CEEH Scientific Report No 7a

Transcription

1 Centre for Energy, Environment and Health Report series ISSN CEEH Scientific Report No 7a Description of the CEEH health effects model - selection of concentration-response functions Aarhus 2011

2 Colophon Serial title: Centre for Energy,Environment and Health Report Series Title: CEEH Scientific Report No. 7a - Description of the CEEH health effect model Author: Jakob H. Bønløkke 1 Other contributors: Torben Sigsgaard 1, Jørgen Brandt 2, Lise M. Frohn 2, Esben M. Flachs 3, Henrik Brønnum-Hansen 3, Marie-Louise Siggaard-Andersen 4 1 Aarhus University, Department of of Public Health, Bartholins Allé 2, Building 1260, 8000 Århus C, Denmark. 2 Aarhus University, National Environmental Research Institute, Department of Atmospheric Environment, Frederiksborgvej 399, 4000 Roskilde, Denmark. 3 National Institute of Public Health, University of Southern Denmark. 4 University of Copenhagen, Niels Bohr Institute, Juliane Maries Vej 30, 2100 København Ø, Denmark Responsible institution: Aarhus University Language: English Keywords: Air pollution, concentration-response functions, health effect model, Energy system analysis, integrated modeling, optimization, energy, environment, atmospheric pollution, meteorology, climate, health, externality, CEEH, Denmark, energy scenario, Balmorel, DEHM, Enviro-HIRLAM. Url: ISSN: ISSN Version: November 2011 Website: Copyright: Any use of the content of this report should be cited as: Bønløkke, J.H. (2011) CEEH scientific report No 7a, Centre for Energy, Environment and Health (CEEH) report series. Cover image: Clouds with CRF showing two typical concentration-response functions created only for illustrative purposes by Torben Sigsgaard and Jakob Bønløkke. Page 2 of 76

3 Content: Introduction... 5 Selection of air pollution concentration-response function (CRF) in the CEEH health effects models... 5 Summary... 6 CRF in existing models... 6 Other pollutants proposed in the literature... 7 Recommendations for CEEH: Pollutants and health effects to include in the model... 7 Recommended CEEH Concentration-Response functions (CRF)... 9 Concentration-response functions used in the CEEH health effects model CEEH review on health effects of fine particles (PM 2.5 ) BACKGROUND SIZE CHEMISTRY AND COMPOSITION LENGTH OF EXPOSURE How has particles been treated in recent modelling projects, cost-benefit analyses, reviews and reports? CEEH review on health effects of ozone CEEH review on health effects of gases other than ozone Nitrogen oxides (NO/NOx) Carbon monoxide (CO) Sulphur dioxide (SO 2 ) CEEH review on health effects of heavy metals References Appendix I International Classification of Diseases codes for the diseases relevant to air pollutants Side 3 af 76

4 Appendix II Comprehensive list of publications Appendix III Proposed use in CEEH of changes in incidences of health effects associated with air pollution, i.e. concentration-response functions (CRF) for incidences Page 4 of 76

5 Introduction This report is based on a number of up-to-date reviews of the existing literature on health effects from air pollution at the population level and concludes with recommendations for CEEH. Selection of air pollution concentration-response function (CRF) in the CEEH health effects models The CRF in the CEEH health effects models have been selected based on the following criteria: The pollutants stem from combustion sources either directly or via chemical transformations. This may include evaporation or dust from the energy sources themselves (i.e. from wear and tear). Other pollutants derived from energy production but not liberated to the air is not included (e.g. heavy metals in soil deposits). However, heavy metals, dioxins and possibly PAHs that are emitted to the air, deposit on soil or in water and end up by being ingested should ideally be covered by CEEH. The pollutants can be modelled in the CEEH settings both in terms of emissions, chemical transformations and transport. The pollutants have been sufficiently investigated in terms of documented health effects at a population level and CRF can be calculated with a high degree of confidence. Combustion sources and accordingly combustion products are highly diverse and widespread. In practice, therefore, not all sources that contribute to health effects in humans can be included in the CEEH models. Other sources than combustion contribute some of the pollutants and as a result a brief discussion of the relevance and effect of such non-combustion derived air pollutants is included in this report. For a more comprehensive discussion of secondary pollutants from e.g. agricultural activities the reader is referred to CEEH report no. 3. Side 5 af 76

6 Summary CRF in existing models Different models of societal externalities related to combustion sources and energy production exist and include varying combustion products, CRF functions and cost calculations. This review does not claim to review and compare with all existing models used previously but emphasizes some models considered particularly relevant for the CEEH modelling of health effects from combustion products in Denmark. These include EVA (Economic Valuation of Air pollution) (Frohn et al., 2005; 2007; Andersen et al., 2006; 2007; 2008; Brandt et al., 2010), an integrated system developed by AU-NERI, which simulates atmospheric dispersion, chemical transformation, human exposure, health impacts and costs related to health effects of air pollution. The EVA model was built following the impact-pathway approach, which is also the basis of other existing European models such as EcoSense 4.0 from ExternE. A more recent version of the ExternE methods is now available (ExternE, 2005). The pollutants contained by ExternE are provided in table 1. Pollutants included in the original version of the EVA model system are the following: SO 2, O 3, CO, PM 2.5, nitrate, sulphate, dioxin, Pb, and Hg. In addition to the EVA model system and ExternE the following national and international reports that have been reviewed in the process of selecting CRF for CEEH health effects model include US-EPA, WHO, OECD, DEFRA (UK), COMEAP (UK), APHEA, NEEDS and CAFE (EU) reports. Table 1. ExtenE air pollutants and associated health impacts (ExternE, 2005). Primary Pollutants Particles (РМ 10, PM 2.5, black smoke) Secondary Pollutants Impacts mortality cardio-pulmonary morbidity (cerebro-vascular hospital admissions, congestive heart failure, chronic bronchitis, chronic cough in children, lower respiratory symptoms, cough in asthmatics) SO 2 SO 2 Sulphates like particles NO x Nitrates like particles NO x +VOC Ozone mortality cardio-pulmonary morbidity (hospitalization, consultation of doctor, asthma, sick leave, restricted activity) mortality morbidity (respiratory hospital admissions, restricted activity days, asthma attacks, symptom days) Page 6 of 76

