8.5 The analysis of human exposures to contaminants in the environment
Oxford Textbook of Public Health
The analysis of human exposures to contaminants in the environment
Paul J. Lioy, Amit Roy, and Natalie Freeman
Approaches used in exposure analysis
Media and routes of exposure
Data analysis and models
Exposure assessment modelling
Exposure probabilities for individuals and populations
The presence of chemical and physical agents in the environments where people live, work, and play may cause illness. Therefore it is essential to develop and employ reliable methods that define the intensity and duration of contact with such agents and assess the likelihood of any cause–effect relationship. The field of exposure analysis and its assessment is associated with epidemiology, risk assessment, and disease intervention and prevention (Lippmann and Lioy 1986; Graham et al. 1992; Sexton et al. 1992; Jayjock and Hawkins 1993; Ott 1995; Lioy 1999) and the scientists and engineers who conduct these studies now are called exposure analysts (NRC 1985, 1991b,c) or exposurologists (Ott 1995). The kinds of exposures examined by the exposure analyst are illustrated in Fig. 1. The traditional terms, industrial hygienist and radiation health physicists, refer specifically to those individuals who conduct exposure assessments in the various workplaces and who provided much of the fundamental technical bases for the first sets of field studies.
Fig. 1 Types of exposures that may be experienced by the general population.
The exposure measurement techniques may be indirect or direct (NRC 1991c). Indirect techniques include sampling locations (microenvironments) where contact may occur with a contaminant, and/or the administration of survey instruments; such as, time/activity questionnaires. Direct techniques include personal monitors worn by individuals, and samples of blood, urine, and other bodily fluids, which permit measurements of exposure and dose for specific individuals.
The measurement of the concentration of physical, chemical, radio-active, and/or biological agents in air, water, food, and so on, and the estimation of human behaviours, using instruments such as time/activity pattern diaries, has led to the development of models that can predict exposure and dose. Currently, it is feasible to trace an agent from its source through pathways into exposed people. Figure 2 illustrates the ‘flow’ of a contaminant through the points of contact, to the exposure, and to the dose that can appear inside the body. The domain-related scientific and professional disciplines: environmental science, exposure assessment, toxicology, and epidemiology, are also illustrated in Fig. 2.
Fig. 2 Continuum for the emission of and exposure to a contaminant and the expression of a health effect. (Source: Lioy 1990.)
This chapter describes the features of exposure assessment starting from concepts and theory. The types of exposure measurements and estimates needed for the applications to environmental health will also be illustrated using, in varying detail, information about lead, benzene, trihalomethanes, pesticides, airborne particulate matter, infectious agents, and alternate fuels. Finally, some observations will be made concerning the future of exposure assessment in public health practice.
Over the past 15 years, the theoretical and conceptual bases for exposure assessment have evolved from simple mathematical expressions that consider exposure and dose, to complex mechanistic descriptions of exposure/dose equations and concepts that can describe multiple routes of contact with a toxic agent. The aim, as described by Fig. 2, is to establish a relationship between the release of a toxicant, and a dose that may cause an adverse health outcome (Lioy 1990, 1999; Lioy and Pellizzari 1995).
Some pollutants, such as ozone, are not emitted into the environment but are formed from precursors. In such cases it is necessary to establish a relationship between the release of precursors of the toxicant, the conditions under which the toxicant is formed, and the health effect. The study of the effect of exposure on ecological receptors can have important implications for human health. Ecological effects can have indirect effects on human health, and can also serve as sentinels of human exposure. Ecological effects can also have significant environmental justice implications, because the lives of some tribal and disadvantaged human populations are more closely linked to ecological health than that of the general population.
Exposure to a contaminant is defined as the ‘contact at a boundary between a human and the environment at a specific concentration over an interval of time’ (NRC 1991c). Based on this definition the types of integral or summation equations needed to estimate or describe exposure were identified and are presented in Box 1. The integral equation is an exact expression of an individual’s exposure, over the course of time, while the summation provides an approximate representation of exposure. The integral equation requires knowledge of the instantaneous concentration of the toxicant, which is generally available only from modelling studies, whereas the summation requires a knowledge of time-averaged concentrations, which are often available from the application of direct and indirect measurement techniques. Generalized versions of these equations and concepts can be used to estimate exposures in modelling studies. These can be found in Georgopoulos and Lioy (1994), Ott (1995), and Zartarian et al. (1997b).
Box 1 Equations governing exposure
where Ei is the integrated exposure of the ith individual to a concentration Ci of a contaminant for time period t1 to t2 associated with a biological response.
Microenvironmental increments of exposure
DEji = Cji(Dt) Dti
where DEji is the exposure of person i to a contaminant measured over an interval Dt associated with the jth activity (or location). These expressions are appropriate for inhalation and dermal exposure, based on which dose can be estimated. For uptake by ingestion, dose is generally estimated directly based upon media concentrations and uptake rate.
In principle, the uptake of a toxicant can occur via three routes: inhalation, dermal absorption, and ingestion. The definition of exposure provides guidance for the selection of locations for sampling. For the purposes of exposure assessment, any material passing a hypothetical boundary surrounding the body is termed potential dose. Inhaled dose is the amount of toxicant that enters the nose or mouth, dermal dose is the amount that enters the skin, and ingested dose is the amount that enters a hypothetical boundary over the mouth. As suggested by the term potential, the amount of toxicant that actually enters tissue can be significantly lower than the potential dose. These occur at boundaries external to the body; thus, exposure is determined at the points of entry into the body, that is, the nose, mouth, and skin. The data collected can then be used to establish criteria identifying exposure within a population and constructing predictive models (Duan 1982; Ott 1982, 1990; Sexton and Ryan 1988; McKone 1991; Ryan 1991; Georgopoulos et al. 1997). Figure 3 illustrates three major routes of uptake: inhalation of airborne agents, ingestion of food and liquids, and skin contact with air, soil, and all types of materials and products.
Fig. 3 Routes of exposure. A comprehensive exposure assessment invokes consideration of each possible route through which a chemical can enter the body.
Once a contaminant has crossed a boundary and entered the body, it is then considered a dose, which is routinely described in one of three ways.
Potential dose is the amount of material that enters an hypothetical external boundary around a receptor deposited on a surface, which can potentially cause an effect on the surface or can be transferred to another organ or tissue. The entire mass (100 per cent) is assumed to cause a biological response.
Internal dose is the amount of material that actually enters one or more tissues; and target tissue dose is the amount of toxicant that actually reaches sensitive tissue in which the toxic response occurs, or estimated to be available to be absorbed by an organ or tissue or absorbed to a surface and is available to undergo biological processes, which can cause altered physiological function.
Biologically effective dose is the amount of target tissue dose required for manifestation of a toxic effect. It is the amount of the contaminant or metabolite that interacts with cellular macromolecules and alters physiological function (Lioy 1990).
The units most frequently encountered when calculating exposure and dose are shown in Table 1.
Table 1 Examples of units used to express exposure
Biological markers of exposure/dose can be measured in body fluid specimens and then be associated with time of occurrence and/or persistence of the contaminant (NRC 1989a). Henderson et al. (1992) and Shulte and Perera (1993) published a conceptual framework for describing the persistence of the different types of markers in the body. The general time course of elimination of each type of marker is illustrated in Fig. 4. The results indicate that exhaled parent compounds yield the highest levels of biomarker concentrations relative to exposure concentration. Adducts spend the longest time in the body. The term ‘markers of exposure’ is currently used to describe most of the above, but in actuality the level of a contaminant or transformed product, present in the body is defined as a dose (NRC 1989b, 1992).
Fig. 4 Hypothetical relationships between different biological markers of exposure and time after a single exposure. (Source: Henderson et al. 1992.)
The accurate measurement or estimation of exposure requires baseline information about the plausibility of human contact with the contaminant of concern and assists in establishing the data quality objectives for analysing any particular problems. In the selection of the appropriate measurement ‘tools’, the analyst also needs to account for factors such as the sensitivity and specificity of a technique for each medium or route studied, and ease of sample handling and collection. In some situations simple techniques such as survey instruments are extremely valuable in acquiring semiquantitative data for the characterization exposure (Carpenter and Huston-Stein 1980; USEPA 1988; Lebowitz et al. 1989; Robinson et al. 1989; Freeman et al. 1991, 1997; NRC 1991c; Schwab et al. 1991; Zartarian et al. 1997b, 1998). In other cases more complex techniques such as microenvironmental or personal monitors are used to establish the primary route by which human contact occurs with a contaminant (Seifert and Abraham 1983; Akland et al. 1985, Spengler et al. 1985; Stetter 1986; USEPA 1988; Wallace et al. 1988; NRC 1991c; Clayton et al. 1993; Lioy 1993; Valerio et al. 1997; Pellizzari et al. 1999). It is important to note, however, that all types of techniques do not have to be employed in a single study. Those selected would depend upon the data quality objectives and the hypotheses being tested for a particular study.
