3

Review of Research Progress and Status

INTRODUCTION

In this chapter, the committee reviews the progress made in implementing the particulate-matter (PM) research portfolio over the period from 1998 (the year in which the portfolio was first recommended by the committee) to the middle of 2000. Because that period represents the initial stage of the PM research program, the committee's assessment necessarily focused more on continuing and planned research projects than on published results.

For each of the 10 topics in the research portfolio, the committee first characterizes the status of relevant research and progress, including the approximate numbers of studies in progress on various subtopics (the committee did not attempt to list all relevant research projects but did attempt to capture the major studies across the spectrum of the research in progress), then considers the adequacy of the current research in addressing specific needs as identified in its first two reports, and finally applies the first three evaluation criteria discussed in Chapter 2: scientific value, decisionmaking value, and feasibility and timing. The remaining three criteria—largely cross-cutting—are considered in more general terms in Chapter 4. The committee's next report, due near the end 2002, will consider research in relation to these criteria in more detail.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • 3 Review of Research Progress and Status INTRODUCTION In this chapter, the committee reviews the progress made in implementing the particulate-matter (PM) research portfolio over the period from 1998 (the year in which the portfolio was first recommended by the committee) to the middle of 2000. Because that period represents the initial stage of the PM research program, the committee's assessment necessarily focused more on continuing and planned research projects than on published results. For each of the 10 topics in the research portfolio, the committee first characterizes the status of relevant research and progress, including the approximate numbers of studies in progress on various subtopics (the committee did not attempt to list all relevant research projects but did attempt to capture the major studies across the spectrum of the research in progress), then considers the adequacy of the current research in addressing specific needs as identified in its first two reports, and finally applies the first three evaluation criteria discussed in Chapter 2: scientific value, decisionmaking value, and feasibility and timing. The remaining three criteria—largely cross-cutting—are considered in more general terms in Chapter 4. The committee's next report, due near the end 2002, will consider research in relation to these criteria in more detail.

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • RESEARCH TOPIC 1. OUTDOOR MEASURES VERSUS ACTUAL HUMAN EXPOSURES What are the quantitative relationships between concentrations of particulate matter and gaseous copollutants measured at stationary outdoor air-monitoring sites and the contributions of these concentrations to actual personal exposures, especially for subpopulations and individuals? In its first report (NRC 1998), the committee recommended that information be obtained on relationships between total personal exposures and outdoor concentrations of PM. Specifically, the committee recommended longitudinal panel studies, in which groups of 10-40 persons would be studied at successive times to examine the relationship between their exposures to PM and the corresponding outdoor concentrations. The studies were intended to focus not only on the general population, but also on subpopulations that could be susceptible1 to the effects of PM exposures, such as the elderly, children, and persons with respiratory or cardiovascular disease. It was recommended that some of the exposure studies include measurements of PM with an aerodynamic diameter of 2.5 µm or less (PM2.5), PM with an aerodynamic diameter of 10 µm or less (PM10), and gaseous copollutants. It was expected that the investigations would quantify the contribution of outdoor sources to personal and indoor exposures. The design and execution of studies were to take about 3 years, and the suggestion was made to conduct the studies at various geographical locations in different seasons. Research Progress and Status Substantial research is in progress, and some studies, started before the committee's first report, have been completed. Results of 1   The committee is aware that there are several definitions of “susceptibility” (Parkin and Balbus 2000). in using the term in this report, the committee refers to an increased risk at a particular exposure that is greater for susceptible people than for healthy people.

