Grantee Research Project Results
Final Report: A Source-Oriented Evaluation of the Combined Effects of Fine Particles and Copollutants
EPA Grant Number: R827997Title: A Source-Oriented Evaluation of the Combined Effects of Fine Particles and Copollutants
Investigators: Ito, Kazuhiko , Thurston, George D. , Xue, Nan , Lall, Ramona , DeLeon, Samantha
Institution: New York University School of Medicine , Albert Einstein College of Medicine
EPA Project Officer: Chung, Serena
Project Period: February 18, 2000 through February 17, 2004 (Extended to February 17, 2006)
Project Amount: $478,522
RFA: Airborne Particulate Matter Health Effects (1999) RFA Text | Recipients Lists
Research Category: Air Quality and Air Toxics , Air , Human Health , Particulate Matter
Objective:
The objective of this project was to apply novel approaches to estimate the combined effects of size-specific particulate matter (PM) air pollution (i.e., PM2.5) and their co-pollutants in major U.S. cities where source types, levels of PM co-pollutants, and weather patterns vary considerably. The alternative approaches to evaluate the health impacts of PM are needed because the current (prevailing) regression approach does not address the possible source-specific toxicity of PM, or its interactive toxicity with gaseous co-pollutants that may or may not come from the same source. PM is a chemically non-specific pollutant, and may originate or be derived from different emission source types. Thus, its toxicity should vary depending on its chemical composition and perhaps on the presence of any gaseous co-pollutants. In time-series analyses of the acute effects of PM, the prevailing approach to dealing with gaseous co-pollutants is to treat them as confounders, and to include them simultaneously in regression models. Such an approach can not only lead to misleading conclusions in “identifying” the causal pollutant (e.g., when pollutants are correlated and have varying extents of exposure error), but also cannot address the likely combined effects of PM and gaseous co-pollutants. The regulatory implications of these limitations in the prevailing regression approach could be serious, because the wrong sources may be identified for regulatory control. The expected reduction in risk (i.e., benefits) may also not be optimized, depending on the extent to which the PM and the gaseous co-pollutants share the same source types. The null hypothesis to be tested is that the PM effect size estimate is constant for all source types, regardless of composition or the presence of gaseous co-pollutants.
Summary/Accomplishments (Outputs/Outcomes):
This project resulted in several important findings, but also identified some important issues that need to be further investigated in interpreting source-specific PM2.5 effects or the combined effects of PM2.5 and gaseous air pollutants. The areas of findings include: (1) spatial-temporal variations of PM2.5 chemical components and gaseous pollutants; (2) source-specific effects of PM2.5 and gaseous pollutants; and (3) the relationship between PM2.5, gaseous pollutants, and weather variables.
Spatial-Temporal Variations of PM2.5 Chemical Components and Gaseous Pollutants
In interpreting source-specific effects of PM2.5 or the combined effects of PM2.5 and gaseous pollutants, it is important to assess the extent of spatial/temporal variations of PM2.5 chemical components and gaseous pollutants because, in time-series analysis, the relative strength of associations with health outcomes among the air pollutants or PM components can be biased downward for those pollutants that have larger spatial variations (i.e., exposure measurement errors). Possible exposure characterization errors across PM2.5 components were examined using PM2.5 chemical speciation data collected at three locations in New York City (NYC) during 2001–2002 (Ito, et al., 2004). The species that are associated with secondary aerosols (e.g., SO4, NH4, NO3, organic carbon [OC], etc.) tended to show high monitor-to-monitor correlations, whereas the species that are likely associated with more local sources (e.g., elemental carbon [EC] as a traffic source marker) showed lower monitor-to-monitor correlation. Source-apportionment using these data was also conducted for each monitor’s data using Positive Matrix Factorization (PMF) and Absolute Principal Component Analysis (APCA). The estimated source-apportioned PM2.5 mass generally showed the highest monitor-to-monitor correlation for the secondary aerosol factor (r range: 0.72–0.93). The correlation for the more localized traffic-related factor was more variable (r range: 0.26–0.95). The estimated mean PM2.5 mass contributions by source/pollution type across the monitors varied least for the secondary aerosol factor. These results suggest that the extent of exposure error varies across the PM components and the derived source-apportioned PM2.5 mass. Thus, interpretations of health effects analysis using specific PM2.5 chemical components or derived source-apportioned PM2.5 mass concentrations will need to take into consideration their associated exposure measurement errors.