The issue of multimedia contact with contaminants has increased the awareness and the desire to obtain measurements on multiple routes of exposure, and to insure that exposure–response relationships are constructed for the media or routes of greatest concern (McKone 1991; Georgopoulos et al. 1997). The data gathered will also improve a manager’s ability to prioritizing strategies for intervention and eventual reduction exposure. Experience of exposure analysts with environmental health problems leads to a tacit point: it is not scientifically sound to prejudge which is the most important medium or route of concern for a particular contaminant (Lioy 1990; NRC 1991c). Avoiding this pitfall will make it possible to obtain a broader view of a problem and improve the selection of measurement and analytical techniques. In the past, many studies have focused on a limited number of routes, and frequently have led to poorly identified exposures, and eventually selection of inappropriate remedial solutions.
One example of how a poorly designed assessment can lead to misclassification of exposure involves benzene (Wallace 1989). Two pie charts, shown in Fig. 5, apportion benzene exposure within the general population. The first pie chart identifies the emissions of benzene from major environmental sources and has been used in the past to define exposure reduction strategies. The second pie chart identifies the actual benzene exposure experienced by a statistically representative sample of the general population. The clear message from the emissions pie chart is that motor vehicles represent the major source of benzene to the ambient atmosphere and provide the greatest number of opportunities for members of the general population to experience benzene exposures in non-occupational settings. Thus, one is led to believe that the most important source of benzene is the automobile. In contrast, measurements made within personal monitoring studies, also in Fig. 5, have shown that the predominant source contributing to benzene exposure (>50 per cent) is cigarette smoke, with only 20 per cent of the exposures caused by automobile emissions. Thus, potentially high exposures to benzene would be misclassified or ignored in current strategies to reduce benzene based primarily upon emissions data. Regulators or health officials would benefit from data collected on individual or population activities to help identify the ‘true’ major source of exposure.
Fig. 5 Benzene emissions versus exposures. ‘Personal activity/home’ refers to benzene from materials such as paints, adhesives, and marking pens. For individuals who do not actively smoke, the ‘active smoking’ contribution to exposure is zero, and the other exposure categories increase proportionally. (Source: Wallace 1989.)
Another example of where improved exposure assessment data actually provided better information on how commuters come into contact with a contaminant is the fuel additive, methyl-tertiary-butyl-ether (MTBE). This compound is an oxygenate designed to reduce carbon monoxide and is representative of other chemicals found in new or reformulated fuels. In this instance, many initial studies on oxygenated fuel were conducted to estimate the environmental levels of MTBE or other hydrocarbons caused by automobile tail pipe emissions. However, experiences of the general public and gasoline station workers with gasoline oxygenated by MTBE at 15 per cent by volume have suggested that the highest exposures to the driver or passengers in an automobile, or garage workers were derived from evaporative emissions released by the engine compartment or gas tank into the interior of a car or in microenvironments adjacent to gasoline service pumps. This is in contrast to the typical tail pipe emissions scenario used in exposure assessments for motor vehicle fuels. Experiments conducted using cars that followed a typical commuter route and then had the gasoline tank filled at some point during the trip are illustrated in Fig. 6. The results showed that the highest exposures to MTBE occurred during a tank refill. The approach used in the study demonstrated the importance of both personal and Microenvironmental analyses in providing insight on what can lead to high exposures to evaporative emissions. Other examples exist for dermal and ingestion exposure; however, the main point is not to demonstrate all misclassifications of exposure that can or have occurred, but to recognize that when you design a study, it must include flexibility to evaluate the possibility that a variety of sources and routes can affect exposure. This will improve the detection of and source apportionment of emissions, and how each affect the intensity of human contact with the contaminants of concern.
Fig. 6 MTBE exposure samples during commuter refuelling.
Approaches used in exposure analysis
Determination of human contact or the potential contact with a contaminant is not an integral part of traditional environmental quality measurements. Usually there are criteria available for making environmental quality measurements, which establish a statistically representative sampling scheme for determining the areal extent of contamination, and establishing long-term concentration trends within an environmental medium (Ott 1977; Lodge 1989; Lioy 1990; Anderson-Sprecher et al. 1994). Unfortunately, these measurements do not provide data that can be used to assess exposure directly. Historically, an assessment of exposure was based primarily on the concentration of a contaminant found in an environmental medium at a single sampling site for a prescribed sampling period. However, little or no information was provided on the duration of contact or the probability of contact with the contaminant by people spending time in the location where the measurements were the being made. More representative historical examples of exposure measurements would be the breathing zone samples collected in occupational settings; however, the concentrations were much higher.
In many cases, environmental quality measurements continue to provide the only data available for calculating exposure–response relationships within various public health-related studies, for example hazardous waste sites and air pollution (USEPA 1989a). These data have been used in applications within epidemiology or risk assessment, and have yielded results with a high degree of uncertainty. Also, environmental quality data are rarely collected using strategies that ensure the duration of the measurements is coupled with the relevant biological response time to the presence of a contaminant within the body. As stated previously, the latter point is critical for exposure analysis. All too often environmental quality measurement programmes are based upon a regulatory requirement for determining compliance to a regulation, and/or the limits of detection and collection capacity of a sampler.
The major paradigm shift in framework for exposure analyses, which started at about 1990, has led to the expanded use of personal monitoring and/or microenvironmental monitoring for the development of exposure databases. Both types of monitors are more intrusive than the devices used for monitoring environmental quality as the sample is taken: (a) at or from the individual; (b) in an area occupied by the individual; or (c) from objects used, worn, or eaten by an individual. The samples also require the acquisition of time resolved data on where and how individuals spend their time (Lebowitz 1989; NRC 1991c; Freeman et al. 1997; Zartarian et al. 1997a).
A hierarchy of exposure measurements is presented in Table 2, and emphasizes that personal monitoring provides the best data for completing an exposure assessment (NRC 1991a). The table should be reviewed with some caution, however, as in some studies even weak metrics of exposure may be adequate for examining the exposure–health effects relationships. The weak metrics are clearly useful in situations where an isolated source significantly affects a specific community, or a major event or episode has caused health effects. In addition, some techniques currently used for personal exposure could alter a person’s usual activity patterns. For instance, personal samplers for particles are usually bulky and cannot be worn comfortably during periods of outdoor and indoor exercise (Lioy 1993). Therefore, based upon needs of a study, it can be safely stated that the exposure analyst has a virtual tool box of techniques to ensure measurements can be used successfully to answer a public health question (Lioy 1992, 1999).
Table 2 Hierarchy of exposure measurements, estimates, and surrogates
Another component critical to an exposure analysis is the identification of the study population. In contrast to studies on environmental quality, where minimal information is required on the population of concern, examination of exposure requires the selection of either a probability based sample of the general population with possibilities for oversampling of specific subgroups, or a specific subgroup of the population that exhibits characteristics of susceptibility or is potentially at the high end of exposure to a contaminant. The latter is a major challenge because it is difficult a priori to select the high end exposure groups, greater than the 90th percentile (USEPA 1992a).
Typical sampling plans based on the general population or populations at risk are illustrated in Table 3. The selection of a susceptible subgroup is more difficult as detailed information is required on the physical or physiological characteristics of interest before selections are made for entry into a study. Some of the major questions that need to be addressed in properly selected susceptible or sensitive individuals are shown in Table 4.
Table 3 Summary of sampling for exposure assessment
Table 4 Defining high-exposure populations
Media and routes of exposure
The preceding discussion generally described the environmental media and routes of entry to the body needed to characterize exposure. Each has been examined over the years to provide information on the magnitude and extent of environmental problems. For exposure assessment there is a special need to know how each medium or route is associated with the degree to which individuals or members of a population come into contact with a contaminant. Many types of sources that impact each medium, and some of the more common issues are shown in Table 5. Clearly, inorganic or organic emissions can come from industrial processes, commercial activities, personal use and activities, disposal activities, and nature.
Table 5 Typical sources of contaminants that can be present in various media
It is somewhat obvious that the environmental media that can lead to contact with a contaminant include: air, water, soil, and food as people routinely come into contact with these each daily. However, it may be somewhat of a surprise that all routes of entry to the body, except inadvertent or purposeful injection of a biological or chemical contaminant, may be directly or indirectly impacted by a contaminant originally released in one medium. This important point is illustrated in Table 6 for the contaminants lead and pesticides.
Table 6 Routes of entry of contaminants to the body
Lead can be emitted by sources that directly impact the air, water, soil, and food, then transported to and deposited in another medium, and thereby be made available to indirectly cause exposures via multiple routes of entry to the body. From the standpoint of public health the situation is complicated because there is no easy formula available to determine the route of entry, or to apportion the sources contributing to lead burden. In the case of house dust, lead can be derived from indoor and outdoor sources, and ingestion occurs after dermal adhesion and transfer to the mouth or after the adhesion of the dust to a food or toys (NRC 1993a). For pesticide exposures, there is an added source of exposure, as an important way to increase contamination is the direct application of a pesticide to surfaces by a homeowner or resident (this is in addition to any amount derived from the work of a professional exterminator or crop duster) (NRC 1993b). However, once in the home there can be re-emission and redistribution of semivolatile pesticides to other surfaces (Gurunathan et al. 1998).