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • recent panel studies of personal exposure conducted in Wageningen, Netherlands (Janssen et al. 1999), Boston, MA (Rojas-Bracho et al. 2000), Baltimore, MD (Sarnat 2000; Williams et al. 2000), and other places suggest that 12-15 measurements per person are sufficient to examine relationships between personal exposures and outdoor PM concentrations. These longitudinal panel studies have increased the understanding of the relationships between personal exposures and outdoor concentrations more than did earlier cross-sectional exposure studies. Several additional longitudinal panel studies are going on in other U.S. cities, including New York, NY; Atlanta, GA; Los Angeles, CA; Research Triangle Park, NC; and Seattle, WA. A number of research and funding organizations—including academic institutions, the U.S. Environmental Protection Agency (EPA), the Health Effects Institute (HEI), the Electric Power Research Institute (EPRI), and the American Petroleum Institute (API)—already have been engaged in this effort. Collectively, the studies should provide an understanding of the relationships between personal exposures and outdoor pollutant concentrations in a large number of geographic areas in the United States. Several insights have been gained from the results of completed studies (Janssen et al. 1999, 2000; Ebelt et al. 2000; Rojas-Bracho et al. 2000; Sarnat et al. 2000; Williams et al. 2000). These studies have observed significant differences among study participants in the relationship between personal exposures and outdoor concentrations. When such relationships were analyzed for each person, substantial variability was found. Because outdoor concentrations exhibited little spatial variability, the heterogeneity was attributed to differences in indoor concentrations. Indeed, indoor concentrations were found to be an excellent predictor of personal exposures for most study participants, independently of city (Baltimore or Boston), season (winter or summer), and panel (elderly; chronic obstructive pulmonary disease, or COPD; or children). The finding that indoor concentration is an excellent predictor of personal exposure is not surprising, in that people spend more than 80% of their time indoors (EPA 1996a). Apart from exposures to tobacco smoke and emissions from cooking, which produce long-term increases in PM exposures of around 30 µg/m3 (Spengler et al. 1981) and 15-20 µg/m3 (Ozkaynak 1996), respectively,

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • home activities that were expected to produce particles, such as vacuum-cleaning and dusting (EPA 1996; Ozkaynak 1996), were found to explain very little of the total variability in personal exposures (Rojas-Bracho et al. 2000). In general, indoor sources tend to operate intermittently and, when measured by continuous monitors, can produce indoor concentrations as high as several hundred micrograms per cubic meter (Abt et al. 2000). The impact of these indoor (or other microenvironmental) peak concentrations can be captured only by real-time or semicontinuous personal monitors (Howard-Reed et al. 2000). However, when such short-term increases in concentration are averaged, their contributions to the average 24-hr indoor concentrations or personal exposures are estimated to be small. Analyses of data from the study of elderly people in Baltimore (Sarnat et al. 2000) and the study of COPD patients in Boston (Rojas-Bracho et al. 2000) have demonstrated that ventilation (rate of exchange of indoor with outdoor air) is the measure that most strongly influences the relationship of personal-exposure to outdoor concentration. Personal exposure data were classified into three groups based on reported home ventilation status, a surrogate for the rate of exchange of indoor with outdoor air. Homes were classified as “well,” “moderately,” or “poorly” ventilated, as defined by the distribution of the fraction of time that windows were open while a person was in an indoor environment. When the PM datasets were stratified into these ventilation groups and analyzed cross-sectionally, strong relationships between personal exposures and outdoor concentrations were observed for well-ventilated homes and, to a lesser extent, for moderately ventilated homes. However, a low correlation coefficient was found for the poorly ventilated homes. Those findings suggest that for homes with no smokers and little cooking activity most of the variability in indoor concentrations, as well as in personal exposures of occupants, is due to the varied impact of outdoor sources on the indoor environment. That effect is underscored by the influence of air-exchange rates on the relationship between indoor and outdoor concentrations when no activities are occurring in the homes. For instance, for well-ventilated homes, indoor-to-outdoor particle ratios are close to 1.0, whereas for homes with low rates of exchange and

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • no activities, indoor-to-outdoor ratios can be substantially lower (about 0.4-0.6) (Abt et al. 2000; Long et al. 2000). Home ventilation rates are expected to vary with season, geographical location, and home characteristics; that implies that the relationship of human exposures to outdoor PM concentrations will also vary with these factors. Therefore, PM risk relationships estimated from epidemiological studies might differ by city, season, and overall home characteristics. However, the additional influence of personal activity patterns on the overall relationship between human exposure and outdoor PM concentrations is also relevant to interpretation of the results of observational studies. The pattern of reported findings is still based on a small number of studies, and replication of the results will be needed from current or recently completed studies in other cities before firm conclusions can be drawn. Adequacy of Current Research in Addressing Research Needs Considerable effort is going into examining the relationship between ambient particle concentrations and personal exposures. Several longitudinal panel studies are being conducted in various geographic locations, including New York, NY; Atlanta, GA; Los Angeles and Fresno, CA; and Seattle, WA (see Table 3.1). Collectively, these studies are assessing exposures of healthy subjects and susceptible subpopulations (such as those with COPD; myocardial infarction, or MI; or asthma) to PM and some gaseous copollutants (such as ozone, sulfur dioxide, and nitrogen dioxide). The studies are expected to greatly expand the database on personal exposures, indoor and outdoor concentrations, human activities, and home characteristics. They are also expected to improve understanding of factors that influence the relationship between ambient concentrations and personal exposures. Therefore, as new information from the panel studies accumulates, it appears that, in spite of the time needed to initiate them, many of the elements of research topic 1 are being addressed. Most of the studies have not been completed; their findings are expected to appear in the peer-reviewed literature in the next several years.