The above analysis in NYC was limited to PM2.5 chemical speciation data. To further investigate this issue for gaseous pollutants, we conducted an analysis of monitor-to-monitor correlation for all the monitors within a 20-mile radius from the center of NYC for PM2.5, collected by the 24-hr filter samples using the Federal Reference Method (FRM, 30 monitors), PM2.5 measured by the Tapered Element Oscillating Microbalance (TEOM) procedure (24 monitors), ozone (O3, 16 monitors), nitrogen dioxide (NO2, 15 monitors), sulfur dioxide (SO2, 19 monitors), and carbon monoxide (CO, 18 monitors) for the study period 1999–2002. The corresponding median correlations were 0.91, 0.95, 0.89, 0.87, 0.74, and 0.60, respectively. The coefficients of variation of mean values of these monitors were 11%, 8%, 19%, 17%, 36%, and 36%. Thus, SO2 and CO are also expected to have larger exposure measurement errors compared to PM2.5 (whose major constituent is sulfate) and other gaseous pollutants.
We also extended the analysis of monitor-to-monitor correlation of PM2.5 chemical speciation to 28 metropolitan statistical areas (MSA’s) where multiple monitors were available for the years 2001–2003. Because the NYC analysis (Ito, et al., 2004) described above also found that many of the chemical species that showed low monitor-to-monitor correlation were also those that had a low percentage of measurements above detection limit (ADL), we focused on the components that had high ADL and of interest in terms of source signature and toxicological importance: Ni, V, Pb, Zn, Cr, Mn, Fe, Si, As, Se, sulfate, nitrate, EC, and OC. Again, the species associated with secondary aerosols (e.g., sulfate and nitrate) showed high monitor-to-monitor correlation. However, the monitor-to-monitor correlation for other species varied widely across the MSA’s, likely reflecting the variation in the levels and major source types across the MSAs.
Source-specific Effects of PM2.5 and Gaseous Pollutants
Our initial plan was to utilize the newly available PM2.5 chemical speciation data, but due to the delayed start of data collection in most of the monitors, and less-than-expected sampling frequencies and amount of data collected, the statistical power of health effects analysis was less than desirable for the new speciation data network for our project period, especially for mortality analysis because daily counts were smaller than morbidity. Thus, we chose to use the PM2.5 chemical speciation data collected in Washington, DC, as part of the Interagency Monitoring of Protected Visual Environment (IMPROVE) network. The data were analyzed for source-apportionment in a workshop by multiple investigators (Thurston, et al., 2005; Hopke, et al., 2006). We analyzed the associations between daily mortality and the estimated source contributions of PM2.5 for the period from 1988–1997, as obtained from each participant research group’s source apportionment procedure(s). A Poisson Generalized Linear Model (GLM) was used to estimate source-specific relative risks at lags 0 through 4 days for total non-accidental, cardiovascular, and cardio-respiratory mortality adjusting for weather effects, seasonal/temporal trends, and day-of-week. Several alternative weather models, including the ones similar to those used in recent PM time-series studies were also applied as a sensitivity analysis. Generally, the effect size estimates obtained for each of the source types and their patterns of lagged associations were similar across research groups. However, the varying lag structure of associations across source types, combined with the Wednesday/Saturday sampling frequency in this data set made it difficult to compare source-specific effect sizes in a simple manner. The largest (and most significant) estimated percent excess deaths per 5th-to-95th percentile increment of source apportioned PM2.5 for total mortality was found for secondary sulfate (mean percent excess mortality = 7.1% [95 %CI: 2.0, 12.2]), although its lag structure of association was peculiar (lag 3 day). Primary coal related PM2.5, identified by only three teams, was also significantly associated with total mortality and had the same lag structure as sulfate. Risk estimates for traffic related PM2.5, while significant in some cases, were more variable across investigators. Soil related PM showed smaller effect size estimates, but they were more consistently positive at multiple lags. The patterns of associations for cardiovascular and cardio-respiratory mortality were generally similar to those for total mortality. Alternative weather model specifications generally gave similar patterns of associations, but in some cases they affected the lag structure of associations (e.g., for sulfate). The extent of multi-collinearity (i.e., concurvity) between the source-apportioned PM2.5 and regression covariates also varied across source types. Overall, for a given lag, the variations in relative risks across the investigator/methods were found to be smaller than those across source types or those across weather models. The cross-investigator consistency found in these Washington, DC results suggest the robustness of the application of source apportionment methods to health effects analyses, but also raises analysis issues that will need to be investigated by considering speciation data from multiple cities.
