been likely to slow chlorpyrifos volatilization and removal rates over the years, thus leaving those homes with larger chlorpyrifos reservoirs in the present day. Conversely, homes that used swamp coolers or opened windows were assumed to have higher ventilation rates that would help remove chlorpyrifos from the indoor environment. However, the 95% CIs for all cooling methods included zeropossibly because many households reported using multiple cooling methods (presumably at different times).

Several predictors were included in the models because we expected them to relate to surface wear and resuspension of dust in the home: the number of people living in the home, the presence of warm-blooded pets inside the home, and whether someone in the home swept or vacuumed during the air sample. Having a larger number of people living in the home was consistently associated with a lower probability of detecting chlorpyrifos and lower chlorpyrifos concentrations in indoor air. This result was unexpected, as having a larger number of people living in the home would be expected to result in greater wear to solid surfaces that are potential reservoirs for chlorpyrifos. Mobile homes were also consistently associated with a lower probability of detecting chlorpyrifos in the home and lower airborne chlorpyrifos concentrations (Figures , S4, and S6). The rationale for this finding was also not clear; perhaps mobile homes had higher ventilation rates or tended to contain smaller volumes of materials to which chlorpyrifos sorbed at high rates.

With the temperature of the sampled air controlled for, later campaigns were associated with lower chlorpyrifos concentrations in home air. There was a ≥0.95 probability that both the May–Jun 2024 and Sep 2024 campaigns were associated with lower odds of detecting chlorpyrifos in home air (Figure S6). The Sep 2024 campaign was negatively associated with the chlorpyrifos concentration in home air for the linear model fit to the full data set (Figure ) but the 95% CI included zero in the model fit during the sensitivity analysis (Figure S4). If chlorpyrifos is not continuing to enter homes post ban, in-home levels are expected to decay over time.

For all parameters in all linear and logistic models, the number of effective samples was >2900 and the R̂ convergence diagnostic was <1.002, indicating a sufficient sample size and MCMC convergence. Fitted values vs residual and quantile–quantile plots for the linear mixed models for chlorpyrifos concentrations in home air are shown in Figure S5.

3.5. Implications for Efforts to Monitor and Reduce Exposures

Our results demonstrate persistent human exposure to chlorpyrifos through home indoor airan exposure pathway that would not be detected with the outdoor pesticide sampling approach employed by the California DPR. − This point is underscored by the fact that we did not detect chlorpyrifos in any of our outdoor air samples, whether we sampled for 24 h, 7 days, or 14 days at 1 or 2 L min^–1. Though outdoor sampling is an important and efficient approach to environmental monitoring, efforts to characterize human exposure to chlorpyrifos in air should also include personal or indoor sampling.

The California ban on almost all agricultural uses of chlorpyrifos was rightly lauded as a step toward environmental justice for predominantly Latino/a agricultural communities; still, our results demonstrate that this ban has not yet eliminated nondietary exposures to chlorpyrifos among residents of agricultural communities. Inside homes, chlorpyrifos (and other semivolatile organic compounds) can remain in air, dust, and solid surface materials (e.g., carpet fibers and pads) for many years.

Overall, the airborne chlorpyrifos concentrations measured in this study were very low. All 24 h average indoor and personal exposures were <14 ng m^–3, or less than 3% of the California DPR’s chronic health-screening level of 510 ng m^–3 for chlorpyrifos in ambient air, which suggests that these inhalation exposures do not pose substantial health risks to residents. Persistent chlorpyrifos in indoor environments likely poses the greatest risk to small children, who probably have more frequent contact with materials to which pesticides are sorbed and more frequent hand-to-mouth activity. Additionally, the inhalation exposures that we measured do not occur in isolation; instead, these exposures occur in concert with OP exposures resulting from consumption of treated foods, OP exposures through other routes, and exposures to other air pollutants. Though the low concentrations that we measured relative to health-screening levels might indicate that remediation of home environments to reduce chlorpyrifos levels is not necessary, we discuss below how various approaches might (or might not) benefit this goal.