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • TABLE 3.1 Current Studies Relevant to ResearchTopic 1 Funding Agencya Research Groupb Location Proposed Cohortc EPA New York University New York, NY Anaheim, CA Seattle, WA 16 asthmatic, 16 COPD 16 asthmatic, 16 COPD 16 asthmatic, 16 COPD EPRI and API Harvard University Nashville, TN Boston, MA 10 COPD 18 COPD EPA University of Washington Seattle, WAd 48 COPD and elderly 48 COPD and elderly 48 MI 48 MI 24 COPD and elderly, 24 MI 25 COPD and elderly, 24 MI EPRI and API Harvard University Baltimore, MDd 15 elderly HEI and Mickey Leland EOHSI, UMDNJ, and Rutgers University Elizabeth, NJ Houston, TX Los Angeles, CA 50 healthy adults 50 healthy adults 50 healthy adults HEI Harvard University Baltimore, MDd Baltimore, MDd Boston, MAd 15 children 15 COPD, 15 children 15 children, 15 elderly EPA and EPRI,e Harvard University Boston, MAd 15 MI, 15 MI spouses 15 healthy adults CARBf Emory University Rutgers University and UMDNJ Atlanta, GAd Los Angeles, CAd 15 COPD 15 COPD EPA EPA and RTI Baltimore, MD Fresno, CA Fresno, CA 15 elderly 5 elderly 60 elderly EPAg EPA and RTI Research Triangle Park, NC 15 MI, 15 low-SES healthy adults CARB University of California, Berkeley Fresno, CAh 25 asthmatic children CARB and EPA (planned) Harvard University, Rutgers University, and IES Los Angeles, CAd 16 healthy adults (nonsmoking)

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • HEI Wagenigen University Helsinki, Finland and Amsterdam, Netherlands 50 healthy adults DOE, OCDO, EPRI, API, and EPA Harvard University and CONSOL Energy Inc. Steubenville, OH d 25 elderly 15 children EPRI and EPA Harvard University and Washington University St. Louis, MO Not determined a   API = American Petroleum Institute CARB = California Air Resources Board DOE = U.S. Department of Energy EPA = U.S. Environmental Protection Agency EPRI = Electric Power Research Institute HEI = Health Effects Institute Mickey Leland = Mickey Leland National Urban Air Toxics Research Center OCDO = Ohio Coal Development Office b   EOHSI = Environmental and Occupational Health Sciences Institute IES = Integrated Environmental Sciences UMDNJ = University of Medicine and Dentistry-New Jersey RTI = Research Triangle Institute c   COPD = chronic obstructive pulmonary disease MI = myocardial infarction SES = socioeconomic status d   Copollutants (NO2, O3) also measured e   Funding for Atlanta panel study to be provided by EPRI f   Cofunding for Los Angeles panel provided by CARB g   Information is preliminary h   Copollutants (CO, O3, SO2, NO2) also measured Many of the recently completed and current studies examine the relationship between ambient concentrations of gaseous pollutants and personal exposures. Understanding that relationship will provide profiles of multipollutant exposures that can inform understanding of research topic 7 (combined effects of PM and gaseous pollutants). In

