UBC Faculty Research and Publications

An Integrated Risk Function for Estimating the Global Burden of Disease Attributable to Ambient Fine… Burnett, Richard T.; Pope III, C. Arden; Ezzati, Majid; Olives, Casey; Lim, Stephen S.; Mehta, Sumi; Shin, Hwashin H.; Singh, Gitanjali; Hubbell, Bryan; Brauer, Michael; Anderson, H. Ross; Smith, Kirk R.; Balmes, John R.; Bruce, Nigel G.; Kan, Haidong; Laden, Francine; Prüss-Ustün, Annette; Turner, Michelle C.; Gapstur, Susan M.; Diver, W. Ryan; Cohen, Aaron Apr 30, 2014

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-Burnett_R_et_al_Integrated_risk_function.pdf [ 520.39kB ]
JSON: 52383-1.0220726.json
JSON-LD: 52383-1.0220726-ld.json
RDF/XML (Pretty): 52383-1.0220726-rdf.xml
RDF/JSON: 52383-1.0220726-rdf.json
Turtle: 52383-1.0220726-turtle.txt
N-Triples: 52383-1.0220726-rdf-ntriples.txt
Original Record: 52383-1.0220726-source.json
Full Text

Full Text

Environmental Health Perspectives • volume 122 | number 4 | April 2014 397ResearchAll EHP content is accessible to individuals with disabilities. A fully accessible (Section 508–compliant) HTML version of this article is available at http://dx.doi.org/10.1289/ehp.1307049. IntroductionLong-term exposure to ambient fine par-ticulate matter (≤ 2.5 μg/m3 in aerody-namic diameter; PM2.5) is associated with increased mortality from non accidental and cause-specific diseases (Brook et al. 2010; Committee on the Medical Effects of Air Pollutants 2009; Cooke et al. 2007; Krewski et al. 2009). Epidemiologic cohort studies, conducted largely in the United States, have reported this association for annual ambient average concentrations from approximately 5 to 30 μg/m3, although definitive knowledge of which specific sources or characteristics of PM2.5 are responsible for these associations is currently lacking [U.S. Environmental Protection Agency (EPA) 2009; World Heal th Organizat ion (WHO) 2006, 2007]. No epidemiologic study, however, has estimated the association of long-term exposure to direct measurements of PM2.5 with mortality from chronic cardio vascular and respiratory disease at the higher ambi-ent exposures common in cities and other areas in Asia and other developing countries where annual average exposures can exceed 100 μg/m3 (Brauer et al. 2012; Health Effects Institute 2010). As a result, estimates of dis-ease burden attributable to ambient air pol-lution in these locations have had to be based on extrapolations of the results of epidemio-logic studies from locations with lower ambi-ent PM2.5 exposures (Anenberg et al. 2010; Cohen et al. 2004; Evans et al. 2013).Previous efforts to estimate global bur-den from exposure to ambient air pollution (AAP) in the form of PM2.5 postulated risk Address correspondence to R.T. Burnett, Population Studies Division, Environmental Health Sciences and Research Bureau, Environmental and Radiation Health Sciences Directorate, Healthy Environment and Consumer Safety Branch, Health Canada, Room 134, Environmental Health Center, 50 Columbine Driveway, Ottawa, Ontario, Canada K1A 0K9. Telephone: (613) 952-1364. E-mail: rick.burnett@hc-sc.gc.caSupplemental Material is available online (http://dx.doi.org/10.1289/ehp.1307049).We thank the members of the Ambient Air Pollution Expert Group of the Global Burden of Diseases, Injuries, and Risk Factors Study for their helpful comments and discussion of this material and M. Lipsett for the relative risk estimate for chronic obstructive pulmonary disease mortality from the California Teachers Study. A.C. received support from the William and Flora Hewlett Foundation.A.C. is employed by the Health Effects Institute (HEI). The HEI receives about half of its core funds from the U.S. Environmental Protection Agency and half from the worldwide motor vehicle industry, although other public and private organizations peri-odically support special projects or certain research programs. The views expressed in this article do not represent those of the HEI or its sponsors. A.P.-U. is a staff member of the World Health Organization (WHO), and the views expressed in this article do not necessarily represent the views, decisions, or policies of the WHO. This article should not be reproduced for use in association with the promo-tion of commercial products, services, or any legal entity. The WHO does not endorse any specific organization or products. Any reproduction of this article cannot include the use of the WHO logo.The authors declare they have no actual or potential competing financial interests.Received: 6 May 2013; Accepted: 7 February 2014; Advance Publication: 11 February 2014; Final Publication: 1 April 2014.An Integrated Risk Function for Estimating the Global Burden of Disease Attributable to Ambient Fine Particulate Matter ExposureRichard T. Burnett,1 C. Arden Pope III,2 Majid Ezzati,3 Casey Olives,4 Stephen S. Lim,5 Sumi Mehta,6 Hwashin H. Shin,1 Gitanjali Singh,7 Bryan Hubbell,8 Michael Brauer,9 H. Ross Anderson,10 Kirk R. Smith,11 John R. Balmes,12,13 Nigel G. Bruce,14 Haidong Kan,15 Francine Laden,16 Annette Prüss-Ustün,17 Michelle C. Turner,18 Susan M. Gapstur,19 W. Ryan Diver,19 and Aaron Cohen20*1Health Canada, Ottawa, Ontario, Canada; 2Brigham Young University, Provo, Utah, USA; 3MRC-PHE Centre for Environment and Health, School of Public Health, Imperial College London, London, UK; 4School of Public Health, University of Washington, Seattle, Washington, USA; 5Institute for Health Metrics and Evaluation, Seattle, Washington, USA; 6Global Alliance for Clean Cookstoves, Washington, DC, USA; 7Harvard School of Public Health, Harvard University, Cambridge, Massachusetts, USA; 8U.S. Environmental Protection Agency, Research Triangle Park, North Carolina, USA; 9School of Population and Public Health, University of British Colombia, Vancouver, British Columbia, Canada; 10MRC-PHE Centre for Environment and Health, King’s College London, London, UK; 11University of California, Berkeley, Berkeley, California, USA; 12School of Medicine, University of California, San Francisco, San Francisco, California, USA; 13School of Medicine, University of California, Berkeley, Berkeley, California, USA; 14Department of Public Health and Policy, University of Liverpool, Liverpool, UK; 15School of Public Health, Fudan University, Shanghai, China; 16Exposure, Epidemiology, and Risk Program, Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts, USA; 17World Health Organization, Geneva, Switzerland; 18Institute of Population Health, University of Ottawa, Ottawa, Ontario, Canada; 19American Cancer Society, Atlanta, Georgia, USA; 20Health Effects Institute, Boston, Massachusetts, USA *Senior AuthorBackground: Estimating the burden of disease attributable to long-term