Mark S. Goldberg, 1,2 Richard T. Burnett, 3 Jeffrey Brook, 4 John C. Bailar III, 5 Marie-France Valois, 1 and Renaud Vincent 3

Transcription

1 American Journal of Epidemiology Copyright by the Johns Hopkins University Bloomberg School of Public Health All rights reserved Vol., No. Printed in U.S.A. Cause-specific Daily Mortality and Tropospheric Ozone Goldberg et al. Associations between Daily Cause-specific Mortality and Concentrations of Ground-level Ozone in Montreal, Quebec Mark S. Goldberg,, Richard T. Burnett, Jeffrey Brook, John C. Bailar III, Marie-France Valois, and Renaud Vincent The authors investigated the association between daily variations in ozone and cause-specific mortality. Fixed-site air pollution monitors in Montreal, Quebec, provided daily mean levels of ozone, particles, and other gaseous pollutants. Information on the date and underlying cause of death was obtained for residents of Montreal who died in the city between 8 and. The authors regressed the logarithm of daily counts of cause-specific mortality on mean levels of ozone, after accounting for seasonal and subseasonal fluctuations in the mortality time series, non-poisson dispersion, and weather variables. The effect of ozone on mortality was generally higher in the warm season and among persons aged 6 years or over. For an increase in the -day running mean concentration of ozone of. µg/m, the percentage of increase in daily deaths in the warm season was the following: nonaccidental deaths,.% (% confidence interval (CI):.,.); cancer,.% (% CI:., 6.); cardiovascular diseases,.% (% CI:.,.); and respiratory diseases, 6.6% (% CI:.8,.8). These results were independent of the effects of other pollutants and were consistent with a log-linear response function. Am J Epidemiol ;:8 6. air pollution; cardiovascular diseases; mortality; neoplasms; ozone; respiratory tract diseases The predominant source of tropospheric ozone is from photochemical reactions involving volatile organic compounds and nitrogen oxides, for which internal combustion and the use of fossil fuels are important sources. Exposure to ozone is known to cause a number of deleterious effects on health ( ), including symptoms (e.g., eye discomfort, headache) (), increased reactivity of the airways, inflammation of the lung, decrements in lung function (), and decreased capacity for exercise (6). There are a limited number of time-series studies of the associations between ambient ozone and daily mortality ( ) and admissions to hospitals and visits to the emergency room for cardiovascular Received for publication October,, and accepted for publication May,. Abbreviations: CI, confidence interval; LOESS, locally weighted regression smoothers; MPC, mean percentage of change in daily mortality; PM and PM., particulates having aerodynamic diameters of µm or under and. µm or under, respectively. Department of Medicine, McGill University, Montreal, Quebec, Canada. Joint Departments of Epidemiology and Biostatistics and of Occupational Health, McGill University, Montreal, Quebec, Canada. Environmental Health Directorate, Health Canada, Ottawa, Ontario, Canada. Air Quality Processes Research Division, Meteorological Service of Canada, Environment Canada, and Department of Public Health Sciences, University of Toronto, Toronto, Ontario, Canada. Department of Health Studies and Harris School of Public Policy, University of Chicago, Chicago, IL. Correspondence to Dr. Mark Goldberg, Departments of Epidemiology and Biostatistics and of Occupational Health, McGill University, Pine Ave. West, Room A, Montreal, Quebec, Canada HA A ( mark.goldberg@mcgill.ca). and respiratory conditions ( ). It has been suggested that persons whose health is failing and who have difficulty regulating their physiologic setpoints may be at higher risk from exposure to ambient air pollution (). Thus, it can be hypothesized that potentially susceptible subgroups may include not only those with cardiorespiratory conditions but also those with illnesses that have important systemic effects, such as diabetes and cancer. The purpose of this paper is to investigate this hypothesis by examining the association between cause-specific daily mortality and ozone from a time-series study of air pollution and daily mortality in Montreal, Quebec, a large metropolitan area that experiences relatively low levels of air pollution. MATERIALS AND METHODS We have described the methodology used in this study ( ). Briefly, the study population consisted of all residents of Montreal who died in the metropolitan area during the period from 8 to and who were registered with the universal provincial health insurance plan. Subjects were identified from the provincial database of death certificates that provided information on the decedents date of death, place of death, residence at the time of death, and underlying cause of death (coded to the International Classification of Diseases, Ninth Revision). The data were received in denominalized form, and the study was approved by the Institutional Review Board of McGill University. Measurements of ozone were made using chemiluminescence (Bendix model 8 ozone analyzer; Bendix Downloaded from by guest on January 8

