Dominici et al1 have done an excellent job of laying out the rationale for moving toward a multipollutant scientific approach to address the health consequences of air pollution. They have, as well, provided some innovative statistical ideas in the hopes of accomplishing that. Few can challenge the logic of trying to address the range of pollutants which everyone is exposed, rather than tackling air pollutants one at a time. Given its logic, one might reasonably ask why has this not already happened? Many factors have contributed to the status quo, and it is useful to understand them as we take the first steps toward the next generation of multipollutant science and decision-making. We describe some of these factors below, and some concrete steps we and others are taking.
The single-pollutant approach has been in place since the earliest versions of the U.S. Clean Air Act called for the establishment of National Ambient Air Quality Standards for so-called “criteria pollutants.”2 Although some clean-air actions have addressed multiple pollutants (eg, standards for motor vehicle emissions, which have addressed the emissions of several pollutants at the same time), the overriding impetus in the setting of these standards has been understanding the “independent” effects of individual pollutants so as to determine ambient levels that “protect public health with an adequate margin of safety” (CAA; 42 USC§7401). Among the reasons for this:
* Most of the interested parties advocating for or challenging standard-setting have, for their own reasons, pursued individual pollutants.3 Industries have sought to identify the one pollutant or a limited number of pollutants that they could target for control. Environmentalists have often found it more effective to rally advocates around individual “toxic” pollutants than the more amorphous “air pollution mixture.” Regulators have found it difficult to come up with reliable methods for measuring individual pollutants, let alone a complex mixture of them.
* Science, meanwhile, has been engaged with regulators in answering a form of the age-old question “which came first, the chicken or the egg?” Early on, scientists relied on limited data to conduct health studies that more often addressed mixtures than individual pollutants (eg, total suspended particles). Regulators then set pollutant-specific standards and established monitoring networks for each pollutant. Scientists in turn used the new monitoring data to conduct more pollutant-specific research, which led to even more focused single-pollutant standards (eg, from total suspended particles to PM10 to PM2.54). It has rarely been possible for science to initiate true multipollutant research that could find its way directly into multipollutant rule-making, and so the incentive has been to continue to focus on the pollutant of most interest at the moment (eg, carbon monoxide and ozone from 1970 to 1990; particulate matter from the early 1990s to today).
* Finally, as discussed in detail by Dominici et al,1 research strategies to understand the effects of multiple exposures are very, very difficult.
Given these challenges, how can we make concrete progress toward the next generation of multipollutant approaches?
First, although the kind of single indicator for air quality that Dominici et al suggest for setting ambient standards is likely farther off, regulators can take advantage of near-term opportunities to place consideration of multiple pollutants on a single timeline for all who must address them. Between December 2009 and April 2011—a period of 16 months—US Environmental Protection Administration (EPA) will revisit and finalize the National Ambient Air Quality Standards for the 4 most wide-reaching ambient pollutants—nitrogen dioxide (NO2), sulfur dioxide (SO2), ground-level ozone (O3), and fine particles (PM2.5). Under normal procedures, each of these actions would set off separate timetables at the EPA and in the states to assess current attainment of the standards and to take action to reduce emissions. However, these ambient pollutants share many of the same sources, and a move by the EPA to fully align the timetables for implementation could both facilitate multipollutant implementation strategies and send clear signals to the scientific community about the need for multipollutant approaches.
Second, we in the scientific community should take steps toward designing and implementing studies of multiple pollutants. Dominici and her colleagues1 describe a number of opportunities for such research; the Health Effects Institute (HEI) has been working with scientists at a number of institutions to implement just such studies. In one effort—that has some parallels to the comments made by Dominici et al—the Health Effects Institute has begun a program seeking the development of innovative statistical methods for characterizing the health effects of “real world” air pollutant mixtures. The program includes methods for investigating the joint effects of air pollution constituents (ie, exploring how the effects of a mixture as a whole differ from individual or combined effects of its components), as well as methods proposed to quantify exposure to multiple air pollutants, along with the impact of exposure measurement error on observed correlations among pollutants.
In another effort, the Health Effects Institute is funding a program aimed at developing improved assessments of exposure to multiple constituents of the pollutant mix. This program exploits the information generated in a prior study, the Relationships of Indoor, Outdoor, and Personal Air study, jointly funded with the Mickey Leland National Urban Air Toxics Research Center.5,6 The study collected information on indoor, outdoor, and personal concentrations of a large number of pollutants—volatile organic matter, carbonyls, and PM2.5—for approximately 100 persons in 3 urban areas of the United States. With HEI support, these data have been organized into a database (http://riopa.aer.com) available to anyone in the scientific community—a rich resource for testing new approaches to multipollutant analyses. HEI is now funding studies to examine exposure to the mixture of air toxics and PM2.5 in this dataset. The approaches developed in these studies are likely to provide assessments of exposure and approaches to modeling multiple-pollutant concentrations for future large-scale epidemiologic studies.
