Community exposures to airborne agricultural pesticides in California: ranking of inhalation risks.

We assessed inhalation risks to California communities from airborne agricultural pesticides by probability distribution analysis using ambient air data provided by the California Air Resources Board and the California Department of Pesticide Regulation. The pesticides evaluated include chloropicrin, chlorothalonil, chlorpyrifos, S,S,S-tributyl phosphorotrithioate, diazinon, 1,3-dichloropropene, dichlorvos (naled breakdown product), endosulfan, eptam, methidathion, methyl bromide, methyl isothiocyanate (MITC; metam sodium breakdown product), molinate, propargite, and simazine. Risks were estimated for the median and 75th and 95th percentiles of probability (50, 25, and 5% of the exposed populations). Exposure estimates greater than or equal to noncancer reference values occurred for 50% of the exposed populations (adults and children) for MITC subchronic and chronic exposures, methyl bromide subchronic exposures (year 2000 monitoring), and 1,3-dichloropropene subchronic exposures (1990 monitoring). Short-term chlorpyrifos exposure estimates exceeded the acute reference value for 50% of children (not adults) in the exposed population. Noncancer risks were uniformly higher for children due to a proportionately greater inhalation rate-to-body weight ratio compared to adults and other factors. Target health effects of potential concern for these exposures include neurologic effects (methyl bromide and chlorpyrifos) and respiratory effects (1,3-dichloropropene and MITC). The lowest noncancer risks occurred for simazine and chlorothalonil. Lifetime cancer risks of one-in-a-million or greater were estimated for 50% of the exposed population for 1,3-dichloropropene (1990 monitoring) and 25% of the exposed populations for methidathion and molinate. Pesticide vapor pressure was found to be a better predictor of inhalation risk compared to other methods of ranking pesticides as potential toxic air contaminants.

Agricultural pesticides have historically been used in proximity to rural communities in California. Use near populated areas is increasing nationwide, as population growth expands into formerly rural farmland. Pesticides applied in agriculture can travel in the air through processes such as spray drift and post-application volatilization, sometimes for substantial distances (1)(2)(3). A wide range of agricultural pesticides has been found in ambient air (1,4,5). Agricultural pesticides have also been measured in indoor air, sometimes at increased concentrations (6)(7)(8). There is increasing public health concern regarding potential residential exposures to these agricultural pesticides and limited understanding about the potential for such exposures. Acute health effects, such as eye, respiratory, and gastrointestinal irritation, fatigue, and headaches, have been associated with some instances of agricultural pesticide drift into California communities (9)(10)(11). However, there is a risk of other, nonacute health effects from airborne agricultural pesticides, many of which are less readily apparent than irritant effects.
Some methods for ranking agricultural pesticides by their potential hazard as air contaminants have been proposed based on use, volatility, toxicity, and so on (12,13). Ultimately, the rankings are used to determine exposure reduction or public health priorities. One of the initial uses of the ranking developed by the California Department of Pesticide Regulation (CDPR), called the (pesticide) toxic air contaminant (TAC) ranking, is to direct air monitoring for agricultural pesticides in California (13). The California Air Resources Board (CARB) conducts this ambient air monitoring in agricultural communities, which are selected on the basis of area use of the monitored pesticide, and in regional urban centers (4). For the monitored pesticides, an opportunity exists to calculate inhalation risk.
The conventional approach to risk assessment typically uses single health-conservative exposure values, such as inhalation rate (39,40). The resulting risk estimate, while health conservative, gives little information about the likelihood of risk in an exposed population. In contrast, probability analysis, presented here, uses distributions for exposure variables to estimate a range (likelihood) of risks. These risks expressly apply to the populations in the vicinity of the air monitoring stations. The monitored pesticides can be ranked by inhalation risk in the exposed communities using these risk estimates. Estimates of the total California population with a similar exposure potential can also be made by determining the agricultural pesticide use density near the air monitoring locations and then enumerating the California population living in areas with similar or higher use densities.