2 88 Goldberg et al. Corporation, San Diego, California) at hourly intervals at nine fixed-site monitoring stations, and these data were used to calculate a daily city-wide average. Although other metrics have been used to characterize concentrations of ozone (e.g., maximum 8-hour running average), we decided to use the simple daily mean, as it was well correlated with these other metrics (Pearson s correlation coefficients for daily mean concentrations with a daily maximum concentration of.8, a daily mean concentration between 8 a.m. and 8 p.m. of., a daily mean concentration between a.m. and 6 p.m. of., and a daily highest 8-hour maximum of.). Nitrogen dioxide, sulfur dioxide, carbon monoxide, and the coefficient of haze (measures elemental carbon) were monitored every hours at several fixed-site stations in Montreal. Particle mass (total suspended particles, particles having aerodynamic diameters of µm or under and. µm or under (PM and PM. ), and sulfates in the coarse and fine fractions of the particles) was measured every 6 days (6, ). From July to September, the measurement schedule for PM and PM. was increased at one site to daily sampling (6). We also made use of daily measurements of total sulfate particles (86 ) from an acid rain monitoring station at Sutton, Quebec, a rural community about km southeast from the city. The average correlation between sulfates measured at this station and the two Montreal stations was.. Visibility, barometric pressure, temperature, total precipitation (distinguishing snow from rain), relative humidity, and dew point temperature were measured at Dorval International Airport. Visibility was converted into an extinction coefficient (a measure of light scattering and absorption, due mostly to sulfates) after accounting for relative humidity (8 ). We used the measurement at noon when there was no precipitation or the hour closest to noon without precipitation. We also developed statistical models for the period 86 to predict PM. and sulfates from PM. for days in which they were not measured. We used the coefficient of haze, the extinction coefficient, and total sulfates from the Sutton station as predictor variables (hereafter referred to as predicted PM. ). The R for the prediction model for PM. was. and for sulfates from PM. it was.8. Regarding statistical methods, analyses were conducted for all nonaccidental causes of death and selected other underlying causes of death. We used quasilikelihood estimation within the context of the generalized additive models () to model the logarithm of daily counts of causespecific deaths as functions of the predictor variables. We assumed that the daily counts of death were distributed approximately as a Poisson variate with constant over- or underdispersion. We used locally weighted regression smoothers (LOESS) for nonparametric terms. We regressed the natural logarithm of the daily number of deaths on a LOESS term for day of study, thus providing an adjustment for seasonal and subseasonal variations (temporal filter), another term to account for annual trends in daily mortality, and LOESS terms to adjust for the potential confounding Downloaded from by guest on January FIGURE. Scatterplot of daily number of nonaccidental deaths, Montreal, Canada, 8. The solid line is the locally weighted regression smooth representing the long-term trend in the data (span of % of the data). The total number of days in the time series is. Am J Epidemiol Vol., No.,

