INTEGRATED PROBABILISTIC AND DETERMINISTIC MODELING
TECHNIQUES IN ESTIMATING EXPOSURE TO WATER-BORNE
CONTAMINANTS: PART 1: EXPOSURE MODELING
CR Wilkes1*, JN Blancato2, SC Hern2, FW Power3 and SS Olin4
1 Wilkes Technologies, Inc., Bethesda, MD USA 20814
2 US Environmental Protection Agency, Human Exposure and Atmospheric Sciences Division
3 Anteon Corporation, Las Vegas, NV USA 89119
4 International Life Sciences Institute, Risk Science Institute, Washington, DC USA 20005
ABSTRACT
The Total Exposure Model (TEM) uses deterministic and stochastic methods to estimate the
exposure of a person performing daily activities of eating, drinking, showering, and bathing.
There were 250 personal activity time histories generated for the three exposure routes (oral,
dermal, and inhalation) for the chemicals trichloroethylene (TCE), trichloroacetic acid (TCA),
and dichloroacetic acid (DCA). Uptake models were used to estimate the absorbed dose. In a
second, related paper, these exposures were input to the physiologically based
pharmacokinetic (PBPK) model, ERDEM (Exposure Related Dose Estimating Model).
DISCLAIMER
The United States Environmental Protection Agency through its Office of Research and
Development collaborated in the research described here and the Office of Water partially
funded the work under Cooperative Agreement CX-822663-01. This work is being submitted
for peer review and has not yet been approved for final publication. Mention of trade names
or commercial products does not constitute endorsement or recommendation for use.
INDEX TERMS
Exposure model, Pharmacokinetic model, VOCs, Human activities
INTRODUCTION AND OBJECTIVES
Exposure models have evolved in recent years from simple models that examine a single,
straightforward exposure event to more sophisticated models that represent the multitude of
factors affecting exposure (e.g., activity patterns, source-use behavior, physical and chemical
properties, etc). This paper is the first of a two-part paper that examines enhancing the utility
of an exposure model by linking it to a pharmacokinetic model, with this paper describing the
exposure modeling study and results. The second part (Blancato et al., 2002) will present the
results of the pharmacokinetic modeling study. Such a combined application provides a
wealth of information, including an understanding of path and route specific contributions to
target tissue concentrations and dose.
Modeling exposure to water-borne contaminants in a residential setting requires consideration
of the factors that affect the concentration of the contaminant in each of the two contact
media: air and water. It also requires consideration of factors that affect contact with the
contaminated media, including water-use behavior and the nature and length of the contact. A
model has been developed to represent the complex relationship between the physical
Contact author email: c.wilkes@wilkestech.com
-------
environment, the behavior of the occupants, and the physical and chemical processes that
affect the resultant contact between the chemical and the occupants. The model, initially
developed as MAVRIQ (Model for the Analysis of Volatiles and Residential Indoor-air
Quality, Wilkes et al., 1996), has been further modified with additional capabilities, and is
currently named the Total Exposure Model (TEM). The general model processes and
capabilities have been described in detail elsewhere (ILSI, 1998).
The model estimates exposure on a case-by-case basis by combining deterministic and
stochastic techniques. Once a population group is chosen for an exposure assessment, input
parameters, such as the activity and location behavior are sampled from a database compiled
from National Human Activity Pattern Survey (NHAPS) (Klepeis et al., 2001). The sampled
occupant locations are mapped to appropriate locations in a modeled "representative
residence," with housing characteristics held constant for all simulations. Water use
frequency and duration, as well as source representations and contaminant characteristics, are
defined explicitly, sampled from representative databases, or sampled from representative
distributions. Water use activities are simulated, consistent with the sampled location and
activity, based on population characteristics. Once the input parameters are selected, a mass
balance model is executed to predict air and water concentrations. Combining these predicted
concentrations with the locations of the occupants yields an estimated exposure for one
representative member of the population group. This simulation process is repeated to
compile a distribution of expected exposures to members of the population group.
EXPOSURE MODELING STUDY
The exposure study considers the expected exposure of a population group to water-borne
trichloethylene (TCE), dichloroacetic acid (DCA) and trichloroacetic acid (TCA). Each of
these chemicals can potentially cause adverse health effects in humans. TCE has been widely
used as a household and industrial solvent, and is a widespread environmental contaminant.