exposure to fine  particulate matter (PM2.5) in ambient air requires knowledge of both the shape and magnitude of the relative risk (RR) function. However, adequate direct evidence to identify the shape of the mortality RR functions at the high ambient concentrations observed in many places in the world is lacking.oBjective: We developed RR functions over the entire global exposure range for causes of mortal-ity in adults: ischemic heart disease (IHD), cerebrovascular disease (stroke), chronic obstructive pul-monary disease (COPD), and lung cancer (LC). We also developed RR functions for the incidence of acute lower respiratory infection (ALRI) that can be used to estimate mortality and lost-years of healthy life in children < 5 years of age.Methods: We fit an integrated exposure–response (IER) model by integrating available RR infor-mation from studies of ambient air pollution (AAP), second hand tobacco smoke, household solid cooking fuel, and active smoking (AS). AS exposures were converted to estimated annual PM2.5 exposure equivalents using inhaled doses of particle mass. We derived population attributable frac-tions (PAFs) for every country based on estimated worldwide ambient PM2.5 concentrations.results: The IER model was a superior predictor of RR compared with seven other forms previ-ously used in burden assessments. The percent PAF attributable to AAP exposure varied among countries from 2 to 41 for IHD, 1 to 43 for stroke, < 1 to 21 for COPD, < 1 to 25 for LC, and < 1 to 38 for ALRI.conclusions: We developed a fine particulate mass–based RR model that covered the global range of exposure by integrating RR information from different combustion types that generate emissions of particulate matter. The model can be updated as new RR information becomes available.citation: Burnett RT, Pope CA III, Ezzati M, Olives C, Lim SS, Mehta S, Shin HH, Singh G, Hubbell B, Brauer M, Anderson HR, Smith KR, Balmes JR, Bruce NG, Kan H, Laden F, Prüss-Ustün A, Turner MC, Gapstur SM, Diver WR, Cohen A. 2014. An integrated risk function for estimating the global burden of disease attributable to ambient fine particulate matter exposure. Environ Health Perspect 122:397–403; http://dx.doi.org/10.1289/ehp.1307049Burnett et al.398 volume 122 | number 4 | April 2014 • Environmental Health Perspectivesfunctions for cardiopulmonary mortality as linearly increasing in relative risk (RR) from 7.5 to 50 μg/m3, with no further change in RR at higher concentrations (Cohen et al. 2004). Sensitivity analyses included a model in which RR varied as the logarithm of concentration, producing a more gradual diminution of the marginal increase in RR than the base case model. The logarithmic model was subsequently recommended by the WHO for use in air pollution burden of disease estimates at the national level (Ostro 2004). The coefficients of these models were based on information from a single U.S. cohort study—the American Cancer Society Cancer Prevention Study II (CPS-II) (Krewski et al. 2009; Pope et al. 2002)—with exposure assignments of < 22 μg/m3. The form of the models used for global bur-den assessment was motivated largely by the concern that linear extrapolation using these coefficients would produce unrealisti-cally large estimates of RR compared with other known PM2.5-related mortality risks such as active smoking (AS) and exposure to secondhand tobacco smoke (SHS) (Cohen et al. 2004; Ostro 2004). These RR models were also employed in more recent estimates of global mortality associated with ambient PM2.5 concentrations (Anenberg et al. 2010; Evans et al. 2013).Absent empirical epidemiologic evidence on the magnitude of the association with mortality at high exposures of PM2.5 in ambi-ent environments, Pope et al. (2011b) sug-gested that the integration of epidemiologic evidence on cardiovascular and lung cancer (LC) mortality RR from disparate types of PM2.5 exposure such as AAP, SHS, and AS, may provide insight into the shape of the exposure–response relation over a much wider range of exposures.Here we present the methodology used to estimate the population attributable fraction (PAF) from exposure to ambient PM2.5 in the Global Burden of Diseases, Injuries, and Risk Factors Study 2010 (the GBD 2010 project) (Lim et al. 2012). We selected a mathematical form of the RR function with a PM2.5 con-centration that could describe the observed relationships between RR and exposure for the five outcomes examined. We fit this model for cause-specific adult mortality for four causes of death—ischemic heart disease (IHD), stroke, chronic obstructive pulmo-nary disease (COPD), and LC—using RR information from epidemiologic studies of long-term exposure to particulate matter not only from AAP, SHS, and AS, but also from studies of household air pollution from solid cookfuel (household air pollution; HAP). We used these models to estimate the percentage of PAF associated with exposure to ambient PM2.5 for each of the 187 countries included in the GBD 2010 project. We identified a specific model form that best predicts source-specific RR for all four causes of death. In addition, we examined the relationship between PM2.5 exposure and the incidence of acute lower respiratory infection (ALRI) in infants, another health outcome considered in the GBD 2010 project. Because infants and young children are non(active)-smokers, the largest PM2.5 exposures considered for ALRI are from HAP.MethodsUnderlying assumptions. The model we propose in here was based on the following underlying assumptions:• Exposure to PM2.5 from diverse combustion sources is associated with increased mor-tality from IHD, stroke, COPD, and LC and with increased incidence of ALRI. This assumption is based on systematic review of the available epidemiologic literature conducted by the GBD 2010 Ambient Air Pollution Expert Group as part of the GBD 2010 project (Lim et al. 2012).• The observed RRs from AAP, SHS, HAP, and AS are a function of PM2.5 mass inhaled concentration across all combus-tion particle sources (Smith 1987). The toxicity of PM2.5 is assumed to differ only with regard to inhaled mass (exposure) and not with PM2.5 composition. The toxicity of emissions from different combustion sources may well differ, but current knowl-edge does not allow definitive and quanti-fiable conclusions regarding their relative toxicity and little is known about inter-national variation in source contributions around the world (Stanek et al. 2011; U.S. EPA 2009; WHO 2006).• The relation between PM2.5 exposure and excess mortality RR is not necessarily restricted to a linear function over the range of human exposure to PM2.5 from diverse sources (Pope et al. 2009, 2011b).