Methods
Pesticide monitoring. Air monitoring methods have been discussed in detail by Baker et al. (4). Briefly, pesticides under evaluation by the CDPR as possible TACs are sampled by CARB in the California county and in month(s) with the reported highest use of each pesticide in recent years. In California, all agricultural applications of pesticides are reported, including geographical location and date of application [Pesticide Use Report (PUR) data] (41). Complete agricultural pesticide use reporting has been a California requirement since 1990, with restricted pesticide usage reported pre-1990. These PUR data are error checked and maintained by the CDPR. On average, three to four rural agricultural communities and a regional urban comparison site were selected for monitoring of each pesticide. One to two air monitors were placed on the roofs of community buildings such as schools. Air samples were typically collected by pump-and-adsorbent-cartridge capture methods, using low-and medium-volume flow rates. Monitoring for methyl bromide and 1,3-dichloropropene in 2000 also used evacuated Silcosteel canisters (18,19). Generally, 24-hr samples were collected each week for several weeks. The community air data are descriptive of average pesticide air concentrations in high-use agricultural regions, since community monitors were not positioned near known field applications.
Pesticides among the top 20 potential California TACs and U.S. EPA HAPs with monitoring data are listed in Table 1.
Monitoring data were not available for potential TAC pesticides ranked 6th (p-dichlorobenzene), 7th (cyanazine), 12th (alachlor), and 13th (dimethoate). All pesticides have at least one month of community air-monitoring data, except molinate (1.5 weeks), 1,3-dichloropropene in 1990 (2 weeks), and MITC (2 weeks). The statistical analysis of air data in this report differs from that of Baker et al. (4), who did not include nondetectable analytic results in the statistical estimate of the mean. We entered nondetectable compounds in estimates of the mean as zero values for pesticides detected in < 10% of air samples, and at onehalf the minimum quantitation limit for all other pesticides. 1,3-Dichloropropene monitoring includes community air data collected before its use was suspended in California in 1990 and community data collected following reintroduction in 1996 (16)(17)(18)(19).
PUR data were obtained electronically from the CDPR for 1986-1999, the most recent year available at the time of this report (41). Agricultural pesticide use was evaluated within 1.5-3 miles of each monitoring station, for the year of monitoring. PUR data are available by township, range, and section, a section being approximately 1 mi 2 . The adjacent years 1999 and 1989/1991 were used, respectively, for PUR analysis for 2000 monitoring and for 1990, which was a transition year to complete use reporting. We calculated annual pesticide use density (pounds/square mile) within 1.5-3 miles of each air monitoring station for selected pesticides, based on PUR data for the year of monitoring (1999 proxy year for 2000 monitoring). Population estimates were derived for all 1990 California census block groups with annual average pesticide use (pounds/square mile) greater than or equal to that in the vicinity of the air monitoring sites, using methods previously described (12).
Risk assessment. The risk assessment evaluates inhalation exposures to adults and children ≤ 12 years of age. Noncancer risks are assessed for chronic (> 1 year), subchronic (≥ 15 days), and acute exposures (typically 1-24 hr) (40). Cancer risks assume a lifetime exposure. The following equations were used to estimate inhalation risk: where C air = concentration of pesticide in community air (mg/m 3 ); IR = inhalation rate (liters/kilogram body weight-day); CF = conversion factor (0.001 m 3 /L air); EF = exposure frequency (all chronic; months/12 months); ED = exposure duration (cancer risk; years/70 years). For risk, Noncancer Risk (hazard quotient) = Intake (mg/kg-day)/RfD (mg/kg-day) [2] Cancer Risk = Intake (mg/kg-day) × PF (mg/kg-day) -1 . [3] Noncancer risk is defined as the ratio (hazard quotient; HQ) of the estimated intake to the reference dose (RfD). The RfD is the dose at or below which adverse noncancer health effects are not estimated to occur. Noncancer risks from exposures to pesticides with nonsystemic (portal-of-entry) effects (e.g., respiratory irritation) were assessed by eliminating IR and CF from Equation 1; in other words, the effect is dependent on air concentration. The resulting exposure estimate in milligrams per cubic meter is divided by the reference value in milligrams per cubic meter. Cancer risk estimates (Equation 3) use the potency factor (PF), a numerical estimate of the potency of the carcinogen. The PF multiplied by the estimated intake yields an estimate of the cancer risk over a lifetime from that exposure alone. For probability distributions of risk, we conducted Latin hypercube analysis using commercial software, Crystal Ball v. 2000 (42). We used 10,000 equation solutions, or trials, to define the range of risk for each exposure scenario. The sample size for Latin hypercube sampling was 5,000. Contribution to variance was used for sensitivity analysis. Probability distributions were defined for pesticide air concentrations, inhalation rates, and exposure frequencies. Variables are listed in Tables 2 and 3.