3 Cause-specific Daily Mortality and Tropospheric Ozone 8 effects of relevant weather variables. For each underlying cause of death, we selected the temporal filter with a smoothing bandwidth (span for the LOESS function) that produced a filtered time series that was consistent with a white noise process, using Bartlett s statistic as dicussed by Priestly (). This produced residual time series that had the least amount of serial autocorrelation. We then included in the model for each cause of death LOESS terms for selected combinations of weather variables. These combinations were derived from fitting a series of models that contained different sets of weather variables, evaluated across different lag periods (lags ), and we selected the ones that yielded the minimum Akaike Information Criterion ( ). Filtered and weather-adjusted models using daily concentrations of ozone averaged across the fixed-site monitoring stations were considered. We also estimated mortality at lag day and for the average of lags days (referred to as the -day mean). As there were strong seasonal variations in TABLE. Daily mortality by cause of death and age group, Montreal, Canada, 8 Underlying cause of death Nonaccidental Neoplasms Lung cancer Cardiovascular diseases Coronary artery disease Respiratory diseases Nonmalignant digestive diseases Other nonaccidental causes ICD-* codes Total no. of deaths,,6,8,, 8,,,6,6,6,6,6, 6,,8,,8,66,8,8,,,, No. of days in which there were no deaths 8, 8 6 6,68,,, 6 Mean Variance Variance mean th th Percentiles th th Downloaded from by guest on January Diabetes,6,,6,8, Renal diseases 8,8,6,,, Injuries and poisonings 8,,6, 6,8, * ICD-, International Classification of Diseases, Ninth Revision. Am J Epidemiol Vol., No.,

4 8 Goldberg et al. concentrations of ozone, we also assessed associations according to a cold season (October to March) and a warm season (April to September). Exposure-response functions for each metric of ozone were plotted, and we used an approximate F test to determine whether the fitted nonparametric smooth was consistent with a log-linear exposure-response function (). Assuming linearity, we also estimated the relative increase in the logarithmic number of daily deaths per unit increase in the concentration of ozone. The percentage of change in the mean number of daily deaths for an increase of ozone equal to its interquartile range was calculated (referred to as the mean percentage of change (MPC)). The associated upper and lower percent confidence limits on the mean percentage of change were obtained assuming that the estimated regression coefficient was distributed normally, with the standard error corrected for non-poisson dispersion. RESULTS There were, nonaccidental deaths during the study period 8 (figure ), with neoplasms, cardiovascular diseases, and respiratory diseases accounting for. percent,.8 percent, and 8. percent of all deaths, respectively (table ). Seventy-five percent of the deaths were among persons aged 6 years or over. Some of these cause-specific time series were overdispersed (e.g., cardiovascular diseases), and this arose from strong seasonal fluctuations in mortality. However, the filtering of each of these time series to white noise considerably reduced the overdispersion as well as the serial autocorrelation ( ). Table shows that levels of ozone, other gaseous pollutants, and particles were low compared with those of other North American and European cities (e.g., mean concentration of ozone was. µg/m ). Station-specific mean values of ozone over the -year period ranged from 6.6 to. µg/m, and the range of Pearson s correlation coefficients between pairs of monitoring stations was.. (, ). Ambient concentrations of ozone had strong seasonal variations, peaking in the summer months, and concentrations of ozone increased slightly over time (figure ). Table shows the principal results for ozone for each cause of death, unadjusted for other pollutants. Across all seasons, associations with the -day mean were generally stronger than those at lag and lag days. We did not find any statistically significant associations in any of the three lag periods considered among persons who died of lung cancer, respiratory diseases, nonmalignant digestive diseases, injuries and poisonings, and other nonaccidental causes of death. We did find, TABLE. Distributions of ozone, other gaseous pollutants, selected indices of air particulates, and weather variables, Montreal, Canada, 8 Ozone (µg/m ) Sulfur dioxide (µg/m ) Nitrogen dioxide (µg/m ) Nitrogen oxide (µg/m ) Carbon monoxide (ppm) Coefficient of haze (. unit/.8 linear meters) Predicted PM. *, Mean temperature ( C) No. of days of measurements Mean (SD*). (.).8 (.). (.).8 (.).8 (.). (.).6 (8.8) 6. (.8) Minimum th th Percentiles th th Interquartile range Downloaded from by guest on January Change in maximum temperature from the previous day ( C). (.) Mean dew point temperature ( C). (.6) Relative humidity (%) 6. (.) Change in barometric pressure from previous day (kpa). (.) * SD, standard deviation; PM., particulates having aerodynamic diameter of. µm or less. For the period 86. Am J Epidemiol Vol., No.,

