Predicting Residential Exposure to Phthalate Plasticizer Emitted from Vinyl Flooring: Sensitivity, Uncertainty, and Implications for Biomonitoring

Background Because of the ubiquitous nature of phthalates in the environment and the potential for adverse human health effects, an urgent need exists to identify the most important sources and pathways of exposure. Objectives Using emissions of di(2-ethylhexyl) phthalate (DEHP) from vinyl flooring (VF) as an illustrative example, we describe a fundamental approach that can be used to identify the important sources and pathways of exposure associated with phthalates in indoor material. Methods We used a three-compartment model to estimate the emission rate of DEHP from VF and the evolving exposures via inhalation, dermal absorption, and oral ingestion of dust in a realistic indoor setting. Results A sensitivity analysis indicates that the VF source characteristics (surface area and material-phase concentration of DEHP), as well as the external mass-transfer coefficient and ventilation rate, are important variables that influence the steady-state DEHP concentration and the resulting exposure. In addition, DEHP is sorbed by interior surfaces, and the associated surface area and surface/air partition coefficients strongly influence the time to steady state. The roughly 40-fold range in predicted exposure reveals the inherent difficulty in using biomonitoring to identify specific sources of exposure to phthalates in the general population. Conclusions The relatively simple dependence on source and chemical-specific transport parameters suggests that the mechanistic modeling approach could be extended to predict exposures arising from other sources of phthalates as well as additional sources of other semivolatile organic compounds (SVOCs) such as biocides and flame retardants. This modeling approach could also provide a relatively inexpensive way to quantify exposure to many of the SVOCs used in indoor materials and consumer products.

Because of their substantial and widespread use, phthalates have become ubiquitous environ mental contaminants (Koch et al. 2003;Weschler and Nazaroff 2008;Wormuth et al. 2006). More than 3.5 million tons of phtha lates are used worldwide each year, primarily as plasticizers in flexible polyvinyl chloride (PVC) products (Cadogan and Howick 1996). Di(2 ethylhexyl) phthalate (DEHP) is an impor tant phthalate, with more than two million tons produced globally each year (Lorz et al. 2002). About 90% of phthalates are found in numerous consumer products, including floor and wall coverings, car interior trim, cloth ing, gloves, footwear, wire insulation, artificial leather, and toys Bornehag et al. 2005;Müller et al. 2003). DEHP is mainly used in PVC products such as vinyl flooring (VF), where it is typically present at concentrations of about 20-40% (wt/wt) Deisinger et al. 1998). Because phthalate plasticizers are not chemically bound to the product materials, they are emit ted slowly into the surrounding environment (Müller et al. 2003;Wormuth et al. 2006) and have become widely recognized as major indoor pollutants (Bornehag et al. 2005;Clausen et al. 2003;Fromme et al. 2004;Jaakkola and Knight 2008;Wensing et al. 2005;Xu and Little 2006).
The ubiquitous human exposure to phtha lates (Wormuth et al. 2006) is of concern because toxicologic studies in animals have demonstrated considerable adverse effects of phthalates and their metabolites (National Toxicology Program 2006). Because of the extensive environmental contamination with phthalates, a need exists to identify the most important sources and pathways of exposure [National Research Council (NRC) 2006]. Levels of phthalate metabolites measured in the general population using biomoni toring methods provide direct evidence of widespread human exposure (Calafat and McKee 2006;Centers for Disease Control and Prevention 2005;Heudorf et al. 2007). Biomonitoring data suggest that more than 75% of the U.S. population is exposed to phthalates ). For phthalates with short alkyl chains, monoesters represent the major human metabolites, although in the case of DEHP, diisononyl phthalate, and dii sodecyl phthalate, the monoesters are further metabolized. Exposure estimates based on uri nary monoester concentrations might under estimate the population's actual exposure to these specific phthalates (Wormuth et al. 2006). When urinary concentrations of sec ondary metabolites are measured, the esti mate increases to 95% (Kato et al. 2004). Results of recent biomonitoring studies, in which phthalate metabolites were measured, are reviewed in Heudorf et al. (2007). Using mean body burden of DEHP expressed as urinary excretion of DEHP metabolites, they estimated that the effective intake of DEHP is higher in children than in adults and may occur at levels of significant concern. Data are not available for children < 3 years of age.
