Particles size and composition in Mediterranean countries: geographical variability and short-term health effects MED-PARTICLES Project 2011-2013 Under the Grant Agreement EU LIFE+ ENV/IT/327 Particles size and composition in Mediterranean countries: geographical variability and short-term health effects MED-PARTICLES ACTION 13. Exploratory analysis of the role of PM components on mortality and morbidity Summary: Report on the methods and results of the short-term effects of particulate matter constituents on daily mortality and hospitalizations in five Mediterranean cities: results from the MED-PARTICLES project. Introduction Dozens of studies link increases in daily particulate matter (PM) levels to increases in mortality and hospital admissions (Brook et al. 2010; Brunekreef and Holgate 2002; Ruckerl et al. 2011; Samoli et al. 2013; Stafoggia et al. 2013). However, particulate matter is a location- and season-dependent complex mixture of different elements with potentially different toxicity. Since the different chemical composition of PM may be one of the reasons explaining the heterogeneity of the effects of air pollution in different studies, several recent investigations have examined the relationship between concentrations of individual chemical species and mortality or hospital admissions (Bell et al. 2009; Burnett et al. 2000; Franklin et al. 2008; Ito et al. 2011; Lall et al. 2011; Lippmann et al. 2006; Mar et al. 2000; Ostro et al. 2007; Ostro et al. 2010; Ostro et al. 2011; Peng et al. 2009; Sarnat et al. 2008; Son et al. 2012; Zanobetti et al. 2009; Zhou et al. 2011). Most of the studies were conducted in North America and studies in Europe are scarce (Andersen et al. 2007; Atkinson et al. 2010; Ostro et al. 2011). Elemental carbon (EC), nickel, and silicon are among the components most commonly associated to adverse health events in previous epidemiological studies. However, the main conclusion derived from these studies is that although many components contribute to health effects of PM, there is no sufficient evidence to differentiate the constituents more closely related to health outcomes (Environmental Protection Agency 2009; Rohr and Wyzga 2012; World Health Organization 2013). In this study, we examine the relationship between PM constituents and mortality and hospital admissions in five cities in the Mediterranean area, a region characterized by high vehicle and population densities, a high proportion of diesel cars, high sea traffic and intense solar radiation. Methods Study population The study included five cities, three in Spain (Barcelona, Madrid and Huelva) and two in Italy (Rome and Bologna). The overall period of study was from January 2003 to April 2013, although the different cities contributed to different periods according to air pollution data availability. Mortality data Daily mortality counts for all non-external causes [International Classification of Diseases, 9th Revision (ICD-9) codes 001–799; 10th revision (ICD-10) codes A00–R99] (ref: WHO 1999), cardiovascular causes (ICD-9 codes 390–459, ICD-10 codes I00–I99) and respiratory causes (ICD-9 codes 460–519, ICD10 codes J00–J99) were collected for all cities. Hospital admissions data Daily counts of emergency hospital admissions were collected from national or regional health information systems. Hospitalizations only of residents ≥ 15 years of age were considered. Two study outcomes were defined on the basis of the primary discharge diagnosis: cardiovascular hospitalizations [International Classification of Diseases, 9th Revision (ICD-9) codes 390–459; and 10th Revision (ICD10) codes I00–I99] (WHO 1999), and respiratory hospitalizations (ICD-9 codes 460–519; ICD-10 codes J00–J99). All data were extracted and collected according to a common protocol. Because data were anonymous and collected as daily counts, no informed consent was needed. And because the analysis on daily counts was conducted by a public health institute, there was no need for approval by an institutional review board. Exposure data PM10 and PM2.5 data for speciation analysis were available from a single station in each city. Madrid’ station was urban while all other were urban background stations. More details about the stations are provided in the Supplemental Material. PM2.5 data were not available for Rome while PM10 data were not available for Bologna. The period and frequency of sampling was (Table 1): Barcelona, 2003-2010, approximately every third day; Madrid, 2007-2008, two measurements per week; Huelva, 2003-2010, one measurement per week; Rome, 2005-2008, daily data; Bologna, 2012-2013, daily data. PM data collected during intense African dust outbreaks, identified using the methodology of Pey et al. (2013), were excluded from the analysis. In the three Spanish cities, daily samples of PM2.5 and PM10 were collected by high-volume samplers (MCV, S.A., Barcelona, Spain) with a flow rate of 30 m3/hr using quartz filters. PM10 and PM2.5 concentrations were determined gravimetrically. Subsequently, the quartz filters were analyzed for water-soluble ions (NH4+, Cl-, SO42-, NO3-), major elements (Al, Ca, K, Mg, Fe, Na) and 46 trace elements by ion chromatography, inductively coupled plasma atomic emission, ICP-AES, and inductively coupled plasma mass