Air concentrations. Log-normal distributions were identified for the pesticide air monitoring data based on preliminary histogram analysis (not shown). Pesticide concentrations listed in Table 1 in micrograms per cubic meter were converted to milligrams per cubic meter, natural log transformed, and means and standard deviations calculated on the transformed data set for the log-normal distribution (not shown). Air data from all rural communities combined (Table 1) were used to estimate chronic exposures. We used the highest 15consecutive-day air concentration to estimate subchronic exposures (Table 1), except for EPTC and methyl bromide. These used 22day and 6-week intervals, respectively, corresponding to the intervals used to establish their subchronic RfDs (Table 4). As with the air data used for chronic exposure estimates, means and standard deviations were calculated on natural log-transformed data for subchronic exposures (not shown). Acute exposure estimates used the sample maximum concentration measured in community ambient air ( Table 1). Because of this, the air concentration in the equation for all acute exposures is a single value rather than a distribution.
Inhalation rate. Inhalation rate distributions are based on analysis of ventilation rate data for a population cross-section (43). Inhalation rates follow a gamma distribution. They are defined for a child (≤ 12 years) and an adult (> 12 years) for noncancer risk, and a lifetime (0-70 years) for cancer risk.
Exposure frequency. Exposure frequency refers to the fraction of a year over which an exposure occurs (e.g., 3 months/12 months = 0.25). It applies only to chronic exposures, which are, by definition, a year or more (40). EF estimates were based on analysis of pesticide use report (PUR) data for the agricultural sections immediately surrounding each air monitoring site, typically within a 1.5mile radius (41). For methyl bromide, MITC, and 1,3-dichloropropene, the monitoring data showed elevated air concentrations in background urban sites, and/or there was no reported use in a 1.5-mile radius at the time when air monitoring showed the presence of the pesticide. For these pesticides, we expanded the radius for PUR analysis to 3 miles. "Triangular distributions," used when estimates of the minimum, most likely mean, and maximum points in the distribution are available, were used to describe exposure frequency. In the triangular distribution, the minimum chronic exposure was assumed to be 1 day/year for all pesticides. The most likely exposure period included months with agricultural pesticide use at least 50% of that in air monitoring month(s), in the radius around each monitoring site. In the case of DEF, the most likely exposure period equaled the monitoring period, approximately 10 weeks of the year (28). The maximum exposure period included all months where use of each pesticide was ≥ 10 pounds in the defined radius.
[In the absence of indoor air monitoring data, daily exposures assume a 24-hr exposure at the measured ambient air concentrations. Supporting this are other study findings of comparable or higher concentrations of a range of agricultural pesticides, including MITC, indoors compared to outdoors (6)(7)(8)].
Exposure duration. An ED equal to 1 (a lifetime) is assumed for the cancer risk assessment. It was chosen by default because factors  Log normal µ, σ of ln-transformed data such as mobility of the population are not well defined. In the risk assessment, we averaged air concentrations over all rural communities monitored for a pesticide as a proxy for chronic "regional" exposures. These regions include distances within which an average resident might reasonably be expected to move, using national data on number of miles moved by home buyers (44). Table 4 lists noncancer RfDs and cancer PFs. These are from the U.S. EPA (45)(46)(47)(48)(49)(50)(51)(52)(53)(54), the CDPR (55-64), the California Office of Environmental Health Hazard Assessment (OEHHA) (65,66), and the Agency for Toxic Substances and Disease Registry (ATSDR) (67,68). RfDs are listed in Table 4 by exposure duration (acute, subchronic, or chronic) and corresponding target organ/toxicity. Note that the term "RfD" is used in Table 4 headings to indicate all noncancer reference values cited from various sources, avoiding the use of multiple terms developed by various agencies, such as reference concentration (45), minimum risk level (67), or reference exposure level (65). Reference values shown in air concentration units of milligrams per cubic meter, rather than milligrams per kilogram body weight, are based on portal-of-entry effects. PFs are listed, with the cancer classification of the U.S. EPA Office of Pesticide Programs (50). When available, values based on inhalation studies were always chosen over oral values. Values based on oral studies are indicated in Table 4 by (o), or (o→i), indicating oral to inhalation extrapolation by the listing agency. Where agencies/programs listed values that differed from one another by > 2-fold, the high and low values are both listed in Table 4. This occurred for the following: chlorpyrifos, 1,3-dichloropropene, dichlorvos, EPTC, and molinate with chronic RfDs; diazinon, EPTC, and molinate with subchronic RfDs; 1,3-dichloropropene, methyl bromide, and MITC with acute RfDs; and 1,3-dichloropropene and dichlorvos with cancer PFs. In these cases, risks were estimated separately using each of the two values.