The need to complete a source apportionment for lead, pesticides, and other chemicals provides a message for public health officials and the exposure assessor: the obvious answer (source) may not always provide the correct way to solve a problem. For instance, a person may live in a residence that has lead-based paint the walls. The first thought would be that the lead paint was the major source of blood lead levels measured in the occupants. If the painted surface is isolated or intact, however, the source that could cause an increase in blood lead may be street dust and/or soil in the neighbourhood. Therefore source apportionment plays a crucial part in linking the point of emission through the route of exposure to an internal or biologically effective dose (previously illustrated in Fig. 2 and Lioy (1999)).
An example of how exposure and risk derived from multiple routes of entry can be underestimated is associated with regulations or public health advisories for potable water supplies. In the 1970s, potential exposures and health risks were based solely on the quantity of the contaminant ingested by drinking the water (USEPA 1980), and the assumed consumption of 2 litres of water per day, which is unusual for most members of the general population. Public health advisories on the use of contaminated water supplies, for example wells with water containing contaminants leached from hazardous waste sites, stated that individuals using a particular water supply ‘should not drink the water’. If scientists and regulators had seriously considered all the opportunities for contact with potable water prior to estimating the risk, they would have included two other routes of exposure: dermal and inhalation (Brown et al. 1984). Only after studies by Andelman (1985), which focused on the shower as a route for inhalation exposure, and Jo et al. (1990) and Weisel et al. (1993), which demonstrated that significant human exposures to volatile organics occurred via inhalation and the dermal route during showering or bathing, did the public health practice and environmental regulations embrace the concept of total exposure. The results have led to the concept that individuals should ‘just not use contaminated water’. According to Jo et al. (1990), at least half of an individual’s internal dose derived from chloroform found in a public water supply could be from just one 10-min shower per day. Obviously more and varied uses of the water would lead to higher or lower daily internal doses.
The importance of both the dermal and inhalation routes is demonstrated in Fig. 7, for the concentration of chloroform found in exhaled breath after using a swimming pool. Integration of the area under the curve indicates that both routes made similar contributions to an internal dose, even though there is a much slower rate of chloroform accumulation by the dermal route (Georgopoulos et al. 1997; Kim and Weisel 1998).
Fig. 7 Exhaled breath concentration of chloroform following inhalation and dermal absorption in a swimming pool (air concentration 100 µg/m3, water concentration 150 µg/l).
The measurement process used to establish the presence of a contaminant in one or more of the above media or routes can become complex and require the application of a variety of techniques. Two primary categories exist for exposure measurements: direct and indirect techniques, Fig. 8 (NRC 1991c). These categories correspond to the types of methods being used to collect data or estimate exposure within field studies and modelling simulations respectively. The main differences between these two types of techniques are associated with the proximity of the measurement or estimate of exposure to the individual, and the qualitative or quantitative nature of the information. At present, there is no uniformity in the quality and quantity of techniques available for any category of indirect or direct measurements. In fact, there are major instrumentation needs for each environmental medium and route of exposure. This is not to say that there is a lack of sampling and analytical equipment to measure chemical, physical, and biological agents in various environmental media. However, many have been designed to provide environmental quality measurements rather than human exposure measurements. This is an important point and is substantiated by the fact that many of the currently available techniques are too bulky to be used in applications requiring either microenvironmental and/or personal measurements (Lioy 1993).
Fig. 8 Possible approaches for analysis of air contaminant exposures. (Source: NRC 1991c.)
In many cases personal monitors still require optimization for a number of parameters. A list of general technical criteria that must be met to develop these monitors is shown in Table 7 (NRC 1991c). It is apparent that the techniques must be compact, and each must have low detection limits and sufficient time resolution in order to obtain human exposures for the chemical(s) under consideration. These three features are difficult to achieve simultaneously in a single device. For example, as an instrument is miniaturized the substrate available for sample collection or the detection volume available for instrumental analyses (e.g. a photocell) is reduced. Consequently, the detection limits will rise and/or the time necessary to collect an adequate sample will increase, which can preclude their use in specific applications, for example low-level short-term exposure or acute exposures or acute exposures. Such incompatibilities can only be eliminated by conducting basic research prior to the development of a sampling programme that will employ a device in any particular application. Sometimes there may be ‘off the shelf’ devices available for use in an application or study, but the possibilities have been limited for most media and routes of exposure. During the mid-1970s, because of the increased concern about health problems associated with indoor air pollution, technology developed rapidly and microenvironmental and personal air measurements were available for the detection of traditional contaminants (e.g. volatile organics, fine particles, and carbon monoxide). Others are being developed for agents such as microbiological aerosols. Some of the personal monitors are based on passive sampling techniques while other sampling systems use an active pump (Seifert and Abraham 1983; Ryan et al. 1986; Samet et al. 1987, 1988).
Table 7 Methods criteria
Before the concept of total exposure became a starting point for the design of field studies and risk assessments, the following measurement issue received little attention: How comparable is the data collected by various techniques used across more than one medium or route of entry into the body? Unfortunately, there is no complete answer because of the many different types of physical, chemical, and biological agents that exist and the nature of the emissions, transport, accumulation, and transformation processes that can affect the occurrence of biologically significant exposures. A summary of the types of techniques employed for collecting microenvironmental and personal samples is shown in Table 8. To illustrate the problem with obtaining comparable measurements Table 8 can be used to examine the situation as it currently exists for microenvironmental monitors. For microenvironmental studies of air pollution there exist both continuous and integrating monitors; the devices have high resolution and low detection limits for specific compounds, for example metals and volatile organics. The data can easily be used to estimate the direct inhalation exposure. However, for other media there are very few comparable techniques available for completing events with even quasi-continuous monitors. In fact, there are no continuous monitors currently available for these sampling media that operate without the constant use of a technician as a personal shadow. For example, take a surface soil sample at every location, he or she came into contact with soil during the day (Hawley 1985). Alternatively, the person would be required to have a trained technician directly shadow his or her movements during the sampling in order to collect the appropriate surface soil samples or provide dermal contact samples similar to those experienced by the subject. In either case the approach is very cumbersome and can lead to many errors.
Table 8 General availability of monitors for measuring exposure
For integrated samples, the situation is somewhat better. You can periodically obtain integrated samples from soil or dust available for dermal contact or consumption, food available for consumption, and water or fluids for consumption. However, it is quite difficult to obtain integrated samples that represent the contact that can occur with water in all microenvironments and activities during a given day (e.g. swimming, bathing, cooking, etc.).
At present the best opportunities for obtaining comparable multimedia microenvironmental samples are associated with the periodic collection of integrated samples as in studies estimating total exposure in a residential setting (Lioy and Pellizzari 1995; Pellizzari et al. 1995). The general concept involves capturing a day or week in the life of a statistical representative sample of a population or particular subgroup of a population at risk. In such a study, an investigator periodically collects a set of short-term samples and/or grab samples over a representative sampling interval. Figure 9 and Table 9 indicate the types of samples that can be collected to estimate a residential exposure for a week in the life of a family. The output from such a measurement study will be a series of microenvironmental samples that are analysed for the chemicals of concern. The data are then used as an input to exposure scenarios to specific environmental contaminants in residential settings. They can also be used to construct total exposure estimates using variants to the summation equations given in Box 1.
Fig. 9 Types of integrated microenvironmental sentinel for home exposure to metals, pesticides, and/or volatile organics.
Table 9 Sampling strategy
One major component of this type of study is the application of questionnaires, which must include a time/activity log that is completed by the members of the household during the time of the field sampling. The data are essential for reducing uncertainties that are inherent with the application of generic exposure scenarios to site- or person-specific assessments.
Activity logs have become customized to address the exposures that can occur for specific chemicals (Robinson et al. 1989; Freeman et al. 1991; Schwab et al. 1991), in addition to logs that address generic issues on contact with environmental contaminants (e.g. frequency of personal product use and contact with volatile components, frequency, and duration of outdoor activities). For instance, in a study of residential exposure typical questions for a week-long study of chromium exposure would include the following.
Were any of the following used in the house today? (a) vacuum; (b) carpet sweeper; (c) broom; (d) dust cloths/mops; (e) wet mops; (f) other house cleaning; (g) laundry.
Did you notice any green, yellow, red, or orange deposits or stains on the walls or floors of your home?
If you noticed these deposits, were you or members of your family in the room or rooms with these deposits for more than 10 min at a time?
Results obtained by these types of methods can be validated by video records, technician observations, and fluorescent tracer studies (Fenske et al. 1991). During the chromium study validation was obtained via observations made by a trained technician.
Although not shown in Fig. 9, a residential microenvironmental study can easily be expanded to include personal monitoring and biological monitoring. Some of the more common are used to measure organic/inorganic chemicals in blood and urine. These samples will provide personal integrated or time series data for the duration of the sampling period (e.g. a day or a week). Biological monitoring data provides baseline information on the residents and can be used to determine if they have been ‘truly’ in contact with a contaminant. If a residential experiment is repeated one or more times, biological marker data can be valuable in pharmacokinetic model simulations for some contaminants. Follow-up biological monitoring samples can also allow the analyst to establish any incremental changes in dose.