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • addition, understanding of differences between personal exposure and ambient concentrations for a suite of gaseous pollutants and PM will provide input into analyses of measurement error in a multi-pollutant context (see research topic 10, analysis and measurement). Application of Evaluation Criteria Scientific Value The current panel exposure studies are straightforward and have expanded on findings from previous investigations. They have used well-established research tools for conducting personal and micro-environmental measurements. They have also relied on field protocols developed as part of previous exposure studies, (such as the Particle Total Exposure Assessment Methodology (PTEAM) study (Pellizzari et al. 1993). The studies are generally designed to assess the range of exposures including those that occur in the home, in the workplace, and while traveling. To a large extent, the scientific value of these investigations will be judged by the appropriateness of their design. It appears that the study designs, (such as repeated measurements of a small number of people) can adequately address the scientific questions in research topic 1. Completed studies have indicated key factors that influence outdoor-personal relationships. Preliminary results suggest that for homes with no smokers and little cooking activity, home ventilation rate (or air-exchange rate) is the most important modifier of personal exposure. To a great extent, ventilation rate controls the impact of both outdoor and indoor sources on the indoor environment, where people spend most of their time. If correct, this observation implies that such entities as home characteristics, season, and location could be more important determinants of personal exposure than activities and type of susceptible subpopulation studied. The panel studies will also produce a large set of data on human activities and home characteristics. These data will substantially enrich the existing information and will be available to other researchers

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • involved in human-exposure assessment investigations (such as EPA 's National Human Exposure Assessment Survey). Decisionmaking Value Exposure assessment is of paramount importance for understanding the effects of ambient particles and for developing cost-effective exposure-control strategies. The current studies should allow the scientific community and decisionmakers to understand the factors that affect the relationship between personal exposure and outdoor concentrations. That will be accomplished through the continued development of personal-exposure monitoring tools that allow a better understanding of the sources of exposure, physical and chemical properties of PM, and sampling durations that could be relevant to the subpopulations being studied. Although the panel studies are based on small numbers of participants (10-50 per panel), they are addressing factors that influence relationships between outdoor air and personal exposures. This is the first step in attempts to develop a comprehensive exposure model, which is a key research tool in the source-exposure-dose-response paradigm. Feasibility and Timing Sampling and analytical procedures, time-activity questionnaires, and other related methods necessary for conducting the panel studies have been adequately tested. They have been implemented successfully in various geographical locations by various research groups (such as Janssen et al. 1999, 2000; Ebelt et al. 2000; Rojas-Bracho et al. 2000; Sarnat et al. 2000; Williams et al. 2000). Therefore, it is expected that the current longitudinal panel studies will be completed without great difficulty. Although there was some delay in initiating some of the studies, abundant personal and microenvironmental measurements have been collected. Reporting of results from research related to this topic began during the summer of 2000, and the re-

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • maining studies should be reported.within about 2 years, a year later than originally planned. RESEARCH TOPIC 2. EXPOSURES OF SUSCEPTIBLE SUBPOPULATIONS TO TOXIC PARTICULATE-MATTER COMPONENTS What are the exposures to biologically important constituents and specific characteristics of particulate matter that cause responses in potentially susceptible subpopulations and the general population? The committee recommended that after obtaining and interpreting results of studies from research topic 1 human exposure-assessment studies examine exposures to specific chemical constituents of PM considered relevant to health effects. To make research topic 2 investigations more practicable, it will be necessary to characterize susceptible subpopulations more fully, identify toxicologically important chemical constituents or particle-size fractions, develop and field-test exposure-measurement techniques for relevant properties of PM, and design comprehensive studies to determine population exposures. Methods of measuring personal exposures to particles of various physical properties (such as particle number and size) or chemical properties (such as sulfate, nitrate, carbon, and other elements) are available and are being field-tested. Methods of measuring personal exposures to some gaseous copollutants—such as ozone, nitrogen dioxide, and sulfur dioxide—are also used. As interest in personal-exposure measurements increases, new sampling and analytical techniques will probably emerge. The results of the longitudinal panel studies discussed under research topic 1 should facilitate the design of cost-effective protocols for future exposure studies that focus on PM components considered in determining toxicity. These studies will be based on toxicity and epidemiological studies that are successful in identifying particle properties of interest over the next few years; because they will prob-