• The RR of mortality from chronic disease experienced by people exposed to AAP, SHS, HAP, and AS is a function of long-term, cumulative exposure quantified in terms of daily average exposure concentra-tion and does not depend on the tempo-ral pattern of exposure (Pope et al. 2011a, 2011b). This assumption is required because the temporal nature of PM2.5 exposure dif-fers for AAP, SHS, HAP, and AS.• The RR associated with each type of expo-sure does not depend on the other types of exposure. That is, we are assuming no interaction among the different exposure types for any cause of mortality. We are aware of no empirical epidemiologic evi-dence that tests that assumption; however, the direct epidemiologic evidence from the cohort studies we used to estimate the burden attributable to ambient PM2.5 shows that active cigarette smokers are also affected adversely by exposure to ambient PM2.5, and these studies do not provide support for sig-nificant heterogeneity of the relative excess AAP RR across smoking categories.Model form. We selected a mathemati-cal form of an integrated exposure–response (IER) model that could describe several pat-terns in RR thought to be a priori applicable to exposure–response models. We wanted the IER to be able to take shapes similar to models previously used for burden assessment, such as linear and log-linear (Cohen et al. 2004) and a power function (Pope et al. 2009, 2011b). In addition to these shapes, we also required the IER to have a property that it flattens out at high exposures, consistent with evidence of the relationship between IHD mortality and smoking intensity (Pope et al. 2009).The form must equal 1 when PM2.5 values are below some concentration that rep-resents a counterfactual low exposure where below this level there is no excess risk. We also desired a model that increases mono-tonically with increasing PM2.5 exposure con-centration and could take a variety of shapes, such as near linear, sub linear, and supra linear. Our IER model has the following form:for z < zcf ,  RRIER(z) = 1for z ≥ zcf ,  RRIER(z) = 1 + α {1 – exp[– γ (z – zcf)δ]}, [1]where z is the exposure to PM2.5 in micro-grams per meter cubed and zcf is the counter-factual concentration below which we assumed there is no additional risk. For very large z, RRIER approximates 1 + α. We included a power of PM2.5, δ, to predict risk over a very large range of concentrations. Further, RRIER (zcf + 1) approximates 1 + αγ. Thus, γ = [RRIER (zcf + 1) – 1]/[RRIER (∞) – 1] can be inter-preted as the ratio of the RR at low-to-high exposures. We term our model an “integrated-exposure response” model because its develop-ment requires the integration of exposures to PM2.5 from different combustion types (i.e., AAP, SHS, HAP, and AS).In formulating our RR model, we relied on information on the RR of mortality at specified PM2.5 exposure concentrations from the available literature. Suppose we have a set of RR estimates { r^1(s),…,r^Ks(s), s = 1,…,S} and corresponding confidence intervals (CIs) based on PM2.5 concentrations {z1(s),…,zKs(s), s = 1,…,S}, for S different types of PM2.5 sources, where Ks is the number of RR esti-mates available from for source type S. The unknown parameters (α, γ, δ) are estimated by non linear regression methods. We then weighted the RR estimates by the inverse of Fine particulate mass–based risk modelEnvironmental Health Perspectives • volume 122 | number 4 | April 2014 399the variance estimate of the logarithm of the RR in order to reflect the uncertainties in each estimate.We compared the IER model  to seven other models that have been previ-ously suggested for burden assessment. These include an RR model that is linear in exposure throughout the global concentra-tion range (Lin), a model that is linear up to 30 μg/m3 and constant above 30 μg/m3 (Lin30), a model that is linear up to 50 μg/m3 and constant above 50 μg/m3 (Lin50), and a model that is a function of the logarithm of exposure (Log). These models were used in a previous assessment of global burden of disease due to AAP exposure (Cohen et al. 2004). We also postulated a model in which we added an unknown parameter to concen-tration in the Log model to allow more flex-ibility in fitting the type-specific RRs (Log2). The sixth model examined related RR to a power of exposure as proposed by Pope et al. (2009, 2011b), with the seventh model equivalent to the IER with δ = 1(Exp). For the mathematical forms of the models, see the Supplemental Material (Sensitivity of RRs and PAFs to Model Form, pp. 9–11). We then calculated both the Akaike and Bayesian information criteria (AIC, BIC) for each of the eight models examined and five health  outcomes as  measures of goodness of fit.The method of constructing uncertainty bounds on model predictions is described in detail in the Supplemental Material (Characterizing Uncertainty, pp. 28–29). Briefly, we simulated 1,000 sets of source type–specific RRs based on their point esti-mates and CIs and fit the IER model to these simulated values, obtaining 1,000 sets of parameter estimates of (α,γ,δ). Using these parameter estimates, we then gener-ated 1,000 IER functions over the global concentration range. Estimates of uncertainty were also generated for the PM2.5 concentra-tions. Uncertainty in the PAFs is a function of the uncertainty in the IER model predic-tions and the exposure estimates and is deter-mined by simulation methods as described in the Supplemental Material (Characterizing Uncertainty, pp. 28–29).Specifics of the selection of source type–specific RR and PM2.5 exposure for each type are described below for the four mor-tality outcomes. The logarithm of the RR per micrograms per meter cubed, its SE, and associated PM2.5 concentration for the five outcomes is given by type of PM2.5 in Supplemental Material, Table S1.AAP. To fit the risk models, we used cause-specific mortality AAP RR estimates from available published cohort studies. We evaluated each RR estimate at its study-specific PM2.5 mean concentration minus a less-polluted counter factual level (Lim et al. 2012). Most RRs were obtained from published reports; however, in some cases new analyses were conducted for the pres-ent study. These estimates are identified in Supplemental Material, Table S1. We had eight studies reporting RR estimates for IHD mortality, five for stroke mortality, three for COPD mortality, and four for LC mortality.SHS. We selected RRs for both IHD (8 studies reporting separate estimates for males and females) and LC (46 studies) mortality from studies included in the U.S. Surgeon General’s Report, The Health Consequences of Involuntary Exposure to Tobacco Smoke (Office on Smoking and Health 2006). We associated the RR of death due to