spectrometry, ICP-MS, respectively (Querol et al. 2001). In Rome, PM10 sampling was conducted by a beta attenuation monitor SM200 (Opsis AB Furulund-S) operating at flow rate of 1 m3 h-1 and teflon filters used as a substrate while a MICRODUST sequential sampler (AQUARIA, Lacchiarella, MI-I) operating at a flow rate of 2 l min-1 was used for PM sampling on quartz filters. The Teflon filters were subjected to energy-dispersion X-ray fluorescence analysis (XLab2000, Spectro Analytical Instruments, Kleve-D) for the determination of major and minor elements: Al, Si, Fe, K, Ca, As, Cr, Cu, Mn, Ni, Pb, Ti, V, Zn. Then, the filters were extracted in deionized water and analyzed by ion chromatography (ICS90 Dionex Co) for the determination of Cl-, NO3-, SO4=, Na+, NH4+, K+, Mg++ and Ca++. Elemental EC, and organic carbon OC were analyzed in the quartz filters by thermo-optical analysis (OCEC Sunset Analyser, Sunset Laboratory) using the NIOSH temperature protocol. In Bologna, two Dual Channel Monitors (SWAM, Fai Instrument) were used to sample PM on quartz filters. EC and OC were determined by thermo-optical analysis (OCEC Sunset Analyser, Sunset Laboratory) using the EUSAAR_2 protocol. Ion chromatography (Dionex) was used for the determination of Cl-, NO3-, SO4=, Na+, NH4+, K+, Mg++ and Ca++. Major and trace metals were determined. Only 17 pre-specified PM constituents were included in the mortality analyses: silicon (Si) (calculated as 3 times the aluminum concentration in the Spanish cities), calcium (Ca), aluminum (Al), iron (Fe), potassium (K), magnesium (Mg), zinc (Zn), copper (Cu), titanium (Ti), manganese (Mn), vanadium (V), nickel (Ni), nitrate (NO3–), sulfate (SO42-), elemental carbon (EC), organic carbon (OC) and total carbon (TC), calculated as EC+OC. The list of 17 species was based on their detectability in the included cities and on a review of the previous literature linking PM constituents and health. If one of the above constituents had more than 20% of values missing or below the limit of detection or quantification in a certain city, the city was excluded from the analysis of that particular constituent. Otherwise, non-detectable values were replaced by half the limit of detection. Daily average temperature and relative humidity were collected from each city, and they were combined to calculate apparent temperature (O'Neill et al. 2003). Information on influenza epidemics was obtained from hospital admissions records. Information on bank holidays was also collected for each city. Study design and data analysis We used a time-stratified case-crossover study design (Levy et al. 2001). In cities with daily data, all days from the same year, month and day of the week were grouped in the same stratum. In cities without daily data, strata were composed of days from the same year and month. Since the casecrossover analysis conducts comparisons within strata, seasonal and long-term effects are eliminated and they do not confound the results. Analyses were performed using conditional Poisson regression (Whitaker et al. 2006) and were further adjusted for holidays, influenza epidemics, day of the week (in cities without daily data), and apparent temperature. To account for non-linearity and for the different lag structure of cold and hot temperatures, we included two separate spline terms for apparent temperature, with 3 degrees of freedom each, one for the cold period (temperatures below the annual median) using average temperature for lags 0-5, and the other for the warm period using average temperature for lags 0-1. All analyses were conducted separately for each city, and results were combined using randomeffects meta-analysis. Lags 0, 1 and 2 of the effect of each pollutant were examined in different models, as models accounting for multiple lags could not be built in cities without daily data. Results are reported for and interquartile range (IQR) increase in pollutant concentrations. In order to account for potential confounding for total PM levels, while accounting for the fact that PM constituents are components of PM, we applied the recently proposed constituent residual method (Mostofsky et al. 2012). First, a linear regression model for the constituent as a function of total PM levels is fitted. Then, the residuals for this model, which represent the variation in constituent levels independent of PM, are included as an explanatory variable in the case-crossover model, along with total PM levels. The coefficient for the residuals represent the effect of the constituent while holding PM constant, i.e. the independent contribution of the constituent; and the PM coefficient represents the independent effect of total PM. Results Study populations Table 1 summarizes the daily number of deaths by natural, cardiovascular and respiratory causes and of cardiovascular and respiratory hospital admissions in the five participating cities. Madrid, Rome and Barcelona are large cities with an average of at least 40 daily deaths by natural causes, while Bologna and Huelva had only 12 and 3 daily deaths on average, respectively. Cardiovascular causes accounted for between 30% and 40% of all deaths, while respiratory causes accounted for between 5% and 15% of them. We observe the same for hospital admissions, where