471 20 Models of Exposure to Pesticides Robert A. Canales Harvard University James O. Leckie Stanford University CONTENTS 20.1 Synopsis 471 20.2 Introduction 472 20.3 Activity Data 472 20.4 Routes of Exposure 473 20.4.1 Dermal Exposure 473 20.4.2 Nondietary Ingestion Exposure 475 20.4.3 Dietary Ingestion Exposure 475 20.5 Exposure Models 475 20.5.1 The CalTOX Approach 476 20.5.2 Calendex™ 476 20.5.3 LifeLine™ 477 20.5.4 The Cumulative and Aggregate Risk Evaluation System (CARES) 477 20.5.5 The Stochastic Human Exposure and Dose Simulation (SHEDS) 478 20.6 Issues in Modeling Exposure 478 20.6.1 Combining Data for Model Inputs 478 20.6.2 Temporal and Spatial Variability 479 20.6.3 Co-Occurrence of Environmental Contaminants 480 20.6.4 Estimating Dose from Exposure 480 20.7 Questions for Further Review 481 References 481 20.1 SYNOPSIS This chapter is devoted to the exploration of models of exposure to pesticides, their complexities, and their emergence after the passing of the Food Quality Protection Act of 1996. A discussion begins with a basic component of exposure models — human activity patterns — and establishes a foundation for understanding the many models that have been proposed for estimating exposure. The quantification of dietary ingestion, nondietary ingestion, and dermal exposure is surveyed. Several exposure models are explored in terms of inputs, modeling objectives, and methods. The chapter ends with a number of other issues important in modeling exposure to pesticides. These © 2007 by Taylor & Francis Group, LLC 472 Exposure Analysis issues include temporal and spatial variability, dose estimates, the co-occurrence of contaminants, and the use of literature data for model inputs. 20.2 INTRODUCTION Interest in measuring and modeling exposure to pesticides through all important pathways has increased since the development of the Food Quality Protection Act (FQPA) of 1996. The impetus for the Act was a report by the National Research Council (NRC) entitled Pesticides in the Diets of Infants and Children (NRC 1993). The report examined policies of government agencies in regulating pesticide residues in foods. The influence of the NRC publication is evident in the Act’s resolution that there should be a consistent, health-based standard for all pesticides used on foods, pesticide regulation should address the vulnerability of potentially sensitive groups (e.g., children), and older pesticides should be reassessed and perhaps replaced with safer substitutes. The report also recognized the importance of accounting for exposures from all non-occupational sources and from all routes of exposure (i.e., “aggregate exposure”), and the effects of exposure to multiple pesticides with “common mechanisms of toxicity” (i.e., “cumulative exposure”) (USEPA 1997). 20.3 ACTIVITY DATA Several types of aggregate exposure models exist and they may be characterized in a number of ways. For instance, models may be categorized by their purpose (e.g., screening analysis, heuristic exploration, regulatory compliance), the form of necessary input data (e.g., probability distributions, point estimates), or their output (e.g., lifetime risk, short-term exposure, population estimates). A crucial distinction unique to human exposure models is the type and detail of activity data. The U.S. Environmental Protection Agency (USEPA) recognizes that models for assessing exposure can be structured around different forms of activity pattern data (Cohen Hubal et al. 1998). The different approaches dictate the level of detail, the level of accuracy, the inherent uncertainty, and the computational structure of an exposure model. The form of the output and the adequacy of the exposure and dose estimates for risk determinations are also affected by the qualitative nature of the activity pattern data. The common forms of activity pattern data include macro-, meso-, and micro-level activities (Figure 20.1). Macro-level activities describe human activities by general location and general activity. Nor- mally collected via diaries or interviews, examples of this form of activity include playing outdoors, reading indoors, or working in an office. This information is typically used to estimate inhalation FIGURE 20.1 Forms of activity patterns used in modeling exposure. working 4 hr eating 1 hr working 5 hr commuting 1 hr time segment 1 mouth 9 contacts soil 3 contacts time segment 2 grass 7 contacts soil 5 contacts soil outdoor 5 s grass outdoor 18 s soil outdoor 29 s open patio 3 s table patio 14 s pet patio 2 s Macro-Level Activity Patterns Meso-Level Activity Patterns Micro -Level Activity Time Series © 2007 by Taylor & Francis Group, LLC Models of Exposure to Pesticides 473 exposures, but it has been used in assessing children’s residential exposure to pesticides (Cohen Hubal et al. 1998). The quantification of exposure from the use of macro-level activities, however, inherently lumps several details, such as dermal contacts and objects inserted into the mouth, that may be important in estimating nondietary ingestion and dermal exposure (Cohen Hubal et al. 1998; Michelson and Reed 1975). Using meso-level activities, exposure can be modeled with more specific data on microenvi- ronments and contact behavior. Meso-level activity data track some characteristic of activity over a certain time frame (Figure 20.1). As an example, an individual may complete 9 hand-to-mouth contacts and 3 contacts with soil some time within a 30-minute period, and 7 contacts with grass and 5 contacts with soil some time in the following 30 minutes. These data may be further reduced to an hourly or daily basis. Data collected in this manner may provide an opportunity to preserve the sequence of activities, at least at some level, and characterize the variability of the data (Reed, et al. 1999). It is obvious that data requirements for the meso-level activity approach are more extensive when compared to the collection of macro-level activities (Cohen Hubal et al. 1998). Furthermore, since there is little regard to order within the arbitrary time segment, this activity form may not adequately describe sequential contacts necessary for including certain phenomena when estimating dermal and nondietary ingestion exposure. A more detailed activity pattern form can account for sequential micro-level contact events and microenvironments, and allow for the estimation and inclusion of other time-varying factors (Figure 20.1). In estimating dermal exposure this information is an improvement over other activity forms since detailed data on sequential contacts are necessary for appropriately handling removal mech- anisms and loading from multiple media. This more specific representation of activities is also useful for understanding the fundamental mechanisms and behaviors influencing exposure. Data collection and modeling tasks associated with micro-level activity, however, are more extensive when compared to other forms of activity data. Collecting such data may be infeasible for large populations. 20.4 ROUTES OF EXPOSURE Aggregate exposure models use activity data to assign environmental concentrations and related transfer parameters to route-specific equations. Separate equations then exist for inhalation, non- dietary ingestion, dietary ingestion, and dermal exposure. Since inhalation exposure is treated exhaustively in several chapters in this book, including Chapter 4 and Chapter 19, it will not be discussed here. This chapter will focus on dermal exposure and dietary and nondietary ingestion. 20.4.1 D ERMAL E XPOSURE The dermal route is perhaps the most complex of the more important exposure routes and the most difficult to model (Zartarian and Leckie 1998). A number of slightly different equations have been developed by different groups to estimate dermal exposure. All the equations have in common a term for the concentration of the substance in the medium (e.g., soil, water) and a term for the “transfer factor” governing how the substance moves from the medium to the skin. Some include a term for the area of the skin involved. Many include a term for the frequency of contact with the medium, using activity pattern data to describe such frequencies (Moschandreas et al. 2001). Other models (Zartarian et al. 2000; Price, Young, and Chaisson 2001) consider dermal exposure via dislodgeable residues, which may be some combination of liquid residues, house dust, and soil. The OP Case Study Group Non-Dietary Subcommittee (1999) and Price, Young, and Chaisson (2001) utilize a version of what the USEPA’s National Exposure Research Laboratory (NERL) terms the “macroactivity approach.” The general version of this approach uses the following equation: © 2007 by Taylor & Francis Group, LLC 474 Exposure Analysis (20.1) where E der = dermal exposure resulting from the completion of an activity (mg) ED = duration of activity (hr) TC der = dermal transfer coefficient (cm 2 /hr) C surf = dislodgeable contaminant loading on surface (mg/cm 2 ) The key to this approach is the transfer factor or transfer coefficient, which provides a lumped measure of contaminant transfer within a microenvironment or after some general activity. Other models, such as the one by Moschandreas et al. (2001), use rates of dermal contact to incorporate activities into dermal exposure estimates. A unitless term representing the proportion of contaminant on a surface transferred to the skin is utilized rather than a lumped transfer coefficient. This unitless term, or transfer efficiency, may be a function of the type of contact event, surface, or environmental media. NERL’s general equation for such a model is: (20.2) where E der = dermal exposure associated with a given event (mg/day) C surf = dislodgeable contaminant loading on surface (mg/cm 2 ) TF = fraction available for transfer from surface to skin (unitless) SA = surface area contacted (cm 2 /event) EV = event frequency (events/day) Models have also been developed that consider sequential dermal contacts (Zartarian 1996; Canales 2004). While these models are relatively more complicated, they tend to account for phenomena such as removal from the skin (e.g., hand-to-mouth contacts, negative transfers, wash- ing) and accumulation due to a number of media (e.g., liquids, air, residues, soil, house dust). The use of sequences of dermal contacts also permits the development of detailed exposure time profiles and greater insight into mechanisms and behaviors that result in dermal exposure (Zartarian 1996). The general equation below is applied to sequential contacts and the concentration, surface area, and exposure factor terms change as a function of the contact or the type of medium. (20.3) where DE = spatially averaged dermal exposure after contact event [M/L 2 ] C medium = concentration in medium [M/L 3 ], [M/L 2 ], or [M/M] EF medium = exposure factor for particular medium [L], [–], or [M/L 2 ] (A medium /A skin )= fraction of skin area covered with medium [–] Total dermal exposure is then the sum of exposure from each contact. EEDTCC der der surf =× × EC TFSAEV der surf =××× DE C EF A A medium medium medium skin =× × © 2007 by Taylor & Francis Group, LLC Models of Exposure to Pesticides 475 20.4.2 N ONDIETARY I NGESTION E XPOSURE Nondietary ingestion exposure, a generally overlooked route of ingestion exposure, occurs when an agent on surfaces (e.g., toys, food, fingers) or in nondietary media (e.g., soil, house dust) is ingested via hand-to-mouth or object-to-mouth contact, rather than through the ingestion of food. Several models created to meet the needs of the FQPA of 1996 consider the route, but few agree on methodologies. Pang et al. (2002), for instance, account for daily incidental ingestion (ng/day) of carpet dust and soil, using the number of carpet contacts per day multiplied by an ingestion rate per contact. Pang et al. also consider an absorption factor, so that their approach actually results in a daily dose, an amount crossing the gut surface, rather than an exposure. Cohen Hubal et al. (2000), however, consider contaminants on hands and on objects, and use surface concentrations multiplied by transfer fractions and contact frequency in quantifying nondietary ingestion exposure. 20.4.3 D IETARY I NGESTION E XPOSURE Dietary ingestion exposure is a function of food consumption patterns and the concentration of chemical in the food (Cohen Hubal et al. 2000). The product of these two factors is calculated for each item of food and summed over all food consumed over some time period. Typically the result is a daily ingestion exposure in units of micrograms per day. While this approach is very simple, complications arise when considering the variability in chemical concentrations in food, person-to-person differences in consumption patterns, and varia- tions in dietary profiles across age, gender, ethnic groups, and geographic regions. The primary issue when estimating dietary exposure then is not the exposure equation but rather the source of the inputs to the equation. If the study population is small, or if adequate resources are available relative to the size of the study, 24-hour duplicate diet sampling is often employed. This method involves collecting a duplicate portion of all food and beverages consumed by an individual over a 24-hour period. Upon analysis, both the concentration of the contaminant in the food and the total mass of food are known, and an estimate of the daily ingestion exposure can be calculated. If an estimate of ingestion exposure is required over longer time periods or for a larger population, this daily information can then be extrapolated via modeling. In studies examining a relatively large population, databases of food and beverage consumption are often utilized. One such database consists of the combined data from the Continuing Survey of Food Intakes by Individuals (CSFII) and the National Health and Nutrition Examination Survey (NHANES). The NHANES consists of a multistage sample representative of the United States. Both surveys recorded consumption data for a single day per person for close to 10,000 people. Several groups were oversampled, such as adolescents, minorities, and pregnant women, to allow for more precise estimates. The CSFII was conducted by the U.S. Department of Agriculture (USDA) as several 1-year surveys to provide 1-day and 3-day dietary intake data. The data have been mined to produce reports specifying eating patterns by sex, age, race, and geographic region. The integrated database is titled What We Eat in America — National Health and Nutrition Examination Survey (Dwyer et al. 2001). 20.5 EXPOSURE MODELS Prior to the 1990s there were few models that specifically addressed aggregate and cumulative exposure assessment. Of those existing models that explored aggregate exposure, the majority had divergent overall goals with exposure as an intermediate step, or they examined exposure with an inadequate level of detail. The Food Quality Protection Act’s new requirements have led to an evolution in exposure assessment methodologies. Several aggregate models are reviewed below. While others exist, these © 2007 by Taylor & Francis Group, LLC 476 Exposure Analysis models were chosen for review based on their range of methods and the availability of documen- tation (i.e., either in peer-reviewed journals or as workshop summaries). As these models are utilized and evaluated they may change in equation form, structure, or level of detail. In time some may be deemed more useful than others and may become the predominant models used in assessing exposure to pesticides and other contaminants. This review then represents a snapshot of these aggregate models during this evolution. Model inputs, overall objectives, general methods, imple- mentation, and special features are discussed. 