TCE affects the central nervous system, the liver and kidneys, and is a suspected carcinogen
(Lewis, 1998). DCA and TCA are metabolites of TCE and are chlorination disinfection by-
products (DBFs), found in many U.S. municipal water supplies. DCA and TCA are classified
as probable and possible human carcinogens, respectively, by the USEPA (Cohn et al, 1999;
Krasner et al., 1989). For this study, the water supply is assumed to be contaminated with 100
ug/L of TCE and 30 ug/L each of DCA and TCA. The exposure can occur through the three
primary routes, ingestion, inhalation, and dermal.
Source Representation
Each of the water-using appliances or fixtures, when operated, represents an opportunity for
emission of water-borne chemicals. The emission behavior during a given water use is a
function of a variety of chemical and physical factors, including water temperature, surface
area, concentration, chemical diffusivities, and Henry's Law constant. To facilitate prediction
of water and air concentrations, the emission behavior is idealized using two types of models:
the plug flow model (PFM) and the completely mixed flow model (CMFM). The derivations
of these models are presented elsewhere (ILSI, 1998).
The equations representing emissions require a volatilization coefficient (Ky), which is based
on the overall mass-transfer coefficient (KoiA), and Henry's Law constant (H). Corsi and
Howard (1992) conducted a series of laboratory experiments to determine mass-transfer
coefficients. The experiments were conducted for 5 reference chemicals (acetone, ethyl
acetate, toluene, ethylbenzene, and cyclohexane) and for 5 water-use types (sinks, showers,
-------
bathtubs, washing machines, and dishwashers). The results of this study were used to
estimate the source model parameters for TCE, which are tabulated in Table 1.
Table 1. Mass Transfer Parameters for TCE Used in the Exposure Model
Water Appliance
Shower
Bathtub: Fill
Pool
Toilet
Dishwasher
Clothes Fill
Washer Agitat
Faucets
Model
PFM
CMFM
CMFM
CMFM
quilibriu
CMFM
CMFM
PFM
Water
o
Temp. C
35
35
35
20
35
35
35
20
Water Flowrate
L/min
10
NA
NA
NA
NA
NA
NA
6
KOLA
L/min
13.5
2.5
1.0
0.1
NA
4.2
13.0
1.5
Kv
L/min
7.4
2.5
1.0
0.1
NA
4.2
13.0
1.33
H
TCEA
0.68
0.68
0.68
0.68
0.68
0.68
A. Reference: Lincoff and Gosset, 1984.
Activity Patterns and Water Use Activities
Activity patterns of the occupants need to account for activities that result in water use and
subsequently contaminant release, and activities that bring the occupant into contact with the
contaminant in the air or water. Since the objective of this study is to generate population
based exposure and dose estimates, NHAPS is used as the basis for generating population-
based activity patterns. The database contains 24-hour records of the activities and locations
for surveyed individuals in the United States. The population group being evaluated is single-
family households comprised of two adults, both between the ages of 18 and 33, one male and
one female. The occupant activity patterns are sampled from the subset of NHAPS database
matching gender and age. Water uses are subsequently simulated consistent with the sampled
activity pattern. The sampled activity pattern was translated from the original domain (the
sampled individual's home) to the model domain (the "representative" modeled house). For a
complete discussion of this procedure, see Wilkes, 1998.
Although limited information about water appliance use is contained in the database, the
information is not complete, and cannot be used as a part of the sampled activity pattern. For
this reason, the water use behavior is simulated based on an analysis of the water use behavior
reported in NHAPS and other surveys. The frequency and length of showering and bathing
for this population group taken from their responses to the NHAPS questionnaire are given in
Table 2. The frequency and duration of water use behavior for dishwashers, clothes washers,
toilets and faucets are based on values reported in the Exposure Factors Handbook and other
sources, as described in Wilkes, 1998.
Table 2. Exposure Model Water Use Characteristics
Appliance
Shower
Bath
Toilet
Dishwasher
Clothes Washer
Faucets
Mean Event 1
Male
13.7
20.5
NA
Duration, min
Female
13.8
20.5
NA
73
45
0.5
0.5
Stand. Dev.
