Unhealthy particles At the beginning of the twenty-Ærst century we still have only a partial understanding of the effects on human health of one of the most basic pollutants particulate matter. Yet recent epidemiological studies suggest the social costs of air pollution are far higher than previously thought. Air pollution has widespread impacts on the environment on quality of life and on human health. The economic costs of air pollution have been studied for many years. We now have excellent data on for example the costs of industrial pollution controls developing cleaner or substitute fuels deposition on forests and vegetation and repairing buildings from acid rain damage. It is increasingly apparent however that the costs of air pollution are measured not only in monetary terms but also in human lives.Yet only recently have we started to quantify the health impacts. Of the many pollutants of potential signiÆcance to health most attention has Box 1 Validating health data EPA's 1997 National Ambient Air Quality Standards for Particulate Matter cited two major studies of air pollution and mortality undertaken in the early 1990s. These studies have been the basis for nearly every worldwide assessment of the beneÆts of reducing particulate air pollution by the World Health Organisation EPA and others. In the light of intense pressure from Congress industry the scientiÆc community and others for public scrutiny of the supporting data the Health Effects Institute undertook an intensive reanalysis of the two studies.The reanalysis afÆrmed the validity of the original methodologies and results and provided important new insights from the original data. These included (1) The relation between air quality and mortality appears to be modiÆed by level of education in that estimated mortality effects were higher in the subgroup with less than school education. (2) The use of spatial analytic methods suggested that when the analyses controlled for correlations among cities located near one another the associations between mortality and Æne particles or sulfate remained but were diminished. (3) An association between sulfur dioxide mortality was observed and persisted when other possible confounding variables were included. Furthermore when sulfur dioxide was included in models with Æne particles or sulfate the associations between these pollutants (Æne particles and sulfate) and mortality diminished.Adapted from ref. 3. This journal is # The Royal Society of Chemistry 2000 focused on particulate matter (PM). This is both the most prevalent air pollutant and it turns out also one of the most complex in health terms. Epidemiological studies generally support an association between PM and adverse health outcomes but many questions remain about how PM may cause such effects and who is most at risk. Progressive regulation The sources of particulates are many and varied. They include natural sources such as dust storms and forest Æres and man-made sources such as power plants motor vehicles and industry.1 In addition to the primary particles emitted directly secondary particles are formed from atmospheric reactions of sulfur dioxide nitrogen oxides and certain organic compounds.Diesel exhaust has been singled out as a major culprit. Moves to improve fuel efÆciency and promote public transport Focus in developed countries have led to a large increase in the number of diesel-powered vehicles. Diesel exhaust is a complex mixture of hundreds of chemical compounds in either a gaseous or particle phase and unlike petrol exhaust contains a high concentration of Æne and ultraÆne particles.2 Being highly respirable and having a very large surface area these particles are an excellent carrier for inorganic and organic exhaust products.Many of the organic compounds adsorbed onto diesel PM such as polycyclic aromatic hydrocarbons (PAHs) and oxidised PAH derivatives are known carcinogens. Potential health consequences from air pollution Ærst came into the public consciousness in the 1950s and 60s as major cities in Europe North America and elsewhere began to experience smogs resulting from the growth of industry and vehicle trafÆc. Governments responded by passing clean air legislation which removed the worst i.e. the most visible pollution. While these 71N J. Environ. Monit. 2000 2 Focus largely allayed public concern at the time scientiÆc evidence began to emerge that PM should remain a key focus for regulation.Over the last thirty years regulators have moved to control successively smaller particles from PM up to 40 mm in diameter (referred to as total suspended particles [TSP]) to less than 10 mm (PM10) and most recently to less than 2.5 mm (PM2.5). In 1997 the US Environmental Protection Agency decided to retain the existing PM10 standards and to add both a daily and an annual standard for PM2.5 to provide increased public health protection.1 EPA's decision was controversial and was criticised by industry Congress and some in the scientiÆc community who said that the health case for regulating ever-smaller particles was unproven. A recent reanalysis of supporting studies has strengthened the EPA's case however (see Table 1).3 In 1998 the European Union adopted limit values for PM10 and called for new research on Æne particles to determine whether this fraction should also be regulated.Counting the costs A variety of studies over recent years have begun to