SHS exposure with an equivalent ambient PM2.5 concentration of 20 μg/m3 for low-to-moderate SHS exposure and 50 μg/m3 for moderate-to-high exposure based on the analysis of Pope et al. (2009) for IHD mortality because RRs were reported by the Office on Smoking and Health (2006) for these two descriptive exposure categories. We assigned a concentration of 35 μg/m3 based on the midpoint of the range 20–50 μg/m3 for LC mortality because no specific description of the level of SHS exposure was provided by the Office on Smoking and Health (2006). We selected 29 RRs from studies examined by Oono et al. (2011) for stroke mortality on the basis of prospective cohort studies with an associated PM2.5 concentration of 35 μg/m3. There was insufficient evidence to estimate a RR due to SHS exposure for COPD mortal-ity. We assumed that the SHS RRs are associ-ated with a change in PM2.5 exposure based on non smoking subjects living with a smoker compared with those not living with a smoker. We have not incorporated other potential sources of PM2.5 exposure for these study subjects, such as from indoor sources, near-roadway conditions, or occupational exposures by subject.AS. Following Pope et al. (2009, 2011b), we estimated the RR of each of the four causes of death for current cigarettes smoked per day compared with never smokers from the CPS-II. We estimated the RR and 95% CIs associated with 10 cigarettes-per-day groupings: 1–3, 4–7, 8–12, 13–17, 18–22, 23–27, 28–32, 33–37, 38–42, and > 42 cigarettes/day. We estimated that smoking a single cigarette was equivalent to breath-ing a daily ambient concentration of PM2.5 of 667 μg/m3, assuming an average breath-ing rate of 18 m3/day and an inhaled dose of 12,000 μg PM2.5 mass per cigarette (Pope et al. 2009). We then estimated the equiva-lent ambient concentration of PM2.5 by multi plying the average cigarettes/day smoked in each interval by 667 μg/m3. The shape of the curve fitted by Pope et al. (2009, 2011b) was not sensitive to the estimate of equivalent ambient PM2.5 concentrations for AS.HAP. Smith et al. (2014) conducted a meta-analysis of studies examining COPD and LC incidence rates among men and women exposed to air pollution from burn-ing coal or biomass for cooking. There were no studies relating IHD or stroke mortality or incidence to HAP at the time of the GBD 2010 project analyses, and thus this PM2.5 type cannot contribute to the fit of our RR function. The equivalent long-term PM2.5 exposure from HAP was estimated for study subjects using coal or biomass for cooking (Balakrishnan et al. 2013) and for those study subjects using cleaner fuels to integrate this information into our IER risk model. PM2.5 exposure estimates for women (300 μg/m3) were higher than for men (200 μg/m3). For the COPD meta-analysis, the relevant female control group was assumed to be using a mix-ture of gas and chimney stoves (an estimated PM2.5 exposure of 100 μg/m3). The PM2.5 exposure for males was estimated to be 65% of that for females (65 μg/m3). For LC, the female control group was assumed to be using only gas stoves with an estimated PM2.5 expo-sure of 70 μg/m3. For males, the exposure was again assumed to be 65% of females, resulting in an equivalent exposure of 45.5 μg/m3. The meta-analytic summary risk estimate for male COPD incidence in association with HAP PM2.5 was 1.90 (95% CI: 1.56, 2.32) and for females was 2.70 (95% CI: 1.95, 3.75). For LC incidence among males, the summary risk estimate was 1.26 (95% CI: 1.04, 1.52) and among females was 1.81 (95% CI: 1.07, 3.06).The lower exposure estimates in the HAP studies are substantially higher than counter-factual exposure due to the nearby use of less clean fuels; therefore, these RRs are not directly comparable to those obtained from AAP, SHS, or AS types compared with either the counterfactual (i.e., AAP) or a 0-μg/m3 exposure (i.e., SHS, AS). This information was included in the curve-fitting process by equating the observed RRs to the ratio of the IER model evaluated at the respective two PM2.5 concentrations.The HAP studies estimated effects on incidence rather than mortality. For building the IER, we assumed that the RRs of mortal-ity and incidence are equal.Age-modification risk models for IHD and stroke mortality. Epidemiologic studies of risk factors for both IHD and stroke indicate that the RR declines with the logarithm of age, reaching 1 between 100 and 120 years of age (Singh et al. 2013). We thus modified the type-specific RR for both IHD and stroke mortality using a linear regression model of the logarithm of the median age at death for each study with the intercept equal to 1 at 110 years of age. The slope of the regression line was estimated from a meta-analysis of several risk Burnett et al.400 volume 122 | number 4 | April 2014 • Environmental Health Perspectivesfactors (Singh et al. 2013). We applied this age-modification to the RRs and fit the IER model for each age group separately.Selecting the counterfactual exposure. For each risk factor examined in the GBD 2010 project (Lim et al. 2012), the distribution of exposure was compared with an alterna-tive (counterfactual) distribution termed the theoretical- minimum- risk exposure distribu-tion (TMRED). For AAP, zero exposure is not a practical counterfactual level because it is impossible to achieve even in pristine envi-ronments (Brauer et al. 2012). Furthermore, the lowest level of exposure to PM2.5 that is deemed beneficial has not been clearly identi-fied. Defining the TMRED was based on two criteria (Lim et al. 2012): a) the availability of convincing evidence from epidemiologic studies that support a continuous reduction in risk of disease to the chosen distribution, and b) a distribution that is theoretically possi ble at the population level.Lim et al. (2012) suggested that a positive counterfactual concentration be used. Their counterfactual concentration is bounded by the minimum concentrations observed in the studies used to estimate risk and some low percentile of the PM2.5 distribution. There is clearly no evidence of an association below observed levels, and it is impractical to esti-mate the shape of the curve at the extremes of the exposure distribution. Lim et al. (2012) suggested that the fifth percentile be used and that the lower and upper bounds on the counterfactual concentration be determined by the corresponding minimum and fifth percentiles, respectively, of the AAP PM2.5 exposure distribution for the CPS II cohort (Krewski et al. 2009), the largest cohort study of air pollution. The minimum was 5.8 μg/m3 and the fifth percentile was 8.8 μg/m3. Uncertainty in the counterfactual concentra-tion was modeled as a uniform distribution between the