Male
7.9
13.4
NA
Duration, min
Female
8.3
13.4
NA
6.8
6
0.3
0.3
Frequi
Male
1.28
0.24
5
^ncy/Day
Female
1.16
0.37
5
0.4
0.3
12
12
-------
For the model simulations, the occurrence of water uses by each individual are simulated
consistent with the sampled activity pattern. This is accomplished by identifying all intervals
in the activity pattern where a particular water use is allowed to occur, based on the reported
activity and location. For example, if an individual reports being in the bathroom engaged in
personal hygiene, the shower, bathtub, faucet, and toilet are eligible for use. The Poisson
process is used to simulate the occurrence of the water-use events. Subsequently, a lognormal
distribution is used to simulate the length of the water use. This process is discussed in
Wilkes, 1998.
Residence
Wilkes et al. (1996) showed that the most important predictors of inhalation exposure in
decreasing order were the time spent in the shower, the time spent in the bathroom, and the
time spent in the bath. For these reasons, the idealization of the residence focused on
explicitly representing each major water-using location in the home as a well-mixed zone with
the non-water-using zones aggregated into a common zone. The resultant idealized residence
provides a representation of the physical environment for calculating air concentrations. The
volumes of the various locations within the home are based on the 50* percentile values for
US housing, as given in Table 3 (Hoke 1988, USEPA. 1998). The airflows shown in Figure 1
are assigned to represent airflows typically found
in U.S. housing, with a whole house air exchange
rate of approximately 0.4. The airflows between
the bathroom and the shower are set at 50 m3/hr
based on an unpublished analysis of shower
ventilation rates in the Giardino and Andelman
study in Vanport, PA (Giardino, 1990). Temporal
effects such as opening and closing of doors and
windows were not considered.
Table 3. House Representation
House
Bathroom
Shower
Laundry
Kitchen
Volume,
m3
310
12
2.8
21.5
20
Vent. Rate,
m3/hr
^124
9.9
50
15
13.3
Uptake Models
Uptake models are implemented to estimate the amount of contaminant absorbed into the
bloodstream. For inhalation exposure, an equilibrium lung model is used. The breathing rates
are based on measured values from various studies, as reported in the Exposure Factors
Handbook (EPA, 1997). Two breathing rates were used, one for resting periods and another
for active periods. For resting periods, a breathing rate of 0.43 and 0.33 m3/hr are used for the
adult male and female, respectively, and 0.64 and 0.50 m3/hr for active periods, respectively.
For dermal exposure, a membrane model is used. For a complete description of these models,
refer to Olin et al. (1998). The skin permeability coefficients for TCE, DCA, and TCA used in
the model are 0.0157, 3.58E-6, and 1.84E-6 cm/hr, respectively. For a discussion of these
values, see (Blanato et al., 2002). Both direct and indirect consumption were represented in
the ingestion exposure calculation. The water concentration was adjusted to account for
assumed processing (e.g., cooking, pouring, etc.) prior to consumption. Consumption of tap
water was represented based on analysis of USDA data gathered between 1994-1996 (Jacobs
et al., 2000). The consumption behavior was represented as a lognormal distribution, with a
geometric mean of 321.5 ml/day and 195 ml/day, respectively for direct and indirect
consumption, with geometric standard deviations of 1.02 and 1.19, respectively. For realistic
input into the pharmacokinetic model, the direct consumption was randomly broken into 6
consumption events and simulated using a Poisson process. In a similar manner, the indirect
consumption was distributed across 8 events. The ingestion uptake was assumed to be 100%
of the amount of mass entering the stomach. The amount of TCE loss due to volatilization
-------
prior to consumption was estimated as 22% for direct consumption and 75% for indirect
consumption. For DCA and TCA, no volatilization occurs,
RESULTS
The exposures and potential doses resulting from the scenario described above are calculated
for 250 simulations of the two-person household. The resulting simulated water uses were
analyzed and compared to the water use characteristics specified in Table 2. The resultant
water uses were slightly lower than those specified because of limitations in the sampled
activity patterns (i.e., the reported activity pattern does not always have an opportunity for all
appropriate water uses). The results for the inhalation, dermal, and ingestion routes and the
total cumulative dose for each of the three chemicals are summarized in Figure 1. The line
labeled "Total Dose" represents the cumulative sum of the three routes. The shaded area
represents the fraction attributed to each route as a function of the total dose. In addition, the
percentiles as a function of route of exposure are presented on the tables inset in Figure 1.