quantify the health consequences of air pollution and especially PM at local national and international levels. In the UK the Ærst ofÆcial quantitative estimates were published in 1998.4 Looking at estimated health impacts of three common pollutants an expert committee concluded that air pollution accounted for up to 24 000 premature deaths each year in the UK and a similar number of hospital admissions. Particulates contributed to around 8100 premature deaths and 10 500 hospital admissions compared to 3500 deaths and 3500 admissions for sulfur dioxide and 12 500 deaths and 9900 admissions for ozone.Nitrogen dioxide and carbon monoxide both of which are thought to contribute signiÆcantly to short-term pollution effects were not considered due to insufÆcient data. Commenting on the experts' Ændings at the time Committee Chairman Dr Jon Ayers said the report almost certainly ``underestimated the true health effects of air pollution'' since it failed to take account of the effects of short-term pollution episodes on vulnerable groups such as the elderly and those suffering from respiratory 72N J. Environ. Monit. 2000 2 diseases.5 The long-term effects on these groups and on healthy people had also not been quantiÆed.Important questions remained unanswered according to Dr Ayres. ``What we would really like to know is if you live in a town like London is your life signiÆcantly shortened compared to living in a rural areaºIn a large population there may be a very signiÆcant effect'' he said. The British research was followed soon after by very different results from a study in France.6 Looking at the relationship between short-term air pollution excursions and mortality rates in nine French cities researchers from the French Institute of Health (INSERM) estimated that air pollution accounted for 265 premature deaths each year. They found that mortality rates increased by between 3 and 4% for a 50 mg l21 rise in concentration of any of the four pollutants studied– particulates sulfur dioxide nitrogen dioxide and ozone.Cardiovascular deaths increased by between 2 and 5% and respiratory deaths by between 1 and 6%. No major variations were found between geographical and demographic areas. The analysis was based on continuous monitoring of both epidemiological and air quality data between 1990 and 1995 and conÆrmed the results of a previous study for the Paris region. Further studies from Norway and Greece put the scale of estimated health effects similar to those of the British survey and much larger than those estimated by INSERM. The Norwegian study put the total national deaths at up to 2000 with fatalities dying on average seven years prematurely.7 Social costs nationwide mainly lost working days and hospital treatment could cost up to NKr 28 billion (e3.4 billion) per year.For diesel engines the study contrasted the costs of health impacts NKr 3± 10 l21 with the NKr 4 l21 collected in taxes. In Greece using ofÆcial information and WHO methodology Greenpeace estimated that air pollution accounted for around 1300 deaths per year.8 The group called for stronger government action to reduce emissions including a ban on diesel taxis which it says are responsible for more than 20% of Athens' total vehicle mileage. Earlier this year the Health Effects Institute in the US published the results of a major investigation into links between particular matter and mortality.9 The National Morbidity Mortality and Air Pollution Study (NMMAPS) looked at both the health effects of air pollution and at methodological issues in this type of research.In analyses for both the 20 and 90 largest US cities NMMAPS found overall mortality increased by an average of 0.5% for every 10 mg m23 increase in PM10 measured the day after death. This effect was slightly greater for deaths due to heart and lung disease than for total deaths. The morbidity analysis showed an approximate increase of 1% in admissions for cardiovascular disease and about 2% in admissions for pneumonia and chronic obstructive pulmonary disease for each 10 mg m23 increase in PM10. Further analyses are currently being undertaken on regional differences mortality±morbidity effects and concentration±response relationships.In August The Lancet published what is claimed to be the Ærst study to use comparable data covering more than one country.10 Dr Nino Ku» nzli and colleagues at the University of Basel estimated the impact of trafÆc-related air pollution on public health in Austria France and Switzerland. The researchers estimated cases of illness and death attributable to quantiÆed increases in PM10. They concluded that in the three countries studied 6% of deaths (or more than 40 000 deaths per year) were caused by air pollution half of which was trafÆc-related. Pollution from motor vehicles also accounted for more than 25 000 new cases of chronic bronchitis more than 0.5 million asthma attacks and more than 16 million person-days of restricted activities.Other recent research from the UK has presented evidence for a substantial rise in adult asthma.11 While large increases in childhood asthma are well documented this is one of the Ærst studies to establish a similar trend in adults. The results show a doubling of adult allergic asthma in the two decades to 1996 and although causation was not investigated the researchers suggest emissions of Æne particles from motor vehicles and other sources are the most