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • ably not get under way for several years, the committee is planning to evaluate their progress in its next report. RESEARCH TOPIC 3. CHARACTERIZATION OF EMISSION SOURCES What are the size distribution, chemical composition, and mass-emission rates of particulate matter emitted from the collection of primary-particle sources in the United States, and what are the emissions of reactive gases that lead to secondary particle formation through atmospheric chemical reactions? In its second report, the committee created a separate set of research recommendations that address measurement of the size distribution and chemical composition of PM emissions from sources. Characterization of the emission rates of reactive gases that can form particles on reaction in the atmosphere was also emphasized, including the need to maintain emission data on sulfur oxides, nitrogen oxides, ammonia, and volatile organic compounds (VOCs) (specifically those components that lead to particle formation). The committee noted that traditional emission inventories have focused on representing PM mass emissions summed over all particles smaller than a given size, without detailed accounting of the particle-size distribution or chemical composition. Health-effects research recommended by the committee emphasized identification of the specific chemical components or size characteristics of the particles that are most directly related to the biological mechanisms that lead to the health effects of airborne particles. Detailed information on the size and composition of particle emissions from sources is important for this process of hazard identification and effective regulation. In the near term, toxicologists and epidemiologists need to know the size and composition of particles emitted from key emission sources to form hypotheses about the importance of particle characteristics and to give priority to their evaluation in laboratory- and field-based health-effects studies. In the longer term, detailed information on

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • able when they are based on high doses, as is the case in most of the current studies. A major gap is a lack of testing of the validity of conclusions for specific mechanisms by using relevant low doses; this is due in large part to the lack of a demonstrated causal relationship between relatively low PM exposures and adverse effects in controlled in vivo studies. Thus, in vitro studies have their greatest scientific value when they are designed on the basis of results of controlled whole-animal or clinical studies, involve relatively realistic exposures, and test specific mechanistic hypotheses. Decisionmaking Value Mechanistic information at the cellular and molecular levels obtained from well-designed in vitro studies can contribute to the weight of evidence regarding a causal relationship between PM exposure and health effects. That will reduce uncertainties related to the plausibility of observed adverse PM effects. Knowledge gained about mechanisms of PM toxicity will contribute greatly to the scientific justification of the PM standards. Feasibility and Timing In vitro studies clearly are feasible in many laboratories. It is important for special attention to be directed toward the use of relevant doses. Moreover, the development of appropriate new methods for in vitro studies should be encouraged, including airborne-particle exposures of cell cultures, use of cells from compromised lungs, and use of genetically modified cells. Because the developmental phase of these models is potentially long, useful results might not become available very soon. Research Topic 9c. Clinical Models What are the appropriate clinical models to use in studies of particulate matter toxicity?

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • Clinical studies are controlled exposures of humans. In the case of PM, such studies are designed to use laboratory-generated surrogate particles or concentrated ambient-air particles. The use of human subjects avoids the need to extrapolate results from other species. Both normal and susceptible subpopulations can be studied, and physiologic, cellular, immunologic, electrocardiographic and vascular end points, as well as symptoms, can be assessed. Elucidation of responses in humans is key to understanding the importance of ambient pollution and determining the nature of adverse health effects of PM exposure. Research Progress and Status Review of the HEI database and proceedings of the PM 2000 meeting identified about 10 active human-exposure studies. All are using particles of concern, which include CAP, ultrafine carbon, ultrafine acidic sulfates, diluted diesel exhaust, and smoke from burning of vegetable matter. Studies are under way in healthy volunteers, asthmatics, and atopic people. Studies in people who have COPD or cardiac disease are planned. The clinical studies focus on evaluation of pulmonary and systemic responses, such as pulmonary inflammation and injury to epithelial cells; cardiac rhythm, rate, and variability; initiation of the coagulation cascade; and symptoms. Few laboratories are equipped to perform clinical studies of PM. However, the similarities in their protocols enhance the likelihood of obtaining useful data. For example, studies with CAP and ultrafine particles have incorporated prolonged electrocardiographic monitoring after exposure. All studies include physiologic assessments of lung function, and indicators of airway inflammation in nasal or bronchoalveolar lavage fluid, induced sputum, or exhaled air (such as nitric oxide). In addition, coagulation indexes in blood are examined in some of the studies. In selected cases, efforts have been made to centralize analytical studies in a core laboratory for standardization of techniques. There are a number of difficulties in establishing clinical models to study PM. Although the particle concentrators allow exposure to