Pereentue Dermal Ingestion Inhalation Total
0.002 0.0092 0.039
0.003 0.0197 0.113
0.009 0.0325 0.235
0.012 0.0641 0.415
0.022 0.1658 0.832
0.078
0.179
0.299
0.492
0.903
o
o
Dermal, Percentage of Total
I luges don, Percentage of Total
I | Inhalation, Percentage of Total
. Total Dose, Cumulative Distribution Function
OG
r~~
oo
O
-------
DISCUSSION AND CONCLUSIONS
The results of this modeling study show that, for volatile chemicals such as TCE, the
inhalation route dominates the exposure and uptake of the chemical, and increases in
proportion as the dose increases across the population. For the three chosen chemicals, the
dermal route appears to be of least concern, but it should be noted that there is considerable
uncertainty in the uptake parameters for dermal and ingestion. The results of this paper in
conjunction with the related Part 2 paper (Blancato et al., 2002) demonstrate the usefulness of
merging exposure and pharmacokinetic modeling. Although there is considerable uncertainty
in both the exposure and pharmacokinetic modeling parameters, the results demonstrate the
effect of individual chemicals, pathways, human behavior, and a variety of other factors on
exposed populations, and provide insight into the relationship between environmental
contaminants and their effect on exposed populations.
REFERENCES
Blancato JN, Power FW, Wilkes CR, et al. 2002. Integrated Probabilistic and Deterministic
Modeling Techniques in Estimating Exposure to Water-borne Contaminants: Part 2:
Pharmacokinetic Modeling. Proceedings of this conference.
Cohn PD, Cox M, and Berger PS. 1999. Health and Aesthetic Aspects of Water Quality. In
Water Quality & Treatment, 5th Ed. AWWA. McGraw-Hill.
Corsi R. and Howard CH. July 1998.Volatilization Rates from Water to Indoor Air, Phase II.
U.S. EPA Report.
Giardino NJ. 1990. Assessment and Modeling of Shower Emissions of Volatile Organic
Chemicals. University Pittsburgh, Doctoral Thesis.
Hayduk W and Laudie H. 1974. Prediction of Diffusion Coefficients for non Electrolytes in
Dilute Solutions. A1CHEJ, Vol. 26, No. 3, pp 611-615.
Hoke JR (Ed). 1988. Architectural Graphic Standards. 8th edition, John Wiley and Sons, New
York, NY.
Klepeis NE, Nelson WC, Ott WR, et al 2001. The National Human Activity Pattern Survey
(NHAPS): a resource for assessing exposure to environmental pollutants. J. Exposure
Analysis and Environmental Epidemiology. Vol 11, pp 231-252.
Lincoff AH, and Cosset JM. 1984. The Determination of Henry's Constant for Volatile
Organics by Equilibrium Partitioning in Closed Systems. In Gas Transfer at Water
Surfaces. Brutsaert, W and Jirka, GH, eds. D Reidel Publishing Company.
Krasner SW, McGuire MJ, Jacangelo JG, et al. 1989. The Occurrence of Disinfection By-
products in US Drinking Water. J. A WWA.
Lewis RA. 1998. Lewis' Dictionary of Toxicology. CRC Press, Inc.
USEPA. 1998. Residential Building Characteristics, pp 17-1 - 17-32, In Exposure Factors
Handbook. Report No. EPA-600/P-95/002BA, U.S. Environmental Protection Agency.
Verschueeen K, 1983. Handbook of Environmental Data on Organic Chemicals. Van
Nostrand Reinhold, NY.
Westrick JJ, Mello M, and Thomas RF. 1984. The Groundwater Supply Survey. J AWWA.
Wilkes CR, Small MJ, Davidson CI, et al. 1996. Modeling the effects of water usage and co-
behavior on inhalation exposures to contaminants volatilized from household water. J.