Biological monitoring is currently being used to measure selected heavy metals in blood or urine, volatile organics in blood, and pesticides in blood and urine in studies at hazardous waste sites and within the National Human Exposure Assessment Survey (Pellizzari et al. 1995; Pirkle et al. 1995). There are also some techniques available for the measurement of metabolites and DNA adducts (Fiserova-Bergerova 1987; Perera et al. 1987; NRC 1989a,b, 1991c; Ashley et al. 1992). The most notable are associated with the polycyclic aromatic hydrocarbons (Perera 1987). A first-order analysis of the data would be to determine the change in contaminant level for a bodily fluid that could be associated with a change in the type or intensity of exposure that occurred at the residence. A second-order analysis would involve the application of pharmacokinetic models (Gerlowsky and Jain 1983; Caudill and Pirkle 1992).
In 1993, the National Academy of Sciences published a report (NAS 1993) arguing that children are a highly susceptible population for exposure to pesticides. Both in terms of surface-to-volume ratio and physiological function children are different from adults, and may be more susceptible to exposure to environmental contaminants. Their way of interacting with the world is different from adults, and they spend more time on the floor, and take baths rather than showers. Further, infants and toddlers are more likely to mouth objects and exhibit hand-to-mouth behaviours, and have substantially greater food and fluid consumption when expressed as grams or litres per kilogram of body weight than adults (NAS 1993; Tsang and Klepeis 1996, Freeman et al. 1997, 1999). Because of their close contact with floors and carpets, the concept of a well-mixed air environment may not be appropriate for the air they breathe. The air inhaled close to a carpet may have very different concentrations of chemicals than the air inhaled 4 or 5 feet from the floor, where the inlets of air samplers are typically placed. The prolonged hot baths of toddlers and school children in combination with the greater surface-to-volume ratio produces potentially greater exposure to volatile organic compounds in water than an adult receives in his or her 5- to 10-min shower. The mouthing behaviours of children become a constant means of incidentally ingesting contaminants on their hands or the objects that they mouth.
In response to the issues raised by the National Academy of Sciences report (NAS 1993), the methods used in exposure assessment have had to change in new directions as interest in children’s exposure to environmental contaminants has evolved. Previously, when children were the target population of interest (primarily lead exposure studies), information about the children was obtained from parents or carers. This allowed acquisition of global knowledge about exposure activities such as identifying the microenvironment in which the child spent time or macroactivities such as whether or not the child took a bath or play in a sandbox. The amount of time spent in a microenvironment, submerged in a bath, or in contact with the sand could not effectively be obtained from parents as the parents are often not present with the child, much less timing the events. The temporal information obtained from parents are at best ‘guesstimates’.
Collecting information from children also has problems as often the target population is so young that self-reports cannot be obtained. Even children as old as 10 or 12 years have difficulty with concepts of time, and reportage using real-time diaries have not been entirely successful (Schwab et al. 1990, 1991). Younger children not only have limited concepts of time, but also may not have the ability to read and complete diaries, or verbally express themselves in response to an interviewer.
Additional problems with understanding children’s exposure to environmental contaminants have emerged as the source of exposure has shifted from outdoor and/or indoor air pollutants to water-borne and dust/soil-borne contaminants and the routes of exposure have shifted from inhalation to dermal contact, and dietary and non-dietary ingestion. To understand these sources and routes of exposure information is needed about not only for microenvironments but microactivities, such as, contact with dust and soil, or mouthing objects and fingers (Cohen-Hubel et al. 1999; Reed et al. 1999). Additional sources of exposure in the child’s environment may be the toys the child plays with and mouths (Gurunathan et al. 1998). The dynamic character of semivolatile chemicals such as pesticides means that surfaces and objects not directly sprayed may become reservoirs and future sources of exposure. Understanding the potential exposure from these surfaces and objects requires collection of information about microactivities that has seldom been collected.
While parental reportage of time/activity information about their children continues to be used, parental responses are now being supplemented, if not supplanted, by observational methods (Zartarian et al. 1997a, 1998; Reed et al. 1999). Videotaped observations can be used to verify parental responses, quantify use of microenvironments, and collect frequency and duration data about microactivities. Reed et al. (1999) found that even a simple event such as hand washing was not accurately reported by either parents or day-care teachers. The adult reports were perhaps influenced by expectations rather than reality. Mouthing behaviours of children that contribute to children’s exposure to dust and soil-borne contaminants can only be accurately quantified by an observational technique. The independently conducted observational studies by Zartarian et al. (1997a, 1998) and Reed et al. (1999) found very similar frequencies of activities in toddlers in a Californian farm community and in urban and suburban New Jersey. The children in these studies made hundreds of hand contacts with surfaces and objects in their environment every hour, maximizing the opportunity of contact with contaminants. Part of the evolution in exposure assessment prompted by the NAS report is to think of exposure to a contaminant or family of contaminants from all potential media, i.e. to aggregate the individuals exposure from air, food, soil, dust, water, and other contaminated media. For the child, the ‘other media’ may be a major pathway, but one for which there are presently few data, and ones for which activities may have a large influence on exposure. This example of children just illustrates the needs of one subgroup of the general population. In the future, investigators will need to fill in major blanks for cultural, gender, and age-specific behaviours that can influence individual or subpopulation exposure.
Data analysis and models
Once microenvironmental and/or personal exposure data have been acquired in a field study or estimated by a model there are a number of analyses that can be used to place the data in a form that is helpful for examining a public health issue. The levels of analysis are dependent upon the types and amount of data available from a particular study or a series of companion or comparative studies (USEPA 1989a). A parallel issue is the form of the data necessary for the application of interest, for example epidemiology or risk assessment. For instance the data can be reported as exposure using the units of concentration and time (µg/m3 per h) or as a time-weighted average (µg/m3 per day). Then, depending upon the amount of data available, a distribution of exposure can be constructed and particular statistical quantities calculated from that data. Information that is derived from a distribution of exposures are shown in Fig. 10 (USEPA 1992a), and include the mean exposure (50th percentile), the high end exposure, greater than the 90th percentile, the form of the exposure distribution curve, and the worst possible case estimate (bounding estimate) of exposure.
Fig. 10 Major parameters to be determined from a distribution (known/default) of population exposure.
If the database includes information that can be examined across pathways or routes of exposure, the result will be estimates of total exposure across each medium or each route of entry into the body. The data collected that represent a day or week of a family, Table 9, and Fig. 9 can be used to determine the microenvironmental increments to total exposure. Theoretically, an integrated exposure can be derived from microenvironmental exposures by the summation formula found in Box 1.
Risk assessment applications require at least one further level of analysis: a dose calculated from the exposure level. The result can then be used in a risk characterization analysis, and, as stated earlier, these calculations can be in one of three forms: potential, internal, or biologically effective dose. The general form of the equations needed to calculate dose from exposure data are shown in Box 2. Based upon the ancillary information and parameters needed to complete such calculations (e.g. absorption rate) the value most frequently calculated is the potential dose. In rare instances the biologically effective dose can be calculated, but there are large uncertainties in the values used for factors to complete such calculations (e.g. repair rate etc.) (Lioy 1990).
Box 2 Generalized equations governing exposure and dose
where E is exposure, C(t) is time-variant concentration, and t1, t2 are time periods of exposure associated with a specific biological response.
where Dp is the potential dose and f(t) is the contact rate.
where DI is the internal dose and gab is the absorption function (e.g. skin, lung membrane, gut).
Target tissue dose
where DT is the target tissue dose and gpk is the pharmacokinetic model (accounts for absorption, distribution, and elimination processes).
Biologically effective dose
where p(as,rd,me,el) is a function based on nature of assimilation, repair, elimination, and/or metabolism.
A second reason for calculating a dose from the exposure data is to place units of measurement in a form that are consistent for comparisons among each route of entry. A typical form for dose is micrograms of contaminant per kilogram of body weight per unit of time. The format makes it easier to compare intensity of the contact with the amount that has been deposited within the body for different routes of entry. These data can also be used to determine which of the exposures encountered were at levels that may cause a biological effect.
As shown in Box 2, unless the investigator has acquired biological marker data, the determination of a dose requires information on a series of variables or factors that may only be measurable in detailed exposure assessment studies (USEPA 1989a; AIHC 1994). An update on such factors was published by the United States Environmental Protection Agency in 1999 on a CD-ROM. Examples of factors needed to complete dose calculations include breathing rate, skin absorption rate, ingestion rate, internal absorption rate, elimination, and repair rates. Obviously, it is easier to acquire data on breathing or ingestion rate than on organic cellular repair mechanisms. In fact, there are no methods available at the current time that can quantify cellular repair.
A type of data that is not usually available for dose calculations, but could be obtained, is the bioavailability of a contaminant in the matrix that contains it (e.g. soil) (Umbreit et al. 1986; Kitsa et al. 1992; Ruby et al. 1993; Wainman et al. 1994; Hamel et al. 1998). This value is dependent upon the amount of a contaminant that can be extracted from the matrix (e.g. soil) by bodily fluids found within the digestive system or the lung.
As it is not possible to acquire data routinely in a field study on accumulation or elimination rates, or absorption factors, dose calculations employ what have been conventionally described as generic exposure factors (single values or a distribution of values). Based upon the type of dose calculation the number of exposure factors selected could be minimal or extensive. These are driven by the data quality objectives, the amount of data available, the anticipated variability of the activities affecting the dose, and the types of individual or population characteristics considered to be of importance. Once these types of information have accumulated and the purpose and objectives of the analyses have been established, the analyst can complete either a point estimate of a dose or a distributional estimate of dose.