7 CO PAH diesel soot, benzene, 1,3-butadiene, dioxins As, Cd, Cr-VI, Ni Hg,Pb mortality (congestive heart failure) morbidity (cerebro-vascular) cancers cancers other morbidity morbidity (neurotoxic) Other pollutants proposed in the literature Other pollutants than those in EVA and EcoSense contribute to the adverse health effects of (combustion-derived) air pollution. These include separate effects of particles greater than 2.5 μm (PM 10 -PM 2.5 ) and black smoke (which correlate highly but do not correspond completely with PM 2.5 and PM 10 ) and ultrafine particles. Numerous scientific studies either document or are highly suggestive of separate effects of these components. The composition of particles is important too and there is evidence of effects from secondary particles such as ammonium, nitrates and sulphates as well as from primary components such as carbon (both elemental and organic) and from metals, PAHs, and biologic material that aggregate with the particles and are carried into the lower airways by them, such as endotoxins. Several publications suggest effects of NO 2 or other NO x. The dominating opinion is, however, that the contribution from NO x is either via contribution to O 3 - a pollutant with effects separate from those of particles or in case of possible separate effects of NO x these are indistinguishable from those of fine particles due to the high correlation with other combustion derived pollutants such as particles. Recently, aldehydes from bioethanol have been shown to add importantly to the toxicity from this type of fuel. The effects of aldehydes appear to be mediated almost entirely through O 3 -generation. Some of these additional pollutants have short life-spans in the atmosphere or are rapidly diluted and therefore only contribute to health effects within short distances of the source. This appears to be the case for ultrafine particles and CO. The relevance of modelling these is therefore limited except in models of health effects of pollution from traffic. In addition to combustion products another type of pollution that is particularly relevant in studies of traffic is noise which adds to the observed morbidity and mortality but which has been studied too little for meaningful assessment in the CEEH chain. Recommendations for CEEH: Pollutants and health effects to include in the model As there is consensus that the most important health effects (also in terms of costs) stem from particles, CEEH should include effects of particles composed of nitrates and sulphates as well as primary particles Side 7 af 76

8 in the size fractions PM 10 and PM 2.5 as both size fractions are used in CRFs for a range of health effects. Effects of particulate matter include all-cause or cause-specific morbidity (as hospital admissions) and mortality for both adults and infants. Ozone should be modelled in CEEH for use in a CRF on long-term effects on all respiratory mortality. The gaseous pollutants SO 2 and NO 2 should also be modelled. In contrast to other recent reviews, but from the point of view that independent effects of SO 2 do seem to exist and that they are plausible from a mechanistic point of view and despite the risk of double counting effects in CEEH we propose a shortterm CRF of SO 2. This is particularly warranted as we suggest a slightly conservative CRF of particles on cause-specific mortality (see the chapter on particles) which miss some of the effect of air pollution. Recent Danish evidence (Andersen Z, et al., 2010) of an effect of long-term NO 2 on COPD incidence has prompted us to use NO 2 rather than PM for the CRF for this disease, Separate effects of other NO x, VOC, and CO are presently not well-enough documented and too small for reasonable inclusion in the overall health effects models. However NO x emissions contribute to secondary particles and to ozone and are important in the atmospheric modelling and generate important health effects through PM and ozone. New studies regularly appear in which NO x has been used as the primary (only) pollutant adding new information about health effects of air pollution. The entire list of recommended CRFs is provided in table 3. Ideally noise should also be included in models of traffic-generated air-pollution but our current health effect models rarely include noise. From a health perspective, PAHs, dioxin and the metals As, Cd, Cr, Hg, Ni and Pb should be included in a health effects model. Metals such as Fe, present in combustion-derived particles and potentially responsible for specific toxicity, remain to poorly documented for specific modelling of health effects. Similarly, the health effects of other chemicals and metals cannot be separated meaningfully form PM effects. Although heavy metals originate to a large extent from combustion sources, and health effects of lead and mercury are well documented, heavy metals are not considered in CEEH. Cost evaluations of their effect (e.g. on intelligence and mental health) are largely disputed. However their effect may be contributing to the general toxicity of fine particles and is therefore covered to a large extent. Their effect from uptake via food and water is a major effect but is not considered in CEEH. Further we propose to Use estimates for fine particulate matter (PM 2.5 ) effects when available as the contribution from coarser fractions cannot be separated sufficiently and would lead to double counting. If the evidence is insufficient for fine but sufficient for coarse particles estimates for the latter can be used instead. Regarding mortality, only use estimates from population (prospective) studies as in other recent reviews (primarily NEEDS 2007 and COMEAP 2009). Page 8 of 76