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • relevant atmospheres, the mixtures vary from day to day and, typically, minimal chemical analyses of the particles are performed. If responses to CAP are variable, it is not possible to determine whether the variability resulted from differences in human susceptibility or in particle chemistry. In contrast, studies with surrogate particles result in reproducible exposures but mimic only selected aspects of ambient particulate pollution. Furthermore, the epidemiological data suggest that the most severely ill are at risk of pollutant effects; these subgroups cannot be used in controlled clinical studies. Because clinical studies by design are limited to short-term exposures, they will rarely be able to contribute to an understanding of development of chronic disease secondary to exposure to particles. The particle-exposure systems used in clinical studies include environmental chambers, facemasks, and mouthpieces. Each design offers specific advantages, but the mouthpiece studies with ultrafine particles have incorporated measurements of total particle deposition. One clinical study will investigate the interaction of particles with ozone, another plans to incorporate metals into the particles, and virtually all include some level of exercise to enhance minute ventilation, thus increasing the inhaled dose of pollutants. Adequacy of Current Research in Addressing Research Needs The current and planned clinical studies are designed to investigate CAP and several specific components of PM (such as size, acids, metals, and diesel exhaust) with a number of pulmonary and systemic end points. Studies are under way in susceptible subpopulations and are planned in other subgroups with pre-existing disease. Despite the limited facilities available for clinical research, the array of studies under way should provide valuable information on PM toxicity. Application of Evaluation Criteria Scientific Value Clinical studies present an opportunity to examine responses to

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • PM in both healthy and susceptible subpopulations. Carefully designed controlled exposures provide information on symptomatic, physiologic, and cellular responses in both healthy and at-risk groups. They also provide important insights into mechanisms of action of PM. Such studies can provide needed information on PM deposition and retention in healthy and susceptible subpopulations (see research topic 6). Decisionmaking Value Clinical studies often provide important information for regulatory decisions. Assessing acute responses in groups that have chronic diseases will establish important insights into plausible mechanistic pathways. In addition, they provide crucial data on relative differences in responsiveness between healthy and potentially at-risk populations. Feasibility and Timing Studies are under way in several laboratories. They should provide highly relevant information for the next review of PM for regulatory decisions. RESEARCH TOPIC 10. ANALYSIS AND MEASUREMENT To what extent does the choice of statistical methods in the analysis of data from epidemiological studies influence estimates of health risks from exposures to particulate matter? Can existing methods be improved? What is the effect of measurement error and misclassification on estimates of the association between air pollution and health? The first report of this committee (NRC 1998) outlined several methodological issues that needed further study. These included the

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • choice of statistical methods for analyzing data obtained from other studies, especially epidemiologic studies. Because more than one method can be used to analyze data, it will be important to understand the extent to which alternative approaches can influence analytical results. In addition, new study designs will require new approaches to analyze the data. These include development of analytical methods to examine several constituents and fractions of PM in an effort to understand their associations with health end points and design of models and approaches to incorporate new biological insights. Specific attention was given to measurement error, an issue inherent in most epidemiological studies that use ambient-air data to characterize subjects' exposure. The committee's second report (NRC 1999) reiterated those needs and noted the existence of relevant research and papers nearing completion. Review of scientific literature, meeting abstracts, and the HEI database identified extensive progress on several methodological subjects. The review was intended to evaluate the extent to which the research needs previously identified by the committee are being addressed and to stimulate further targeted research. General Methodological Issues Model Development and Evaluation Over the last several years, there has been considerable development of time-series data-analysis methods, which have provided much of the evidence on the association between PM exposures and health effects. The methods assess the variation in day-to-day mortality or morbidity counts with variation in PM concentrations on the same or previous day. Although systematic and comprehensive comparisons of alternative methods have not been reported, limited comparisons have suggested that results are relatively robust to the statistical approach used. However, the choices of input variables and data have been shown occasionally to influence results (Lipfert et al. 2000). That is particularly true with respect to the choice of pollution variables in