Point estimates of exposure require the application of an equation similar to those found in Box 2 for each route of exposure and each microenvironment that can lead to an individual having contact with chemicals. For example, selection of ingestion exposure, inhalation exposure, and dermal exposure scenarios can provide an estimate of the potential or, with additional data, an internal dose of a contaminant by completing a calculation similar to that illustrated in Box 3 (USEPA 1989a). Results can then be summed for all microenvironments and media to obtain point estimates of exposure for a hypothetical or representative member of the local population. One can also develop a distribution of dose point estimates based upon exposure measurements (e.g. personal monitoring) and/or estimates of exposure using exposure factors characteristic of the population of concern.
Box 3 Point estimate of potential dose
Ingestion of chemicals in water or beverages:
where CW is the chemical concentration in water, IR is the ingestion rate (l/day), EF is the exposure frequency (days/year), ED is the exposure duration (years), BW is the body weight (kg), and AT is the averaging time (period over which exposure is averaged) (days).
Chemicals in soil:
where CS is the chemical concentration in soil (mg/kg), IR is the ingestion rate (mg soil/day), CF is the conversion factor (106 kg/mg), FI is the fraction ingested from contaminated source (unitless), EF is the exposure frequency (days/years), ED is the exposure duration (years), BW is the body weight (kg), and AT is the averaging time (period over which exposure is averaged) (days).
Inhalation of airborne (vapour-phase) chemicals:
where CA is the contaminant concentration in air (µg/m3), IR is the inhalation rate (m3/h), ET is the exposure time (h/day), EF is the exposure frequency (days/year), ED is the exposure duration (years), BW is the body/weight (kg), and AT is the averaging time (period over which exposure is averaged) (days).
Dermal contact with chemicals in soil:
where CS is the chemical concentration in soil (mg/kg), CF is the conversion factor (106 kg/mg), SA is the skin area (cm2), AF is the soil-to-skin adherence factor (mg/cm2), ABS is the absorption factor (unitless), EF is the exposure frequency (days/years), ED is the exposure duration (years), BW is the body weight (kg), AT is the averaging time (period over which exposure is averaged) (days).
There has been a distinct move away from relying exclusively on point estimates of exposure and dose. This is done primarily to reduce the uncertainties that surround identifying a ‘most exposed individual’ (USEPA 1992a), which was frequently described as the person exposed to everything over a lifetime. In fact, exposure assessors are now being encouraged to employ distributional analyses by the frequency distributions of all or selected factors needed to estimate particular exposures or doses. This has led to the use of Monte Carlo techniques for combining the selected distributions of parameters or variables (Rubinstein 1981; Marnicio et al. 1991; USEPA 1992a, b; Hattis and Burmaster 1994). On the surface this appears to be a step forward in the development of exposure/dose data bases especially for risk assessment applications. However, there are some ‘land mines’ buried in the analysis of distributions that employ the random selection of points to establish a distribution of exposure. Figure 11 illustrates the general concept of combining distributions of independent variables to establish an overall distribution of one dependent variable; in our case exposure or dose. At first glance this seems to be a relatively simple task, as Monte Carlo techniques, available in many computer program, combine the points along each known or approximated variable distribution, and produces a final distribution that represents the exposure or dose. There are inherent statistical limitations to Monte Carlo analyses that must be examined prior to selecting the distributions used in applications of a particular set of exposure data. These have been outlined by numerous individuals (Rubinstein 1981; Marnicio et al. 1991; USEPA 1992a; Hattis and Burmaster 1994). Beyond the statistical constraints, there are other informational issues that must be evaluated to ensure that the estimates are plausible and realistic. Included is the evaluation of the usefulness of the values combined across distributions to simulate either the high end exposures or low end exposures. An example of a distribution of an exposure factor, fish ingestion, is shown in Fig. 12. It is clear that there is a tendency toward biomodality (AIHC 1994). The shape of the curve indicates different consumption patterns for subgroups of a population. Thus, proper utilization of the data requires knowledge of consumption activities within a potentially affected population.
Fig. 11 Representation of Monte Carlo analysis used to construct a dependent variable distribution Y.
Fig. 12 Distribution of fish ingestion. (Source: AIHC 1994.)
Evaluations of distributional data must also ascertain whether or not all projected exposures or doses can occur and can they occur for the situation or activity being under investigation. At a minimum sensitivity analyses should be conducted on the tails of the variable distributions used to estimate the exposure/dose. For example, an acute toxin (such as cyanide or ozone) at sufficient concentration to induce a biological response (death or asthma attack respectively) over a short period of time would not logically be coupled with a contact period equivalent to a week or more. An 82-year-old grandparent or unathletic person would not be spending too much time engaging in activities with a high ventilation rate 1.5 m3/h, when the outdoor ozone concentrations exceed 150 ppb. Finally, a child would not be spending 24 h a day over a 12-year period sitting on the grounds of a hazardous waste site. These examples may seem somewhat absurd, but if the appropriate constraints are not placed on a distributional analysis of exposure, these types of results and worse could be propagated through a computer program and reported as part of the estimated distribution of exposure or dose.
Although distributional analyses are more likely to be conducted for risk assessments, they are of value in epidemiological studies. A specific case is a comparison of a biological marker data for a contaminant or metabolite with a dose estimated from external exposure measurements. In intervention studies distributional data are of immense value for comparing a point measurement of exposure or dose (individual or affected subgroups) with the values observed and/or estimated for a much larger population (Lioy 1992).
Exposure assessment modelling
Predictions of an exposure or potential dose have been based on emissions, environmental transport and fate modelling (Thibodeaux 1979; Javandel et al. 1984; Cohen 1989; Georgopoulos 1990) and population time/location and activity pattern modelling combined with microenvironmental quality modelling (Ott 1980; Duan 1982, 1991; Schwab et al. 1991; Pardi 1992; Patrick 1994). This is called a prognostic assessment. Prediction of exposure can also be done based on modelling of biomarker data (Georgopoulos and Lioy 1994), which is called the diagnostic assessment. Whenever possible, both microenvironmental and biomarker data should be used, because these data are from independent sources, and should therefore result in the reduction of overall uncertainty (Roy and Georgopoulos 1998).
It was noted above, in the section on basic principles, that the overarching aim is to relate environmental releases to adverse health effect. Although it is possible to relate exposure with toxic effects, increasingly more direct relationships can be obtained by using potential dose, internal dose, and target tissue dose respectively. Calculation of target tissue dose requires the application of a pharmacokinetic model (sometimes referred to as a toxicokinetic model) that describes the uptake, distribution, metabolism, and elimination the toxicant. Pharmacokinetic models used in exposure assessments are generally compartmental models, which are empirically based, or physiologically based pharmacokinetic (PBPK) models that have a mechanistic basis and represent the major tissues of the body as separate compartments, linked by anatomically correct blood flows.
The fraction of an internal dose that reaches the target tissue can be highly dependent upon the route of uptake, and PBPK models, are a natural choice for estimating the target tissue dose for each route of uptake. For example, the fraction of an ingested internal dose that reaches the liver will generally be much greater than that of an inhaled internal dose, which in turn will be greater than that of a internal dermal dose. A further advantage of the PBPK formulation is that it is amenable to interspecies scale-up. This is an important attribute, as ethical and practical reasons generally preclude the intentional dosing of humans with toxic substances. Thus a PBPK model for humans can be developed on the basis of a PBPK model in laboratory animals (Ramsey and Anderson 1984). Moreover, PBPK models can be adapted to reflect the inherent variability in human populations. Model parameters are generally formulated as functions of body weight, and in principle this can be extended to other covariates such as age, height, and sex. Physiologically based models that relate exposure to internal and target tissue dose have been successfully applied to predict doses for a variety of toxicants. Both traditional ‘lumped’ parameter (ordinary differential equation) formulations as well as ‘refined’ distributed parameter (partial and ordinary differential equation) schemes, have been used for the inverse problem of dose to exposure medium to reconstruction (Georgopoulos and Lioy 1994; Georgopoulos et al. 1994). This approach utilizes time profiles of biomarker concentrations found in excreted fluids following exposure to reconstruct the single and/or multimedia/multiroute exposures experienced by an individual (e.g. simultaneous inhalation and dermal absorption of a volatile organic present in air and water).
A detailed exposure assessment may also require resource-intensive data collection studies or model-based simulations to characterize one or more of the following: source attributes, toxicant properties, geographical domain of influence attributes, population composition/stratification, population time/location pattern and activity patterns, macroenvironmental media properties/concentrations, microenvironmental media properties and concentrations, and the exposure routes and pathways. Consequently, the complexity of the exposure system and the wide range of information requirements necessitate simulation that can describe the exposure to dose or the dose to exposure. Finally, case-specific requirements of available mechanistic information must be available to link each component of the exposure continuum (Fig. 2), and then estimate doses potentially received by a particular population. The overall types of analyses and data needs required to complete an exposure simulation successfully is shown in Table 10 (Patrick 1994).