5 Cause-specific Daily Mortality and Tropospheric Ozone 8 FIGURE. Scatterplot of daily concentrations of ozone, Montreal, Canada, 8. The solid line is the locally weighted regression smooth representing the long-term trend in the data (span of % of the data). however, positive and statistically significant associations across all lag periods for all nonaccidental causes of death combined (e.g., MPC -day mean.6 percent, percent confidence interval (CI):.,.), neoplasms (lag and lag days; MPC lag.6 percent, percent CI:.,.), cardiovascular diseases (MPC -day mean. percent, percent CI:.,.), and coronary artery disease (MPC -day mean. percent, percent CI:.6,.). In general, these associations were restricted to persons 6 years of age or over, although associations were found among persons under the age of 6 years who died of coronary artery disease (lag day, -day mean). Similar effects were found in men and women (data not shown). The above estimates were based on the assumption that the fitted exposure-response function for ozone was consistent with a linear function. Using a nonparametric test for linearity (), we found, however, that all nonaccidental mortality for the three lag periods (test for linearity p value <.; figure ) and neoplasms at the -day mean (p <.) showed a jshaped relation, with lower effects occurring at intermediate concentrations of ozone. These nonlinear relations are due to the negative correlations between ozone and mortality in the cold season and positive ones in the warm season. (The mean level of ozone in the cold season was.8 µg/m and in the warm season it was. µg/m.) Table and figure show this explicitly: Associations for all nonaccidental causes, neoplasms, and respiratory diseases were positive and statistically significant in the warm season; lung cancer showed a similar pattern but its association was of borderline significance. Figure shows the results for ozone evaluated at the - day mean adjusted simultaneously for other pollutants (carbon monoxide, nitrogen dioxide, sulfur dioxide, and coefficient of haze) for both age groups combined. The adjustments did not greatly alter the findings that were reported in table, but the magnitude of the associations for deaths from neoplasms and respiratory diseases was increased somewhat. Results similar to these were also found when predicted PM. and total sulfates (from the Sutton monitoring station) were used as metrics for particle pollution instead of the coefficient of haze (data not shown). DISCUSSION To compare the findings for nonaccidental causes of death with those of other studies, we rescaled our estimates for an increase in ozone of µg/m. This yielded a. percent increase (from the results for the -day mean), and this is comparable with findings from most other studies (a. percent increase in Philadelphia, Pennsylvania (), a. percent increase in Amsterdam, the Netherlands (), a. percent increase in eight European studies) but is larger than that found in Mexico City, Mexico (.6 percent) (). In just a few studies have cause-specific daily mortality and ground-level ozone been investigated ( ). Positive associations with deaths from cardiovascular diseases have been found in four of five studies (, ), and associations with deaths from respiratory illnesses were found in three of six studies ( ). There have been rather more investiga- Downloaded from by guest on January Am J Epidemiol Vol., No.,

6 8 Goldberg et al. TABLE. Summary estimates of the mean percentage of change in cause-specific daily mortality across the interquartile range of lagged exposures to ozone, Montreal, Canada, 8 * Cause of death Mean % of change Lag days % CI Mean % of change Lag day % CI Mean % of change -day mean Warm season: -day mean Mean % of % CI % CI change Nonaccidental....,..,..6, ,..6,.8., ,..,..66, ,..,.8.,.6 Neoplasms Lung cancer Cardiovascular diseases Coronary artery disease Respiratory diseases Nonmalignant digestive diseases Injuries and poisonings Other nonaccidental causes Diabetes mellitus ,..,..6,.8.,.., 6..68,..,..68,..,.8.,..,.8.,..,..8,..,..,..6,.8., Renal diseases..8,..8.,...,..88., 6. * The statistical model for each cause of death was E(log(y i )) = α + LOESS(i, span = x) + LOESS(year) + multiple weather variables + β pollutant, where i is an indicator for day and x is the selected span (in days, divided by the