9 In contrast to other recent reviews which use all-cause mortality we recommend the use of cause-specific estimates because these form a better basis for specific cost calculation although from a health perspective the total effect appears to be better covered by all-cause estimates. Only include estimates if based on several independent sources or based on single high quality and large studies endorsed by other reviews, i.e. assuring that the evidence is solid and the risk of double counting is negligible. CRF that include 1 in the 95% CI are thus not considered even if endorsed by recent review such as COMEAP or NEEDS. Treat secondary particles as primary but conduct sensitivity analysis as suggested by NEEDS. Recommended CEEH Concentration-Response functions (CRF) The general approach to estimating the effects of pollutants on health uses the relative risk (RR) found in the epidemiological studies, expressed as % change in end-point per unit pollutant (concentration) (CRF). If the CRF is multiplied on the background rates of the health end-point in the target population and the population at risk the CRF can be expressed as new cases (or events) per year (or day) per unit of this population per relevant pollution increment. Results are then expressed as estimated new or extra cases, events or days per year (or day) attributed to the pollutant. Table 2. CRFs used in EVA. PM2.5 from (Andersen MS, 2010) sulphate and nitrate from (Andersen MS, 2004) Health effect PM 2.5 Sulfate Nitrate Unit Valuation Euros (2006-prices) MORBIDITY (Particulate matter) Chronic Bronchitis 8.2E E E-05 cases/μgm -3 (adults) 52,962 per case Restricted activity days Hospital admissions 8.4E E E-02 days/ μgm -3 (adults) (- hospital admissions) 131 per day - respiratory 3.46E E E-06 cases/ μgm -3 7,931 per case - cerebrovascular 8.42E E E-06 cases/ μgm -3 10,047 per case -congestive heart failure 3.09E E E-05 cases/ μgm -3 (> 65 yr) 16,409 per case Lung cancer 1.26E E E-03 cases/ μgm -3 (adults) 21,152 per case Side 9 af 76

10 Asthma children (prevalence: 7,6 % < 16 years) - bronchodilator use 1.29E E E-02 cases/ μgm per case - cough 4.46E E E-01 days/ μgm per day - lower respiratory symptoms Asthma adults (prevalence: 5,9 % > 15 years) 1.72E E E-01 days/ μgm per day - bronchodilator use 2.72E E E-01 cases/ μgm per case - cough 2.8E E E-01 days/ μgm per day - lower respiratory symptoms MORTALITY 1.01E E E-02 days/ μgm per day Acute mortality (SO 2 ) 7.85E-6 cases/ μgm -3 2,111,888 per case Chronic mortality 1.138E E E-04 YOLL/ μgm -3 (>3077,199 per (PM) YOLL years) Infant mortality (PM) 4.68E-6 cases/ μgm -3 (< 9 3,167,832 per months) case Acute mortality (O 3 ) 3.27E-6 cases/μgm -3 2,111,888 per *SOMO35 case YOLL=Years of Lost Life. The concentration-response functions are expressed as number of new occurrences in the relevant age group per µg/m 3 /year. SOMO35 denotes the number of days where max. 8-hour mean of 35 ppb/m 3 is exceeded. Concentration-response functions used in the CEEH health effects model The reviews of the CRFs in the literature and the selection of CRFs for CEEH is provided in the chapters on specific pollutants below. The majority of existing reports from other centres are discussed only in detail in the chapter on particles as they either agree to a large extend on the remaining pollutants or differ to much from the CEEH model on those for relevant use of CRFs. As discussed in the reports, recent literature has added specific effect estimates for different cardiovascular endpoints possibly allowing for a more detailed estimation of effects than presently in EVA; long-term effects of ozone and effects on incidence of COPD (Danish results). SCHER does not offer any dose-response information (SCHER, 2005) and is not treated further. Table 3 provides the recommended CEEH CRF. Some effects are included more than once and are suggested for sensitivity analyses only (double counting will result if all CRF are used). The precise selection of which CRF to include and which to use only for sensitivity analysis depends on the method Page 10 of 76

11 applied in WP4 of CEEH in which costs are calculated. The method recommended based on the health review alone is the following: All-cause mortality should be based on PM 2.5 long-term and all-cause infant mortality based on the PM 10 CRF. Cause-specific mortality of PM exposure should be used only for sensitivity analyses. Respiratory mortality should also be based on the long-term effect of ozone. The specific CRF for respiratory and cardiovascular hospital admissions should be used for respiratory and the cardiac subgroups (and the overall estimate for cardiac only for sensitivity analysis). The NO 2 CRF on COPD should be used; the one based on PM should be considered for sensitivity analysis. Secondary particles should be treated as primary but sensitivity analysis treating nitrates and sulphates differently should be conducted. Additional sensitivity analyses should include the 95% CI. In addition we suggest a sensitivity analysis by using just the all-cause mortality function (excluding all other functions) and include the range 0-15% as proposed by COMEAP - keeping in mind that recent American prospective studies have suggested effects in the range of 16% (Laden et al., 2006). Heavy metals have well documented and quantifiable long-term effects on intelligence which should be considered from a health perspective but which are not quantifiable cost wise in the CEEH model chain. They are thus not treated any further. Side 11 af 76