Table 10 Exposure modelling: concepts and data
A general modelling framework can guide the collection and analysis of new data while, on the other hand, the quality and quantity of available data limits the sophistication of any model. Priorities in data collection and model development must be established and the options must be explored for analysing available information and for modelling various components of the exposure system. Components of both single-medium and multimedia environmental and environmental exposure models, such as, for example the Human Exposure Model HEM II (USEPA 1991) used for the assessment of population exposures from air releases, and the STREAM model (Donigian and Mulkey 1992), used for the exposure assessment of pesticide run-off, should be considered in the development and expansion of models.
Microenvironment models should be evaluated prior to their use in assessing exposure for the application under consideration (Ott et al. 1988; USEPA 1989b). The relative advantages and limitations of stochastic simulation models (such as SHAPE, TRIM, and BEAM), cartesianization, or convolution models, and the most general double covariance models have been summarized elsewhere (Georgopoulos and Lioy 1994). An assumption of log normality for integrated exposures and doses provides a starting step for conducting probabilistic exposure analyses (USEPA 1992a); however, such an assumption, combined with the use of the off-the-shelf Monte Carlo simulation software, that typically assume non-correlation among the variables, can lead to erroneous results. Log-normal exposures are usually claimed as a direct result of log normality observed in ambient environmental concentrations; however, deviations to this assumption occur for the impact of isolated strong sources. Exponential concentration probability densities have been shown to apply in such systems. From a practical perspective a two- or three-parameter log-normal distribution is flexible enough to satisfactorily fit the main range of most right-skewed data sets, a reason for its popularity in practice.
Potential problems are associated with the additional requirements for accuracy of data needed to describe high exposures and doses. It is exactly in that range where assumptions on independence (typical in Monte Carlo simulations) are less valid. One solution in any analysis is to use asymptotic distributions of extremes, such as Gumbel’s double-exponential distribution, for the high ends (‘distribution tails’) of concentration and exposure time.
Practical application of exposure assessment has been mainly driven by generic or ‘typical’ assumptions (e.g. the person eating large quantities of waste all his or her life). However, as data evolve, management requirements for information obtained from large-scale, comprehensive exposure assessment programmes such as the National Human Exposure Assessment Survey, will be overwhelming by comparison with today’s standards for routine exposure data management (Sexton et al. 1995). Consequently, state-of-the-art information management tools must be evaluated and used to organize, utilize, and interpret exposure-related data efficiently. These include geographical information systems, interactive scientific visualization systems, distributed relational database management systems, and object-oriented environments for data and model integration.
Exposure probabilities for individuals and populations
Exposure distributions (probability density functions and cumulative distribution functions of exposure) for an individual expresses the probability that an individual will experience a given level of exposure over a specified duration (such as a day, year, or lifetime). The exposure distribution for a population, expresses the probability that a fraction of the population will experience a given level of exposure. Exposure distributions can vary significantly among individuals in a population, resulting in multimodal distributions for a population. For example, the distribution of exposure in a population can be bimodal when a fraction of the population is occupationally exposed at levels much greater than environmental levels experienced by the other fraction of the population. Consequently, population strata need to be characterized to achieve the data quality objectives. Exposure distributions that should be developed include individuals expected to experience the highest long-term exposures, individuals expected to experience the highest short-term exposures, and special or susceptible segments of the population.
As mentioned in the previous section, log-normal distributions of exposures are commonly employed, and they also have been suggested as a ‘default’ when case-specific information is not available (USEPA 1992a). The log-normal distribution possesses many advantageous properties, such as positivity (the probability of a negative exposure is zero), left-skewedness (implying that the average exposure is less than the median exposure), and its mathematical properties are well known as it is closely related to the normal distribution. Multimodal exposure distributions can sometimes be describing by superposition of two or more log-normal distributions. However, its adoption in a particular study should be with caution because log normality of a random variable implies that the randomness in the underlying processes are multiplicative. Other alternatives, such as asymptotic distributions for extreme values (e.g. bi-exponential), could potentially provide more appropriate information for risk analyses.
Attributes related to the potential target population and sensitive subpopulations should include (a) plausible contact patterns with the contaminants for different routes of exposure, (b) spatial population distribution stratification by age, sex, etc., and (c) identification of subgroups of people sharing potentially similar exposure patterns. As stated earlier, identification of time–activity patterns for potentially exposed populations, school children versus adults, men versus women, office workers versus outdoor workers, and so on and of spatial distributions of target population groups are essential for exposure assessments.
Exposure assessments are inherently uncertain, due to limitations in the precision with which nature can be observed, and due to the randomness inherent in nature. Uncertainty, and the closely related concept of variability, are means of quantifying the lack of knowledge regarding a quantity of interest, which in exposure assessment can be any variable affecting the estimation of exposure. Uncertainty generally refers to a lack of knowledge of a quantity due to limitations in available quantification techniques, whereas variability is a means of representing the lack of knowledge of a quantity due to unavailability of a measurement on the specific instance of the quantity. For example, an exposure assessment involving contaminated soil will be uncertain, because of a lack of knowledge regarding the relevant concentration of contaminant in soil resulting from (a) imprecision in contaminant concentrations measured in soil samples, and (b) variability in measured concentrations in several randomly selected representative samples. The variability in soil concentrations results in uncertainty in exposure assessment because it is not possible a priori to predict the exact concentration in soil that actually causes the exposure. Although it is useful to conceptualize these two sources of uncertainty in exposure assessment, ultimately however, it does not matter whether the lack of knowledge is due to uncertainty or variability, because they are both represented using probability distribution functions, and their effect on the exposure assessment is estimated by propagating the uncertainty through a exposure model in an identical manner. However, it is important to acknowledge explicitly that exposure assessments are inherently uncertain, and therefore exposure assessments should be probabilistic wherever possible. One of the main benefits of conceptualizing uncertainty as arising due to imprecision and variability is the reduction uncertainty by identifying and filling data gaps. The identification of data gaps usually involves a sensitivity analysis to determine the contribution of individual variables to the overall uncertainty in the exposure assessment. Reduction of uncertainty due to imprecision can only be effected by improving instrumentation, whereas it may be possible to reduce the uncertainty in exposure assessments due to variability in underlying factors by stratifying the population from which the samples are drawn (see Table 3). Knowledge of the population probability distribution functions can be used to judge the appropriateness of stratification of the population into smaller groups. For example, bimodal distributions are an indication that there are at least two subpopulations that are more homogeneous. This type of information is important in identifying subgroups by age, sex, race, and so on, and locating susceptible subgroups exposed to a contaminant.
Frequency distributions estimated from frequency distributions can be affected by a small sample size. In some cases only a few data points are available for quantities such as the mean, variance, and distribution. However, a confidence interval may only be calculated when the mean and variance of the distribution are known with certainty (e.g. based on large numbers of samples or data). A small sample size will increase the uncertainty in the mean and the variance. Calculation of tolerance intervals is one method for identifying sources of uncertainty (Mandel 1969).
Uncertainty about the underlying distribution of a variable can limit the application of standard statistical tests. Most tolerance and confidence intervals assume a normal distribution for all measurements. In cases where measurement error dominates the observed variance, this assumption may be reasonable; however, when there is significant interindividual variability, a skewed distribution can result. In this case, tolerance and confidence intervals based on an assumption of normality will not provide valid information error. Thus, it is important to view statistical tests as only one component in determining the accuracy of the exposure data.
The field of exposure analysis and its application to public health practices provides information and an understanding of the variety of ways an individual or population comes into contact with a contaminant. The approach must be framed within a conceptual framework that can involve multiple disciplines and interdisciplinary studies. Calculations of exposure and dose are data intensive, and often require situation-specific or site-specific data to characterize exposure accurately. Finally, the scientific approaches employed to establish measurement and modelling procedures must consider information on biological mechanisms or health outcomes.
AIHC (American Industrial Health Council) (1994). Exposure Factors Handbook, Sections 1–7. AIHC, Washington, DC.
Akland, G.G., Hartwell, T.D., Johnson, T.R., and Whitmore, R.W. (1985). Measuring human exposure to carbon monoxide in Washington, DC and Denver, Colorado during the winters of 1982–83. Environmental Science and Technology, 19, 911–18.
Andelman, J.B. (1985). Inhalation exposure in the home to volatile organic contaminants of drinking water. Science of the Total Environment, 47, 443–60.
Anderson-Sprecher, R., Flatman, G.T., and Borgman, L. (1994). Environmental sampling: a brief review. Journal of Exposure Analysis and Environmental Epidemiology, 4, 115–32.
Ashley, D.L., et al. (1992). Determining volatile organic compounds in human blood from a large sample population by purse and trap gas chromatography/mass spectrometry. Analytical Chemistry, 64, 1021–9.
Brown, S.H., Bishop, O.R., and Rowan, C.A. (1984). The role of skin absorption as a route of exposure to volatile organic compounds in drinking water. American Journal of Public Health, 74, 479–492.
Carpenter, G.J. and Huston-Stein, A. (1980). Activity, structure and sex-typed behavior in preschool children. Child Development, 51, 862–72.