20.5.1 T HE C AL TOX A PPROACH Developed at the University of California at Berkeley and the Lawrence Berkeley National Labo- ratory, the multimedia fate and transport fugacity-based model, CalTOX, has been extended to explore exposure and dose by investigating pathways that contribute significantly to risk. This approach calculates chemical concentrations in exposure media and estimates the dose of a single chemical through multiple routes. Pathways and routes considered include dietary ingestion of produce, meats, tap water, and mother’s milk; nondietary ingestion of soil and surface water while swimming; inhalation of vapors and particles; and dermal contact with soil and water. Additional extensions of CalTOX address uncertainty, variability, and sensitivity analysis (Bennett, Kastenberg, and McKone 1999). Inputs for the fate and transport portion of CalTOX include bioconcentration factors, landscape characteristics, partitioning coefficients, and other chemical-specific factors. Toxicological variables such as cancer potency factors and lethal dose values are utilized for an assessment of risk. Human uptake estimates require statistically averaged physiological exposure factors (e.g., breathing rate, surface area, food intake rate, soil adherence to skin) and macro-level activity patterns (e.g., time spent sleeping, time spent outdoors) (Bennett, Kastenberg, and McKone 1999). The CalTOX approach first creates characteristics of individuals. With these hypothetical individuals and the predicted environmental media concentrations, a distribution of potential dose is estimated. This methodology incorporates nested Monte Carlo simulation and results in CalTOX’s unique ability to assess the variability and uncertainty of outputs. Another unique feature of CalTOX is its accessibility and ease of use. Portions of the model are freely available and the majority of calculations require only Microsoft Excel ® software. The interface is user-friendly and there are several available documents describing the model’s construct, variables, and use (Bennett, Kasten- berg, and McKone 1999). One could argue that attempting to estimate exposure and dose from an uncertain fate and transport model propagates error, whereas using known exposure media concentrations could help to eliminate this error. Since CalTOX is a fate and transport model, a single discharge or source term must be defined. For certain chemicals or scenarios where the source is unknown or multiple sources are present, CalTOX may need to be modified. Another issue is that the model uses averaged macro- and meso-level activity patterns and therefore may not be equipped to handle the details of dermal and nondietary ingestion exposure. The model is also better suited to estimate exposures over a year or an entire lifetime. CalTOX is then limited in its ability to assess exposures for acute or intermediate scenarios (Fryer et al. 2004). 20.5.2 C ALENDEX ™ The Novigen Calendex™ System performs aggregate and cumulative estimates of consumer and occupational exposures to chemicals. In assessing human exposure to and risks from pesticides used in residential settings exposure levels are calculated for each day and for each individual in the target population. The system relies upon the U.S. Department of Agriculture’s Continuing Survey of Food Intakes by Individuals (CSFII) to provide both dietary consumption data and demographic variables of the population. In calculating an individual’s dose, the model also uses © 2007 by Taylor & Francis Group, LLC Models of Exposure to Pesticides 477 exposure factors (e.g., body weight, breathing rate, activity patterns), use pattern information of pesticides, and environmental concentration data prior to, during, and after pesticide applications (Petersen et al. 2000). This calendar-based probabilistic model then links information on the frequency of pesticide use with the probability of an exposure occurring and calculated exposures for different routes to estimate a daily aggregate dose (Novigen Sciences, Inc. 1998). The model also permits the inclusion of temporal aspects by modeling changes in exposure and dose due to changes in environmental concentrations over time. A macro-activity approach is taken when estimating post-application dermal dose rates. For nondietary ingestion exposures, meso-level activities are used in the form of hand-to-mouth events per hour. Calendex has been criticized for its lack of transparency, extrapolations of short-term data from the CSFII to simulate longitudinal exposures, and lack of tracking mechanisms to analyze contributions to the model output (Kendall et al. 2000). Additionally, the model relies on the CSFII data for both consumption and demographic data. This survey, however, is known to focus on lower socioeconomic classes that are not representative of the entire U.S. population. In addition to the lack of transparency, significant professional judgment is necessary to apply the model (Fryer et al. 2004). 