Interpretation of biomonitoring data for public health decision making requires con textual information to understand the poten tial for adverse health impacts and to identify effective interventions (Albertini et al. 2006;Bahadori et al. 2007). Just as additional infor mation is required to relate a measured con centration of a chemical in a human tissue or fluid to the administered doses used in animal toxicity studies (Clewell et al. 2008), additional information is required to relate biomonitoring data to measures of the parent compound in environmental media Georgopoulos et al. 2008).
Although information on predominant sources, pathways, and routes of exposure is required to protect human health and the environment (NRC 2006), exposure to phthalates is difficult to evaluate because phthalates are so ubiquitous and because phthalate concentration measurements are hampered by contamination (Koch et al. 2003). To complicate matters, phthalates are sorbed strongly to surfaces, as do other semi volatile organic compounds (SVOCs) such as biocides and flame retardants (Weschler and Background: Because of the ubiquitous nature of phthalates in the environment and the potential for adverse human health effects, an urgent need exists to identify the most important sources and pathways of exposure. oBjectives: Using emissions of di(2-ethylhexyl) phthalate (DEHP) from vinyl flooring (VF) as an illustrative example, we describe a fundamental approach that can be used to identify the important sources and pathways of exposure associated with phthalates in indoor material. Methods: We used a three-compartment model to estimate the emission rate of DEHP from VF and the evolving exposures via inhalation, dermal absorption, and oral ingestion of dust in a realistic indoor setting. results: A sensitivity analysis indicates that the VF source characteristics (surface area and materialphase concentration of DEHP), as well as the external mass-transfer coefficient and ventilation rate, are important variables that influence the steady-state DEHP concentration and the resulting exposure. In addition, DEHP is sorbed by interior surfaces, and the associated surface area and surface/air partition coefficients strongly influence the time to steady state. The roughly 40-fold range in predicted exposure reveals the inherent difficulty in using biomonitoring to identify specific sources of exposure to phthalates in the general population. conclusions: The relatively simple dependence on source and chemical-specific transport parameters suggests that the mechanistic modeling approach could be extended to predict exposures arising from other sources of phthalates as well as additional sources of other semivolatile organic compounds (SVOCs) such as biocides and flame retardants. This modeling approach could also provide a relatively inexpensive way to quantify exposure to many of the SVOCs used in indoor materials and consumer products.  Nazaroff 2008). A relatively small gasphase concentration, such as 0.1 ppb, is sufficient for meaningful vapor transport of a phthalate ester and its consequent partitioning between the gas phase and indoor surfaces, including airborne particles and settled dust (Weschler 2003). Adibi et al. (2008) measured phthalate metabolite concentrations in urine samples from 246 pregnant women and correlated these with indoor air concentrations. They concluded that a single indoor air sample may be sufficient to characterize phthalate expo sure in the home. In the recent Children's Total Exposure to Persistent Pesticides and Other Persistent Organic Pollutants (CTEPP) study, the U.S. Environmental Protection Agency (U.S. EPA; 2005) measured concen trations of > 50 target compounds in multi media samples from the homes and daycare centers of 260 preschoolage children. The two phthalates targeted in the CTEPP study were detected in residential air and house dust and on various interior surfaces and dermal wipes. The measured phthalate concentrations were among the highest of any of the target compounds, including pesticides, polycyclic aromatic hydrocarbons, and polychlorinated biphenyls. Based on an analysis of data from the CTEPP study, Xu et al. (2009) developed a model to predict emission and transport of DEHP and to estimate the potential exposure through different pathways.
Using DEHP in VF as an illustrative example, we extended the Xu et al. (2009) model to predict DEHP emissions and potential exposures via inhalation, dermal absorption, and oral ingestion of dust after the installation of VF in a family residence. Rather than conduct an exhaustive exposure assessment, we illustrate an approach that can be used to identify the important sources and pathways of exposure associated with phtha lates in indoor materials and consumer prod ucts. As a result, we conducted sensitivity and uncertainty analyses to identify which model parameters have the greatest influence on exposure and to show why biomonitor ing alone cannot easily be used to identify individual sources of exposure in the general population. Finally, we briefly discuss how the modeling approach could be generalized to include other sources of SVOCs, as well as emissions, transport, and exposure in other environmental media.