12 Table 3. Concentration response functions proposed for use in the CEEH health effects model Pollutant Effect Age ICD-10 CRF (RR) Notes Ozone Mortality of any All ages J ( ) per 10 No threshold. respiratory ppm in summer months mean of 1-hr max Only April September. PM 2.5 Cardiopulmonary 30+ I ( ) mortality J00-99 PM 2.5 Lung cancer mortality 30+ C ( ) PM 2.5 Respiratory hospital all ages J ( ) admissions PM 10 Ischemic heart all ages I ( ) disease hospital admissions PM 10 Dysrhythmia hospital admissions all ages I ( ) PM 10 Heart Failure all ages I ( ) hospital admissions PM 10 Infant mortality ( ) PM 10 Lower respiratory symptoms symptomatic adults 1.3 days/yr/person* PM 10 Lower respiratory days/yr /person* symptoms PM 2.5 Restricted activity days/yr /person* days PM 2.5 Work Loss days days/yr /person* NO 2 COPD incidence adult J % ( ) SO 2 All-cause mortality Adult A00- Y ( ) per 24h previous 10 μg/m 3 day mean PM 10 New cases of chronic bronchitis 27+ J J * sensitivity analysis PM 2.5 All-cause mortality 30+ A00- Y ( ) sensitivity analysis PM 10 Cardiac hospital admissions all ages I ( ) sensitivity analysis Page 12 of 76

13 Pollutant Effect Age ICD-10 CRF (RR) Notes PM 2.5 Incidence of fatal cardiovascular disease adult I females 100% males I %* sensitivity analysis Recommended CRF (for yearly means of a 10 μg/m 3 if not specified otherwise) in CEEH to be used as described in the text. 95% CI in parentheses. * no CI because they are derived from reports or calculations from more than one study without sufficient details for CI calculation. ICD-10: International Classification of Diseases codes for the diseases relevant to PM-exposure. Side 13 af 76

14 CEEH review on health effects of fine particles (PM2.5). BACKGROUND The evidence for negative health effects from particles (particulate matter; PM) stemming form laboratory, animal, short term (time-series) and long term (prospective) studies is massive. Hundreds of studies have observed associations with short-term peaks in particle concentrations and adverse health effects (typically based on same-day exposure or exposure from previous 1-3 days but occasionally longer and up to 40+ days post exposure). The list of health effects observed in such studies is long, ranging from cough to death. Health effects associated with ambient air pollution are divided into two broad categories - premature mortality and morbidity. The two key morbidity effects quantified by the CBA reports are respiratory hospital admissions and cardiovascular hospital admissions. Only a few prospective cohort studies have been published with associations between long-term average exposure to particles and health effects. Yet the latter type of studies has significantly strengthened the likelihood of a direct link between air pollution and severe health outcomes. Based on this evidence, reports from national and international health authorities have stated that such health effects are beyond any reasonable doubt and have tried to quantify it. In this review meant for documentation in the CEEH chain of energy scenario modelling and optimization based on costs, the evidence for health effects will be briefly reviewed and compared with reviews in the most important national and international reports and cost-benefit analyses from different projects. Recent and more detailed reviews can be found in COMEAP 2009 and Jalaludin SIZE Particulate air pollution is divided into size fractions based on estimates of penetration into the human airways as this conceivably influences the health effects they may elicit. The rationale being that particles that penetrate deeper into the narrower airways and closer to where gas diffusion takes place pose a greater threat. Table 4 lists the most commonly used fractions, their sizes and abbreviations. PM 10 are also referred to as coarse particles, PM 2.5 as fine, and particles 0.1 μm are usually referred to as ultrafine (though some controversy exist about the exact cutoff). Table 4. Mass deposition fraction according to the International Standards Organisation Convention Mass fraction Size (aerodynamic diameter) Page 14 of 76

15 Inhalable Inhaled through the nose or mouth TSP (total suspended particles) Extrathoracic Failing to penetrate beyond the larynxapprox μm Thoracic Respirable Penetrating beyond the larynx Penetrating to the unciliated airways approx. PM 10 ( 10 μm) approx. PM 2.5 ( 2.5 μm) The convention names are often used synonymously with the fraction sizes given although they are only approximations of sizes of which the >90% will penetrate as described in normal healthy subjects and although penetration into the airways is not simply an effect of size. Particulates in ambient air have been studied based on the division given in table 4 and samplers constructed in order to cut-off at the given diameters. The vast majority of early studies of health effects were based on the concentration of TSP or PM 10 and only in the mid 90ies concentrations of PM 2.5 (fine particles) started to become available, with studies on ultrafine particles still being rare. Approximate conversion factors between size fractions exist but they vary with source and composition of the particles. This obviously causes difficulty when comparing health effects from different studies. The major driver in the change of focus from the larger to the smaller sized particles has been epidemiologic (population based) studies that showed stronger associations between important negative health effects and fine particles than with coarse particles. Where earlier studied focused more on respiratory effects these more recent studies have demonstrated cardiovascular effects and, accordingly, stronger effects on mortality. Conclusion: Based on the current knowledge estimates of health impact of particulate air pollution should be taken from studies on fine particles. In the absence of such estimates for specific health outcomes estimates should be based on the coarse fraction, preferably modelling this fraction rather than estimating effects based on converting factors from coarse to fine. CHEMISTRY AND COMPOSITION Evidently, in addition to size the chemical composition may be an important determinant of health effects. Several short term studies have demonstrated large and significant differences in health impact between particles of different composition. This literature points to a number of (transition) metals as well as sulphate as the most important in determining effects on mortality. However, differences exist between studies regarding which metals are most important and the contribution from other components. In addition several of the studies do suggest effects of substances such as sea-salt, potassium or sand. What has emerged is that the toxicity seems to depend on the source of the Side 15 af 76