20.5.3 L IFE L INE ™ The LifeLine approach, a collaborative effort between the Hampshire Research Institute, TAS- ENVIRON, and ChemRisk, aims to characterize inter-individual variations in pesticide doses and aggregate exposures received from multiple sources. Databases used in the LifeLine approach include the National Home and Garden Pesticide Use Survey, the Exposure Factors Handbook, U.S. birth records, mobility surveys, and the USDA’s Food Consumption Survey. Other inputs include sample data of pesticide residues in the home, tap water, and food (Muir et al. 1998). LifeLine constructs characteristics (e.g., age, gender, socioeconomic status, body weight, diet, location of home, frequency of pesticide use) of individuals, and simulates each day in these hypothetical individuals’ lives (i.e., from birth to 85 years) using probabilistic rules. Macro-level activity patterns and transitions between activities are assigned probabilistically, while meso-level activity patterns are assigned on a yearly basis dependent upon the individual’s characteristics. These characteristics and patterns then define the likelihood of exposure to a source and are used in conjunction with pesticide concentration data to determine a total dose. Distributions of dose can be viewed as a function of year or season (Muir et al. 1998). Notable features in the LifeLine approach include the ability to enter inputs as deterministic values or probability distributions, the use of transition rules for an individual’s mobility and pesticide use, and the consideration of a number of indoor and outdoor microenvironments when estimating residential exposure. The developers are also committed to making LifeLine user-friendly and available for public and professional use. Since this scheme was created to estimate chronic and lifetime exposures, it may not be well suited to estimate short-term exposure events (i.e., less than a day). Furthermore, since LifeLine was expressly created for the FQPA, it may not be easily adapted to consider other chemicals besides pesticides and nonresidential exposures (Fryer et al. 2004). According to a Federal Insec- ticide, Fungicide and Rodenticide Act (FIFRA) advisory panel, the model may also be more complex and more difficult to use than necessary (Kendall et al. 2000). 20.5.4 T HE C UMULATIVE AND A GGREGATE R ISK E VALUATION S YSTEM (CARES) CARES, funded by CropLife America, represents a collaborative effort between EXP Corporation, Novigen Sciences, Alceon Corporation, Cambridge Environmental, Inc., Summit Research Ser- vices, the University of Georgia, and Sielkin & Associates Consulting, Inc. CARES aims to estimate © 2007 by Taylor & Francis Group, LLC 478 Exposure Analysis cumulative and aggregate risk from dietary, drinking water, and residential exposures to pesticides (Baugher et al. 1999). Model inputs include data from the U.S. Census, Public Use Microdata Samples (PUMS), the Continuing Survey of Food Intakes by Individuals (CSFII), peer-reviewed literature, product attributes, and the Exposure Factors Handbook. The model uses census and survey data to create a reference population, and constructs daily profiles of individuals. These daily profiles are then accumulated to estimate short-term, intermediate, or chronic exposures. Additional features include the ability to accept deterministic or probabilistic inputs, the propagation of population variability, and a module to identify important factors contributing to exposure (Baugher et al. 1999). After a review of the CARES model conducted by the FIFRA advisory panel, several criticisms emerged. For example, the system focuses heavily on dietary exposures and realistically represent- ing population demographics, but uses unrealistic activity patterns extracted from Jazzercise ® scenarios and lacks data for residential exposures. The model construct is fairly inflexible and its calculations are not transparent. Additionally sensitivity analysis is difficult to conduct and the framework does not allow for the assessment of variability between replications of individuals (Roberts et al. 2002). Similar to the LifeLine model, the application of CARES may not be easily adapted beyond residential exposures to pesticides (Fryer et al. 2004). 