Grantee Research Project Results
Final Report: Characterization of Factors Determining Personal Exposure to Volatile Air Toxics in Urban Environments
EPA Grant Number: R826786Title: Characterization of Factors Determining Personal Exposure to Volatile Air Toxics in Urban Environments
Investigators: Esmen, Nurtan A. , Wang, D. , Lynch, Robert A. , Hall, Thomas A. , Johnson, D. L. , Phillips, M. L.
Institution: University of Oklahoma Health Sciences Center
EPA Project Officer: Chung, Serena
Project Period: October 1, 1998 through September 30, 2001 (Extended to September 30, 2002)
Project Amount: $559,352
RFA: Urban Air Toxics (1998) RFA Text | Recipients Lists
Research Category: Air , Air Quality and Air Toxics
Objective:
The primary objective of this research project was to investigate how external factors influence personal exposures to environmental air toxics. Using a multi-city, multi-season factorial design, we sought to study the distribution of personal exposures in relation to eight dichotomous macroenvironmental and household contrasting factors, which were hypothesized to influence personal activity and exposure patterns. The factors considered were: (1) size of urban area, (2) industrialization level of urban area, (3) cold/mild seasonal temperature, (4) hot/mild seasonal temperature, (5) precipitation/no precipitation, (6) workday/off-day, (7) presence/absence of children in household, and (8) blue collar/white collar occupational status.
We sought to quantify benzene, toluene, xylene isomers, ethylbenzene, styrene, n-hexane, and 2,2,4-trimethyl pentane as the index exposure chemicals. Quantification of styrene proved to be impracticable; instead, other common indoor air contaminants were added: acetone, tetrachloroethylene, dichlorobenzene, and trimethylbenzene. Of these chemicals, only benzene, toluene, o-xylene, p-xylene, and ethylbenzene provided consistently measurable concentrations that were useful for the stated purposes of the study.
In support of the primary objective, two subsidiary objectives of the project were to: (1) test and improve the reliability of self-reported personal activity data, and (2) characterize environmental exposure to airborne particulate matter.
Summary/Accomplishments (Outputs/Outcomes):
Of the 11 volatile organic compounds (VOCs) measured in this study, only benzene, toluene, o-xylene, and p-xylene were quantified in a sufficient number of samples to support statistical analysis. The minimal risk level for benzene, toluene, and p-xylene was exceeded in a small proportion of the samples, indicating that ambient exposures to these air contaminants can reach levels at which adverse health effects may occur. The discussion will focus on the implications of the benzene, toluene, o-xylene, and p-xylene results.
Effect of Observational Variables
The distributions of outdoor exposures were consistently and significantly lower than the indoor and personal exposure distributions. Except for daytime p-xylene concentrations, simultaneous indoor and outdoor exposures were largely uncorrelated. Therefore, it may be concluded that indoor exposures were dominated by indoor sources. This is consistent with the finding of the Total Exposure Assessment Methodology (TEAM) Study (Wallace, et al., 1987; Wallace, et al., 1991).
If the sources of exposure are indoor, it might be predicted that exposures would tend to be higher in tight houses than in relatively leaky houses. However, the results of this study did not support this prediction; normalized leakage, a measure of the permeability of a house, was found to be uncorrelated with indoor-sourced exposure, except in the case of nighttime p-xylene exposure, where permeability surprisingly showed a modest positive correlation with exposure. This result would be consistent with infiltration of p-xylene from attached garages. The absence of any significant correlation between permeability and indoor exposure in the predicted (negative) direction suggests that the observed indoor concentrations might reflect mostly localized, short-term emissions inside the house, which would be largely independent of the leakage ventilation rate, as opposed to steady-state concentrations of contaminants, which would be inversely related to the leakage rate.
The distributions of personal exposures and indoor residential exposures were fairly similar, in contrast to the finding of the TEAM Study that daytime personal exposures tended to be higher than indoor exposures (Wallace, et al., 1991). Simultaneous personal and indoor residential concentrations were correlated in daytime samples for all four VOCs, and in nighttime samples for toluene and o-xylene. This suggests that personal activities, which occurred mostly in the daytime, simultaneously influenced personal and indoor exposures. Also, the large portion of the day spent away from home did not appear to contribute substantially to the overall distribution of personal daytime exposures, possibly because few of the participants reported exposure to chemical products or environmental tobacco smoke while away from home. The lack of correlation between nighttime personal and indoor concentrations of benzene and p-xylene could reflect spatial variability in ambient exposure levels within the home, given that the nighttime "personal" samples were collected mostly in the subject's bedroom and the indoor samples were collected in the main living area.