To gain another—and more comprehensive—perspective on the mixtures issue, in January 2007, the Health Effects Institute launched the HEI National Particle Component Toxicity Initiative to answer key questions about the relative toxicity of components of particulate matter, a set of issues also raised by Dominici et al.1 Two teams of investigators are conducting multifaceted and coordinated exposure assessment, toxicology and epidemiology studies. HEI had, along with the US EPA Particulate Matter Centers Research program, also earlier supported short-term epidemiology studies of effects and their relationship to multiple pollutants.
The National Particle Component Toxicity initiative is making a systematic effort to investigate cardiovascular effects of PM components and pollutant gases in long-term cohort studies. In the Women's Health Initiative Observational Study and the Multi-Ethnic Study of Atherosclerosis and Air Pollution, associations of individual PM components and gases with the development of atherosclerosis and incidence of cardiac events are being studied. The National Particle Component Toxicity initiative is also examining cause-specific mortality associated with exposure to PM components in the American Cancer Society cohort for which PM components data are available in over 100 cities. The Initiative also builds on earlier work in Atlanta (eg, Peel et al7), by including time-series analyses of hospital admissions and mortality in the elderly in multiple cities across the US where PM components and sources differ. Complementary toxicology studies include 6-month exposures to concentrated PM2.5 at several sites with widely varying PM mixtures to examine potential cardiovascular effects in the Apo E knockout mouse (a model of atherosclerosis). At the same time, the initiative includes studies of cardiovascular effects in the same mouse model of extended exposures to laboratory-generated atmospheres—motor vehicle emissions, secondary inorganic aerosol, and road dust. These comprehensive studies are slated for completion in 2012.
Finally, HEI is now in the midst of crafting its strategic plans for 2010–2015, in which next-generation multipollutant research will be a central theme. In one project, HEI will soon announce a request for applications to study the effects on the human cardiovascular system of exposure to near-ambient levels of ozone and other pollutants. Controlled human exposures to ozone alone in people ages 55–70 years, will be compared with environments where ozone concentrations are similar to those studied in the laboratory, but with other pollutants—particularly PM—are also present. This study will attempt to tease out the effect of ozone alone and in the presence of other ambient air pollution components.
In conclusion, we applaud Dominici and colleagues1 for bringing this important set of issues and challenges to the fore, and for helping to lay the groundwork to address them. Given the history of single-pollutant approaches, however, it seems clear that neither the policy changes nor the scientific changes necessary to address them will be easy or quick. To move forward we need:
* True multipollutant policy approaches, such as those that the EPA is beginning to take, in order to send a clear signal to the larger policy and scientific communities about what form multipollutant regulation may take, and what types of science will be most valuable in informing those regulatory decisions.
* As Dominici et al state, multidisciplinary scientific approaches—such as those they propose and those being undertaken by the Health Effects Institute—will be essential. However, given the magnitude of the challenges, the work underway or contemplated may be just the proverbial “tip of the iceberg.” Cross-fertilization from pharmacology, broader statistical sciences, and other fields—among them the emerging field of systems biology—are likely to contribute as well.
Taking these actions will not be easy, but given that populations are never exposed to one pollutant at a time, and that efforts to control pollutants at lower and lower levels have become increasingly contentious as well as difficult, getting the science “right” on reducing exposure to the mixture seems to be the only reasonable way forward.
1. Dominici F, Peng RD, Barr CD, Bell ML. Protecting human health from air pollution: shifting from a single-pollutant to a multi-pollutant approach [commentary]. Epidemiology
3. Greenbaum DS. Epidemiology at the edge. Epidemiology
4. Greenbaum DS, Bachmann JD, Krewski D, Samet JM, White R, Wyzga RE. Particulate air pollution standards and morbidity and mortality: a case study. Am J Epidemiol
. 2001b;154(suppl 12):S78–S90.
5. Turpin BJ, Weisel CP, Morandi M, et al. Relationships of indoor, outdoor, and personal air (RIOPA): part II. Analyses of concentrations of particulate matter species. Res Rep Health Eff Inst
. 2007;130(pt 2):1–77.
6. Weisel CP, Zhang J, Turpin BJ, et al. Relationships of indoor, outdoor, and personal air (RIOPA). Part I. Collection methods and descriptive analyses. Res Rep Health Eff Inst
. 2005;130(pt 1):1–107.
7. Peel JL, Tolbert PE, Klein M, et al. Ambient air pollution and respiratory emergency department visits. Epidemiology