20.5.5 T HE S TOCHASTIC H UMAN E XPOSURE AND D OSE S IMULATION (SHEDS) The SHEDS models were developed by the USEPA’s National Exposure Research Laboratory. The models aim to improve risk assessments by considering variability in exposure and dose calculations for at-risk populations and by helping to prioritize measurement needs (International Life Sciences Institute [ILSI] 2001). All versions of SHEDS are physically based, probabilistic models designed to estimate exposure beyond the screening level. The primary version to support the FQPA is SHEDS-Pesticides. Several case studies were explored in developing SHEDS-Pesticides. With each study a stand-alone model was constructed. Residential-SHEDS focuses on children’s dermal and nondietary ingestion exposure and dose to the pesticide chlorpyrifos applied in and around the home (Zartarian et al. 2000). SHEDS-PM models daily inhalation exposures to particulate matter from ambient and indoor sources (Burke, Zufall, and Özkaynak 2001). Children’s aggregate exposures to arsenic and chromium from treated play sets and decks are explored in the SHEDS-Wood model. The SHEDS-Air Toxics model aims to model aggregate exposures to urban hazardous air pollutants (HAPS) such as benzene, formal- dehyde, and metals from industrial sources (USEPA 2003). SHEDS-Pesticides evaluates ingestion, inhalation, and dermal exposure and dose specifically for children aged 1–6 years (Fryer et al. 2004). The underlying engine of SHEDS uses Monte Carlo methods and the USEPA’s Consolidated Human Activity Database (CHAD). The macro-level activity diaries contained in CHAD are utilized in the inhalation models and are supplemented with consumption and contact data for ingestion and dermal exposure estimates (Burke, Zufall, and Özkaynak 2001; Zartarian et al. 2000). As with other models, some believe the SHEDS models are only applicable for a limited range of scenarios and chemicals. Any customization of the models requires the user to have relatively sophisticated data as inputs. Additionally the models may not properly evaluate longitudinal expo- sure trends (Fryer et al. 2004). 20.6 ISSUES IN MODELING EXPOSURE 20.6.1 C OMBINING D ATA FOR M ODEL I NPUTS As needed for the FQPA, the aim of many modeling projects is to predict population exposures for policy development. Unfortunately measuring parameters relevant for a large population may © 2007 by Taylor & Francis Group, LLC Models of Exposure to Pesticides 479 be difficult and expensive. It may be necessary to combine or extrapolate data from smaller studies. Several issues may arise, however, in using these smaller studies. For example, in the developing field of exposure assessment, measurement techniques to collect relevant environmental media and contaminants often differ from study to study given that few standard techniques have been established. In measuring contaminants in house dust, for instance, sampling devices include vacuum cleaners, moistened wipes, hand presses, rollers, dust fall plates, and dust settling mats. The amount of dust captured by the various methods most likely differs even though each technique may be trying to capture the same true value. Nonetheless, a lack of data often persuades the modeler to combine data from different sources — perhaps sources collecting data using different measurement techniques. The effects of combining such datasets on the resulting estimates have not been adequately studied. Combining time activity data from various sources also poses problems. Activity data are typically collected as categorical data of microenvironments or objects contacted. From study to study, however, these categories may differ, making it difficult to group discrete activities. Adding to the difficulty, activities may be reported as frequency data or as a time series. Data may also be collected at the micro-level for dermal contacts and nondietary mouth insertions or at the macro- level for use in inhalation exposure assessment. The Consolidated Human Activity Database (CHAD) is a collection of activities from existing human activity pattern studies. In order to combine the data, however, information in some studies was modified according to a predefined format (e.g., a limited number of microenvironments). Given the potential loss of information, the original raw activity data are also provided if more detail is needed. A separate concern is the extrapolation of data. In particular, there is a need for human activity data for use in modeling longitudinal exposures. These estimates require human activity pattern data spanning several years. Collecting longitudinal activities, however, requires extensive resources, time, and subject participation. Existing activity pattern data, therefore, typically span only a few hours to a few days (e.g., via diaries, recall, videography techniques). Ideally existing short-term data could be used to extrapolate to longer periods of time. Unfortunately methods for extrapolation are not well defined. Partially this is because short-term activities have not been studied to assess inter- and intra-variability, distinct changes in activities over long periods are unknown, and there is little knowledge as to how well the existing activities represent specific populations or age groups. Past methods to simulate longitudinal activities from short-term data have included the use of probabilistic models, random sampling of activities, and repetition of activities representing particular days of the week, seasons, or age groups. Whether or not the activities resulting from these methods realistically represent longitudinal activities is still unknown. 