The influence of urban, temporal, and household factors on exposure distributions was evaluated by two different methods. The Wilcoxon rank-sum test was used as a non-parametric analog to the t-test to determine whether the two distributions of a contrast were offset from each other. The excursion fraction was used to assess the influence of the factor on the probability of exceeding a health-effects-based "Operational Exposure Limit." Although the rank-sum test weighed all portions of a distribution equally, determination of the excursion fraction was typically dominated by the upper portions of the distribution. Overall, the excursion fraction test, using a cut-off of 10 percent of the MRL, was more powerful than the rank-sum test, but its use was limited to data sets with at least seven uncensored values, and it had a relatively low power for detecting differences between small excursion fractions.
Taking into account the different powers of the two methods, the outcomes were consistent. In both tests, dry weather, absence of children in the household, and absence of a refinery were found to be associated with higher exposures in personal or indoor exposures. In view of the finding that indoor concentrations were driven primarily by indoor sources rather than by outdoor sources, it is somewhat surprising that external factors (precipitation and refineries) had an apparent effect on indoor concentrations. The direction of the refinery effect also is surprising. A reasonable explanation of these results is that the factors: precipitation, refineries, and children were confounded with household characteristics or activities that were the true determinants of exposure. Fortuitous confounding with unrecognized factors is a problem inherent in small sample sizes. The borderline significance of the findings of the Wilcoxon rank-sum tests also suggests that none of the contrasts represented strong determinants of exposure.
Assessment of the Factorial Approach
The combinations of factors represented in the actual experiments deviated considerably from the planned fractional factorial set for a number of reasons: (1) weather during the experiment often did not conform to the forecast which was the basis for scheduling an event; (2) some combinations of temporal factors were relatively rare; and (3) representation of blue collar occupations, households with children, and residents of Ponca City within the volunteer pool fell short of the recruitment goals, which called for 24 households in each city, including an equal number of blue collar and white collar volunteers and an equal number of households with and without children. Although the factorial design could not be adhered to in practice, the attempt at a factorial approach was effective in preventing severe collinearity between factors. A relatively small correlation occurred among high/low temperature, mild/extreme temperature, and wet/dry weather, reflecting Oklahoma's climate with its hot, dry summers and mild winters. The factor "attached garage" was not part of the original factorial design. Its moderate correlation with the factors "workday" and "blue collar" provides an example of the type of happenstance collinearity that can occur in the absence of a factorial design.
The efficiency of the fractional factorial design can actually be a disadvantage in field-based studies because the small data sets generated are susceptible to high sampling variability and to confounding by unrecognized factors. This problem can be addressed-at considerably increased cost-by conducting replicate experiments. A 2-p fractional factorial design in n factors would then become, in essence, a stratified random sampling design with 2n-p strata.
Relevance of Distinction Between Daytime and Nighttime Exposures
Dividing the 24-hour monitoring period into nighttime and daytime samples made the monitoring procedure more complicated because the study participant had to change sample tubes in the morning. Collection of a single 24-hour sample would simplify the procedure, while also increasing the total volume of air sampled, resulting in a lower limit of detection. Therefore, it is reasonable to question whether it was productive to collect separate daytime and nighttime samples. In response, it should be noted that: (1) activity patterns differed between day and night; (2) although sequential nighttime and daytime exposures tended to be correlated, one was not strongly predictive of the other; and (3) some observed effects were not consistent across daytime and nighttime data sets. Combining daytime and nighttime exposures into a single 24-hour sample thus could reduce the resolution of a study. Therefore, although 24-hour samples are reasonable for surveillance studies, 12-hour or shorter term samples would be preferable for studies seeking to elucidate determinants of exposure.
An additional consideration is evidence from this study that suggests that indoor concentrations might be dominated by short-term events within the house. This possibility warrants further investigation involving multiple short-term samples.
Use of Excursion Fraction
Estimates of central tendency are of limited usefulness for many environmental data sets, which tend to be censored and to exhibit very broad multi-modal lognormal distributions. The 95th percentile value of a distribution is sometimes used as an estimate of the "worst case" exposure because this measure takes into account the spread of the data. It is not clear, however, what degree of difference between two 95th percentile values would be considered statistically significant. Another potential problem with the use of the 95th percentile point is that it implicitly discounts the most exposed portion of the population.