Reference doses and potency factors.
With two exceptions, all of the RfDs and PFs listed in Table 4 are based on administered doses, with no adjustment for absorption by the listing agency. The exceptions, for DEF and dichlorvos, are footnoted in Table 4, and the absorption factor is included in Equation 1 of their exposure assessment. Acute RfDs have not been published for endosulfan, propargite, and simazine. In these cases, we identified the no-observed-adverse-effect level (NOAEL) from the most sensitive teratology study on file with the CDPR (58). The teratology studies were chosen because they are well-reviewed, short-term studies, albeit oral, on a potentially sensitive subpopulation of pregnant animals. The NOAEL was divided by the standard default uncertainty factor of 100 (10 for extrapolation from an animal to human population and 10 for potentially sensitive human subpopulations) to estimate an acute RfD for these three pesticides.
The federal Food Quality Protection Act (FQPA) of 1996 directs the U.S. EPA to use an additional safety factor of up to 10-fold, if necessary, to account for data uncertainties when evaluating pesticide risks to infants and children (38). The U.S. EPA is in the process of assigning these FQPA safety factors to pesticides that may pose additional risks to children. Available FQPA factors are listed in Table 4. In this risk assessment, the "adult" RfDs shown in Table 4 were divided by the available FQPA factor when assessing all risks to children.

Results
Noncancer and cancer risks are estimated for the 50th, 75th, and 95th percentiles of likelihood of risk (probability estimates, Tables 5-7) in the monitored communities. Two risk entries per pesticide exposure scenario indicate use of two different RfDs or PFs (Table 4). These pairs denote a range of RfD or PF values used by different agencies or programs, and, consequently, a greater range in the risk estimates for these pesticide exposure scenarios. The range was most notable for MITC acute noncancer risks.   (50) Noncancer risks are presented in Table  5 for children ≤ 12 years of age and in Table  6 for adults. Risks are ranked in approximately ascending order. Results are presented as HQs, that is, intake divided by the reference dose in milligrams per kilogram-day, or exposure divided by the reference value in milligrams per cubic meter (Equation 2).
[MITC and chloropicrin acute risk estimates are presented as point estimates. For these, the only exposure distribution (inhalation rate) is eliminated from the equation because MITC and chloropicrin acute target health effects are nonsystemic.] The risks to children are consistently greater than for adults because children have a greater inhalation-tobody weight ratio and, in some cases, because the FQPA factor for children lowered the reference dose (Table 4). Four pesticides have HQs > 1 for an estimated 25-50% of the exposed populations of children (75th and 50th percentiles of risk). These include MITC for subchronic and chronic exposures; chlorpyrifos for acute and subchronic exposures; 1,3-dichloropropene for subchronic exposures in 1990; and methyl bromide for subchronic exposures in 2000a and 2000b ( In 2000, joint air monitoring was conducted for methyl bromide and 1,3-dichloropropene in two regions in California, identified in Tables 1 and 3 and Tables 5-8 as  2000a and 2000b. The 2000a monitoring was in the county with high use of 1,3-dichloropropene (and secondary use of methyl bromide), while 2000b was in counties with high use of methyl bromide (and secondary use of 1,3-dichloropropene). While the available methyl bromide air monitoring data do not reflect the history of regulatory actions, this may be due to sampling limitations in the 1986 air monitoring, which used a method with a much higher minimum quantitation limit (4.2 µg/m 3 ) compared to later sampling (Table 1) (4,18,19).