minimum and fifth percentile.Estimation of PAF. We estimated the PAF associated with ambient PM2.5 exposure for all 187 countries separately for 2005. We first estimated surface PM2.5 concentrations on a 0.1° × 0.1° grid for the globe using a com-bination of remote sensing and atmos pheric models calibrated to ground moni toring data (Brauer et al. 2012). For each grid cell within a given country, we estimated the RR based on the IER model at the estimated PM2.5 concentration. We then constructed a population-weighted average RR for each country using the corresponding population count 0.1° × 0.1° grid cell (Brauer et al. 2012). Both the gridded PM2.5 and population values can be obtained from Brauer et al. (2012). The country-specific PAF = 1 – 1/WRRIER, where WRRIER is the population- weighted average of the RRIER values at each PM2.5 grid cell within the country.IER model for ALRI. Mehta et al. (2013) reviewed the evidence for an association between exposure to ambient PM2.5 and ALRI. Four cohort studies were deemed appropriate to include in an IER model (Mehta et al. 2013). We included 23 studies of parental SHS and ALRI reported by the Office on Smoking and Health (2006) with each study-specific odds ratio (OR) assigned a PM2.5-equivalent ambient exposure of 50 μg/m3, assuming a moderate-to-high level of exposure. Smith et al. (2011) examined the relationship between exposure to carbon monoxide (CO) from the burning of solid biomass for heating and cooking and the inci-dence of ALRI in Guatemala and reported incidence rates by decile average of CO per-sonal exposures. These decile CO averages were converted to PM2.5 concentrations using the following equation: PM2.5(mg/m–3) = 0.10(0.093, 0.12) × CO(mg/m–3)  + 0.067 (0.0069, 0.13), [2]with 95% CIs displayed in parenthesis (Northcross et al. 2010). This equation had good predictive power (R2 = 0.76).Incidence rates, I(zi), corresponding to the 10 decile values of PM2.5, denoted by zi for 1 = 1,…10, can be compared with the risk model by taking the ratio of incidence rates for all unique pairs of PM2.5 deciles, a total of 45 pairs, and equating them to the ratio of the corresponding risk model evaluated at the appropriate decile average. That is,RRALRI(zi, zj) = I(zi)/I(zj)   = [1 + α{1 – exp[–γ (zi – zcf)δ]}]   ÷ [1 + α{1 – exp[–γ (zj – zcf)δ]}] [3]for all 45 unique pairs of concentrations (zi, zj), ∀i > j = 1,…10. The 45 incidence rate ratios were combined with the 4 AAP cohort study ORs and the 23 SHS ORs in order to fit the IER model for ALRI. We assumed the same counterfactual uncertainty distribution as with the mortality IER models.ResultsThe average of the RRIER predictions among the simulations are displayed for the four causes of death in Figure 1 in addition to the 95% CI and the type-specific RR estimates and cor-responding 95% CIs used to fit the curves. The HAP RRs for COPD and LC are presented for males and females in Figure 1 as pink-shaded boxes with the height of each box represent-ing the uncertainty in the RR estimates and the width representing the exposure contrast at which the RRs was assumed to pertain. Each box is centered at the RR estimate and the mid-point of the two exposure values. This alternate depiction of the HAP information was neces-sary because the lowest exposure levels were substantially higher than the counterfactual Figure 1. Predicted values of IER model (solid line) and 95% CIs (dashed line) and type-specific RRs (points) and 95% CIs (error bars) for IHD (A), stroke (B), COPD (C), and LC (D) mortality. Shaded boxes for COPD and LC mortality represent uncertainty (height) and exposure contrast (width) of RR HAP estimates for males (smaller boxes) and females (larger boxes) separately.Predicted and RRPredicted and RRPredicted and RRPredicted and RRLog PM2.5 (µg/m3) Log PM2.5 (µg/m3)Log PM2.5 (µg/m3) Log PM2.5 (µg/m3)IHD● ●●●●●●●●●Stroke●●●●●●●●●●COPD●●●●●●●●●●LC●●●●●●●●●●43215025101502510143215 10 100 300 1,000 10,000 55,0005025 5 10 100 300 1,000 10,000 55,00050255 10 100 300 1,000 10,000 55,0005025 5 10 100 300 1,000 10,000 55,0005025●AAPSHSASPredictedFine particulate mass–based risk modelEnvironmental Health Perspectives • volume 122 | number 4 | April 2014 401exposure and, therefore, not directly compa-rable to the RRs from the other sources. The pooled estimate of RR and its corresponding CI for SHS is displayed in placed of the study-specific SHS RRs for each unique PM2.5 value because the study-specific RRs and CI could not be visually distinguished. Results are pre-sented similarly for ALRI in Figure 2. In addi-tion to the RR, the incidence of ALRI is also displayed on the right-hand y-axis.The RRIER function fits well the RRs for all types of PM2.5 and causes of mortality, except for COPD and HAP, in which the IER model underestimates the observed RRs (Figure 1). This may be due to the use of the ratio of incidence rates rather than RR based on mortality data for this outcome. However, the IER curve fits the LC incidence data rea-sonably well. The time between diagnosis of COPD and mortality is much longer than that for LC, and thus the LC incidence data may better reflect mortality patterns than the COPD incidence data.We compared the country-specific esti-mated PAFs using the age-modified models to those models using age-independent data. Age-modified RRIER curves are displayed for IHD and stroke mortality in Supplemental Material, Figure S15 (top panels), with gener-ally decreasing risk with increasing age. The country-specific PAFs based on risk mod-els not modified by age and those in which age-modification models were used for both IHD and stroke mortality are presented in Supplemental Material, Figure S15 (bot-tom panels). Incorporating age-modification risk models tends to slightly decrease the PAF estimates.The distribution of population-weighted country-average PM2.5 concentrations and PAFs are displayed in Figure 3. The country average PM2.5 concentrations ranged from 2–70 μg/m3 for 2005 (Figure 3A), whereas the country-level PAFs were < 0.4 for ALRI, IHD, and stroke and < 0.25 for LC and 0.2 for COPD (Figure 3B).Plots similar to Figures 1 and 2 are displayed for the other seven model forms examined in Supplemental  Material , Figures S1–S14 for both the four causes of  death (Figures  S1–S7) and ALRI (Figures S8–S14). In addition, both the AICs and BICs are given in Supplemental Material, Table S2, for all eight models and five out-comes. The IER model was a better predictor of the type-specific RRs than the other seven  models examined for ALRI and three of the Figure 2. Predicted values of IER model (solid line) and 95% CIs (dashed line) and type-specific RRs (points) and 95% CIs (error bars) for ALRI in