Madrid, Rome and Barcelona show more hospitalizations while Rome is the city that contributes the most in days in the analyses for PM10 and Barcelona for PM2.5. Air pollution The concentrations of PM10 and PM2.5 and of their selected constituents in the five cities are summarized in Table 2. The highest PM10 levels were observed for Barcelona and Madrid, with levels above 40 g/m3, while Huelva and Rome showed levels between 30 and 35 g/m3. In terms of PM2.5, the highest levels were found in Barcelona and Bologna, with values above 25 g/m3, while Madrid and Huelva had values around 20 g/m3. The selected species represent between 37% and 62% of the total PM levels depending on the city and fraction. Total carbon, sulfates, nitrates, silicon and calcium were among the species with highest contribution in terms of mass, while the other species had much smaller contributions that did not exceed 4% of the mass in any city. Correlations between species and total PM10 levels were all in the range 0.4-0.8, with a few exceptions: correlations exceeded 0.8 only in Madrid for total carbon, calcium, iron, potassium and manganese; correlations were lower than 0.4 in some cities for zinc, copper, titanium, manganese, sulfates and magnesium (Tables 2). Correlations between species and total PM2.5 levels were generally smaller than those found for PM10, only showing high values for carbon and nitrates (Supplemental material, Tables S1-S5). The correlations between the same constituent in PM10 and PM2.5 were high for all constituents in Barcelona, while some low correlations were found for some constituents in Madrid or Huelva (e.g. for nickel, iron, magnesium, manganese, calcium, zinc and copper) (Tables 2). This comparison was not possible in the Italian cities as they only contributed with one PM fraction. The correlations between PM10 species were only high (around 0.8 or higher) for the group composed by silicon, calcium, aluminum, iron and titanium, which in some cities were also highly correlated with potassium, magnesium, manganese and vanadium. The correlations between PM2.5 species were only consistently high in the different cities for aluminum and silicon (correlations greater than 0.9). Mortality results The meta-analysis results linking mortality and PM10 constituents are shown in Figure 1a. Results for PM2.5 constituents are shown in Figure 2a. As shown in the leftmost panels of Figures 1a and 2a, there was no statistically significant elevation in mortality risk by total PM10 or PM2.5 levels for any of the lags or mortality causes. Several PM10 and PM2.5 constituents showed positive and significant associations with mortality, but there was little consistency across lags and mortality causes. Results for PM10 and PM2.5 are not entirely comparable, as they were based on different cities. Total carbon was associated with total mortality at lag 1 and only for PM10, while elemental carbon was associated with increased total, cardiovascular and respiratory mortality risk at lags 1 or 2, again for PM10 but not for PM2.5. For organic carbon, only a significant protective effect for cardiovascular mortality at lag 0 for PM10 was found. Sulfates were only associated with respiratory mortality, with increased risks for both PM10 and PM2.5 at different lags. Significant protective effects at lag 0 were found for nitrates in PM10. Both silicon and aluminum showed a positive association with total mortality at lag 2 for PM2.5. Iron showed a significant positive association with total mortality at lag 1 for PM10 and with cardiovascular mortality at lag 2 for PM2.5. Potassium was associated with total mortality at lag 1 for PM10, and magnesium was only associated with cardiovascular mortality at lag 0 for PM2.5. Titanium was associated with total mortality at lags 1 and 2, and with cardiovascular mortality at lag 2, and these relationships were only present for PM10. Manganese showed positive and significant associations with total mortality at lag 1 for PM01, while for PM 2.5 associations were significant for cardiovascular mortality at lag 0 and for respiratory mortality at lags 0 and 2. Finally, nickel was associated with respiratory mortality for PM10 at lag 2. In terms of the size of the associations, most of the significant results represented around a 1% increase in mortality for an IQR increase in the PM constituent (Figures 1a and 2a). Notable exceptions to this pattern were iron, with percent changes of around 2%, magnesium, with percent changes around 3%, and all the significant results found for respiratory mortality, which reached values of 4% and 5% for elemental carbon and sulfates, and were higher than 3% for manganese and nickel. Figure 3a illustrates the between-city variation in effects and the weight of each city in the final analysis for the particular case of iron and manganese (in PM10 at lag 1). Barcelona and Rome, and also Madrid for iron, had a similar weight in the analysis, while Huelva had a small contribution. Results were quite homogeneous between cities. Table 4 presents the results of the case-crossover models with and without adjustment for total PM levels using the constituent residual method for the specific case of the effects of lag 1 on total mortality. Note that the models that only