16 particles, with e.g. sea-salt toxicity probably reflecting contribution from ship emissions and that at present it is not possible to use the data for assigning specific RR to even the major components. Furthermore, because chemical analyses for source apportionment over periods of years has rarely been done in studies of health effects, no detailed data exist on differences in health impact associated with composition in long term studies. Long term effects reflect development of chronic disease to some extent and may differ substantially from short term effects when it comes to the role of individual particle components. Only for sulphate good data exist on long term effects. Both short-term and long-term studies point to combustion particles from road traffic as probably the most important contributor to particle associated health impact. Other combustion sources such as ship emissions, power plants and wood stoves appear to be important too as suggested from the associations with mortality that appear in studies with sea-salt in coastal regions, sulphate in studies of long-range transport and potassium in studies including residential areas with large wood stove contributions to winter air pollution. Long-term studies finding effects associated with NOx concentrations - sometimes stronger than effects associated with particles - have been interpreted as suggestive of the importance of road traffic for the negative health effects such as mortality. Particles come as primary (emitted) and secondary (generated in the atmosphere). The vast majority of fine particles inhaled by people are secondary except when being at very short distances from emitters such as in a busy street or in a number of occupational settings. The long-term effects are based on yearly averages on residential addresses and conceivably reflect exposure to the regional background + the local mix of pollutants and are thus dominated by secondary particle effects. People spend most of their time away from strong emission sources (i.e. at home and during evenings and nights) where they respire a mixture of primary and secondary particles most of which carry components of different origin, e.g. from combustion and all of which therefore probably contribute to the toxicity of particles observed in long-term studies. Sulphate-, nitrate- and ammonium-ions contribute a substantial number of the secondary particles. Several laboratory, animal and experimental exposure studies with humans have failed to document any important toxic effects of the pure substances. In the atmosphere, however, they form in the presence of a range of other components, e.g. metals ions and organic molecules with which they are likely to react or agglomerate. Therefore they can t be treated as pure and non-toxic contributions to the particle mixture respired by people, not even in rural areas. Assuming lower toxicity from sulphate and nitrate, some previous reviews and models including EVA have assigned less toxicity to these particles based on educated guesses. Table 5 lists some of these. Conclusion: Based on the current knowledge, it is not possible with certainty to assign different RR to primary or secondary particles or components thereof in calculations of negative health effects from long-term exposure. It may be suggested to use an educated guess approach in consideration of less effect of secondary sulphates and nitrates on mortality in sensitivity analyses. Table 5. Toxicity assigned to sulphates and nitrates in previous reviews. Page 16 of 76

17 Table extracted from NEEDS P: particle effect. O: oxidising effect. LENGTH OF EXPOSURE Extensions of short-term studies have not documented that increased mortality following episodes of particulate air pollution is due to harvesting of the most vulnerable followed by a decrease in mortality associated with only the less vulnerable being left in the population. On the contrary, when comparing estimates of associations between increased particle concentrations and death, stronger effects have been observed with longer observation times - the strongest stemming from long-term studies. However poorly peak episodes leading to sudden death from, e.g. myocardial infarction are reflected in yearly averages of particle concentrations, the long-term studies per definition include such acute effects in their estimates. No studies have combined short-term followed of daily particle concentration effects with long-term follow-up of the same population and at present it is not possible to disentangle any additional effect of short-term episodes from the long-term effects. A number of studies have been performed in regions with significant decreases in air pollution either temporarily, e.g. during Olympic Games, or permanently due to bans or regulations. They confirm the findings from time-series studies of short term effects and the findings from earlier prospective studies on long-term effects. It is difficult, however to derive estimates from them for use in modelling such as in CEEH unless they give risk estimates in the text and they have not been considered systematically in this report. Conclusion: Based on the current knowledge the best (i.e. those that are closest to accurately capturing the range and size of effects in a population) estimates of health impact from particulate air pollution stem from long-term studies. Only in the case of lack of such studies and the presence of good shortterm studies of specific health outcomes should these be used. Table 6. Relative risk estimates of all-cause mortality associated with a 10 μg/m 3 increase in the yearly average particle concentration. Data from major long-term studies. Reference Agegroup PM 2.5 PM 10 SO 4-2 Dockery * 1.1* 1.33* Side 17 af 76

18 Laden Krewski 2000 Pope Pope ** Krewski Jerrett Abbey 1999 males NS NS Enstrom males NS females Filleul (TSP) RR = relative risk. NS: not statistically significant. * Recalculated from the numbers in the paper assuming linear associations. ** number not in text but read from figure 5. TSP: totally suspended particles. Comment on the estimates given in the tables: In general the overall estimate with the a priori chosen adjustments of the authors has been selected - usually corresponding to what is presented in the abstracts and discussion. In the case of several estimates from, e.g. different follow-up periods or different adjustments, the ones using the most recent air pollution data and most adjustments have been selected unless these are characterized by the authors as sensitivity analyses or similarly. In Pope (2002) the authors follow up to 1998, using either PM data for , or a combination to calculate concentration exposures. They prefer reporting estimates based on the former exposure data, but as pointed out by Laden (2006) most of the effect of particles can be ascribed to the past year exposure and the data should therefore reflect better the more recent exposure in the population. The estimates used here are thus the ones from the combined exposure data. The same argument was used for selection of data from table 6 column 1 in Krewski (2009). In the case of Enstrom (2005) results from the first of the two follow-up periods have been selected because there were no more recent air pollution data to the later of the follow-up periods in which the population were at a very high age compared with other studies. Page 18 of 76