Chlorpyrifos can be removed by ventilating indoor environments with outdoor air that does not contain chlorpyrifos; however, there are two caveats associated with the prospect of removing chlorpyrifos from indoor air via ventilation. First, as the concentration of chlorpyrifos in indoor air is reduced by ventilation, additional chlorpyrifos will desorb from solid surface materials where the vast majority of chlorpyrifos in homes is stored. Nevertheless, higher ventilation rates will increase the rate at which the total level of chlorpyrifos pollution in a home decays over time. Second, residents of California’s Central Valley, in particular, might be reluctant to open windows more frequently and bring in more outdoor air given: (a) the high levels of PM_2.5, PM₁₀, and ozone pollution that are typical of the region; (b) the high temperatures associated with most seasons; as well as (c) ongoing application of other pesticides to agricultural fields near study homesall of which pose health risks. These considerations highlight how efforts to reduce exposure to pollutants of indoor origin, in general, can conflict with efforts to reduce exposure to air pollutants and other hazards (like heat) that exist outdoorsespecially in the context of a warming climate that is expected to exacerbate outdoor air pollution and extreme heat events. ,

In-home levels of chlorpyrifos pollution can also be reduced by removing chlorpyrifos-containing dust. McCaule et al. demonstrated that steam cleaning reduced the amount of chlorpyrifos-containing dust in carpet; however, they did not evaluate whether steam cleaning reduced the amount of chlorpyrifos sorbed onto carpet fibers.

Authors of past studies in which chlorpyrifos was measured and/or modeled in air, dust, carpet fibers, and carpet pads inside homes reported that only a small proportion of the total mass of chlorpyrifos inside a home would be found in air or dust while the majority (>98%) would be sorbed onto carpet fibers, carpet pads, and on other solid surfaces. , Consistent with these prior studies, our linear mixed-model results suggested that having more rooms with carpets and area rugs was associated with a higher probability of detecting chlorpyrifos in home air and higher in-home chlorpyrifos concentrations. Consequently, the most effective approach to eliminating chlorpyrifos in the home might be to replace or remove carpet, rugs, and other fabric- and foam-based furnishings.

Glossary

Abbreviations

AC

air conditioning

AMN

air monitoring network

AST

Access Sensor Technologies

CI

credible interval

CSU

Colorado State University

DAP

dialkyl phosphate

DPR

Department of Pesticide Regulation

EPA

Environmental Protection Agency

LC-MS/MS

liquid chromatography-tandem mass spectrometry

NIOSH

National Institute for Occupational Safety and Health

OP

organophosphate

OSHA

Occupational Safety and Health Administration

OVS

OSHA Versatile Sampler

PM

particulate matter

PM_2.5

fine particulate matter

PTFE

polytetrafluoroethylene

PUR

pesticide use reporting

SVOC

semivolatile organic compound

TCPy

3,4,6-trichloro-2-pyridinol

US

United States

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsestair.5c00379.

• Additional methodological details on sample collection, sample analysis, and predictors used in mixed models as well as additional results for acephate, malathion, diazinon, and naled (PDF)

The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript. J.T. and G.E. contributed equally and should be considered as cofirst authors.

This work was funded by the National Institute of Environmental Health Sciences (NIEHS) under grant R44ES024041.

The authors declare the following competing financial interest(s): The Ultrasonic Personal Air Sampler (UPAS) v2.1 PLUS, which was used to collect air samples in this study, is manufactured and sold by AST. Jessica Tryner and Mollie Phillips are employed by AST. John Volckens is an inventor of the UPAS, a founder of AST, and has an equity interest in the company. The terms of John Volckens arrangement with AST has been reviewed and approved by CSU in accordance with CSU conflict-of-interest policies.

Published as part of ACS ES&T Air special issue “Indoor Chemistry in the Context of a Changing Climate”.