20.6.2 TEMPORAL AND SPATIAL VARIABILITY There are also temporal characteristics to consider. In measuring contaminant concentrations in air, for example, data sets with measurements taken at different sampling periods may be difficult to combine. While you can always average, say, 1-hour measurements, over a day to conform to other daily measurements, there is typically a loss of detail. This detail in hourly variability may be important to evaluate extreme exposures or to explore contaminant sources. Similarly, attention must be paid to specific designs and goals of individual studies. The study design may be aiming to collect values that represent a continuous time series, seasonal variation, or annual averages. Values from studies with different goals therefore may not be representative for a particular project/exposure assessment. Similar to temporal characteristics, spatial characteristics must be examined. The first issue is one of scale. Are measurements meant to capture the effects of global, regional or local sources? Once again using the example of concentrations in air, measurements may represent personal exposure, microenvironmental concentrations, or ambient concentrations. Each of these different © 2007 by Taylor & Francis Group, LLC 480 Exposure Analysis scales is typically used in a unique way in inhalation exposure models, and combining raw data from different scales could result in unrealistic estimates. Another issue is one of how well the data represent the exposure scenarios of interest. The National Human Exposure Assessment Survey (NHEXAS), for example, aimed at collecting mea- surements of multiple chemicals to identify predictors of exposure in multiple geographic locations (Callahan et al. 1995). Sites included the upper Midwest, Arizona, and Maryland. But are data from these sites useful for other locations, or are they representative of a larger population? Can indoor surface concentrations be used to assess dermal exposures in South Texas? Can microenvi- ronmental air concentrations from all sites be combined to estimate inhalation exposures across the entire United States? 20.6.3 C O -O CCURRENCE OF E NVIRONMENTAL CONTAMINANTS Exposure modeling efforts typically focus on single contaminants. Although dealing with one chemical at a time makes aggregate exposure assessment more feasible, the reality is that few data exist on the co-occurrence of chemicals and their interactions. A particular case of interest is that of estimating exposure from the co-occurrence of environ- mental concentrations of pesticides. Such considerations are required by the FQPA. In the absence of quantitative data from sampling media, modelers consider information regarding pesticide properties, transport, and timing of applications. Co-occurrence may then be determined via mod- eling techniques where probabilities are assigned to the coincidental occurrence of pesticides or where fate and transport models are applied to a region to examine the resulting pesticide spatial distributions. More research should be aimed at collecting the concentrations of multiple contam- inants per sample, understanding the phenomena of co-occurrence, and in modeling such phenom- ena for use in human exposure models. 20.6.4 ESTIMATING DOSE FROM EXPOSURE Modeling methods for assessing exposure are, for the most part, mathematically simple. Often the most elaborate portions of exposure models entail data storage and management. Other pieces of the human health risk model are historically more complex. For example, models of the fate of contaminants in microenvironments and the transport between microenvironments often involve a system of differential equations and perhaps a number of uncertain parameters. Mechanistic models that specifically estimate dose from exposure can be very complicated as well. Although the focus of the chapter has been exposure modeling, methods for estimating dose are also of concern, since it is the dose that is important in assessing risk and health effects. Furthermore several models claiming to explore pesticide exposure result in a final estimate representing intake or dose. There are a number of complicated models for estimating dose from dermal exposure. A portion of these are box models with separate compartments representing the layers of skin (i.e., stratum corneum, viable epidermis, papillary dermis). These models are typically based in mass transfer theory, and include chemical-specific properties and parameters, and transfer under steady- or unsteady-state conditions. The majority of dermal exposure models account for dermal dose by simply assuming a fraction of the skin surface concentration is absorbed into the skin. But how different are dose estimates when using a complex model vs. a simple absorption fraction? Is there a need to combine the most detailed exposure models with the most detailed dose models? To complicate matters, the differences in modeling approaches may be hidden due to the variability within populations and the uncertainties in parameters. Further research needs to explore these issues. © 2007 by Taylor & Francis Group, LLC [...]