included total PM levels resulted in non-significant results, with percent changes of 1.2 (95% confidence interval (CI): -0.2, 2.7) for PM10 and of 0.4 (95% CI: -1.3, 2.1) for PM2.5. The adjusted effect represents the effect of higher levels of the constituent when holding total PM constant. For PM10, models for constituents not adjusted for PM levels resulted in significant mortality increases associated to total carbon, iron, potassium, titanium and manganese. When adjusting for PM levels, all effects were diluted and no longer statistically significant, except for iron, which showed stronger results with 3.2% increase in mortality associated to an IQR change in the pollutant. Interestingly, in the models for total carbon and nitrates, the effect of PM reached statistical significance. For PM2.5, no significant effects were found in models not adjusted for PM levels. When total PM was also included in the model, significant associations were found for calcium, iron, and manganese. Hospitalization results Figures 1b and 2b show the associations between total PM and each component and respiratory and cardiovascular admissions. We observed a significant increase of cardiovascular admissions associated with an increase of a one IQR of total PM2.5 (lag 0 and 2) and its sulfates (lag 0), EC (lag 0) and Mn (lag 0) components. For PM10, we found a positive and significant association for total PM (lag 0), NO3- (lag 0), Ca (lag 0), Fe (lag 0), Fe (lag 0), Cu (lag 1), Ti (lag 1), Mn (lag 1). For respiratory admissions, a positive and significant association was observed with total PM (lag1), Ca (lag2), Fe (lag 1), V (lag 0 and 1) and Ni (lag 0 and 1) from PM2.5 and with sulfates (lag 0), silicon (lag 2), V (lag 0) and Ni (lag 0) from PM10. While most of the effects on cardiovascular hospitalizations are observed at lag 0, the pattern is less clear for respiratory admissions. Results from PM10 and PM2.5 are not directly comparable, as they do not include the same cities. Most of the significant positive associations were around 1% increase for one IQR, reaching 2% for Fe and Mn from PM10 for cardiovascular admissions and for total PM and EC from PM2.5 for both cardiovascular and respiratory admissions, and for Ni from PM2.5 for respiratory admissions. Table 4 presents the results of the case-crossover models with and without adjustment for total PM levels using the constituent residual method for the specific case of the effects of lag 0 on cardiovascular admissions. We observed that the total PM10 effect on cardiovascular admissions persisted regardless of the specie added in the model. For total PM2.5, the PM effect in some models lost its significance but the estimate was always positive and in the same range (around 2% increase). The effect of Fe from PM10 decreased and became not significant when adjusting for total PM. The percent increase went from 2.1 (95%CI: 0.5-3.8) to 0.6 (-3.5, 5.0). The same thing was observed for Ti, but the Mn effect persisted, the percent increase was 2.0 (1.0-3.1) without adjusting for total PM and 1.9 (0.53.4) adjusting for total PM. For PM2.5 species, adjusting or not for total PM dids not have an impact on the Mn effect, but it decreased the sulfate and EC effects that in addition became not significant. Figure 3b shows the between-city variation in effects and the weight of each city in the final analysis for the particular case of iron and manganese (in PM10 at lag 0) on cardiovascular admissions. We observe that Rome and Barcelona were the cities that contribute the most to the total effect. In addition, except for Fe in Huelva, all the effects were positive and significant for Barcelona and Rome, while in Madrid the confidence intervals are very large. Conclusions: in the report we have examined the effects of PM constituents on mortality and hospital admissions in five Mediterranean cities. Our study suggests that several PM constituents are associated to increases in daily mortality, including elemental carbon, sulfate, silicon, aluminum, iron, potassium, magnesium, titanium, manganese and nickel. Several of the metals are highly correlated and therefore it is difficult to attribute effects to a particular constituent. Studies with larger datasets and daily data are needed to confirm these results in the Mediterranean area. References Andersen ZJ, Wahlin P, Raaschou-Nielsen O, Scheike T, Loft S. 2007. Ambient particle source apportionment and daily hospital admissions among children and elderly in Copenhagen. J Expo Sci Environ Epidemiol 17(7):625-636. Atkinson RW, Fuller GW, Anderson HR, Harrison RM, Armstrong B. 2010. Urban ambient particle metrics and health: a time-series analysis. Epidemiology 21(4):501-511. Bell ML, Ebisu K, Peng RD, Samet JM, Dominici F. 2009. Hospital admissions and chemical composition of fine particle air pollution. Am J Respir Crit Care Med 179(12):1115-1120. Brook RD, Rajagopalan S, Pope CA, 3rd, Brook JR, Bhatnagar A, Diez-Roux AV, et al. 2010. Particulate matter air pollution and cardiovascular disease: An update to the scientific statement from the American Heart Association. Circulation 121(21):2331-2378. Brunekreef B, Holgate ST. 2002. Air pollution and health. Lancet 360(9341):1233-1242. Burnett RT, Brook J, Dann T, Delocla C, Philips O, Cakmak S, et al. 2000. Association between particulate- and gas-phase components of urban air pollution and daily mortality in eight Canadian cities. Inhal Toxicol 12 Suppl 4:15-39. Environmental Protection Agency. 