Reference doses are based on studies identifying the most sensitive target organ(s) and critical health effect(s) for a length of exposure. For MITC, the critical effect for subchronic and chronic exposures is respiratory: nasal epithelial atrophy in animal studies (62). The critical subchronic effect for methyl bromide exposure is neurologic: decreased responsiveness in animal studies (61). Chlorpyrifos acute and subchronic critical effects are also neurologic: enzyme cholinesterase inhibition in animals (47). The 1,3-dichloropropene critical subchronic effect is respiratory: nasal epithelial changes, also in animal studies (56). An HQ > 1 is generally a trigger for regulatory scrutiny. However, because uncertainty (safety) factors, typically 100-fold, are incorporated into the reference values, this does not necessarily mean that an individual will become ill from such an exposure.
In some cases, particularly chloropicrin and MITC, the sampling intervals were greater than the RfD interval. The chloropicrin samples were taken over 4 hr, while the acute RfD is for a 1-hr exposure. MITC samples were over 24 hr, while the acute MITC RfDs are for 1-to 8-hr exposures (Tables 1 and 4). These acute RfDs were used here without modification, as it was beyond the scope of our analysis to rescale them. Risks for these acute exposures consequently may be underestimated.
Lifetime cancer risks for pesticides with cancer potential and available cancer potency factors are presented in Table 7. Increased regulatory scrutiny of cancer risk often occurs when the estimated risk reaches 1 × 10 -6 to 1 × 10 -5 , 1/1,000,000 to 1/100,000, excess lifetime cancer risk. Lifetime cancer risks that reach or exceed 1 × 10 -6 for an estimated 25-50% of the exposed populations include 1,3-dichloropropene for 1990, methidathion, and molinate ( Table 7). The uncertainties are relatively greater for methidathion and molinate estimates compared to 1,3-dichloropropene. Methidathion and molinate are listed by the U.S. EPA as possible human carcinogens (limited evidence), while 1,3dichloropropene is a probable human carcinogen and, unlike methidathion and molinate, has cancer PFs specific to the inhalation route ( Table 4). As with noncancer risks, the true cancer risks to an individual are likely to be lower, due to upper bound estimates established for potency factors and other health-conservative assumptions.
1,3-Dichloropropene use permits were suspended in California in 1990 after high concentrations were found in community air (16). The reinstatement of 1,3-dichloropropene permits in 1996 included a number of California-specific regulatory controls. Cancer risks for 1,3-dichloropropene are reduced for the subsequent monitoring years in 1996 and 2000 (Table 7).
In this risk assessment, the variability in ambient air concentrations contributed the largest part of the variance in probability distributions for chronic and subchronic exposures (acute exposure estimates used the maximum air concentration). Several pesticides have air concentrations that span two to Articles • Agricultural pesticides in community air Environmental Health Perspectives • VOLUME 110 | NUMBER 12 | December 2002 three orders of magnitude (Table 1). For chronic exposures, the order of percent contribution to variance (mean ± SD) was air concentration (75 ± 16), exposure frequency (22 ± 14), and inhalation rate (3 ± 3). The percent contribution to variance for subchronic exposures was air concentration (96 ± 4) and inhalation rate (4 ± 4; mean ± SD).
(Exposure frequency applies only for chronic exposures.) Conventional point estimates of risk are typically more conservative than 50th percentile estimates of risk presented here. This is due to the use of conservative exposure assumptions (e.g., arithmetic mean air concentrations, upper bound inhalation rate estimates). In comparison to stochastic risk estimates, use of conventional assumptions generally resulted in point estimates of risk at or above the 75th percentile (not shown).
In this risk assessment, community exposures and risks were characterized for the populations within a few miles of the air monitoring stations. We also estimated the total California population living in census block groups with a similar or greater pesticide use density, compared to the monitored communities (Table 8). Annual pesticide use density (pounds per square mile) was calculated in the vicinity of the air monitoring sites for chlorpyrifos, metam sodium, methyl bromide in 2000 monitoring (2000b), and 1,3dichloropropene in 2000 monitoring (2000a). Methyl bromide had the largest estimate of the total exposed California population, 208,757, followed by metam sodium (MITC), 185, 441, and 1,3-dichloropropene, 43,246. Chlorpyrifos had the lowest estimate of the total exposed California population, 2,523. Table 9 shows pesticide use in the county of air monitoring. Use in the year of monitoring is compared to average use during 1991-1999, for both total use and use per square mile of agricultural land in the county (69). For many pesticides, use in the year of monitoring is generally representative of average use over the past several years. For naled (dichlorvos parent), a year-by-year analysis shows an apparent steady decline in use (not shown). The greatest increase in county use occurs for metam sodium (the parent of MITC). A yearly analysis shows an increase of 3-to 5-fold every year from 1995 onward, compared to 1993 (not shown).