infants.5 10 25 50Log PM2.5 (µg/m3)100 300 600ALRI●● ●●Predicted and RRHAP incidence rate4. (µg/m3)DensityALRI COPD IHD LC Stroke05101520PAF (%)0 25 50 75 100 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40 0 10 20 30 40Figure 3. Density plots of country-specific, population-weighted PM2.5 concentrations (μg/m3) (A) and PAFs (B) by risk model and health outcome. Dashed lines represent smooth fit of density function.Burnett et al.402 volume 122 | number 4 | April 2014 • Environmental Health Perspectivesfour causes of death. For COPD mortality, the Power model provided a better fit than the IER model on the basis of lower AIC and BIC values (see Supplemental Material, Table S2). This was likely due to the better prediction of the HAP RR, for which the IER model clearly underestimated the RR. Graphical comparisons of the predicted val-ues to the type-specific RRs in Supplemental Material, Figures S1–S14, verify the conclu-sions drawn from the AIC/BIC results.DiscussionExposure to PM2.5 in ambient air has been linked to an increased risk of death from chronic cardiovascular and respiratory disease and LC in cohort studies in the United States and Europe (Chen et al. 2008; U.S. EPA 2009). Unfortunately, few long-term cohort studies have been reported for these diseases in other regions such as East and South Asia and the Middle East, where ambient expo-sures are much higher and where the relative contribution of specific sources of air pollu-tion differ from those in North America and Europe (Brauer et al. 2012; Heath Effects Institute 2010).To derive the shape of the exposure–response curve at higher ambient concentra-tions, we incorporated information on risk due to exposure to SHS, HAP, and AS in order to extend the risk estimates to higher exposures. The IER model combines informa-tion on mortality RR from separate types of combustion, unified by equating the deliv-ered dose from all types in terms of equiva-lent ambient PM2.5 exposures. Although we assumed that the toxicity of PM2.5 exposure, as characterized by RR, changes with the magnitude of exposure, we also assumed that at any fixed exposure level, toxicity is roughly equivalent among all types and temporal pat-terns of PM2.5 exposure. These are impor-tant assumptions because estimated PM2.5 exposure throughout the world, whether from ambient origin or household indoor combus-tion, has not been differentiated by the com-ponents or sources of fine particulate matter.Only evidence from multiple epidemio-logic studies of long-term exposure to PM2.5 in highly polluted settings can pro-vide definitive estimates of the shape of the exposure– response function for mortal-ity from chronic cardiovascular and respi-ratory diseases. However, these are starting to appear. For example, Cao et al. (2011) reported an increased risk of mortality from cardio vascular and respiratory disease and LC associated with long-term exposure to total suspended particulates (TSPs) in 71,000 resi-dents of 31 Chinese cities. Their study offers an opportunity to assess the ability of our RRIER model to estimate the observed RRs in situations with very high levels of outdoor air pollution. In order to estimate PM2.5 RRs in the cohort, the authors used a 3:1 ratio to convert TSP to PM2.5, based on current and historical Chinese data (Cao et al. 2011). Estimated PM2.5 (converted from TSP) concen trations ranged among cities from 38 to 166 μg/m3. Increases of 2.1% (95% CI: –0.3%, 4.6%), 3.3% (95% CI: 0.9%, 5.4%), and 3.3% (95% CI: –0.3%, 6.9%) in IHD, stroke, and LC mortality, respectively, were associated with a 10-μg/m3 change in esti-mated equivalent PM2.5 exposures in this cohort (Kan H, personal communication).Because the cohort members did not experience exposures near the lowest con-centrations applicable to our RR model (i.e., the counterfactual concentration), we cannot determine RRs estimated from the cohort and directly compare them to our RR model, which is relative to a much lower counter factual concentration. However, we can determine RR between concentra-tions observed in the cohort itself. We first determined the mean of the four quartiles of PM2.5 concentrations as 40, 91, 106, and 127 μg/m3, respectively (Kan H, per-sonal communication) and calculated the RR between consecutive quartile averages assuming the exponential risk model form as was used by the study authors. The geo-metric average of these three RRs was then determined as a summary measure of change in risk over the PM2.5 exposure distribu-tion. A similar calculation was undertaken for the RRIER model. The RRs observed in the Chinese cohort and those predicted by RRIER were similar for the three causes of death examined [IHD: China RR = 1.06 (95% CI: 0.99, 1.14) and IER RR = 1.05 (95% CI: 1.03, 1.1); stroke: China RR = 1.10 (95% CI: 1.03, 1.17) and IER RR = 1.08 (95% CI: 1.01, 1.14; LC: China RR = 1.10 (95% CI: 0.99, 1.22) and IER RR = 1.09 (95% CI: 1.06, 1.12)], suggesting that our IER model yielded reasonable predictions in the change in risk over a range of concentra-tions that prevail in China and other highly polluted settings that were not observed in cohort studies conducted in North America and Western Europe.There are, however, some limitations in this comparison. First, TSP was a poorer pre-dictor of cardiovascular mortality than PM2.5 in U.S.-based cohort studies (Pope et al. 2002). Second, uncertainty about the tempo-ral and spatial consistency of the TSP/PM2.5 conversion ratio of 3:1 added uncertainty to our interpretation of the results from the Chinese cohort.Additional uncertainties are due to a lack of information on actual exposure to PM2.5 for some source-specific RRs used to fit the model, notably a) scarce informa-tion on actual exposure from SHS in the relevant epidemiologic studies (Pope et al. 2009, 2011b), which required the estimation of PM2.5 concentrations from other studies; b) potential misclassification of exposure for SHS estimates due to possible co-exposure from AAP of the exposed group; and c) the duration of exposure, which differs when it comes to exposures from AAP, SHS, HAP, and AS—the lifetime duration of exposure in AS may be much shorter than in the other exposures and the received doses may, there-fore, not be proportional to concentrations according to type of exposure. Uncertainties may be reduced by improving precision in the actual exposure estimates of the RRs from the epidemiologic literature used for developing the proposed model.Multiple studies were used to estimate RRs associated with exposure to AAP, SHS, and HAP. For AS, we estimated RRs for active cigarette smokers from a single cohort, the CPS II. This cohort was also used in the GBD 2010 project to estimate