2009. Integrated science assessment for particulate matter (final report). Washington, DC: United States Environmental Protection Agency. Franklin M, Koutrakis P, Schwartz P. 2008. The role of particle composition on the association between PM2.5 and mortality. Epidemiology 19(5):680-689. Ito K, Mathes R, Ross Z, Nadas A, Thurston G, Matte T. 2011. Fine particulate matter constituents associated with cardiovascular hospitalizations and mortality in New York City. Environ Health Perspect 119(4):467-473. Lall R, Ito K, Thurston GD. 2011. Distributed lag analyses of daily hospital admissions and sourceapportioned fine particle air pollution. Environ Health Perspect 119(4):455-460. Lippmann M, Ito K, Hwang JS, Maciejczyk P, Chen LC. 2006. Cardiovascular effects of nickel in ambient air. Environ Health Perspect 114(11):1662-1669. Mar TF, Norris GA, Koenig JQ, Larson TV. 2000. Associations between air pollution and mortality in Phoenix, 1995-1997. Environ Health Perspect 108(4):347-353. Ostro B, Feng WY, Broadwin R, Green S, Lipsett M. 2007. The effects of components of fine particulate air pollution on mortality in california: results from CALFINE. Environ Health Perspect 115(1):13-19. Ostro B, Lipsett M, Reynolds P, Goldberg D, Hertz A, Garcia C, et al. 2010. Long-term exposure to constituents of fine particulate air pollution and mortality: results from the California Teachers Study. Environ Health Perspect 118(3):363-369. Ostro B, Tobias A, Querol X, Alastuey A, Amato F, Pey J, et al. 2011. The effects of particulate matter sources on daily mortality: a case-crossover study of Barcelona, Spain. Environ Health Perspect 119(12):1781-1787. Peng RD, Bell ML, Geyh AS, McDermott A, Zeger SL, Samet JM, et al. 2009. Emergency admissions for cardiovascular and respiratory diseases and the chemical composition of fine particle air pollution. Environ Health Perspect 117(6):957-963. Rohr AC, Wyzga RE. 2012. Attributing health effects to individual particulate matter constituents. Atmospheric Environment 62:130-152. Ruckerl R, Schneider A, Breitner S, Cyrys J, Peters A. 2011. Health effects of particulate air pollution: A review of epidemiological evidence. Inhal Toxicol 23(10):555-592. Samoli E, Stafoggia M, Rodopoulou S, Ostro B, Declercq C, Alessandrini E, et al. 2013. Associations between Fine and Coarse Particles and Mortality in Mediterranean Cities: Results from the MED-PARTICLES Project. Environ Health Perspect 121(8):932-938. Sarnat JA, Marmur A, Klein M, Kim E, Russell AG, Sarnat SE, et al. 2008. Fine particle sources and cardiorespiratory morbidity: an application of chemical mass balance and factor analytical source-apportionment methods. Environ Health Perspect 116(4):459-466. Son JY, Lee JT, Kim KH, Jung K, Bell ML. 2012. Characterization of fine particulate matter and associations between particulate chemical constituents and mortality in Seoul, Korea. Environ Health Perspect 120(6):872-878. Stafoggia M, Samoli E, Alessandrini E, Cadum E, Ostro B, Berti G, et al. 2013. Short-term Associations between Fine and Coarse Particulate Matter and Hospitalizations in Southern Europe: Results from the MED-PARTICLES Project. Environ Health Perspect 121(9):1026-1033. World Health Organization. 2013. Review of evidence on health aspects of air pollution – REVIHAAP Project: Final technical report. Bonn: WHO Regional Office for Europe. Zanobetti A, Franklin M, Koutrakis P, Schwartz J. 2009. Fine particulate air pollution and its components in association with cause-specific emergency admissions. Environ Health 8:58. Zhou J, Ito K, Lall R, Lippmann M, Thurston G. 2011. Time-series analysis of mortality effects of fine particulate matter components in Detroit and Seattle. Environ Health Perspect 119(4):461-466. Table 1: Data availability and summary of daily deaths by city. Daily deaths (Mean Standard deviation) Date beginning – Date end Number of days City Daily hospitalizations (Mean Standard deviation) Total Cardiovascular Respiratory Cardiovascular Respiratory Barcelona 2003/01/01 – 2010/12/31 736 41.5 8.6 13 4.1 4.4 2.6 55.8 10.8 55.8 19.4 Madrid 2007/01/01 – 2008/02/29 104 72.1 11 21.7 5 12.2 4.7 129.6 27.4 142.3 42.4 Huelva 2003/01/01 – 2010/12/31 406 2.9 1.7 11 0.3 0.6 1.8 2.3 0.9 1.2 Roma 2005/01/01 – 2007/12/31 1059 57 10.5 23.5 6.3 3.5 2.2 81.9 18.7 35.5 11.3 Bologna 2011/11/15 – 2013/02/27 472 12.2 3.7 - - 16.8 4.6 9.7 3.7 Table 2: Description (mean standard deviation) of PM species levels (µg/m3) by city. Barcelona Madrid Huelva Rome Bologna Common (2003-2010) (2007-2008) (2003-2010) (2005-2007) (2011-2013) IQR PM10 PM2.5 PM10 PM2.5 PM10 PM2.5 PM10 PM2.5 PM10 PM2.5 40.2 17.4 26.3 12.2 41.5 19.5 20.7 9.2 35.2 15.2 20.8 10.7 30.1 12.6 27.6 19.5 19.6 16.0 Total Carbon 6.9 3.7 4.9 2.6 10.4 5.6 7.6 3.9 5.4 3.4 3.9 2.3 10.6 5.2 6.7 4.7 5.2 3.8 OC Total PM 3.3 1.8 2.9 1.8 na na na na 8.5 4.2 5.1 3.9 3.4 3.0 2- 3.8 2.6 3.4 2.3 3.2 1.4 2.4 1.1 3.8 3.2 3.3 2.7 3.4 2.0 2.4 1.6 2.7 2.3 NO3 - 4.3 3.5 2.4 3.2 2.5 2.1 1.5 1.7 2.7 1.8 1.1 1.2 2.1 1.4 5.1 7.9 2.2 3.0 SiO2 3.8 3.6 1.4 1.8 nd 1.0 0.8 4.9 4.4 1.4 2.0 1.0 0.9 2.7 