likely cause. Probing the PM±health link These and other epidemiological studies have advanced considerably our understanding of the health effects of ambient PM. Overall the evidence suggests there is an association between ambient PM exposures and adverse health outcomes–either death or illness.However the magnitude of the effects depends on environmental and biological factors which have yet to be explained. Or as EPA says in its 1999 draft of the Air Quality Criteria for Particulate Matter ``ºcurrent levels of ambient PM may be harmful to human health but not necessarily equally harmful everywhere or at all times''.1 Current work is advancing on several fronts.12 Some researchers are looking at health endpoints using ambient PM10 and closely related mass concentration indices such as PM13 and PM7. Others are evaluating the effects of different size fractions of ambient PM from PM10 through to ultraÆne particle mass concentrations of PM1 and smaller.The relationship between health endpoints and ambient particle number concentrations is also being addressed while other studies focus on the sensitivity of estimated health effects to air pollution mixtures including PM components. Finally there is much new research into the potential susceptibility of infants and children and into cardiovascular disease as well as asthma and other respiratory conditions. One of the most important priorities is to provide a satisfactory explanation for geographical variations. In the NMMAPS study for example the researchers observed some differences in PM10 effects on mortality between the 90 cities studied with the largest effect being within the Northeast region.9b The investigators were unable to explain this but most likely it is due to differences in the nature of particulate matter from one city to another which is known to vary considerably.The HEI researchers conclude that ``The heterogeneity of effect across cities offers the potential to identify factors that could inØuence the effects of PM10 on health and thus provide valuable insights into the mechanisms by which PM10 causes adverse health effects''. Size matters The variation of ambient particulate matter by size and chemical composition is increasingly recognised as signiÆcant in explaining and estimating health impacts. The terms ``Æne'' and ``coarse'' were originally applied to the two major atmospheric particle distributions which overlap in the size range between 1 and 3 mm diameter.1 Now Æne has come to be often associated with the PM2.5 fraction and coarse is often used to refer to PM10-2.5.However PM2.5 may also contain in addition to the Æne particle mode some of the lower-size tail of the coarse particle mode between about 1 and 2.5 mm. Conversely under relatively high humidity conditions the larger Æne particles in the accumulation mode may also extend into the 1±3 mm range. By deÆnition each of the larger PM indices contains elements of the smaller indices and hence it is difÆcult to determine the relative effects of different PM metrics. In its 1999 draft of the Air Quality Criteria for Particulate Matter EPA concluded that ``distinctions between Æne and coarse mode particles (in terms of sources of emissions formation mechanisms atmospheric transformation and transport distances and air quality patterns) warrant Æne and coarse PM being viewed as separate subclasses of ambient PM''.1 Thus better characterisation of particulate pollution is critical to efforts to prove beyond reasonable doubt the adverse health effects of particles and the enforcement of appropriate environmental quality standards.In the UK EPAQS an expert panel set up to advise government on air quality issues highlighted the difÆculties in a report issued earlier this year.13 Referring back to a 1997 decision the panel noted ``Since the EPAQS recommendation on PM10 was made evidence has emerged that PM10 might not be the measurement most representative of the fraction of ambient aerosol that is responsible for its harmful effects on health.Evidence has accumulated that this toxicity might lie in a Æner fraction of the particles perhaps below 2.5 mm or smaller. Moreover it has become apparent that PM10 measurements may include an amount of resuspended dust of probable low toxicity leading to difÆculties in compliance with the standard in areas where coarse particles such as wind blown dust or sea spray may make an important contribution.'' For the UK EPAQS estimated the contribution of coarse particles to the PM10 mass as being around 20 to 50% of the urban background mass.While regulators and their advisers have been pondering whether to act against PM2.5 researchers have been turning their attention to ever smaller particles. Some now consider very small particles–those less than 0.1 mm (100 nm) in diameter–to be particularly toxic. Although they constitute only around 1±8% of the mass of PM in ambient air these ultraÆne particles are present in very high numbers have greater total surface area than larger particles and may deposit in greater Focus numbers in the lungs. To date there has been relatively little research on particles in this sub-micron range although one recent animal study found limited evidence that inhaled ultraÆne particles cause inØammation.14 Pollutant cocktails The chemical composition of PM is another aspect which has received limited attention in epidemiological studies.PM is a