total number of days in the study period (n = )). Mean percentages of change were calculated for an increase of ozone equal to the interquartile range (. µg/m ). CI, confidence interval. tions of hospital admission data. In a combined analysis of data from the largest cities in Canada, including Montreal (), positive associations for cardiovascular diseases were found for concentrations of ozone lagged by days. A positive association between ozone and ischemic heart disease was also found in Helsinki, Finland (), but no associations were found in studies conducted in Tucson, Arizona (), London, England (6), and Edinburgh, Scotland (8). The results for respiratory illnesses are more consistent, with positive associations found in a series of multihospital studies in the Province of Ontario (,, 6) and in a combined analysis of 6 Canadian cities (). As well, positive associations have been found in a number of American cities: Buffalo, Albany, and New York, New York (); Spokane, Washington (); New Haven, Connecticut, and Tacoma, Washington (); Minneapolis, Minnesota (8); ,..,..,.8.6,.8.,..,.6.8,..,.8.,..,.6.,.6.6,.6.,..,..6, 6..,..8,.., ,..66,..6,..8,..,.6.6, 6.6.,..,..6,.8.6,..,..,.6.8,.8.6, 6..8,..,.6.,.8., ,..6,.86., 6.., 8..,..,..,..,.88.,.8.,..,..,..,..,.6.,.., 8..8, 8..8,.6 Birmingham, Alabama (); Detroit, Michigan (); Seattle, Washington (); and Atlanta, Georgia (for childhood asthma) (). Statistically significant positive associations have been reported in studies in London, England (), and Valencia, Spain (), but not in Amsterdam and Rotterdam, the Netherlands (), and Edinburgh, Scotland (8). The effects of ozone in Montreal have also been investigated in time-series studies of hospital admissions and emergency room visits ( ). During the period 8 88, Delfino et al. () found a positive association during the summer between hospital admissions for respiratory conditions, especially asthma, and increasing levels of ambient ozone. They also found among persons 6 years or over positive associations in the summer for emergency room visits for respiratory conditions (6, ); the estimated effect was about. emergency room visits per unit µg/m increase in ozone. Downloaded from by guest on January Am J Epidemiol Vol., No.,

7 Cause-specific Daily Mortality and Tropospheric Ozone 8 FIGURE. Exposure-response function for the mean percentage of change in daily nonaccidental mortality evaluated at the -day mean for an increase in ambient ozone, Montreal, Canada, 8. The estimated mean percentage of change in daily cause-specific mortality across the interquartile range is shown by the solid line, and the dotted lines represent twice the pointwise standard error. Downloaded from by guest on January FIGURE. Mean percentage of change in daily mortality from selected causes of death evaluated at the -day mean for an increase in the interquartile range of ozone, according to season, Montreal, Canada, 8. The estimated mean percentage of change in daily causespecific mortality across the interquartile range is shown by the large filled circle, and the horizontal bars represent the % confidence limits. CAD, coronary artery disease. Am J Epidemiol Vol., No.,

8 8 Goldberg et al. FIGURE. Mean percentage of change in daily mortality from selected causes of death for an increase in the interquartile range of ozone, adjusted for concentrations of carbon dioxide, carbon monoxide, nitrogen dioxide, sulfur dioxide, and the coefficient of haze, Montreal, Canada, 8. The estimated mean percentage of change in daily cause-specific mortality across the interquartile range is shown by the large filled circle, and the horizontal bars represent the % confidence limits. All pollutants were evaluated at the -day mean. CAD, coronary artery disease. Our findings need to be interpreted in light of the assumptions, strengths, and limitations of the study design. Although outdoor measurements of ozone were used as a surrogate for population exposures, this metric is likely to be valid because during the warm periods of the year ozone permeates through open windows into indoor environments. This penetration occurs because few homes in Montreal are centrally air conditioned or have noncentralized air conditioners. We chose to split the year into only two seasons rather than four to minimize the loss of statistical power due to the reduction in the number of days in each seasonal time series. We also drew on the results of Burnett et al. () who found associations with ozone from April to December for hospital admissions for