Model Description and Results
As shown in Figure 1, DEHP is emitted from VF to the air in a typical residence that we divided into three compartments: kitchen, bathroom, and the main house. The emission rate is controlled by partitioning between the VF and the adjacent air, as well as the mass transfer coefficient within the boundary layer above the VF. The gasphase DEHP is sorbed on interior surfaces, including walls, ceiling, wood floor, carpet, furniture, windows, tile, ceramic fixtures, and particles through parti tioning mechanisms.We obtained the infiltra tion/exfiltration rates and ventilation rates between rooms shown in Table 1 from meas urements made by Wilkes et al. (1992) in a fiveroom house. We estimated the interior surface area of furnishing and materials using typical surface:volume ratios for American houses established by Hodgson et al. (2005) ( Table 1). VF comes in two main types. The one used in homes is softer and has a higher phthalate content than the more rigid one used in commercial applications. For model ing purposes, we use the commercial type because the emission characteristics and DEHP content have been comprehensively investigated in previous studies Xu and Little 2006;Xu et al. 2008Xu et al. , 2009.
We obtained sorption isotherms for phthalates on different interior surfaces from data collected in a residential field study and a laboratory chamber study (Xu et al. 2009). In the CTEPP field study (U.S. EPA 2005), 48hr integrated samples were col lected simultaneously from children's daycare centers and from their homes in either North Carolina or Ohio. The samples were collected from residential air, house dust, interior sur faces, and dermal wipes. Clausen et al. (2004) conducted laboratory experiments to study DEHP uptake by dust on PVC flooring in a

Vinyl flooring
Sinks chamber for laboratory investigations of mate rials, pollution, and air quality (CLIMPAQ). We used the DEHP concentrations in the dust and gas phase to determine the DEHP partition coefficient between dust and air. Loglinear relationships between equilibrium parameters and chemical vapor pressure were obtained, and the partition coefficients for DEHP on different interior surfaces calcu lated based on the vapor pressure of DEHP (Xu et al. 2009). We estimated the value of the masstransfer coefficient for the boundary layer adjacent to the various surfaces using correlation equations (Axley 1991).
The model was used to estimate DEHP emission and transport after VF was installed in a residence ( Figure 2). The three com partments reached steady state within about 1.5 years. The steep initial rise in DEHP con centration occurred because the rate at which it is emitted from the VF is initially faster than the rate at which it is taken up by the interior surface sinks. Compared with the other two compartments, the main house had the lowest gasphase concentration because of the larger ratio of sorption surface area (e.g., carpet and furniture) to emission surface area. The lower the gasphase concentration, the higher the concentration gradient in the boundary layer above the VF and the higher the emission rate. As shown in Table 2, the predicted steadystate concentrations are similar to those measured in homes in the United States and Europe.
Based on these results, we evaluated expo sures to gasphase DEHP in air, particlebound DEHP in air, and DEHP in settled dust. The exposure pathways of interest were inhalation of vapor, inhalation of particles, dermal sorp tion of DEHP, and oral ingestion of household dust. Both children and adults were considered in the assessment. We quantified the magni tude, frequency, duration, and time pattern of contact with DEHP using the screeninglevel assessment described by Xu et al. (2009). Figure 3 shows the change in time in exposure for adults and children (between the first and third year of life) through inhalation, dermal sorption, and oral ingestion of dust. Exposure reaches a steady level after about 1.5 years. Children experience 2-10 times greater exposure than do adults. The results are similar to those of Heudorf et al. (2007) who modeled ambient exposure data and concluded that children may be more highly exposed than adults. The reference dose (RfD) is 20 µg/kg/day according to the U.S. EPA. For children, exposure through oral intake via dust is two times higher than the RfD, although the assumed dust intake rate of 10.3 mg/kg/day may be high (Xu et al. 2009). For DEHP, the primary route of exposure is oral ingestion of dust; inhalation and dermal sorp tion do not appear to be dominant exposure pathways, which is consistent with observa tions of Clark et al. (2003).
Sensitivity analysis. We conducted a sen sitivity analysis to identify the critical model variables for total exposure and for each exposure pathway. Here, we computed expo sure after each of the three compartments had reached steady state. The sensitivity of the model variables were assessed by com puting the percent change in exposure per unit increase in an input variable. The base line conditions are those used for the results shown in Figure 3. Table 3 shows the results of the sensitivity analysis, along with the baseline values of selected model variables. Sensitivity to all model parameters is pro vided as Supplemental Material (doi:10.1289/ ehp.0900559.S1 via http://dx.doi.org/).