Secondary sulfate was found to be an important predictor of mortality in the Washington, DC analysis described above, but gaseous pollutants were not analyzed in that analysis, and left some uncertainty regarding potential confounding with gaseous pollutants, especially ozone which is also high during the warm season when sulfates are high. Conversely, ozone mortality effects may be confounded by sulfate and associated fine particles during summertime. Our meta-analysis of ozone mortality effects of existing studies (Ito, et al., 2005) also raised this question and also identified additional uncertainty regarding sensitivity of risk estimates to weather model specifications. Thus, in addition to the meta-analysis, we also conducted time-series analysis of mortality with both ozone and PM in the model for all-year, warm season, and cold season in six U.S. cities: Chicago (1985–1994, with PM10), Detroit (1992–1994, with PM2.5), Houston (1985–1994, with PM10), Minneapolis–St. Paul (1985–1994, with PM10), Philadelphia (1992–1995, with PM2.5), and St. Louis (1985–1994, with PM10) (Ito, et al., 2005). Our specific aims were to: (1) characterize ozone–PM relationships across seasons; (2) examine the sensitivity of mortality risk estimates to four alternative weather models applied in the past literature and the extent of temporal adjustments; (3) obtain ozone–mortality risk estimates by season; and (4) examine the sensitivity of ozone–mortality risk estimates to adjustment for PM by season. The correlation between PM and ozone were positive for summer but negative in winter in these cities, except in Houston, where the correlation was positive in both seasons. A large city-to-city variation in the PM and ozone risk estimates was found. These risk estimates were generally not sensitive to the extent of degrees of freedom to adjust for season. However, the difference in the weather adjustment model could result in a 2-fold difference in risk estimates. In Chicago, Detroit, and Philadelphia, PM in single pollutant models was also associated with mortality in both all-year and warm months. Including both ozone and PM in the regression models in these cities tended to attenuate both pollutants’ coefficients, but not substantially. The models with both ozone and PM in these cities often showed better fits (i.e., lower Akaike’s Information Criteria) than those with either pollutant alone. These results suggest that ozone and PM contribute independently to mortality. In the other 3 cities, PM was less strongly associated with mortality and did not influence ozone–mortality associations.
Relationship Between PM2.5, Gaseous Pollutants, and Weather Variables
Most of recent air pollution epidemiological studies focused on PM and ozone. These studies often also considered other gaseous co-pollutants as potential confounders, including nitrogen dioxide (NO2), sulfur dioxide (SO2), and carbon monoxide (CO). However, because each of these air pollutants can have different seasonal patterns and chemical interactions, the interpretation of each pollutant’s individual risk estimates may not be straightforward. Multi-collinearity among the air pollution and weather variables also leaves the possibility of confounding and over- or under-fitting of meteorological variables, thereby potentially influencing the effect estimates for the pollutants. To investigate these issues, we examined the temporal relationships among air pollution and weather variables in the context of air pollution health effects models (Ito et al., submitted, 2007). We compiled daily data for particulate matter less than 2.5 μm (PM2.5), ozone, NO2, SO2, CO, temperature, dewpoint, relative humidity, wind-speed, and barometric pressure for New York City for the years 1999–2002. We conducted several sets of analyses to characterize air pollution and weather data interactions, in order to assess different aspects of these data issues: (1) spatial/temporal variation of PM2.5 and gaseous pollutants measured at multiple monitors; (2) temporal relationships among air pollution and weather variables; and (3) extent and nature of multi-collinearity of air pollution and weather variables in the context of health effects models. The air pollution variables showed a varying extent of inter-correlations with each other and with weather variables, and these correlations also varied across seasons. For example, NO2 exhibited the strongest negative correlation with wind speed among the pollutants considered, while ozone’s correlation with PM2.5 changed signs across the seasons (positive in summer, negative in winter). The extent of multi-collinearity problems also varied across pollutants and choice of health effects models commonly used in the literature. These results suggest that the health effects regression may need to be run by season for some pollutants to provide the most meaningful results. Model choice and interpretation also needs to take into consideration the varying pollutant concurvities with the model co-variables in each pollutant’s health effects model specification.
Conclusions:
- Different extent of exposure measurement error was observed across PM2.5 chemical components, derived source-apportioned PM2.5, and gaseous air pollutants. Therefore, interpretation of the relative strength of association with health outcomes across these pollution indices will need to take into consideration the corresponding exposure error. Such exposure errors likely vary across cities even for the same pollutant, depending on the prevailing source types.
- The mortality analysis using source-apportioned PM2.5 in Washington, DC, found that PM2.5 associated with sulfate was most strongly associated with mortality. Risk estimates for traffic related PM2.5, while significant in some cases, were more variable across investigators. Alternative weather model specifications generally gave similar patterns of associations, but in some cases they affected the lag structure of associations (e.g., for sulfate). Since the extent of multi-collinearity (i.e., concurvity) between the source-apportioned PM2.5 and regression covariates also varied across source types, caution is required in interpreting relative importance of source-apportioned PM2.5 risk estimates.