risk specifically for AS (Lim et al. 2012). However, the pattern of the association between the number of  cigarettes smoked per day and cause-specific mortality observed in the CPS-II cohort may not reflect the patterns observed in other cohort studies of AS (e.g., Pirie et al. 2013). Similarly, the IER for ALRI is fit through RR from studies of AAP and SHS conducted in a limited number of mostly high-income countries, and a single developing country RR estimate for HAP PM2.5 exposure and ALRI (Smith et al. 2011). We thus recommend that future work on the IER function include addi-tional sensitivity analyses of the type-specific RRs to which the curve is fit. Future work could also include the uncertainty in the esti-mate of PM2.5 from CO and new information in this relationship (McCracken et al. 2013).The key assumptions that underlie the IER, discussed above, largely serve to justify the integration of risk estimates for different types of PM exposure. These assumptions, and their tenability, have been addressed elsewhere (Pope et al. 2009, 2011a, 2011b). Unfortunately, for several of the most critical assumptions, those concerning the relative tox-icity per unit mass of PM2.5 of different types (e.g., AAP and AS), not accounting for the temporal pattern of exposure, and the absence of interaction among types of combustion, there is little empirical evidence against which to evaluate those assumptions or to evaluate in detail specific implications of their violation. Each warrants additional research.Although we set the counterfactual con-centration to be drawn from a uniform distri-bution with a lower bound of 5.8 μg/m3 and an upper bound of 8.8 μg/m3, we are not sug-gesting that there is convincing evidence that PM2.5 mortality and ALRI risk is zero below any specific concentration based on biological Fine particulate mass–based risk modelEnvironmental Health Perspectives • volume 122 | number 4 | April 2014 403considerations (Brook et al. 2010). Absence of such evidence from epidemiologic studies does not necessarily imply evidence of the absence of such a counterfactual concentra-tion. We thus take the conservative approach and set a positive counterfactual concentra-tion. However, our approach can be adapted to a different counterfactual if new evidence supporting a positive association at lower con-centrations becomes available. One such piece of evidence was observed in Canada, where positive associations as low as 2 μg/m3 were noted (Crouse et al. 2012).The Lin50 and Log models proposed by Cohen et al. (2004) were used for the previ-ous GBD estimates, and the Log model is currently recommended by the WHO (Ostro 2004). However, the unknown parameters in these models were estimated from a sin-gle cohort study of AAP, the CPS-II, which required analysis of the original data. The IER model uses RR estimates available in the open literature, allowing periodic updating of risk functions based on systematic review of the literature, and it does not require analyses of primary data not in the public domain. As new epidemiologic studies and evidence on type-specific PM2.5 exposure appear, the models can be reestimated by any interested member of the scientific community using publically available information.ConclusionFine particulate mass–based RR models can be developed that cover the entire global range of ambient exposure to PM2.5 by integrating RR information from different combustion sources that generate emissions of particulate matter. The specific RR model form we iden-tified in the present study can provide superior predictive power for leading global causes of mortality for air pollution compared with a range of alternative model forms.RefeRencesAnenberg SC, Horowitz LW, Tong DQ, West JJ. 2010. An estimate of the global burden of anthropogenic ozone and fine particulate on premature human mortality using atmospheric modeling. Environ Health Perspect 118:1189–1195; doi:10.1289/ehp.0901220.Balakrishnan K, Ghosh S, Ganguli B, Sambandam S, Bruce NG, Barnes  DF, et  al. 2013. State and national household concentrations of PM2.5 from solid cookfuel use: results from measurements and modeling in India for estimation of the global burden of disease. Environ Health 12:77; doi:10.1186/1476-069X-12-77.Brauer M, Amann M, Burnett RT, Cohen A, Denterner F, Ezzati M, et al. 2012. Exposure assessment for estimation of the global burden of disease attributable to outdoor air pollution. Environ Sci Technol 46:652–660.Brook RD, Rajagopalan, S, Pope CA III, Brook JR, Bhatnagar A, Diez-Roux A, et al. 2010. Particulate matter air pollution and cardiovascular disease: an update to the scientific state-ment from the American Heart Association. Circulation 121:2331–2378.Cao J, Yang C, Li J, Chen R, Chen B, Gua D, Kan H. 2011. Association between long-term exposure to outdoor air pollution and mortality in China: a cohort study. J. Hazard Mater 186:1594–1600.Chen H, Goldberg MS, Villeneuve PJ. 2008. A systematic review of relation between long-term exposure to ambi-ent air pollution and chronic disease. Rev Environ Health 23:243–296.Cohen A, Anderson HR, Ostro B, Pandey K, Krzyzanowski M, Kunzli N, et al. 2004. Urban air pollution. In: Comparative Quantification of Health Risks: Global and Regional Burden of Disease Attributable to Selected Major Risk Factors (Ezzati M, Lopez A, Rodgers A, Murray CJL, eds). Geneva:World Health Organization, 1353–1433.Committee on the Medical Effects of Air Pollutants. 2009. Long-Term Exposure to Air Pollution: Effect on Mortality. A Report by the Committee on the Medical Effects of Air Pollutants. Chilton, Oxfordshire, UK:Health Protection Agency. Available: http://www.comeap.org.uk/images/stories/Documents/Reports/mortality%20report%202009.pdf [accessed 6 March 2014].Cooke RM, Wilson AM, Tuomisto JT, Morales O, Tainio M, Evans JS. 2007. A probabilistic characterization of the relationship between fine particulate matter and mortality: elicitation of European experts. Environ Sci Technol 41:6598–6605.Crouse DL, Peters PA, van Donkelaar A, Goldberg MS, Villeneuve PJ, Brion O, et al. 2012. Risk of nonaccidental and cardiovascular mortality in relation to long-term exposure to low concentrations of fine particulate matter: a Canadian national-level cohort study. Environ Health Perspect 120:708–714; doi:10.1289/ehp.1104049.Evans J, van Donkelaar A, Martin RV, Burnett R, Rainham DG, Birkett NJ, et al. 2013. Estimates of global mortality attrib-utable to particulate air pollution using satellite imagery. Environ Res 120:33–42.Health Effects Institute, International Scientific Oversight Committee). 