19 Table 7. Relative risk estimates of specific-cause mortality associated with a 10 μg/m 3 increase in the yearly average particle concentration. Data from major long-term studies. Reference Agegroup Death cause RR: PM 2.5 RR: SO 4-2 Dockery cardiopulmonary a 1.2 lung cancer b 1.2 NS Laden cardiovascular c 1.16 Krewski respiratory d 1.28 NS 25 lung cancer b 1.27 NS Pope cardiopulmonary e ** 30 lung cancer b ** Krewski cardiopulmonary e ischemic heart disease f lung cancer b NA Jerrett ischemic heart disease f cardiopulmonary g 1.12 NS 30 lung cancer b 1.44 NS Abbey 1999 males 27 lung cancer b 1.98* a International Classification of Disease, 9th edition (ICD-9) codes and ; b ICD-9: 162; c ICD-9: ; d ICD-9: ; e ICD-9: and ; f ICD-9: ; g ICD-9: and * in a non-smoking population. ** number not in text but read from figure 5. NS: not statistically significant. NA: not available How has particles been treated in recent modelling projects, cost-benefit analyses, reviews and reports? Side 19 af 76

20 World Health Organization The WHO has dealt with health effects of air pollution in several publications. When searching for health effects of exposure to air pollution publication on the WHO homepage the suggested publications are the (WHO 2001) and (WHO 2003) Working Group Reports. Several of the national reports and costbenefit analyses reviewed in the following refer to the former. No estimates of concentration-response functions for health effects are given in the publications, though, as they deal with methodology issues alone. The estimates suggested by WHO in their 2005 air quality guidelines (WHO 2005) are provided in table 8. It is purely a list on which estimates are scientifically sound and not a precise recommendation for which ones to use in cost-benefit analyses. In their 2004 report for WHO on European data, Anderson et al. (2004) suggested the use of the estimates of health effects of short-term PM exposures that are listed in table 9. Table 8. WHO air quality guidelines (WHO 2005). Table 9. Relative risk estimates of short-term health effects of PM suggested by WHO in (Anderson et al. 2004). Health outcome Agegroup RR: PM 2.5 RR: PM 10 RR: PM 10 revised Page 20 of 76

21 All-cause mortality all NA Respiratory mortality all NA Cardiovascular mortality all NA Respiratory hospital admissions 65+ NA Revised: revised for possible publication bias. NA: not enough European data available for calculation. OECD The Organisation for Economic Co-operation and Development published a report in 2007 on health costs related to air pollution (OECD 2007). In this they used the estimates developed by CAFE (AEA Technology Environment, 2005). See below for details on these. Europe: CAFE AEA Technology Environment was appointed by the European Commission DG Environment for a project of cost-benefit analysis of health effects of air pollution and effects of policy changes. The result was the so-called CAFE report (the Clean Air for Europe Programme) (Watkiss et al. 2005). Details of the methodology including responses to reviewers can be found in the publications by Holland et al. (2005a, 2005b) and Hurley et al. (2005a, 2005b). As did COMEAP, the CAFE report derived their CRF for adult premature mortality due to PM from Pope et al. (2002). The infant mortality was quantified by use of the CRF derived from a cohort study by Woodruff et al. (1997). The chronic bronchitis function was quantified based on the ExternE methodology (Extern E, 1999). CRFs for cardiovascular and respiratory hospital admissions were obtained from the Air Pollution and Health: A European Information System APHEIS project (APHEIS 2004). CRFs for restricted activity days, minor restricted activity days and work loss days due to PM were derived from studies by Ostro and Rothschild (Ostro 1987; Ostro and Rothschild 1989). Medication use by children and adults with asthma was included from a WHO meta-analysis (Anderson et al., 2004) despite the fact that they were not statistically significant. The CRF for PM related lower respiratory symptoms, including cough in children, was derived from a systematic review of panel studies of effects of PM on lower respiratory symptoms and on cough in children (Ward and Ayres, 2004). The CRF for PM related lower respiratory symptoms in adults was derived from a random effects meta-analysis of 4 panel studies conducted in Europe. Estimates are provided in table 10. Table 10. Relative risk estimates of health effects of PM suggested by CAFE in (AEA Technology Environment, 2005). Estimates per 10 μg/m 3 PM. Side 21 af 76