Caudill, S.P. and Pirkle, J.L. (1992). Effects of measurement error on estimating biological half-life. Journal of Exposure Analysis and Environmental Epidemiology, 2, 463–76.
Clayton, C.A., et al. (1993). Particle Total Exposure Assessment Methodology (PTEAM) Study; distribution of aerosol and elemental concentrations in personal, indoors, and outdoor air samples in a southern California community. Journal of Exposure Analysis and Environmental Epidemiology, 3, 220–30.
Cohen, Y. (1989). Multimedia and intermedia transport modeling concepts in environmental monitoring. In: Intermedia pollutant transport: modeling and field measurements, (ed. D.T. Allen, I.R. Kaplan, and Y. Cohen). Plenum, New York.
Cohen Hubal, E., Sheldon, L., McCurdy, J., et al. (1999). Exposure assessment for children: a review of the factors influencing exposure of children, and the data available to characterize and assess that exposure. Environmental Health Perspectives, 108, 475–86.
Donigian, A.S. and Mulkey, L.A. (1992). STREAM, an exposure assessment methodology for agricultural pesticide runoff. In: Fate of pesticides and chemicals in the environment, (ed. J.L. Schnoor). Wiley, New York.
Duan, N. (1982). Models for human exposure to air pollution. Environment International, 8, 305–9.
Duan, N. (1991). Stochastic microenvironment models for air pollution exposure. Journal of Exposure Analysis and Environmental Epidemiology, 1, 235–57.
Fenske, R.A., Curry, P.B., Wandelmaier, F., and Ritter, L. (1991). Development of dermal and respiratory sampling procedures for human exposure to pesticides. Indoor Environment, 1, 11–30.
Fiserova-Bergerova, V. (1987). Development of biological exposure indices (BEIs) and their implementation. Applied Industrial Hygiene, 2, 87–92.
Freeman, N.C.G., Waldman, J., and Lioy, P.J. (1991). Design and evaluation of a location and activity log used for assessing personal exposure to air pollutants. Journal of Exposure Analysis and Environmental Epidemiology, 1, 327–38.
Freeman, N.C.G., Ettinger, A.S., Berry, M., and Rhoads, G.G. (1997). Hygiene and food related behaviors assoicated with blood lead levels of young children from lead contaminated homes. Journal of Exposure Analysis and Environmental Epidemiology, 7, 103–18.
Freeman, N.C.G., Ettinger, A.S., and Rhoads, G.G. (1999). Mouthing and hygiene behaviors of lead-burdened toddlers. Family and Community Health, submitted for publication.
Georgopoulos, P.G. (1990). The EOHSI Compendium of Air Quality Models—1990. Ozone Research Center Technical Report ORC-TR901101. Environmental and Occupational Health Sciences Institute, Piscataway, NJ.
Georgopoulos, P. and Lioy, P.J. (1994). Conceptual and theoretical aspects of exposure and dose assessment. Journal of Exposure Analysis and Environmental Epidemiology, 4, 253–85.
Georgopoulos, P.G., Roy, A., and Gallo, M. (1994). Reconstruction of using physiologically based pharmacokinetic models. Journal of Exposure Analysis and Environmental Epidemiology, 4, 1–20.
Georgopoulos, P.G., Walia, A., Roy, A., and Lioy, P.J. (1997). An integrated exposure and dose modeling and analysis system: Part 1—formulation and testing of microenvironmental and pharmacokinetic components. Environmental Science and Technology, 31, 17–27.
Gerlowsky, L.E. and Jain, R.K. (1983). Physiologically based pharmacokinetic models: principles and applications. Journal of Pharmaceutical Sciences, 72, 1103–27.
Graham, J., et al. (1992). Role of exposure databases in risk assessment. Archives of Environmental Health, 47, 408–21.
Gurunathan, S., et al. (1998). Accumulation of chlorpyrifos on residential surfaces and on/in toys accessible to children. Environmental Health Perspectives, 106, 9–16.
Hamel, S.C., Buckley, B., and Lioy, P.J. (1998). Bioaccessibility of metals in soils for different liquid for soil ratios in synthetic gastric fluids. Environmental Science and Technology, 32, 358–62.
Hattis, D.B. and Burmaster, D.E. (1994). Some thoughts on choosing distributions for practical risk analyses. Risk Analysis, 14, 713–30.
Hawley, J.K. (1985). Assessment of health risk from exposure to contaminated soils. Risk Analysis, 5, 282–302.
Henderson, R.F., Bechtold, W.E., and Maples, K.R. (1992). Biological markers as measure of exposure. Journal of Exposure Analysis and Environmental Epidemiology, 2 (Supplement 2), 1–14.
Javandel, I., Doughty, C., and Tsang, C.F. (1984). Groundwater transport: handbook of mathematical models. Monograph 10, AGU Water Resources, Washington, DC.
Jayjock, M.A. and Hawkins N.C. (1993). A proposal for improving the role of exposure modeling in risk assessment. American Industrial Hygiene Association Journal, 54, 733–41.
Jo, W-K., Weisel, C., and Lioy, P.J. (1990). Routes of exposure and body burden from showering with chlorinated tap water. Journal of Risk Analysis, 10, 575–80.
Kim, H-K. and Weisel, C.P. (1998). Dermal absorption of dichloro-and trichloroacetic acids from chlorinated water. Journal of Exposure Analysis and Environmental Epidemiology, 8, 555–75.
Kitsa, V., et al. (1992). Particle size distribution of total and hexavalent chromium in inspirable, thoracic and respirable soils from contaminated sites in New Jersey. Aerosol Science and Technology, 17, 213–29.
Lebowitz, M.D., Quackenboss, J.J., Kollander, M., l’Oczer, M.L., and Colome, S. (1989). Standard questionnaire for estimation of indoor concentrations. Journal of the Air Pollution Control Association, 39, 1411–19.
Lioy, P.J. (1990). The analysis of total human exposure for exposure assessment: a multi-discipline science for examining human contact with contaminants. Environmental Science and Technology, 24, 938–45.
Lioy, P.J. (1992). Exposure analysis and the biological response to a contaminant: a melding necessary for environmental health sciences. Journal of Exposure Analysis and Environmental Epidemiology, 2 (Supplement 1), 19–24.
Lioy, P.J. (1993). Measurement of personal exposure to air pollution: status and needs, measurement challengers in atmospheric chemistry, pp. 373–92. Advances in Chemistry Series 232. American Chemical Society, Washington, DC.
Lioy, P.J. (1999). Exposure analysis: reflections on its growth and aspirations or its future. Journal of Exposure Analysis and Environmental Epidemiology, in press.
Lioy, P.J. and Pellizzari, E. (1995). Conceptual framework for designing a national survey of human exposure. Journal of Exposure Analysis and Environmental Epidemiology, 5, 425–44.
Lioy, P.J., Weisel, C.P., Jo, W-K., Pellizzari, E., and Raymer, J.H. (1994). Microenvironmental and personal measurements of methyl-tertiary butyl ether associated with automobile use activities. Journal of Exposure Analysis and Environmental Epidemiology, 4, 427–41.
Lippmann, M. and Lioy, P.J. (1986). Critical issues in air pollution epidemiology. Environmental Health Perspectives, 62, 243–58.
Lodge, J.P. (1989). Methods of air sampling and analysis, (3rd edn). Lewis Publishers, Chelsea, MI.
McKone, T.E. (1991). Human exposure to chemicals from multiple media and through multiple pathways: research overview and comments. Journal of Risk Analysis, 11, 5–10.
Mandel, J. (1969). The statistical analysis of experimental data. Wiley, New York
Marnicio, R.J., Hakkinen, P.J., Lutkenhoff, S.D., Hertzberg, R.C., and Moskowitz, P.D. (1991). Risk analysis software and databases: review of Riskware ’90 Conference and Exhibition. Risk Analysis, 11, 545–60.
NAS (National Academy of Sciences) (1993). Pesticides in the diets of infants and children. National Academy of Sciences, Washington, DC.
NRC (National Research Council) (1985). Epidemiology of air pollution. National Academy Press, Washington, DC.
NRC (National Research Council) (1989a). Subcommittee report on biologic markers in pulmonary toxicology. National Academy Press, Washington, DC.
NRC (National Research Council) (1989b). Report on biologic markers in reproductive toxicology. Subcommittee on Reproductive and Neurodevelopmental Toxicology Committee on Biological Markers. National Academy Press, Washington, DC.
NRC (National Research Council) (1991a). Environmental epidemiology, Volume 1: Public health and hazardous wastes. National Academy Press, Washington, DC.
NRC (National Research Council) (1991b). Frontiers in assessing human exposures to environmental toxicants. National Academy Press, Washington, DC.
NRC (National Research Council) (1991c). Human exposure assessment for airborne pollutants: advances and opportunities. National Academy Press, Washington, DC.
NRC (National Research Council) (1992). Biologic markers of immunotoxicology. National Academy Press, Washington, DC.
NRC (National Research Council) (1993a). Measuring lead exposure in infants, children, and other sensitive populations. National Academy Press, Washington, DC.
NRC (National Research Council) (1993b). Pesticides in the diets of infants and children. National Academy Press, Washington, DC.