In this study, we chose to contrast the fraction of each fitted distribution that exceeded a biologically based "operational exposure level" (OEL), which was set at 10 percent of the Agency for Toxic Substance and Disease Registry (ATSDR) Minimal Risk Level (MRL). This measure explicitly recognizes the portion of the population that could be considered at risk of adverse effects from exposure. The OEL was set at 10 percent of the MRL rather than at the MRL to increase the power of the test; given the very small fraction of concentrations that exceeded the MRL, the data set was too small to find significant differences using the higher limit. Significance testing of contrasts was straightforward using the binomial confidence limits on the excursion fraction. The excursion fraction approach proved to be consistent with, but more powerful than, the nonparametric Wilcoxon rank-sum test at evaluating contrasts. More generally, the excursion fraction approach can be an intuitively simple tool for describing and comparing environmental data distributions in a variety of applications.
Conclusions:
With respect to the stated research goals, the primary objective and the first subsidiary objective were fully realized. The details of the study findings are presented in our final technical report to the U.S. Environmental Protection Agency, as well as in the publications appearing in the literature. In the second subsidiary objective, the research efforts fell short of our expectations, but a number of worthwhile findings are reported. Although we were expecting to quantify the small particulate matter exposures (PM2.5 exposures) in the same manner as air toxics, the PM2.5 concentrations were too low to make really meaningful apportioning between the factors we aimed to study. Probably the most important reason for this outcome is the relatively low population density of the cities studied. We suspect that this observation will be valid for most of the metropolitan areas in the United States, with the exception of the two coasts and the Great Lakes Region.
References:
Wallace LA, Pellizzari ED, Hartwell TD, Sparacino C, Whitmore R, Sheldon L, Zelon H, Perritt R. The TEAM study: personal exposures to toxic substances in air, drinking water, and breath of 400 residents of New Jersey, North Carolina, and North Dakota. Environmental Research 1987;43:290-307.
Wallace LA. Personal exposure to 25 volatile organic compounds. EPA's 1987 team study in Los Angeles, California. Toxicology and Industrial Health 1991;7:203-208.
Journal Articles on this Report : 4 Displayed | Download in RIS Format
Other project views: | All 27 publications | 6 publications in selected types | All 4 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Esmen NA, Johnson DL, Agron GM. The variability of delivered dose of aerosols with the same respirable concentration but different size distributions. The Annals of Occupational Hygiene 2002;46(4):401-407. |
R826786 (2001) R826786 (Final) |
Exit Exit |
|
Johnson DL, Esmen NA. Method-induced misclassification for a respirable dust sampled using ISO/ACGIH/CEN criteria. The Annals of Occupational Hygiene 2004;48(1):13-20. |
R826786 (Final) |
Exit Exit |
|
Phillips ML, Hall TA, Esmen NA, Lynch R, Johnson DL. Use of global positioning system technology to track subject's location during environmental exposure sampling. Journal of Exposure Analysis and Environmental Epidemiology 2001;11(3):207-215. |
R826786 (2001) R826786 (Final) |
Exit Exit |
|
Phillips ML, Esmen NA, Hall TA, Lynch R. Determinants of exposure to volatile organic compounds in four Oklahoma cities. Journal of Exposure Analysis and Environmental Epidemiology 2005;15(1):35-46. |
R826786 (Final) |
Exit Exit |
Supplemental Keywords:
exposure error, indoor air, model building., RFA, Health, Scientific Discipline, PHYSICAL ASPECTS, Air, Toxics, ENVIRONMENTAL MANAGEMENT, Geographic Area, Air Quality, particulate matter, Environmental Chemistry, Health Risk Assessment, air toxics, State, HAPS, VOCs, Risk Assessments, Biochemistry, Physical Processes, Children's Health, Atmospheric Sciences, indoor air, Ecology and Ecosystems, 33/50, Risk Assessment, ambient air quality, health effects, personal exposure, urban air toxics, Hexane, exposure and effects, Toluene, air pollutants, air quality models, ambient air, Xylenes, Ethyl benzene, fine particulates, exposure, air pollution, modeling, benzene, children, human exposure, hazardous air pollutants (HAPs), urban air pollution, PM, indoor air quality, fine particle levels, Volatile Organic Compounds (VOCs), Oklahoma (OK), 2, 2, 4-Trimethylpentane, Benzene (including benzene from gasoline), Styrene, Xylenes (isomers and mixture), activity patterns, exposure assessmentProgress and Final Reports:
Original AbstractThe perspectives, information and conclusions conveyed in research project abstracts, progress reports, final reports, journal abstracts and journal publications convey the viewpoints of the principal investigator and may not represent the views and policies of ORD and EPA. Conclusions drawn by the principal investigators have not been reviewed by the Agency.