- In the cities where PM mortality associations were found in both all-year and warm months in single pollutant models (Chicago, Detroit, and Philadelphia), including both PM and ozone in the warm months tended to attenuate both pollutants’ risk estimates, but not substantially. Better model fits in these models, compared to those in single pollutant models, also appear to suggest that both PM (whose major fraction in the warm months is sulfate) and ozone contribute to the excess mortality.
- Estimating the combined health effects of PM2.5 and gaseous pollutants is not straightforward because the exposure error and concurvity in health effects models vary across these pollution indices. Such patterns are likely different across cities. Thus, a comprehensive understanding of source-specific PM effects or combined effects of PM and gaseous pollutants will require a multi-city study in geographic areas that vary in source types and weather patterns with sufficient statistical power.
References:
Thurston GD, Ito K, Mar T, Christensen WF, Eatough DJ, Henry RC, Kim E, Laden F, Lall R, Larson TV, Liu H. Neas L, Pinto J, Stolzle M, Suh H, Hopke PK. Workgroup report: workshop on source apportionment of particulate matter health effects—intercomparison of results and implications. Environmental Health Perspectives 2005;113(12):1768-1774.
Hopke PK, Ito K, Mar T, Christensen WF, Eatough DJ, Henry RC, Kim E, Laden F, Lall R, Larson TV, Liu H, Neas L, Pinto, Stolzle M, Suh H, Paatero P, Thurston GD. PM source apportionment and health effects: 1. Intercomparison of source apportionment results. Journal of Exposure Science and Environmental Epidemiology 2006;16(3):275-286.
Journal Articles on this Report : 5 Displayed | Download in RIS Format
Other project views: | All 12 publications | 5 publications in selected types | All 5 journal articles |
---|
Type | Citation | ||
---|---|---|---|
|
Ito K, Xue N, Thurston G. Spatial variation of PM2.5 chemical species and source-apportioned mass concentrations in New York City. Atmospheric Environment 2004;38(31):5269-5282. |
R827997 (2005) R827997 (Final) R827351 (2003) R827351 (Final) R827351C001 (2003) R827351C001 (Final) |
Exit Exit Exit |
|
Ito K, De Leon SF, Lippmann M. Associations between ozone and daily mortality: analysis and meta-analysis. Epidemiology 2005;16(4):446-457. |
R827997 (Final) |
Exit Exit |
|
Ito K, Christensen WF, Eatough DJ, Henry RC, Kim E, Laden F, Lall R, Larson TV, Neas L, Hopke PK, Thurston GD. PM source apportionment and health effects: 2. An investigation of intermethod variability in associations between source-apportioned fine particle mass and daily mortality in Washington, DC. Journal of Exposure Science & Environmental Epidemiology 2006;16(4):300-310. |
R827997 (Final) R827351 (Final) R827351C001 (Final) R827353C015 (Final) R827354 (Final) R827354C001 (Final) R827355 (Final) R827355C008 (Final) R832415 (2010) R832415 (2011) R832415 (Final) |
Exit Exit |
|
Ito K, Thurston GD, Silverman RA. Characterization of PM2.5, gaseous pollutants, and meteorological interactions in the context of time-series health effects models. Journal of Exposure Science and Environmental Epidemiology 2007;17(Suppl 2):S45-S60. |
R827997 (Final) |
Exit Exit |
|
Lall R, Ito K, Thurston GD. Distributed lag analyses of daily hospital admissions and source-apportioned fine particle air pollution. Environmental Health Perspectives 2011;119(4):455-460. |
R827997 (Final) R827351 (Final) |
|
Supplemental Keywords:
ambient air, PM2.5, PM10, chemical speciation, health effects, epidemiology, source apportionment, human health, modeling,, RFA, Scientific Discipline, Health, Air, Geographic Area, Waste, particulate matter, Environmental Chemistry, Health Risk Assessment, air toxics, State, Epidemiology, Susceptibility/Sensitive Population/Genetic Susceptibility, chemical mixtures, Risk Assessments, Atmospheric Sciences, Ecology and Ecosystems, tropospheric ozone, genetic susceptability, ambient air quality, co-factors, copollutant exposures, PM10, fine particles, human health effects, PM 2.5, toxicology, Pennsylvania, stratospheric ozone, air pollutants, ambient air, exposure, air pollution, Washington (WA), chronic health effects, particulate exposure, gaseous pollutants, Illinois (IL), Texas (TX), fine particle sources, human exposure, mortality studies, copollutanats, epidemiological studies, hospital admissions, elderly, PM, California (CA), mortality, Michigan (MI), New York (NY), PA, environmental hazard exposuresProgress 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.