The properties affecting the source strength (initial DEHP concentration in VF, partition coefficient between VF and air, and surface area of VF) have a significant effect on all the exposure pathways. Increasing the masstransfer coefficient (h m ) will increase the emission rate and significantly increase exposure, whereas increasing the ventilation rate will reduce expo sure. Note, however, that the latter assumes an increase in airexchange rate alone, without increasing the masstransfer coefficients, which would tend to increase as ventilation increases. Increasing either the total suspended particle (TSP) concentration or the particle/air parti tion coefficient total total suspended particle (TSP) concentration or the particle/air parti tion coefficient (K particle/air ) is equivalent, either    of which has a stronger impact on dermal sorp tion and oral ingestion than on inhalation. The reason is that increasing sorption on particles reduces the gasphase concentration, and both dermal sorption and oral ingestion decrease sig nificantly. However, because particles contrib ute 80% of the inhalation exposure, the two effects were cancelled, and inhalation exposure increased only slightly. As expected, exposure duration and body weight also strongly influ enced the resulting exposure. Uncertainty analysis. Model variables can be defined in terms of a probability distri bution function (PDF) that is derived from a limited set of observations. We adopted a simple Monte Carlo analysis to account for uncertainty associated with the model param eters, as well as natural variability. A PDF for each of the important variables identified in the sensitivity analysis was randomly sampled to obtain a value for the variable. This set of model variables was then used to calculate exposure. The uncertainty analysis consisted of 1,000 such exposure computations, which we used to derive a cumulative distribution function describing an estimate of the uncer tainty in exposure.
As shown in Table 4, we developed ranges in selected model parameters from data pre sented in other studies or obtained directly from the literature. We used simple uniform distributions because of the relative paucity of data, even though this may overestimate uncertainty. Figure 4 summarizes the uncer tainty for the individual exposure pathways as well as for total exposure. Overall, expo sure varies from about 5 µg/kg/day at the 5th percentile to about 180 µg/kg/day at the 95th percentile, a roughly 40fold difference. The median value (50th percentile) of about 38 µg/kg/day is almost double the RfD.

Discussion
The high surface concentrations of phthalate on human skin observed in the CTEPP study were generally assumed to have been the result of dermal transfer. Cohen Hubal et al. (2008) studied the dermal transfer of chemicals from contaminated surfaces (e.g., floors and furni ture) to skin, providing a range of measured transfer efficiencies, all of which were < 100%. Closer examination of the CTEPP data shows that the measured concentrations on skin were almost always higher than the measured concentrations on other surfaces. To inves tigate if the high dermal loadings are caused by transfer of chemicals from contaminated surfaces or from partitioning with air, we con ducted a multilinear regression. As shown in Table 5, the skin concentrations are strongly correlated with the concentrations in air and are not correlated with the hard surface con centrations. Because equilibrium is established fairly quickly between surfaces and air, the dermal transfer of phthalate from surface to skin may not have a substantial influence on exposure (Xu et al. 2009). This is supported by the recent results of Adibi et al. (2008), who concluded that a single indoor air sample may be sufficient to characterize phthalate exposure in the home.
In the simple sensitivity analysis described above, we varied only one parameter at a time. However, when the ventilation rate is increased, the masstransfer coefficients will also increase because of the higher air veloc ity near the surfaces. As a result, the emission rate of DEHP from VF will be higher and the rate of DEHP sorption to interior surfaces will be faster. Thus, the predicted exposure will decrease by only 25% compared with the decrease of 46% predicted in the simple sensi tivity analysis. In addition, the boundary layer of air adjacent to the skin will be thinner and the masstransfer resistance will be reduced. Because the external gasphase resistance con trols the overall rate of dermal permeation (Xu et al. 2009), the permeability of DEHP through the skin will be enhanced, meaning that dermal exposure will actually increase by 13%, as opposed to the decrease of 46% found in the simple sensitivity analysis. This   rather surprising result suggests that the use of indoor fans could increase the permeation rate of DEHP through the skin. Many other interior surfaces, including clothing, bedding, rugs, newspapers, books, magazines, human hair, crockery, and cut lery have not been taken into account in our exposure model. To get a rough idea of the effect of including these additional surfaces, we nominally increased all interior surface areas by a factor of 3 from the model baseline conditions. In this case, sorption of DEHP to the much higher surface area doubles the time to reach steady state. Direct dermal sorp tion and ingestion from these other surfaces may increase the risk of DEHP exposure significantly. For example, DEHP would be expected to accumulate in clothes hanging in a cupboard. When these are worn, dermal sorp tion could increase substantially. De Coensel et al. (2008) studied the chemical contami nation of clothes because of their direct or indirect exposure to moth repellent agents, which are similar to SVOCs, and concluded that clothes sorb high concentrations of con taminants, and that they should be considered as secondary sources of indoor air pollution. Although the surface/air partition coefficient for the interior surfaces did not have an effect on the predicted steadystate exposure, it will influence the time to reach steady state. The stronger the partitioning between interior sur faces and air, the longer it will take to reach steady state. For instance, doubling the wall and ceiling/air partition coefficient increases the time to steady state by about 50%.