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • the statistical models. The presence of other variables in the models can influence the association between health measures and particulate air pollution. The application of the time-series studies has been facilitated by recent advances in hardware and software and by the development of statistical approaches that can appropriately account for the data structure of the daily time series. Time-series analyses were initially conducted on a single location that had been selected primarily on the basis of data availability, rather than representing selection from a defined sampling frame. Meta-analysis was then used to summarize the data and to gain a more precise estimate of the effect of PM on mortality or morbidity. Recently, studies of more-formal, multicity designs have been conducted. These approaches have a priori plans for selecting locations and have standardized statistical methods across locations. The European Air Pollution and Health: A European Approach (APHEA) project (Katsouyanni et al. 1995) is a pioneering effort that initially analyzed routinely collected data from 15 European cities in 10 countries with a common statistical protocol, examining mortality and emergency hospitalizations in some cities. In the United States, the HEI has funded the National Morbidity, Mortality and Air Pollution Study (NMMAPS) (Samet et al. 2000, 2001). The NMMAPS includes analyses of mortality and morbidity separately; a joint analysis of morbidity and mortality is planned. For the mortality analysis, the NMMAPS investigators used a sampling frame defined by U.S. counties. The 90 largest urban areas (by population) were selected, and the daily mortality data for 1987-1994 were analyzed to assess associations with PM and other pollutants. The methods used in the APHEA project and the NMMAPS show the potential power of multicity approaches. The potential selection bias of only a single or a few locations is avoided. Combining information across locations, increases power and heterogeneity. In addition, health effects can be compared between regions that have similar air-pollution levels. Other research efforts involving model development are the exploration of distributed-lag models (Schwartz 2000a; Zanobetti et al. 2000), efforts to understand the dose-response relationship between PM exposure and health effects (Schwartz 2000b; Smith et al. 2000;

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • Schwartz and Zanobetti 2000), and examination of alternative ways of analyzing the relationship between air-quality data and health end points (Beer and Ricci 1999; Sunyer at al. 2000; Tu and Piegorsch 2000; Zhang et al. 2000). Other research efforts have also aimed at combining results from several studies, including those by Stroup et al. (2000) and Sutton et al. (2000). Measurement Error The difference between actual exposures and measured ambient-air concentrations is termed measurement error. Measurement error can occur when measures of ambient air pollution are used as an index of personal exposure. For PM, the three sources of measurement error are instrument error (the accuracy and precision of the monitoring instrument), error resulting from the nonrepresentativeness of a monitoring site (reflected by the spatial variability of the pollutant measured), and differences between the average personal exposure to a pollutant and the monitored concentration (influenced by microenvironmental exposures). With regard to assessing the impact of outdoor exposures, the most important source of measurement error is related to the representativeness of the placement of monitors. In acute studies, other sources of error will not vary substantially from day to day. But in chronic studies, the most important errors are those associated with microenvironmental exposures. The presence of indoor sources of PM and the influence of home characteristics on penetration of outdoor particles into the indoor environment can be a source of substantial exposure error. The influence of home characteristics is important because it varies with geographical location, climate, socioeconomic factors, and season. Because those factors could introduce systematic errors, they must be considered in the analysis and interpretation of results of chronic epidemiological studies. They are often taken into account by using not direct measures of exposure, but surrogate measures that would influence the exposures, such as smoking in the household, the presence of gas stoves, and air conditioning. Measurement error is of particular concern in studies intended to

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • isolate the effects of particles from those of gases or to distinguish the effects of individual particle species or size fractions from each other. When several population variables are included in the same analyses and the different variables have different magnitudes and types of measurement error, the issue of estimating the associations between health responses and specific variables is even more complicated. A well-measured but benign substance might serve as the best empirical predictor of community health effects, rather than a poorly measured but toxic substance that is similarly distributed in the atmosphere. The problem is that most pollutants tend to be similarly distributed, so collocated time series of pollutant measurements tend to covary because all pollutants are modulated by synoptic meteorological conditions. Long-term averages of pollutant concentrations tend to covary across cities because the rates of many categories of emissions tend to increase roughly with population. Various methods are available to adjust statistical analyses for the effects of differential measurement error (Fuller 1987; Carroll et al. 1995). Several statistical issues must be considered in addressing measurement error. A full discussion of these issues is found in Fuller (1987) and Carroll et al. (1995). The most important is the type of model in which the measurement error is imbedded. Generally, in linear models, measurement error can be understood if it is assumed that errors are independent of each other and of other variables in the model and that they follow the same statistical distribution. However, it is common for measurement-error distribution and properties not to be readily apparent, for example in ambient-air quality data, because “true” measurements of personal exposure have not been available. Recent studies have generated data that will provide a better understanding of the properties of measurement error. Until its specific properties are understood, its consequences will be unclear. For instance, Stefanski (1997) cites examples from a linear model in which the regression coefficient could be biased in either direction or unbiased, depending on the characteristics of the measurement error. The issues are increasingly complex as one moves to multiple-regression models (Carroll and Galindo, 1998) and then to nonlinear models. Development of a framework or method will be useful in consider-