2010. Outdoor Air Pollution and Health in the Developing Countries of Asia: A Comprehensive Review. Special Report 18. Boston, MA:Health Effects Institute. Available: http://pubs.healtheffects.org/getfile.php?u=602 [accessed 6 March 2014].Krewski D, Jerrett M, Burnett RT, Ma R, Hughes E, Shi Y, et al. 2009. Extended Follow-Up and Spatial Analysis of the American Cancer Society Study Linking Particulate Air Pollution and Mortality. HEI Research Report 140. Boston, MA:Health Effects Institute. Available: http://pubs.healtheffects.org/getfile.php?u=478 [accessed 6 March 2014].Lim SS, Vos T, Flaxman AD, Danaei G, Shibaya K, Adair-Rahani H, et al. 2012. A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: a system-atic analysis for the Global Burden of Disease Study 2010. Lancet 380:2224–2260.McCracken JP, Schwartz J, Diaz A, Bruce N, Smith KR. 2013. Longitudinal Relationship between Personal CO and Personal PM2.5 among Women Cooking with Woodfired Cookstoves in Guatemala. PLoS ONE 8(2):e55670; doi:10.1371/journal.pone.0055670.Mehta S, Shin H, Burnett R, North T, Cohen AJ. 2013. Ambient particulate air pollution and acute lower respiratory infec-tions: a systematic review and implications for estimating the global burden of disease. Air Qual Atmos Health 6:69–83.Northcross A, Chowdhury Z, McCracken J, Canuzc E, Smith KR. 2010. Estimating personal PM2.5 exposures using CO mea-surements in Guatemalan households cooking with wood fuel. J Environ Monit 12:873–878.Office on Smoking and Health. 2006. The Health Consequences of Involuntary Exposure to Tobacco Smoke: A Report of the Surgeon General. Atlanta, GA:Centers for Disease Control and Prevention. Available: http://www.ncbi.nlm.nih.gov/books/NBK44324/ [accessed 28 February 2014].Oono IP, Mackay DF, Pell JP. 2011. Meta-analysis of the asso-ciation between secondhand smoke exposure and stroke. J Public Health (Oxf) 33:496–502.Ostro B. 2004. Outdoor Air Pollution: Assessing the Environmental Burden of Disease at National and Local Levels. Environmental Burden of Disease Series, No. 5 (Prüss-Üstün A, Campbell-Lendrum, D, Corvalán C, Woodward A, eds). Geneva:World Health Organization. Available: http://www.who.int/quantifying_ehimpacts/publications/ebd5.pdf [accessed 6 March 2014].Pirie K, Peto R, Reeves GK, Green J, Beral V, Million Women Study Collaborators. 2013. The 21st century hazards of smoking and benefits of stopping: a prospective study of one million women in the UK. Lancet 381:133–141.Pope CA III, Brook RD, Burnett RT, Dockery DW. 2011a. How is cardiovascular disease mortality risk affected by duration and intensity of fine particulate matter exposure? An integration of the epidemiologic evidence. Air Qual Atmos Health 4:5–14.Pope CA III, Burnett RT, Krewski D, Jerrett M, Shi Y, Calle EE, et al. 2009. Cardiovascular mortality and exposure to fine particulate matter from air pollution and cigarette smoke: shape of the exposure-response relationship. Circulation 120:941–948.Pope CA III, Burnett RT, Thun MJ, Calle EE, Krewski D, Ito K, et al. 2002. Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. JAMA 278:1132–1141.Pope CA III, Burnett RT, Turner MC, Cohen A, Krewski D, Jerrett M, et al. 2011b. Lung cancer and cardiovascular disease mortality associated with particulate matter exposure from ambient air pollution and cigarette smoke: shape of the exposure–response relationships. Environ Health Perspect 119:1616–1621; doi:10.1289/ehp.1103639.Singh GM, Danaei G, Farzadfar F, Stevens GA, Woodward M, Wormser D, et al. 2013. The age-specific quantitative effects of metabolic risk factors on cardiovascular diseases and diabetes: a pooled analysis. PLoS One 8:e65174; doi:10.1371/journal.pone.0065174.Smith KR. 1987. Biofuels, Air Pollution, and Health: A Global Review. New York:Plenum Publishing.Smith KR, Bruce N, Balakrishnan K, Adair-Rohani H, Balmes J, Chafe Z, et al. 2014. Millions dead: how do we know and what does it mean? Methods used in the Comparative Risk Assessment of Household Air Pollution. Ann Rev Public Health 35:27.1–27.23.Smith KR, McCracken JP, Weber MW, Hubbard A, Jenny A, Thompso LM, et al. 2011. Effect of reduction in household air pollution on childhood pneumonia in Guatemala (RESPIRE): a randomised controlled trial. Lancet 378:1717–1726.Stanek LW, Sacks JD, Dutton SJ, Dubois JJ B. 2011. Attributing health effects to apportioned components and sources of particulate matter: an evaluation of collective results. Atmos Environ 45:5655–5663.U.S. EPA (U.S. Environmental Protection Agency). 2009. Integrated Science Assessment for Particulate Matter. EPA/600/R-08/139F. Research Triangle Park, NC:U.S. EPA. Available: http://www.epa.gov/ncea/pdfs/partmatt/Dec2009/PM_ISA_full.pdf [accessed 6 March 2014].WHO (World Health Organization, Regional Office for Europe). 2006. Air Quality Guidelines—Global Update 2005. Particulate Matter, Ozone, Nitrogen Dioxide and Sulfur Dioxide. Copenhagen:WHO Regional Office for Europe. Available: http://www.euro.who.int/__data/assets/pdf_file/0005/78638/E90038.pdf [accessed 6 March 2014].WHO (World Health Organization, Regional Office for Europe). 2007. Health Relevance of Particulate Matter from Various Sources. Copenhagen:WHO Regional Office for Europe. Available: http://www.euro.who.int/__data/assets/pdf_file/0007/78658/E90672.pdf [accessed 6 March 2014].volume 122 | number 9 | September 2014 • Environmental Health Perspectives A 235ErratumErratum: “An Integrated Risk Function for Estimating the Global Burden of Disease Attributable to Ambient Fine Particulate Matter Exposure”In “An Integrated Risk Function for Estimating the Global Burden of Disease Attributable to Ambient Fine Particulate Matter Exposure” by Burnett et al. [Environ Health Perspect 122:397–403 (2014); http://dx.doi.org/10.1289/ehp.1307049], the authors omitted a reference. Balakrishnan et al. (2013) should have been cited in the first paragraph of “Methods” for HAP (household air pollutants). The correct sentence and reference are provided below.The equivalent long-term PM2.5 exposure from HAP was estimated for study subjects using coal or biomass for cooking (Balakrishnan et al. 2013) and for those study subjects using cleaner fuels to integrate this information into our IER risk model. Balakrishnan K, Ghosh S, Ganguli B, Sambandam S, Bruce NG, Barnes DF, et al. 2013. State and national household concentrations of PM2.5 from solid cookfuel use: results from measurements and modeling in India for estimation of the global burden of disease. Environ Health 12:77; doi:10.1186/1476-069X-12-77.The authors regret the error.All EHP content is accessible to individuals with disabilities. A fully accessible (Section 508–compliant) HTML version of this article is available at http://dx.doi.org/10.1289/ehp.122-A235.  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items