Other sources, such as food packaging, may be important DEHP exposure pathways (Koch et al. 2003), and young children can be exposed by mouthing soft PVC toys and teethers (Petersen and Breindahl 2000). In addition, plasticized PVC is the most widely used electrical insulation material on wires and cables, with an estimated length of about 16 million kilometers in U.S. buildings today (Wilson 2009). By varying the DEHP con tent, cable manufacturers are able to produce wiring that remains flexible at low tempera tures. These additional sources will result in higher DEHP concentrations in room air and dust and on skin. Many other sources of phthalates also exist in the environment. Because the model employs a mechanistic approach to predict exposure to DEHP emit ted from VF, it should be relatively simple to generalize the model to include phthalates emitted from these other sources. As shown in the sensitivity analysis, the most influen tial, chemicalspecific model parameters are the various masstransfer and partition coef ficients. The partition coefficients generally correlate well with vapor pressure, whereas the chemicalspecific dependence of the mass transfer coefficients is easy to estimate (Xu et al. 2009).
Implications for biomonitoring. The abil ity to measure chemicals in humans (biomon itoring) is far outpacing the ability to reliably interpret these data for public health pur poses, which has created a major knowledge gap (Bahadori et al. 2007). As discussed in the introduction, the use of biomonitoring data to design and evaluate public health inter ventions for compounds such as phthalates requires additional information on potential sources, temporal and spatial patterns of expo sure, and a mechanistic understanding of the sourcetooutcome continuum. The sensitiv ity and uncertainty analyses presented above suggest that a single phthalate (DEHP) in a single material (VF) could result in a popula tion exposure that varies by as much as a fac tor of 40. This wide range in exposure would confound the interpretation of crosssectional biomonitoring results.
In the context of human health risks, Calafat and McKee (2006) outline research needs for using DEHP biomonitoring data to inform exposure assessment. Their recom mendations include the need to identify vul nerable segments of the population that may be more highly exposed to phthalates than is the general population and to identify sources of exposure to these vulnerable groups. The example we present in this article demonstrates the utility of physically based models for pre dicting concentrations of SVOCs as a function of time and space in residential environments. Such an approach combined with traditional scenariobased exposure algorithms facilitates identification of potentially vulnerable groups such as pregnant women, babies, and young children. Our example shows that the depen dence on source and chemicalspecific prop erties is relatively simple, suggesting that the model could be extended to include other sources of phthalates, as well as other charac teristics of the indoor environment.
A recent report on phthalates and cumulative risk assessment by the National Academies (NRC 2008) recommends that the U.S. EPA should a) determine prenatal exposure to phthalates at relevant times dur ing pregnancy; b) identify the most important sources of phthalate exposure in the general population; c) identify the full spectrum of phthalate metabolites, which are produced when phthalates enter the body, and identify the metabolites that can be used to reliably indicate phthalate exposure; d) understand the reasons for differences in susceptibility to phthalates based on age, species, and expo sure route; and e) explore the potential of phthalates to cause synergisms in combina tion with other antiandrogens. It is clear that biomonitoring alone cannot provide answers to recommendations b and d. In contrast, the approach articulated in this article can be used to identify the most important sources of phthalate exposure and can explain differ ences in susceptibility to phthalates based on age, species, and exposure route. Although our example focuses on emissions from a spe cific source (VF) to a specific environmental medium (air), it can most likely be general ized to many other sources emitting various SVOCs (e.g., insulated wiring, cosmetics, personalcare products, pharmaceuticals, medical devices, children's toys, food pack aging, and cleaning and building materials) into a wide range of environmental media (air, food, water, saliva, and even blood), pro vided that appropriate behavioral and product use factors are incorporated. Assuming that the necessary model development, param eter identification, and model validation are undertaken, the approach could prove to be a relatively inexpensive and efficient way to identify potential exposures associated with many of the SVOCs used in indoor materials and consumer products. Abbreviations: BBP, benzyl butyl phthalate; BPA, bisphenol A; DBP, dibutyl phthalate. a (q in μg/m 2 ), b (x 1 in μg/m 3 ), and c (x 2 in μg/m 2 ), where q = K 1 (x 1 ) + K 2 (x 2 ).