Exposure Analysis and Environmental Epidemiology. 6(4):393-412.
Wilkes CR 1998. Case Study. In Exposure to Contaminants in Drinking Water. Olin SS, ed.
International Life Sciences Institute. CRC Press.
Wilkes CR, Mason AD, and Hern SC. 2002. Probability Distributions for Showering and
Bathing for Various US Subpopulations based on NHAPS. Draft EPA Report.
-------
NF.RT.-RTP-HKASn-0?-07n
TECHNICAL REPORT DATA
1. Report No,
EPA/600/A-02/070
2,
4. Title and Subtitle
Integrated Probabilistic an Deterministic Modeling Techniques in Estimating
Exposure to Water-Borne Contaminants: Part 1 Exposure Modeling
5. Report Date
Submitted 4/02
6. Performing Organization Code
7. Author(s)
Charles R. Wilkes1, Jerry Blaneato2, Stephen C, Hem2, Fred Power2, and
Stephen S. Olin3
8. Performing Organization
Report No,
9. Performing Organization Name and Address
1. Wilkes Technologies, Inc, Bethesda, MD 20814
2. U.S. EPA, Las Vegas, NV 89114
3. International Life Sciences Institute, Washington, DC 20036
10. Program Element No.
3906, 8.2,1, 3-002 & 3-025A, APG28,
APM36
11. Contract/Grant No.
IAG #DW4?93944301
12. Sponsoring Agency Name and Address
U.S. EPA, Las Vegas, NV 89114
13. Type of Report and Period
Covered
14. Sponsoring Agency Code
15. Supplementary Notes
To be presented at Indoor Air 2002, Monterey, CA, June 30 - July 5,2002
-------
TECHNICAL REPORT DATA
16. Abstract
Exposure to contaminants originating in the domestic water supply is influenced by a number of factors, including human
activities, water use behavior, and physical and chemical processes. The key role of human activities is very apparent in
exposure related to volatile water-borne contaminants, since the release of a contaminant and resultant exposure is related
to the activity of the exposed individual. Estimates of human exposure are often plagued by a poor understanding of
many of the factors affecting the predictions. For example, existing activity pattern surveys do not contain integrated
information about relevant water-using activities such as showers, and showering studies do not collect relevant
information about the location, duration, and water temperature. Methods for integrating diverse data resources into a
consistent modeling framework have been developed and implemented for prediction of exposure to water-borne
contaminants. These methods are implemented in a computer model entitled the Total Exposure Model (TEM). TEM
predicts the exposure and dose to an individual resulting from use of a contaminated water supply by modeling the
fundamental physical and chemical processes that occur during interaction between the contaminated media (in this case
water and air) and the exposed individual.
An application of the model to estimate inhalation, dermal and ingestion exposure to several common waterborne
contaminants to several population groups will be presented. The exposure study considers the expected exposure of a
population group to water-borne trichloethylene (TCE), dichloroacetic acid (DCA) and trichloroacetic acid (TCA). The
distributions of exposures are estimated for several population groups to each compound as a function of route of
exposure (inhalation, dermal and ingestion). The results are provided as inputs to a pharmacokinetic model, ERDEM
(Exposure Related Dose Estimating Model) where target organ concentrations and doses are estimated across the
population groups. This paper presents the results of the exposure modeling and analysis. A second, related paper will
also be presented providing the methods and results of the pharmacokinetic modeling.
This work has been funded wholly or in part by the United States Environmental Protection Agency and has been
approved for publication. Mention of trade names or commercial products does not constitute endorsement or
recommendation for use.
17. KEY WORDS AND DOCUMENT ANALYSIS
A. Descriptors
Human activities, VOCs and SVOCs, residences, modeling
indoor pollutant concentrations, exposure assessment
B. Identifiers / Open Ended
Terms
C. COSATI
18. Distribution Statement
Release to the public
19. Security Class (This
Report)
Unclassified
20. Security Class (This
Page)
Unclassified
21. No. of Pages
6
22. Price
Form Available: Network Neighborhood\Knight\Groups\HEASD\Forms\Tech-Form-2220-l
------- |