Ott, W. (1977). Development of criteria for siting air monitoring stations. Journal of the Air Pollution Control Association, 27, 543–8.
Ott, W. (1980). Models of human exposure to air pollution. Technical Report No. 32. Department of Statistics, Stanford University.
Ott, W. (1982). Concepts of human exposure to air pollution. Environment International, 7, 179–96.
Ott, W.R. (1990). Total human exposure: basic concepts, EPA field studies, and future research needs. Journal of the Air and Waste Management Association, 40, 966–75.
Ott, W.R. (1995). Human exposure assessment: the birth of a new science. Journal of Exposure Analysis and Environmental Epidemiology, 5, 459–72.
Ott, W., Thomas, J., Mage, D., and Wallace, L. (1988). Validation of the Simulation of Human Activity and Pollution Exposure (SHAPE) model using paired days from the Denver, Colorado, carbon monoxide field study. Atmospheric Environment, 22, 2101–13.
Pardi, R.R. (1992). IMES: A system for identifying and evaluation computer models for exposure assessment. Risk Analysis, 11, 319–21.
Patrick, D.R. (1994). Toxic air pollution handbook. Van Nostrand Reinhold, New York.
Pellizzari, E., et al. (1995). Population-based exposure measurements in EPA Region 5: a phase I field study in support of the National Human Exposure Assessment Survey. Journal of Exposure Analysis and Environmental Epidemiology, 5, 327–58.
Pellizzari, E., et al. (1999). Particulate matter and manganese exposures in Toronto, CA. Atmospheric Environment, 33, 721–34.
Perera, F.P., et al. (1987). DNA adducts, protein adducts, sister chromatid exchange in cigarette smokers and non-smokers. Journal of the National Cancer Institiute, 79, 449–56.
Pirkle, J.L., Needham, L.L., and Sexton, K. (1995). Improving exposure assessments by monitoring human tissue for toxic chemicals. Journal of Exposure Analysis and Environmental Epidemiology, 5, 405–24.
Ramsey, J.C. and Andersen, M.E. (1984). A physiologically based description of the inhalation pharmacokinetics of styrene in rats and humans. Toxicology and Applied Pharmacology, 73, 159–75.
Reed, K.J., Jimenez, M., Lioy, P.J., and Freeman, N.C.G. (1999). Quantification of children’s hand and mouthing activities. Journal of Exposure Analysis and Environmental Epidemiology, 9, 513–20.
Robinson, J.P., Wiley, J.A., Piazza, J., and Gerrett, K. (1989). Activity patterns of California residents and their implications for potential exposure to pollution. CARB-A6–177–33. California Air Resources Board, Sacramento, CA.
Roy, A. and Georgopoulos, P.G. (1998). Reconstructing week-long exposures to volatile organic compounds using physiologically based pharmacokinetic models. Journal of Exposure Analysis and Environmental Epidemiology, 8, 407–22.
Rubinstein, R.Y. (1981). Simulation and the Monte Carlo methods. Wiley, New York.
Ruby, M.V., et al. (1993). Development of an in vitro screening test to evaluate the ‘in vivo’ bioaccessibility of ingested mine-waste lead. Environmental Science and Technology, 27, 2870–7.
Ryan, P.B. (1991). An overview of human exposure modeling. Journal of Exposure Analysis and Environmental Epidemiology, 1, 453–74.
Ryan, P.B., Spengler, J.D., and Letz, R. (1986). Estimating personal exposures to NO2. Environment International, 12, 395–400.
Samet, J.M., Marbury, M.C., and Spengler, J.D. (1987). Health effects of sources of indoor air pollution I. American Review of Respiratory Disease, 136, 1466–87.
Samet, J.M., Marbury, M.C., and Spengler, J.D. (1988). Health effect of sources of indoor air pollution II. American Review of Respiratory Disease, 137, 221–44.
Schwab, M., Spengler, J.D., Ozhaynak, H., and Terblanche, P. (1990). The time/activity component of the Kanawah County health study. In: Total exposure methodology, pp. 118–29. Air and Waste Management Association, Pittsburgh, PA.
Schwab, M., Teiblanche, A.P.S., and Spengler, J.D. (1991). Self reported exertion levels on time/activity diaries, application to exposure assessment. Journal of Exposure Analysis and Environmental Epidemiology, 1, 339–56.
Seifert, B. and Abraham, H.J. (1983). Use of passive samplers for the determination of gaseous organic substances in indoor air at low concentration levels. International Journal of Environmental Analytical Chemistry, 13, 237–54.
Sexton, K. and Ryan, P.B. (1988). Assessment of human exposure to air pollution: methods, measurements and models, air pollution: the automobile and public health, pp. 207–38. National Academy Press, Washington, DC.
Sexton, K., Seleven, S.G., Wagener, D.K., and Lybarger, J.A. (1992). Estimating human exposure to environmental pollutants: availability and utility of existing databases. Archives of Environmental Health, 47, 398–407.
Sexton, K., Kleffman, D., and Callahan, M.A. (1995). An Introduction to the National Human Exposure Assessment Survey (NHEXAS) and related Phase I field studies. Journal of Exposure Analysis and Environmental Epidemiology, 5, 229–32.
Shulte, P.A. and Perera, E.P. (1993). Molecular epidemiology. Academic Press, San Diego, CA.
Spengler, J.D., Treitman, R.D., Tosteson, T.D., Mage, D.T., and Soczek, D.L. (1985) Personal exposures to respirable particulates and implications for air pollution epidemiology. Environmental Science and Technology, 19, 700–7.
Stetter, J.R., Jurs, P.C., and Rose, S.L. (1986). Detection of hazardous gases and vapors. Pattern Recognition Analysis from electrochemical sensor array. Analytical Chemistry, 58, 860–6.
Thibodeaux, L.J. (1979). Chemodynamics: environmental movement of chemicals in air, water, and soil. Wiley, New York.
Tsang, A.M. and Klepeis, N.E. (1996). Descriptive statistics tables from a detailed analysis of the National Human Activity Pattern Survey (NHAPS) data. US Environmental Protection Agency, Washington, DC.
Umbreit, T.H., Hesse, E.J., and Gallo, M.A. (1986). Bioavailability of dioxin in soil from a 2.4.5-T manufacturing site. Science, 232, 945–7.
USEPA (United States Environmental Protection Agency) (1980). Federal Register 79318–79.
USEPA (United States Environmental Protection Agency) (1988). Total Human Exposure Research Council (THERC). research needs in human exposure: a 5-year comprehensive assessment (1990–1994). Office of Research and Development, Washington, DC.
USEPA (United States Environmental Protection Agency) (1989a). Exposure factors handbook. EPA/600/08–89/043. Office of Health and Environmental Assessment, Washington, DC.
USEPA (United States Environmental Protection Agency) (1989b). Interim on benzene exposure assessment model (BEAM). EPA 600/X89/015, Jan. Environmental Measurements and Surveillance Laboratory, Las Vegas, NV.
USEPA (United States Environmental Protection Agency) (1991). Human exposure model, II: User’s guide (Draft). Office of Air Quality Planning and Standards, US Environmental Protection Agency, Washington, DC.
USEPA (United States Environmental Protection Agency) (1992a). Guidelines for exposure assessment. Federal Register FRL-4129.
USEPA (United States Environmental Protection Agency) (1992b). A Monte Carlo approach to simulating residential occupancy periods and its applications to the general population. EPA-450/3–92–011. Office of Air Quality Planning and Standards, US Environmental Protection Agency, Washington, DC.
Valerio, F., Pela, M., Lazarotto, A., and Balducci, D. (1997). Prelimlinary evaluation, using passive tubes, of carbon monoxide concentrations in outdoor and indoor air and street shops in Greece. Atmospheric Environment, 32, 2871–6.
Wainman, T., Hazen, R., and Lioy P. (1994). The extractability of CR(VI) from contaminated soil in synthetic sweat. Journal of Exposure Analysis and Environmental Epidemiology, 4, 171–81.
Wallace, L.A. (1989). Major sources of benzene exposure. Environmental Health Perspectives, 82, 165–9.
Wallace, L.A., et al. (1988). California TEAM study: breath concentrations and personal air exposures to 26 volatile compounds in air and drinking water of 188 residents of Los Angeles, Antioch and Pittsburgh, California. Atmospheric Environment, 22, 2141–63.
Weisel, C.P., Jo, W-K., and Lioy, P.J. (1993). Utilization of breath analysis for exposure and dose estimates of chloroform. Journal of Exposure Analysis and Environmental Epidemiology, 2 (Supplement 1), 55–69.
Zartarian, V.G., Ferguson, A.C., and Leckie, J.O. (1997a). Quantified mouthing activity data from a four-child pilot field study. Journal of Exposure Analysis and Environmental Epidemiology, 7, 543–53.
Zartarian, V.G., Ott, W., and Duan, N. (1997b). A quantitative definition of exposure and related concepts. Journal of Exposure Analysis and Environmental Epidemiology, 7, 411–38.
Zartarian, V.G., Ferguson, A.C., and Lecke, J.O. (1998). Quantification of mounting activity data from a child pilot study. Journal of Exposure Analysis and Environmental Epidemiology, 3, 543–54.