1.1 EC 1.8 1.2 1.5 1.0 na na na na 2.2 1.3 1.6 1.0 1.4 1.1 Ca 2.13 1.64 0.60 0.62 2.10 1.35 0.26 0.16 1.16 0.94 0.33 0.52 1.29 0.68 0.10 0.06 1.32 0.27 Al2O3 1.28 1.21 0.47 0.60 0.72 0.59 0.36 0.29 1.62 1.46 0.48 0.70 0.27 0.23 0.81 0.39 Fe 0.86 0.54 0.28 0.21 1.67 0.87 0.20 0.15 0.66 0.52 0.18 0.20 0.46 0.26 0.11 0.07 0.63 0.18 K 0.38 0.29 0.20 0.23 0.33 0.22 0.20 0.41 0.41 0.30 0.21 0.20 0.39 0.32 0.18 0.30 0.26 0.15 Mg 0.279 0.165 0.082 0.072 0.225 0.150 0.083 0.084 0.295 0.193 0.084 0.093 0.095 0.098 nd 0.153 0.069 Zn 0.085 0.080 0.059 0.068 nd 0.050 0.026 0.049 0.074 0.045 0.084 0.037 0.035 nd 0.047 0.039 Cu 0.056 0.051 0.027 0.036 0.113 0.069 0.019 0.010 0.047 0.078 0.029 0.033 0.040 0.020 nd 0.051 0.021 Ti 0.040 0.033 0.014 0.013 0.040 0.031 0.008 0.014 0.056 0.084 0.020 0.039 0.021 0.023 nd 0.030 0.010 Mn 0.018 0.011 0.009 0.006 0.020 0.011 0.005 0.003 0.013 0.024 0.004 0.004 0.012 0.008 nd 0.011 0.004 V 0.010 0.007 0.008 0.006 0.002 0.001 0.001 0.001 0.005 0.004 0.004 0.003 0.006 0.004 nd 0.005 0.004 Ni 0.005 0.003 0.004 0.003 0.002 0.001 0.004 0.003 0.003 0.003 0.002 0.003 0.005 0.003 nd 0.003 0.004 SO4 na: not available nd: non-detectable. Table 3. Pearson correlation coefficients between PM and PM constituents The lower triangle includes the correlations between PM10 constituents. The upper triangle includes the correlations between PM2.5 constituents. The main diagonal includes the correlation between the same species in PM10 and PM2.5. Table 3a Barcelona PM TC OC SO42- NO3- SiO2 EC Ca Al2O3 Fe K Mg Zn Cu Ti Mn V Ni PM 0.85 0.73 0.77 0.57 0.67 0.36 0.61 0.5 0.36 0.63 0.34 0.39 0.27 0.32 0.51 0.6 0.5 0.49 TC 0.71 0.85 0.94 0.31 0.52 0.21 0.74 0.43 0.21 0.61 0.26 0.24 0.27 0.32 0.38 0.58 0.38 0.4 OC 0.71 0.92 0.9 0.34 0.66 0.16 0.72 0.4 0.16 0.53 0.28 0.25 0.16 0.31 0.43 0.52 0.34 0.31 SO42- 0.59 0.35 0.31 0.91 0.29 0.2 0.14 0.19 0.2 0.29 0.23 0.29 0.14 0.19 0.29 0.24 0.55 0.43 NO3- 0.64 0.53 0.61 0.41 0.89 -0.02 0.37 0.1 -0.02 0.23 0.16 -0.05 0.22 0.15 0.06 0.36 0.21 0.24 SiO2 0.68 0.39 0.4 0.4 0.21 0.78 0.21 0.61 1 0.61 0.31 0.64 0.07 0.23 0.66 0.41 0.19 0.17 EC 0.62 0.86 0.72 0.2 0.42 0.39 0.85 0.47 0.21 0.62 0.25 0.29 0.19 0.44 0.44 0.57 0.26 0.24 Ca 0.74 0.61 0.59 0.34 0.32 0.71 0.6 0.75 0.61 0.85 0.3 0.73 0.18 0.34 0.85 0.63 0.22 0.21 Al2O3 0.68 0.39 0.4 0.4 0.21 1 0.39 0.71 0.78 0.61 0.31 0.64 0.07 0.23 0.66 0.41 0.19 0.17 Fe 0.79 0.79 0.72 0.37 0.42 0.71 0.71 0.81 0.71 0.81 0.32 0.69 0.24 0.38 0.82 0.73 0.33 0.31 K 0.55 0.36 0.42 0.31 0.25 0.58 0.45 0.55 0.58 0.51 0.86 0.53 0.14 0.18 0.33 0.32 0.15 0.12 Mg 0.69 0.32 0.35 0.45 0.2 0.77 0.34 0.68 0.77 0.63 0.61 0.67 0.1 0.27 0.78 0.44 0.24 0.18 Zn 0.3 0.33 0.24 0.13 0.25 0.19 0.26 0.32 0.19 0.33 0.18 0.11 0.92 0.13 0.13 0.46 0.09 0.16 Cu 0.41 0.53 0.46 0.15 0.26 0.25 0.54 0.36 0.25 0.48 0.23 0.19 0.21 0.76 0.28 0.27 0.17 0.18 Ti 0.76 0.49 0.5 0.47 0.28 0.88 0.5 0.79 0.88 0.81 0.62 0.81 0.21 0.28 0.67 0.58 0.31 0.26 Mn 0.71 0.67 0.63 0.31 0.41 0.64 0.62 0.75 0.64 0.84 0.54 0.52 0.5 0.38 0.75 0.87 0.24 0.25 V 0.61 0.5 0.48 0.53 0.43 0.4 0.41 0.42 0.4 0.49 0.34 0.42 0.2 0.26 0.48 0.42 0.92 0.82 Ni 0.52 0.45 0.42 0.38 0.36 0.33 0.38 0.37 0.33 0.45 0.26 0.29 0.19 0.23 0.4 0.38 0.76 0.84 Table 3b. Madrid PM TC SO42- NO3- SiO2 PM 0.89 0.88 0.44 0.86 0.08 TC 0.9 0.96 0.23 0.68 0.04 SO42- 0.59 0.41 0.96 0.28 0.29 NO3- 0.78 0.73 0.54 0.96 -0.1 SiO2 Ca Al2O3 Fe K Mg Zn Cu Ti Mn V Ni 0.3 0.1 0.38 0.21 0.27 0.04 0.29 0.15 0.12 0.4 0.72 0.17 0.74 0.24 -0.21 0.08 0.43 0.81 0.07 0.74 0.17 -0.15 0.28 0.3 0.17 0.08 -0.07 0.26 0.15 0.23 0.22 0.1 0.22 0.38 0.6 0.07 0.04 -0.06 0.35 0.54 -0.04 0.6 0.07 -0.21 - - - - - 0.64 0.96 0.26 0 0.32 0.27 0 0.29 0.41 0.48 0.22 Ca 0.87 0.71 0.43 0.49 - 0.53 0.74 0.44 0.02 0.22 0.36 0.18 0.25 0.52 0.46 0.25 Al2O3 0.72 0.47 0.44 0.28 - 0.82 0.65 0.37 -0.01 0.28 0.32 0 0.31 0.44 0.5 0.31 Fe 0.93 0.92 0.37 0.64 - 0.84 0.66 0.32 -0.09 -0.22 0.09 0.29 0.22 0.33 0.2 -0.01 K 0.87 0.66 0.49 0.49 - 0.91 0.93 0.81 0.7 0.78 -0.02 0.26 0.89 0.1 0.11 -0.09 Mg 0.75 0.5 0.42 0.35 - 0.83 0.93 0.7 0.89 0.35 0.21 0.17 0.73 0.16 0.28 0.15 Zn - - - - - - - - - - - 0.48 -0.07 0.64 0.3 0.58 Cu 0.64 0.73 0.17 0.5 - 0.53 0.26 0.74 0.43 0.3 - 0.78 0.14 0.68 0.13 -0.14 Ti 0.75 0.51 0.43 0.35 - 0.78 0.97 0.7 0.9 0.92 - 0.34 0.86 0.13 0.24 -0.03 Mn 0.94 0.84 0.45 0.68 - 0.82 0.74 0.92 0.85 0.77 - 0.66 0.8 0.77 0.42 0.03 V 0.69 0.42 0.62 0.37 - 0.74 0.87 0.6 0.85 0.86 - 0.25 0.86 0.69 0.83 0.23 Ni 0.79 0.77 0.42 0.55 - 0.73 0.6 0.79 0.68 0.59 - 0.57 0.63 0.8 0.56 -0.25 Table 3c. Huelva PM TC SO42- NO3- SiO2 Ca Al2O3 Fe K Mg Zn Cu Ti Mn V Ni PM 0.53 0.5 0.5 0.4 0.5 0.55 0.47 0.58 0.58 0.54 0.49 0.29 0.33 0.62 0.42 0.48 TC 0.44 0.53 0.06 0.42 0.33 0.24 0.31 0.46 0.51 0.21 0.13 0.12 0.22 0.41 0.04 0.1 SO42- 0.57 0.22 0.91 0.2 0.1 0.21 0.08 0.17 0.09 0.18 0.27 0.24 0.14 0.25 0.79 0.47 NO3- 0.61 0.32 0.58 0.64 -0.01 0.13 -0.02 0.08 0.37 0.03 0.19 0.14 0.12 0.17 0.09 0.17 SiO2 0.73 0.34 0.23 0.21 0.73 0.51 0.95 0.89 0.53 0.74 0.22 0.14 0.44 0.75 0.17 0.2 Ca 0.77 0.41 0.31 0.35 0.82 0.3 0.53 0.62 0.61 0.8 0.76 0.47 0.29 0.75 0.17 0.69 Al2O3 0.73 0.34 0.23 0.21 1 0.82 0.74 0.85 0.54 0.79 0.22 0.17 0.42 0.69 0.18 0.19 Fe 0.79 0.46 0.26 0.32 0.94 0.87 0.94 0.72 0.65 0.77 0.29 0.21 0.48 0.85 0.22 0.31 K 0.75 0.5 0.23 0.36 0.86 0.78 0.86 0.89 0.64 0.61 0.42 0.27 0.28 