Average annual MITC use, within 3 miles of the community air monitoring locations, has increased more than 2-fold since the 1993 monitoring (not shown).
We evaluated several predictors of the chronic inhalation risks estimated in this report, using Spearman rank correlation coefficients. The California ranking for potential pesticide toxic air contaminants (Table 1) was not significantly correlated with the child chronic risk ranking (r = 0.22, p = 0.43). For example, propargite and chlorothalonil were first and second in the pesticide toxic air contaminant ranking, but these pesticides were found to have among the lowest inhalation risks. The chronic reference dose ranking ( Table 4) was significantly correlated with the child risk ranking (r = 0.63, p = 0.01). The pesticide vapor pressure ranking (70) was the best predictor of the child chronic risk ranking (r = 0.70, p = 0.003). Similarly, vapor pressure (r = 0.60, p = 0.12) was a better predictor of lifetime cancer risk ranking (Table 7) than the cancer potency factor (r = -0.07, p = 0.87). Vapor pressure has been highly correlated with downwind pesticide concentrations in previous studies (71). Among the 15 pesticides in this study with detectable air concentrations, vapor pressure was also highly correlated with the geometric mean air concentrations in rural communities (r = 0.77, p < 0.001).
Risk interpretation examples: 1 × 10 -6 = 1/1,000,000 lifetime excess cancer risk; 2 × 10 -4 = 2/10,000 lifetime excess cancer risk. a Two estimates are calculated because two potency factors were available (see Table 4). bromide. The actual health impacts of exposure to these pesticides may be less than estimated here, due to the use of healthconservative factors common in risk assessment (e.g., uncertainty factors of 10-100 or more incorporated into reference doses and upper-bound estimates of cancer potency factors). This is also a screening risk assessment with a number of uncertainties, including the existence of differing reference values and potency factors which can influence the risk estimates (e.g., the acute hazard quotients for MITC). These differences, however, do not alter the overall risk ranking and conclusions of the report. Our risk estimates suggest caution in the expanded use of these fumigants. Atmospheric dispersion modeling has been used across census tracts in the contiguous United States to rank risks for a broad range of federal HAPs (72). Our report is distinct in using actual community air monitoring data to rank risks from inhalation exposures to agricultural pesticides. Our risk assessment focuses on California communities in high-use pesticide areas, at or above the 90th percentile of density of use for most pesticides (12). Hence, the risk assessment characterizes the most exposed populations of California. It is worth noting that the pesticide air concentrations in this risk assessment are ambient community air measurements, not measurements near field applications. Near-field air concentrations are typically much higher than ambient community air data (4). In the absence of effective regulatory controls, proximity to field applications could contribute significantly to a higher short-term exposure burden to nearby residents.
Pesticide exposures and risks are characterized for the communities around the air monitoring locations. However, the potential for exposures in other residential areas clearly exists. For example, census data indicate that > 185,000 people live in areas in California with a density of use of metam sodium (MITC) greater than the community air monitoring locations. More than 208,000 people live in areas where the density of use of methyl bromide exceeds use around recent air monitoring locations. These data suggest a potential for exposures and risks, similar to those calculated in this risk assessment, for hundreds of thousands of people in California.