cocktail of chemically and physically diverse dusts and droplets some of which may be particularly signiÆcant in overall health risks. Hydrogen ions for example reØect particle acidity and sulfate has a well characterised role in deposition. In addition gaseous pollutants are often present alongside PM and are associated with similar health outcomes. So the key question here is whether health effects are the result of either the PM or a copollutant acting alone in combination with each other or through the overall pollutant mix? And if PM alone is to blame are one or more of its components better predictors of morbidity or mortality than PM as a whole? One recent study focusing on the Detroit area attempted to unravel the effects of these pollution mixtures.15 Using publicly available data for 1985± 1990 and 1992±1994 the team compared day-to-day Øuctuations in air pollution with day-to-day Øuctuations in deaths and hospital admissions.PM was found to be more signiÆcantly associated with health outcomes than the Hz or sulfate. However when evaluated with gaseous pollutants in a two-pollutant model the estimated effects of PM2.5 were often reduced especially for ozone. Interestingly the relative risks for PM10±2.5 were similar to those for PM2.5 and sometimes even higher for example with ischemic heart disease and stroke.This Ænding underlines the fact that while the Æne fraction of PM needs further investigation the health risks of the coarse fraction also cannot be ignored. Methodological issues Drawing public health conclusions in the face of so many uncertainties is fraught with danger. Given the prevalence of PM pollution sources and uncertainties in their mode of action the scope for error or misinterpretation is substantial. Three issues are prominent in the debate here. 73N J. Environ. Monit. 2000 2 Focus Deaths real or imaginary? Firstly there is the question of what is actually meant by ``deaths'' associated with air pollution. Are they ``real'' premature deaths or is it more an issue of advancing death by only a few days (what epidemiologists call ``mortality displacement'').If the association is solely in terms of a short-term displacement for those that are already near death then the data may be showing an effect of limited signiÆcance to public health. The US NMMAPS study investigated the signiÆcance of mortality displacement as one of a number of methodological issues underlying associations between PM and mortality.9a Using two different but related statistical approaches the investigators concluded that more than a short-term displacement was occurring. They noted however that more experience and better models are needed to measure and explain these displacements for example regarding the role of air pollution mixtures in longterm effects.Measuring exposure Establishing human exposure is another difÆculty. People spend much of their time indoors where exposures resulting from day-to-day activities such as cooking and cleaning can be much higher than those due to outdoor exposures to ambient air. And of course lifestyle factors such as smoking habits hobbies residential furnishings and occupation lead to widely different exposures for people within a given community. Furthermore outdoor exposures also vary considerably due to factors such as weather and trafÆc conditions so that a single monitoring station cannot be relied on as a measure of personal exposure. The issue of exposure assessment is addressed in detail in EPA's 1999 reassessment of the NAAQS.1 It concludes that for those living within the vicinity of a monitoring station over time ambient PM measurements are an appropriate proxy for personal exposures both outdoors and indoors to PM of ambient origin.This is especially true for Æne particulates according to EPA based on studies of ambient sulfate concentrations and sulfate exposures. The effects of measurement error on the relationship between personal exposure and ambient exposure were 74N J. Environ. Monit. 2000 2 also investigated in the NMMAPS study.9a Using theoretical and actual analyses the researchers tended to refute the criticisms that the associations between PM and adverse health effects could be explained by exposure measurement error.They note however that the absence of measured exposure data precludes making Ærm conclusions. EPA and other organisations are currently funding research to obtain more exposure data that should help to resolve this issue. Monitoring metrics Thirdly there is the issue of which measurement metric to use for regulatory purposes. The PM10 metric is increasingly seen as unsuitable because the measure is dominated by a small number of relatively large particles. Chemistry suggests that the major part of the toxic component of PM resides in the ultraÆne fraction below 1 mm but epidemiological studies have been unable to separate out health effects between coarse and Æne/ultraÆne fractions. There are also fears that measurement of smaller metrics will be more sensitive than usual to loss of volatile elements.This uncertainty presents a dilemma for regulators. Should they move to a PM2.5 metric on the basis of limited data or should they wait until the health mechanisms are better understood? In its recent report for the UK EPAQS concluded that ``There is no reason to suppose that other metrics based on particle numbers or surface area may better represent the toxic