... National Human Exposure Assessment Survey, The Journal of Exposure Analysis and Environmental Epidemiology, 5(3): 257–282 Canales, R.A (200 4) The Cumulative and Aggregate Simulation of Exposure Framework, Ph.D diss., Stanford University, Stanford, CA Cohen Hubal, E.A.C., Sheldon, L.S., Burke, J.M., McCurdy, T.R., Berry, M.R., Rigas, M.L., Zartarian, V.G., and Freeman, C.G (200 0) Children’s Exposure Assessment:... Assessment: A Review of Factors Influencing Children’s Exposure, and the Data Available to Characterize and Assess That Exposure, Environmental Health Perspectives, 108(6): 475–486 © 200 7 by Taylor & Francis Group, LLC 482 Exposure Analysis Cohen Hubal, E.C., Thomas, K., Quackenboss, J., Furtaw, E., and Sheldon, L (1998) Dermal and Non-Dietary Ingestion Exposure Workshop, Report No EPA 600/R-99/039, U.S... Models of Exposure to Pesticides 483 USEPA (200 3) Stochastic Human Exposure and Dose Simulation Model for Air Toxics, Significant Research Findings: National Exposure Research Laboratory Research Abstract, http://www.epa.gov/nerl/ research /200 3/g1-5.pdf (accessed November 12, 200 5) Zartarian, V.G (1996) A Physical-Stochastic Model for Understanding Dermal Exposure to Chemicals, Ph.D diss., Stanford... (1999) REx, Residential Exposure Assessment, Generic Methods, Case Study: Lawn Care Products, December 15 Pang, Y., MacIntosh, D.L., Camann, D.E., and Ryan, P.B (200 2) Analysis of Aggregate Exposure to Chlorpyrifos in the NHEXAS-Maryland Investigation, Environmental Health Perspectives, 110(3): 235–240 Petersen, B.J., Youngren, S.H., Walls, C.L., Barraj, L.M., and Petersen, S.R (200 0) Calendex™: CalendarBased... Assessment of Atrazine: The Impact of Age Differentiated Exposure Including Joint Uncertainty and Variability, Reliability Engineering and Systems Safety, 63: 185–198 Burke, J.M., Zufall, M.J., and Özkaynak, H (200 1) A Population Exposure Model for Particulate Matter: Case Study Results for PM2.5 in Philadelphia, PA, The Journal of Exposure Analysis and Environmental Epidemiology, 11(6): 470–489 Callahan,... Moshfegh, A.J., and Johnson, C.L (200 1) Integration of the Continued Survey of Food Intakes by Individuals and the National Health and Nutrition Examination Survey, Journal of the American Dietetic Association, 101(10): 1142–1143 Fryer, M.E., Collins, C.D., Colvile, R.N., Ferrier, H., and Nieuwenhuijsen, M.J (200 4) Evaluation of Currently Used Exposure Models to Define a Human Exposure Model for Use in Chemical... crucial depending on the contaminant of interest In general, what factors require an aggregate exposure assessment? Name four compounds for which an aggregate assessment may be justified 4 You would like to estimate the dermal exposure of compound X using Equation 20. 3 Notice that in defining the concentration and exposure factor terms, a number of different dimensions are given You know that compound X... Hand and Mouthing Activities through a Videotaping Methodology, The Journal of Exposure Analysis and Environmental Epidemiology, 9(5): 513– 520 Roberts, S.M., Odiott, O., Thrall, M.A., Adgate, J.L., Durkin, P., Engel, B., Freeman, N., Hattis, D., Heeringa, S., MacDonald, P., Portier, K., Potter, T.L., Reed, N.R., Zeise, L (200 2) FIFRA Scientific Advisory Panel Meeting, a Set of Scientific Issues Being... models exist to estimate exposure for each route separately, there may be value in creating an aggregate model that considers correlations between model inputs What might be some important correlations in environmental concentrations? How might dermal and nondietary ingestion exposure be correlated? 3 Although the importance of aggregate exposure has been discussed, assessing exposure through multiple... (1998) Dermal Exposure: The Missing Link, Environmental Science & Technology, 3(3): 134A–137A Zartarian, V.G., Özkaynak, H., Burke, J.M., Zufall, M.J., Rigas, M.L., and Furtaw, E.J., Jr (200 0) A Modeling Framework for Estimating Children’s Residential Exposure and Dose to Chlorpyrifos via Dermal Residue Contact and Nondietary Ingestion, Environmental Health Perspectives, 108(6): 505–514 © 200 7 by Taylor . Routes of Exposure 473 20. 4.1 Dermal Exposure 473 20. 4.2 Nondietary Ingestion Exposure 475 20. 4.3 Dietary Ingestion Exposure 475 20. 5 Exposure Models 475 20. 5.1 The CalTOX Approach 476 20. 5.2 Calendex™. 471 20 Models of Exposure to Pesticides Robert A. Canales Harvard University James O. Leckie Stanford University CONTENTS 20. 1 Synopsis 471 20. 2 Introduction 472 20. 3 Activity Data 472 20. 4. 476 20. 5.3 LifeLine™ 477 20. 5.4 The Cumulative and Aggregate Risk Evaluation System (CARES) 477 20. 5.5 The Stochastic Human Exposure and Dose Simulation (SHEDS) 478 20. 6 Issues in Modeling Exposure