Children's exposures require particular attention (38). Risks to children are uniformly higher than those of adults due to a greater inhalation rate-to-body weight ratio and other factors. Our report specifically assesses risks to children from a rarely evaluated exposure-inhalation of agricultural pesticides. This pathway is important because an increasing number of children live along the nation's agricultural-urban edge. For example, in California we estimated that > 53,000 children lived in census block groups where methyl bromide use density exceeded the use density near recent community air monitoring locations.
at reducing airborne emissions, particularly for fumigants (74)(75)(76)(77). This may result in lower exposures and risks under California use conditions. Metam sodium use has increased markedly nationwide, an estimated 7-to 12-fold since the late 1980s (78). Methyl bromide use is likely to decrease nationwide as other fumigants, including 1,3-dichloropropene and metam sodium, are adopted as substitutes in the mandated reduction under the Clean Air Act (79). Chlorpyrifos was the most widely used U.S. household pesticide before the 2000 decision by the U.S. EPA to eliminate most residential, school, and park uses (and cancel agricultural use on tomatoes and greatly reduce use on apples and grapes) to reduce exposures to children (80). Other agricultural uses of chlorpyrifos are less impacted by regulatory changes (80). The California air monitoring in 1996 for chlorpyrifos was in communities in citrus-growing areas. The chlorpyrifos concentrations detected in the California monitoring are several-fold higher than those found in urban areas (Table 1) (81), making it less likely that these concentrations were a result of residential use.
Notable uncertainties in this risk assessment occur in hazard identification, doseresponse assessment, and exposure assessment. Stochastic analysis was used to characterize exposure variability. Among the distributions used, annual exposure frequency was the least characterized, relying on pesticide use data to establish a triangular distribution. Ambient air data were used in the risk assessment in the absence of indoor air data. Both higher and lower indoor air concentrations relative to ambient air have been historically reported for agricultural-use pesticides (6)(7)(8).
The risk assessment only considered inhalation exposures. Ingestion and dermal pathways are also likely exposure routes (82)(83)(84). Young children in particular, with a higher ingestion rate of fresh fruits and vegetables and higher contact rates with soil and housedust through hand-to-mouth activities, are at risk for cumulative pesticide exposures by such routes (82). There is also a large subpopulation at potentially higher risk: farmworker/farm children. An estimated 20% of 5 million U.S. farmworkers live or work in California (85). Increased exposures of children of farmworkers and farmers have been repeatedly documented, through occupational take-home exposures and other routes (86)(87)(88)(89).
With respect to existing toxicologic data, there are important gaps in health reference levels specific to the inhalation route. There are also a number of pesticides for which no FQPA safety factor for children has been established. Several pesticides have noncancer or cancer health reference levels that vary between agencies and programs, resulting in a wider range in risk estimates. Uncertainties in hazard identification are also present. Emerging concerns, such as endocrine disruption and neurologic development, may not have been fully evaluated in toxicity testing (90,91). There is also a lack of toxicity data on exposures to multiple pesticides. Combined exposures to pesticides have been shown to cause effects not observed individually (92)(93)(94) and may potentiate toxicity in some pesticide combinations, for example, cholinesterase inhibitors (95)(96)(97).
Several pesticides found in the air of communities (Table 1) are organophosphate (OP) cholinesterase inhibitors, including chlorpyrifos, DEF, diazinon, dichlorvos, EPTC, and methidathion. Children may be exposed to multiple OPs, all sharing a common toxicity, through multiple routes. Exposure studies have shown OP pesticide accumulation on children's toys as a result of prolonged vaporization from other deposits (98), indoor transport from outdoor applications of OPs, with redistribution into indoor air and surfaces (99) and increased OP metabolites in children living near agricultural applications (86,87). Exposure to organophosphate pesticides may potentially impact neurodevelopment, growth, and respiratory health in children (100).
Pesticides with the highest risks in this risk assessment, MITC, methyl bromide, and 1,3dichloropropene, impact some of these same target organs. Methyl bromide is a developmental, neurologic, and respiratory toxin. MITC and 1,3-dichloropropene are also respiratory toxins. The potential for exposure to more than one of these pesticides clearly exists. Methyl bromide and 1,3-dichloropropene were detected together in dual air monitoring. Several of the California communities selected for air monitoring for a pesticide were reselected in later studies because they were in the highest-use area for another pesticide. Toxicity, epidemiology, and exposure studies addressing likely combinations of these pesticides are needed.
Risk ranking effectively identifies the pesticides most in need of further scrutiny from inhalation exposure to agricultural pesticides. Vapor pressure is a significant predictor of this ranking of inhalation risks. Candidate pesticide air contaminants may be most readily identified using a ranking system that places greater weight on vapor pressure.