22 Health outcome Agegroup PM % CI PM 10 95% CI All-cause mortality 6% 2 11% Infant mortality 4% 2 7% Chronic bronchitis 7% % Respiratory hospital admissions 1.14% % Cardiac hospital admissions 0.6% % Restricted activity days 4.75% % Minor restricted activity days 7.4% 6 8.8% Work loss days 4.6% % Lower respiratory symptoms children 4% 2 6% adults 1.7% % Revised: revised for possible publication bias. NA: not enough European data available for calculation. Europe: EcoSense in ExternE Under the European Commission Directorate-General for Research Sustainable Energy Systems the ExternE Externalities of Energy Methodology 2005 Update was published following up the previous ExternE report from This methodology has formed the basis of several later evaluations. The particle related CRFs (sometimes called exposure-response functions in the report) are described in the following. Although not very appropriate the terms acute mortality and chronic mortality are used to refer to the effect of short-term and long-term exposures. The use of this wording infers a risk of valuating acute deaths differently from chronic deaths and assuming differences in loss of life years between the two. Due to the inclusion of all acute deaths in the mortality rates of long-term studies the ExternE report, however, suggests use of ony the long-term effect estimates. Regarding the use of ICD codes they appear to correspond to ICD 9th revision. Chronic CRF = 5% per 10 μg/m 3 increment of PM 2.5. Based on the two estimates in Pope et al. (2002) and without CI. All-cause infant mortality CRF = 4% per 10 μg/m 3 PM 10 (95% CI 2% - 7%). From Woodruff et al. (1997). New cases of chronic bronchitis was estimated but included 1.0 in 95% CI and will not be referred here. Annual rate of attributable emergency respiratory hospital admissions (ICD ) = 7.03 (95% CI ) per 10 μg/m 3 PM 10 per 100,000 people (all ages). Based on Hurley et al. (2005) using all-ages data, derived from APHEIS-3 (Medina et al. 2005). Page 22 of 76

23 Annual rate of attributable emergency cardiac hospital admissions (ICD ) = 4.34 (95% CI ) per 10 μg/m 3 PM 10 per 100,000 people (all ages). Based on CAFE-NEEDS quantification of an effect of PM 10 on cardiac admissions, using a RR based on APHEA-2 results (Le Tetre et al. 2002) and a Europewide annual rate of emergency cardiac admissions estimated as the arithmetic mean of rates from eight European cities derived from the Appendices of the APHEIS-3 report (Medina et al., 2005). Restricted activity days were based on Ostro (1987) among adults aged A weighted mean coefficient for restricted activity days was linked to estimated background rates of, on average, 19 RADs per person per year to give an estimate of: Change of 902 restricted activity days (95% CI ) per 10 μg/m 3 PM 2.5 per 1,000 adults at age Hurley et al. (2005) derived estimates for work loss days from Ostro (1987) and minor restricted activity days from Ostro and Rothschild (1989) to give, respectively, a change of 207 work loss days (95% CI ) per 10μg/m 3 PM 2.5 per year per 1000 people aged in the general population and a change of 577 minor restricted activity days (95% CI ) per 10μg/m 3 PM 2.5 per year per 1000 adults aged The CRF for medication (bronchodilator) usage by people with asthma based were found not to be statistically significant. A random effects meta-analysis of results from five panels was linked to both estimates of the mean daily prevalences of lower respiratory symptoms, including cough, in symptomatic panels, and estimates of the percentage of people qualifying for such panels, using data from ECRHS (1996), to give an estimated CRF for lower respiratory symptoms, including cough, in adults with chronic respiratory disease: Annual increase of 1.30 (95% CI ) symptom days per 10 μg/m 3 PM 10 per adult with chronic respiratory symptoms (approximately 30% of the adult population). Among children, the CRF for lower respiratory symptoms, including cough in children in the general population, was based on a review by Ward and Ayres (2004) suggesting that effects of PM on respiratory symptoms should be quantified for children generally and should not be confined to children with chronic symptoms. Hurley et al. (2005) combined RRs from Ward and Ayres (2004) with an estimate of the mean daily prevalence of lower respiratory symptoms, including cough, based on two Dutch studies of children to give the CRF estimate of: Annual increase of 1.86 (95% CI ) extra symptoms days per year per child aged 5-14, per 10 μg/m 3 PM 10 (not statistically significant). A selection of the ExternE CRF estimates are also provided in table 11 for clarity. Table 11. Relative risk estimates of health effects of PM suggested by ExternE (2005). Estimates per 10 μg/m 3 PM. Side 23 af 76

24 Health outcome Agegroup ICD-9 PM 2.5 (95% CI) PM 10 (95% CI) All-cause mortality (LT) 1.05 Infant mortality (LT) 1.04 ( ) Respiratory hospital admissions (ST?) *( ) Cardiac hospital admissions (ST?) *( ) Lower respiratory symptoms adults** 1.30 ( ) ( ) ST: Short-term (daily average concentration). LT: long-term (yearly average concentration). * No./year/100,000 persons. ** among adults with chronic respiratory symptoms. Europe: APHEIS In the Air Pollution and Health : A European Information System 3rd year project report (APHEIS 2004) a number of short-term, intermediate-term and long-term CRF were used for different PM indicators. Only those functions considered relevant for the CEEH chain are described any further here. Because the CRFs (referred to as exposure-response functions in the report) used are from publications that used gravimetric methods (Künzli et al and Pope et al. 2002), for consistency APHEIS corrected the automatic PM 10 measurements used by most of the cities in the system by a specific correction factor in order to compensate for losses of volatile particulate matter. This problem does not necessarily apply to modelled PM concentrations as in CEEH because they may be calculated corresponding to the yield of gravimetric measurements. CRF estimates from APHEIS (2004) are provided in table 12. Short-term effects were derived by APHEIS themselves and long-term effects derived from Pope et al. (2002). Table 12. Relative risk estimates of health effects of PM suggested by APHEIS in (2004). Estimates per 10 μg/m 3 PM. Health outcome Agegroup ICD-9 ICD-10 PM 2.5 (95% CI) All-cause mortality (LT) > A00-Y ( ) Cardiopulmonary mortality (LT) > I10-I70 + J00- J ( ) Lung cancer (LT) > C33-C ( ) Page 24 of 76