element.'' Reliance on such alternative metrics would be ``premature'' EPAQS said. It continued ``The panel thus recommends that the control of particulate air pollution by measurement of PM10 provides adequate protection of the health of the public. We conclude however that there are likely to be advantages in the use of a metric more representative of the toxic component in the future and the UK research programme should be aimed at better deÆnition of this component.'' Into the policy arena Despite uncertainties in the data and the underlying methodologies policymakers are showing a keen interest in the health statistics and an eagerness to use them as the basis for concrete policy actions.As noted above EPA has already introduced criteria for PM2.5. While there are no such standards yet in the UK the Ægure of 24 000 premature deaths derived in the 1998 study featured prominently in the government's White Paper on Transport published earlier this year. In the EU the European Commission frequently cited mortality data in its protracted negotiations with member states over air pollution legislation.Data released by the Commission in March claimed that ozone and particulate pollution would cost an ``extra'' 4000 lives in the EU annually if the member states refused to endorse its proposals for national emission ceilings.16 This Ægure was the difference between the 2000 deaths that would be prevented under the UNECE's 1999 Gothenburg Protocol and the 6000 deaths that would be prevented by the stricter EU directive. In the end the member states opted for a compromise.17 Research challenges The high proÆle afforded to health impacts makes it all the more important that further research is undertaken to help Æll the gaps in our understanding.The scientiÆc agenda is extensive and includes a substantial role for environmental monitoring and the analytical sciences.1,9a,12 Priorities include investigation of Ambient PM exposures including further correlation of personal and ambient exposures and investigation of the effects of exposure measurement errors in time-series studies. Spatial variation of exposures to explain observations between different locations and to ensure comparability between datasets. Size-related effects from the ambient monitoring of size fractions (coarse Æne and ultraÆne) through to dosimetry relationships and health effects. Exposure to gaseous co-pollutants which could play a signiÆcant role in both short-term and long-term exposure studies.Long-term PM exposure including accounting for effects of different size metrics and co-pollutants. Vulnerable groups especially infants and children which recent studies suggest may represent a subgroup at special risk from ambient PM exposure; Non-respiratory causes of death and hospitalisation such as cardiovascular conditions especially in older adults. References 1 Air Quality Criteria for Particulate Matter National Centre for Environmental Assessment 2000 EPA/ 600/P-99/002a b c. NCEA Research Triangle Park NC. Available at www.EPA.gov/ncea/partmatt.htm 2 Health Assessment Document for Diesel Exhaust Science Advisory Review Board National Centre for Environmental Assessment 2000 EPA/ 600/8-90/057E NCEA Research Triangle Park NC.Available at www.EPA.gov/ncea/diesel.htm 3 Report of the Particle Epidemiology Reanalysis Project Health Effects Institute 2000 Cambridge MA. 4 QuantiÆcation of the Effects of Air Pollution on Health in the UK Committee on the Medical Effects of Air Pollution (COMEAP) 1998 Department of Health London. 5 ENDS Daily 14th January 1998. www.ends.co.uk 6 French chart air pollution deaths J. Environ. Monit. 1999 1 43N. See also www.inserm.fr 7 Health Effects and Socio-economic Costs of Air Pollution Norwegian Pollution Control Authority (SFT) 2000. SFT Oslo. www.sft.no 8 ENDS Daily 7th July 2000. www.ends.co.uk 9 The National Mobidity Mortality and Air Pollution Study (a) Part I Methods and Methodologic Issues; (b) Part II Morbidity and Mortality from Air Pollution in the United States Health Effects Institute 2000 HEI Research Report 94 Cambridge MA.Available online at www.healtheffects.org 10 Nino Ku» nzli et al. The Lancet 2000 356(9232) (31st August 2000). 11 Br. Med. J. 2000 321(7253) (6th July 2000). Reprint available at www.bmj.com/cgi/reprint/321/7253/ 88.pdf 12 Extensive information on current research into particulate matter worldwide is available at the Particulate Matter Research Activities website www.pmra.org 13 Airborne particles what is the appropriate measurement on which to base a standard Expert Panel on Air Quality Standards Focus (EPAQS) 2000 Department of the Environment Transport and the Regions London. Available online at www.environment.detr.gov.uk/airq/aqs/ index.htm 14 Acute Pulmonary Effects of UltraÆne Particles in Rats and Mice Health Effects Institute 2000 HEI Research Report 96 Cambridge MA. 15 Association of Particulate Matter Components with Daily Mortality and Morbidity in Urban Populations Health Effects Institute 2000 HEI Research Report 95 Cambridge MA. 16 Economic Evaluation of a Directive on National Emissions Ceilings for Certain Atmospheric Pollutants. Part A Costeffectiveness. Part B BeneÆts European Commission 2000 Directorate-General Environment Brussels. Available at www.europa.EU.int/comm/environment/ envecol 17 Ministers fudge emission ceilings J. Environ. Monit. 2000 2 57N. Mike Sharpe 75N J. Environ. Monit. 2000 2