0.67 0.12 0.41 Mg 0.66 0.14 0.28 0.19 0.8 0.66 0.8 0.74 0.65 0.65 0.53 0.36 0.38 0.78 0.21 0.48 Zn 0.25 0.08 0.29 0.16 0.16 0.23 0.16 0.16 0.11 0.33 0.38 0.46 0.14 0.5 0.23 0.69 Cu 0.35 0.13 0.27 0.21 0.19 0.39 0.19 0.27 0.17 0.19 0.23 0.28 0.13 0.3 0.27 0.43 Ti 0.31 0.3 0.21 0.23 0.34 0.3 0.34 0.4 0.35 0.23 0.09 0.04 0.55 0.46 0.19 0.19 Mn 0.24 0.32 0.1 0.14 0.29 0.26 0.29 0.36 0.34 0.17 0.02 0.01 0.79 0.18 0.25 0.51 V 0.67 0.24 0.79 0.51 0.43 0.5 0.43 0.47 0.42 0.45 0.38 0.31 0.29 0.18 0.87 0.5 Ni 0.44 0.25 0.47 0.38 0.2 0.26 0.2 0.23 0.25 0.2 0.3 0.18 0.12 0.03 0.5 0.39 Table 3d. Rome (only PM10) PM TC OC SO42- NO3- SiO2 EC Ca Al2O3 Fe K Mg Zn Cu Ti Mn V PM 1 TC 0.74 1 OC 0.73 0.98 1 SO42- 0.37 0 -0.01 NO3- 0.56 0.32 0.37 0.1 1 SiO2 0.48 0.12 0.09 0.47 0.23 1 EC 0.61 0.82 0.7 -0.02 0.2 0.1 1 Ca 0.54 0.24 0.21 0.36 0.3 0.82 0.22 1 Al2O3 0.48 0.11 0.09 0.46 0.23 0.99 0.09 0.81 1 Fe 0.71 0.57 0.51 0.26 0.37 0.69 0.54 0.76 0.68 1 K 0.67 0.53 0.54 0.17 0.53 0.31 0.4 0.37 0.36 0.52 1 Mg -0.02 -0.25 -0.24 0.03 0.34 0.07 -0.21 -0.01 0.09 -0.17 0.07 1 Zn 0.49 0.57 0.54 -0.01 0.26 0.08 0.53 0.24 0.07 0.53 0.43 -0.25 1 Cu 0.58 0.61 0.58 0.01 0.28 0.18 0.58 0.38 0.17 0.68 0.53 -0.23 0.56 1 Ti 0.5 0.17 0.14 0.42 0.28 0.92 0.16 0.76 0.92 0.69 0.48 0.13 0.12 0.25 1 Mn 0.55 0.37 0.35 0.2 0.25 0.51 0.34 0.55 0.5 0.65 0.42 -0.12 0.38 0.49 0.62 1 V 0.43 0.04 0.01 0.73 0.21 0.65 0.07 0.55 0.65 0.45 0.36 0.15 0.05 0.16 0.67 0.44 1 Ni 0.52 0.44 0.42 0.28 0.25 0.28 0.39 0.34 0.25 0.53 0.35 -0.16 0.53 0.48 0.3 0.54 0.39 Ni 1 1 Table 3e. Bologna (only PM2.5). PM TC OC SO42NO3EC Ca Fe K PM TC OC SO42- NO3- EC Ca Fe K 1 0.83 0.84 0.46 0.94 0.59 -0.18 0.52 0.49 1 0.99 0.25 0.81 0.8 0.23 0.54 0.76 1 0.3 0.8 0.69 0.23 0.48 0.74 1 0.32 0.01 -0.03 0.19 0.18 1 0.65 -0.2 0.44 0.4 1 0.19 0.65 0.64 1 0.31 -0.14 1 0.28 1 Table 4a. Association (percent change and 95% confidence interval for an IQR change) between total mortality and PM constituents at lag 1. Single pollutant models included only one PM constituent at a time. Models adjusted for PM levels were obtained using the constituent residual method: first, a linear regression model was fitted for each constituent with PM levels as the only predictor; then both the residuals from this model and PM levels are introduced as predictors in the health model to obtain the percent changes. Results were obtained from a meta-analysis combining the city-specific results of the cities with available data on each constituent. Species Total Carbon OC SO42NO3SiO2 EC Ca Al2O3 Fe K Mg Zn Cu Ti Mn V Ni Single pollutant PM10 PM2.5 Species effect Species effect 1.2 (0.2, 2.2)* -1.6 (-7.0, 4.1) -0.0 (-1.3, 1.3) -0.2 (-1.1, 0.7) 0.9 (-0.3, 2.1) 0.5 (-0.6, 1.6) 1.3 (-0.7, 3.3) 0.8 (-0.3, 1.9) 2.1 (0.3, 3.9)* 1.0 (0.3, 1.7)* 0.2 (-0.9, 1.3) 0.4 (-0.7, 1.5) 0.6 (-1.4, 2.6) 1.0 (0.0, 2.0)* 1.2 (0.0, 2.4)* 0.0 (-1.2, 1.3) -0.1 (-1.0, 0.8) 1.1 (-0.7, 2.9) -0.7 (-4.8, 3.7) -0.6 (-2.2, 1.0) -0.2 (-1.9, 1.6) 0.2 (-0.7, 1.2) -0.2 (-2.6, 2.2) 0.6 (-0.1, 1.2) 0.2 (-0.8, 1.2) 1.0 (-0.3, 2.3) 0.3 (-0.4, 1.0) 0.7 (-0.7, 2.1) -0.5 (-4.2, 3.4) -0.0 (-1.0, 1.0) 0.7 (-0.2, 1.6) 0.9 (-0.2, 2.1) -0.6 (-1.5, 0.4) -0.1 (-2.1, 1.9) Adjusting for total PM PM10 Species effect PM effecta 1.1 (-1.3, 3.6) 0.0 (-1.4, 1.4) -0.9 (-2.3, 0.6) -1.4 (-2.6, -0.2)* 0.9 (-0.6, 2.4) 0.1 (-1.7, 1.8) 0.9 (-1.9, 3.7) 0.8 (-0.5, 2.2) 3.2 (1.0, 5.5)* 0.8 (-0.1, 1.7) 0.1 (-1.2, 1.3) 0.1 (-0.7, 0.9) -0.0 (-1.6, 1.5) 0.8 (-0.5, 2.0) 0.7 (-0.8, 2.3) -0.6 (-1.6, 0.4) -0.8 (-1.8, 0.2) * p<0.05 a The PM10 effect in single pollutant model was 1.2 (-0.2, 2.7). b The PM2.5 effect in single pollutant model was -0.5 (-2.2, 1.2). 1.3 (0.1, 2.6)* 0.2 (-3.7, 4.2) 1.2 (-0.3, 2.7) 1.4 (0.1, 2.8)* 1.1 (-0.9, 3.2) 0.2 (-3.4, 3.9) 1.2 (-0.5, 2.9) 1.3 (-0.1, 2.7) 1.4 (-0.1, 2.9) 1.3 (-0.2, 2.8) 1.3 (-0.1, 2.7) 1.1 (-1.4, 3.7) 1.2 (-0.8, 3.2) 1.3 (-0.2, 2.8) 1.3 (-0.5, 3.1) 1.2 (-0.3, 2.7) 1.3 (-0.5, 3.1) PM2.5 Species effect 2.8 (-0.0, 5.7) 2.0 (-1.9, 6.1) -0.3 (-2.2, 1.6) 1.9 (-2.7, 6.8) 0.4 (-0.6, 1.4) 1.3 (-1.3, 4.0) 0.7 (0.0, 1.5)* 0.4 (-0.7, 1.4) 1.7 (0.2, 3.2)* 0.4 (-0.3, 1.1) 1.0 (-0.5, 2.4) -0.8 (-5.5, 4.2) 0.0 (-1.0, 1.1) 0.9 (-0.0, 1.9) 1.4 (0.2, 2.8)* -0.5 (-1.5, 0.6) 0.3 (-1.9, 2.5) PM effectb -0.4 (-2.4, 1.5) -2.0 (-7.3, 3.6) -0.7 (-2.6, 1.2) -0.7 (-2.6, 1.2) -0.7 (-2.8, 1.4) -1.7 (-7.0, 3.8) -0.5 (-2.4, 1.4) -0.7 (-2.8, 1.4) -0.7 (-2.5, 1.2) -0.8 (-2.6, 1.1) -0.6 (-2.7, 1.5) -0.9 (-3.0, 1.2) -0.7 (-2.8, 1.4) -0.5 (-2.6, 1.6) -0.6 (-2.7, 1.5) -0.7 (-2.8, 1.4) -0.8 (-2.9, 1.3) Table 4b. Association (percent change and 95% confidence interval for an IQR change) between cardiovascular admissions and PM constituents at lag 0. Single pollutant models included only one PM constituent at a time. Models adjusted for PM levels were obtained using the constituent residual method: first, a linear regression model was fitted for each constituent with PM levels as the only predictor; then both the residuals from this model and PM levels are introduced as predictors in the health model to obtain the percent changes. Results were obtained from a meta-analysis combining the city-specific results of the cities with available data on each constituent. Single pollutant Specie Total Carbon OC SO42NO3SiO2 EC Ca Al2O3 Fe K Mg Zn Cu Ti Mn V Ni Adjusting for PM mass PM10 PM2.5 Specie