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • ing the effects of measurement error on population-mortality relative risks (Zeger et al. 2000). The framework demonstrates that for a wide range of circumstances the impacts of measurement error will either lead to underestimates of association or have a negligible effect. Combined with some of the data now being generated, the framework promises considerable progress toward an understanding of measurement error. Harvesting Harvesting is an issue raised by time-series mortality studies. The term “harvesting” refers to the question of whether deaths from air pollution occur in people who are highly susceptible and near death (and die a few days earlier because of air pollution than they otherwise would have) or the air pollution leads to the death of people who are not otherwise near death. Many studies have identified associations between daily mortality and air-quality variables measured at the same time or a few days before deaths, but none of them has been able to address fully the issue of harvesting, although several recent analyses (Zeger et al. 1999, Schwartz 2000c) suggest that the findings of daily time-series studies do not reflect mortality displacement alone. Several analytical approaches have been proposed to address harvesting, and they need to be tried on additional data sets and refined to quantify better the degree of life-shortening associated with PM and other pollutants. Four recent papers examine this issue from different perspectives (Smith et al. 1999; Zeger at al. 1999; Murray and Nelson 2000; Schwartz 2000c). Spatial Analytical Methods An important issue in the analysis of data from studies that examine the association between city-specific mortality and long-term average pollutant concentrations, is whether observations of individual subjects are independent or correlated. Spatial correlation in

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • mortality can result from common social and physical environments among residents of the same city. Air pollution can be spatially autocorrelated as a result of broad regional patterns stemming from source and dispersion patterns. In a recent reanalysis of data from the study by Pope et al. (1995), which examined associations between mortality in 154 cities throughout the United States and fine-particle and sulfate concentrations, Krewski et al. (2000) developed and applied new methods to allow for the presence of spatial autocorrelation in the data. The methods included two-stage random-effects regression methods, which were used to account for spatial patterns in mortality data and between-and within-city specific-particle air pollution levels, and application of spatial filtering to remove regional patterns in the data. Taking spatial autocorrelation into account in this manner increased the estimate of the mortality ratios associated with exposure to PM and led to wider confidence limits than in the original analysis; it was assumed that all individuals in the study represented independent observations. The initial work on the development of analytical methods for the analysis of community-level data that exhibit clear spatial patterns warrants further investigation. Failure to take such spatial patterns into account can lead to bias in the estimates of mortality associated with long-term exposure to fine particles and to inaccurate indications of statistical significance. Adequacy of Current Research in Addressing Information Needs The recent research appears to address the research gaps and needs addressed by the committee. That is especially true for the measurement-error and harvesting issues. Because this research is new, it needs to be digested and applied to several data sets to increase our understanding. Data that are available or being collected allow further testing of the applications and methods. However, several subjects for further research are: elucidation of the statistical properties of the new spatial approaches discussed, consideration of

OCR for page 47
Research Priorities for Airborne Particulate Matter: • III • alternative ways of addressing spatial autocorrelation in the data, and application of such spatial analytical methods to additional data sets. Application of Evaluation Criteria Scientific Value The research has been well conducted with strong statistical tools. In addition, it has taken advantage of the existing literature and statistical tools while applying them to new subjects. However, the statistical tools have been applied to few data sets. The value of the research will increase as it is applied to more data sets and as approaches and results from the various studies are compared, synthesized, and reconciled. Decisionmaking Value The research can contribute substantially to decisionmaking. Understanding of potential influence of model approaches on results is key to adequate use of the research findings. Because measurement error can affect the results, insights into the influence of measurement error will assist in the interpretation of the results and ultimately increase their influence in decisionmaking. Understanding of harvesting will help to place estimates of effects on mortality in a public-health perspective. Feasibility and Timing Feasibility is not a deterrent to the research in this field. It appears that extensive results will be available within the timeframe laid out by this committee.