25 Respiratory hospital admissions (ST) all ages J00-J ( ) Cardiac hospital admissions (ST) all ages I00-I ( ) ST: Short-term (daily average concentration). LT: long-term (yearly average concentration). ICD: International Classification of Disease Europe: NEEDS New Energy Externalities Developments for Sustainability was a project under the EU Sixth Framework Programme. In 2007 the project submitted Deliverable RS1b/WP3 A set of concentration-response functions (NEEDS 2007). The NEEDS project draws heavily on, and in turn extends, work previously carried out for the Commission under the various projects on the ExternE programme (p. 22). Authors of the NEEDS publication were also involved in the ExternE updates (ExternE 1999) and (ExternE 2005) in collaboration with researchers that also developed the report for WHO (WHO 2001) and that of the US EPA (2004). It was clear, however,..., that a third comprehensive review was necessary, with the twin aims of (i) updating the ExternE quantification framework for air pollution and health in the light of new evidence and understanding; and (ii) aligning the ExternE / NEEDS recommendations more fully with those of the World Health Organisation (WHO). Fortunately, the health impact assessment (HIA) for the cost-benefit analysis (CBA) of the Commission s major Clean Air for Europe (CAFE) programme had similar requirements; and the CAFE CBA team was very strongly linked with ExternE, including having one of us... leading on developing the methodology for HIA within CAFE CBA. This allowed joint effort on the two projects, i.e. NEEDS and CAFE CBA,... p. 25. The NEEDS project is thus a thorough and recent review extending previous work in the EU, USA and WHO. A list of CRFs for PM from NEEDS is given in table 13. In addition to PM NEEDS addressed the CRFs of a number of additional pollutants. Regarding mortality related to long-term exposure they strongly suggest the use of the all-cause CRF for a number of reasons including the fact that CRFs for specific causes of death are less reliable (and less universally transferable between countries). From a table on p. 54 in the report it appears that cause-specific mortality rates underestimate the effect by approximately 14% compared with all-cause mortality (using the same rates as APHEIS; table 9). Side 25 af 76

26 Table 13. Relative risk estimates of health effects of PM suggested by NEEDS in Regarding sulfates and nitrates their effects are quantified insofar as they contribute to PM (in μg/m 3 ) as are all other constituents. This is a change from the most recent ExternE (2005) position. The position above is the position of CAFE CBA, WHO, US EPA and COMEAP... NEEDS (2007) p. 67. For sensitivity Page 26 of 76

27 analyses the report suggests treating primary particles at 1.3 times the toxicity of the PM 2.5 mixture and secondary particles are at 0.7 times the toxicity of PM 2.5. DEFRA & COMEAP, UK In the 2006 report from the Department of Environment, Food and Rural Affairs in the United Kingdom (UK) in partnership with the Scottish Executive, the National Assembly for Wales and the Department of Environment for Northern Ireland (DEFRA 2006) the following pollutants and outcomes were included: Particulate matter: Chronic mortality, acute mortality, and respiratory and cardiovascular hospital admissions. Sulphur dioxide: Acute mortality and respiratory hospital admissions. Ozone: Acute mortality and all respiratory hospital admissions. Nitrogen dioxide: All respiratory hospital admissions (only for sensitivity analysis). To our knowledge the DEFRA report is the only report that includes both chronic and acute mortality in the estimates of health effects - an approach that carry an increased and difficult to estimate risk of double counting. The health effects were based on reports published by the Committee on the Medical Effects of Air Pollutants (COMEAP). COMEAP has changed its recommendations with time. In the recent update from 2009 they recommend using a CRF for all-cause mortality from long-term exposure ( chronic mortality ) of 6% (95% confidence interval: 2% to 11%) per 10 μg/m 3 PM 2.5 derived from the (Pope et al. 2002) study. They further recommend using the plausible low and high values of 1% and 12% and further to include the wider interval of 0 to 15% (relative risk 1.00 and 1.15) in any report on quantification of risks from long-term exposure to particulate air pollution represented by PM 2.5. For cardiopulmonary mortality they propose use of the estimate 1.09 with 95% confidence interval (CI) and for lung cancer mortality they propose an estimate 1.08 with 95% confidence interval (CI) Acute mortality estimates were derived from (COMEAP 1998) (no longer available on the website or in print - access tried July 2nd 2010). Acute mortality is not addressed in the COMEAP 2009 report but is currently under review for a future report. Respiratory and cardiovascular hospital admissions were derived from (COMEAP 1995)(also no longer available). Cardiovascular admissions, however, have been reviewed in (COMEAP 2006). The effect on cardiovascular hospital admission was estimated to be 0.3% (95% CI -0.4% to 0.9%) for a change of 10 μg/m 3 24 h for PM 10 average. For cardiac admissions alone (ICD-9: ; excluding cerebrovascular) the estimate was 0.9% (95% CI 0.7% to 1.0%). The report also presented separate estimates for subgroups of cardiac admissions: 0.8% (95% CI 0.5% to 1.0%) for ischemic heart disease (IHD) (ICD-9: /414); 0.8% (95% CI 0.1% to 1.4%) for dysrhythmia (ICD-9: 427); 1.4% (95% CI 0.5% to 2.4%) for heart failure (ICD-9: 428). Side 27 af 76