effect Specie effect Specie effect PM10 PM effect Specie effect PM2.5 PM effect 0.961 (-0.791, 2.745) 0.719 (-1.001, 2.469) 0.915 (-0.2, 2.043) 0.775 (0.01, 1.547)* 0.732 (-0.289, 1.765) 1.436 (-0.662, 3.579) 1.183 (0.209, 2.166) 0.636 (-0.285, 1.567) 2.117 (0.458, 3.803)* -0.18 (-0.873, 0.516) -0.13 (-1.586, 1.347) 1.115 (-0.109, 2.355) 0.535 (-1.997, 3.133) 1.28 (0.341, 2.227)* 2.042 (0.995, 3.1)* 0.594 (-0.14, 1.333) 0.936 (-0.01, 1.893) 1.502 (-0.251, 3.288) 1.851 (-0.157, 3.9) 1.535 (0.251, 2.835)* 0.723 (-0.421, 1.881) -0.083 (-0.893, 0.732) 2.432 (0.521, 4.379)* 0.004 (-0.556, 0.569) -0.01 (-0.815, 0.801) 1.885 (-0.961, 4.814) -0.034 (-0.965, 0.905) -0.22 (-1.442, 1.015) 0.374 (-0.244, 0.997) -0.017 (-0.859, 0.832) 0.151 (-0.721, 1.031) 1.323 (0.376, 2.279)* 0.274 (-0.466, 1.021) 0.432 (-3.291, 4.301) 0.033 (-1.534, 1.625) 0.163 (-1.177, 1.523) 0.051 (-1.249, 1.369) 0.031 (-1.223, 1.302) -0.116 (-1.418, 1.202) 0.715 (-0.81, 2.265) -1.118 (-4.469, 2.35) -0.202 (-1.377, 0.987) 0.639 (-3.53, 4.989) -1.285 (-2.168, -0.394) -1.047 (-2.227, 0.147) 1.239 (-0.896, 3.421) -0.411 (-3.617, 2.9) 0.609 (-0.617, 1.85) 1.922 (0.471, 3.394)* 0.017 (-0.854, 0.897) 0.581 (-0.398, 1.57) 1.625 (0.525, 2.736)* 1.854 (0.573, 3.152)* 1.671 (0.576, 2.778)* 1.649 (0.559, 2.751)* 1.733 (0.582, 2.897)* 1.805 (0.513, 3.113)* 1.653 (0.569, 2.748)* 1.703 (0.61, 2.807)* 1.698 (0.613, 2.794)* 1.6 (0.51, 2.701)* 1.706 (0.616, 2.807)* 1.693 (0.47, 2.93)* 1.528 (0.381, 2.689)* 1.726 (0.636, 2.829)* 1.636 (0.49, 2.796)* 1.711 (0.62, 2.813)* 1.628 (0.482, 2.788)* -0.93 (-4.867, 3.168) 1.282 (-1.737, 4.395) 1.111 (-0.459, 2.707) -0.686 (-2.221, 0.872) -0.276 (-1.117, 0.57) 2.058 (-0.082, 4.245) -0.169 (-0.768, 0.434) -0.199 (-1.044, 0.653) 0.574 (-0.685, 1.851) -0.29 (-1.154, 0.58) -0.515 (-1.783, 0.769) 0.269 (-0.366, 0.91) -0.23 (-1.107, 0.654) -0.095 (-1.026, 0.845) 1.143 (0.066, 2.232)* -0.026 (-0.886, 0.84) 0.267 (-2.952, 3.594) 1.922 (0.305, 3.566)* 1.869 (-0.291, 4.077) 1.689 (0.146, 3.257)* 1.857 (0.329, 3.409)* 1.654 (-0.101, 3.441) 2.56 (0.322, 4.848)* 1.821 (0.286, 3.379)* 1.686 (-0.068, 3.473) 2.009 (0.491, 3.551)* 1.871 (0.344, 3.421)* 1.617 (-0.14, 3.405) 1.673 (-0.082, 3.46) 1.657 (-0.101, 3.447) 1.653 (-0.107, 3.445) 1.864 (0.109, 3.649) 1.72 (-0.032, 3.503) 1.635 (-0.128, 3.431) Figure 1a. Association between total, cardiovascular and respiratory mortality and PM10 constituents for lags 0, 1 and 2. Associations are expressed as the percent change in mortality for an interquartile range increase in PM10 constituent levels with 95% confidence intervals. Results were obtained from a meta-analysis combining the city-specific results of the cities with available data on each constituent. The cities contributing to the results for a particular constituent are indicated on top of each graph (Bc: Barcelona, M: Madrid, H: Huelva, R: Rome). A single star (*) indicates a p-value<0.10, while two stars (**) indicate a p-value<0.05. Figure 1b. Association between cardiovascular and respiratory hospital admissions and PM10 constituents for lags 0, 1 and 2. Associations are expressed as the percent change in hospital admissions for an interquartile range increase in PM10 constituent levels with 95% confidence intervals. Results were obtained from a meta-analysis combining the city-specific results of the cities with available data on each constituent. The cities contributing to the results for a particular constituent are indicated on top of each graph (Bc: Barcelona, M: Madrid, H: Huelva, R: Rome). A single star (*) indicates a p-value<0.10, while two stars (**) indicate a p-value<0.05. Figure 2a. Association between cardiovascular and respiratory mortality and PM2.5 constituents for lags 0, 1 and 2. Associations are expressed as the percent change in mortality for an interquartile range increase in PM2.5 constituent levels with 95% confidence intervals. Results were obtained from a meta-analysis combining the city-specific results of the cities with available data on each constituent. The cities contributing to the results for a particular constituent are indicated on top of each graph (Bc: Barcelona, M: Madrid, H: Huelva, Bg: Bologna). A single star (*) indicates a p-value<0.10, while two stars (**) indicate a p-value<0.05. Figure 2b. Association between cardiovascular and respiratory admissions and PM2.5 constituents for lags 0, 1 and 2. Associations are expressed as the percent change in hospital admissions for an interquartile range increase in PM2.5 constituent levels with 95% confidence intervals. Results were obtained from a meta-analysis combining the city-specific results of the cities with available data on each constituent. The cities contributing to the results for a particular constituent are indicated on top of each graph (Bc: Barcelona, M: Madrid, H: Huelva, Bg: Bologna). A single star (*) indicates a p-value<0.10, while two stars (**) indicate a p-value<0.05. Figure 3a. Forest plots of the association (percent change and 95% confidence interval for an IQR change in PM10 constituent) between total mortality and iron or manganese at lag 1. Figure 3b. Forest plots of the association (percent change and 95% confidence interval for an IQR change in PM10 constituent) between cardiovascular admissions and iron or manganese at lag 0.