respiratory illnesses. There tends to be a high degree of correlation among dayto-day changes in the concentrations of air pollutants (gases and particles) in urban air. This is due to similarities in the sources, atmospheric chemical processes, and influences of weather among urban pollutants. We accounted for the joint effects of nitrogen dioxide, sulfur dioxide, carbon monoxide, and ambient particles in determining the association between ozone and daily mortality. Daily measurements of concentrations of particles were not available for the entire study period and, thus, the adjustments for particles were based on the coefficient of haze and predicted values for fine particles (, ). We found, however, that the associations between ozone and mortality were relatively unaffected after adjusting for the various indices of particles and other gaseous pollutants. This is consistent with other studies in which weak or no confounding effects of other pollutants were observed (, ). We could not control for the effects of infectious disease epidemics (e.g., influenza, which occurs mostly in the fall and winter, when particle levels are increased) because there are no databases that could be used for this purpose. These epidemics occur mostly in the winter months and the effects of ozone are mostly in the warm months of the year, so that these epidemics should not have confounded the warm season-specific estimates. We are unaware of any studies in which adjusting for influenza epidemics removed associations between mortality and air pollution (, 8, ). We assumed that any inaccuracies in coded underlying causes of death had constant probability distributions over the -year period of this study. There are no data reporting the accuracy of underlying causes of death in Quebec. In other jurisdictions, it has been found that the accuracy of coding varies with cause of death ( ). The site of cancer is usu- Downloaded from by guest on January Am J Epidemiol Vol., No.,

9 Cause-specific Daily Mortality and Tropospheric Ozone 8 ally coded reasonably accurately (above 8 percent), but respiratory and cardiovascular diseases are often confused. In particular, when persons have both conditions concurrently and both contributed to death, there may be some uncertainty about which cause should be selected as the primary underlying cause. In other instances, there may be errors in selecting one underlying cause in a complex chain of health events (e.g., cancer leading to pneumonia and then to respiratory failure). Regarding possible mechanisms, although the concentrations of ozone in Montreal are rather low (median value, µg/m ), there were excursions of 8 µg/m about percent of the time (th percentile, µg/m ; th percentile, 8 µg/m ; th percentile, µg/m ). From a toxicologic perspective, these spikes reach values for which biologic responses in animal models or human volunteers have been found. We postulate that ozone may alter the structure of the surfactant or the extracellular lining of the lungs and that there may be interactions with the functioning of macrophages. This hypothesis is supported by data showing that ozone inhaled by rats will decrease the capacity of macrophages to ingest and kill Listeria monocytogenes. Ozone suppresses the development of cellular immune responses and delays hypersensitivity and lymphoproliferative responses to Listeria antigen in cells (), it causes residual damage in rats whose lungs were injured by influenza (), and it increases bacterial infections (6). In macrophages, ozone reduces binding and responsivity to interferon gamma (), decreases clearance of bacteria from the lungs, and interferes with the coordination of other cellular responses (8). Because persons are exposed simultaneously to many air pollutants, the associations with ozone observed in this study and in other studies may be due to the combined effects of all pollutants. Support for this hypothesis comes from studies showing that injuries in the lungs of rats from exposures to ozone will be amplified by inhaled urban particles ( 6). In addition, repeated inhalation of sulfuric acid aerosol reduces uptake and intracellular killing of bacteria by macrophages and alters a number of their immunomodulatory functions (6). The responses of macrophages to a respiratory virus have also been shown to be affected by ambient particulate matter (6). ACKNOWLEDGMENTS This study was supported financially through contracts with the Health Effects Institute, Cambridge, Massachusetts, and with the Toxics Substances Research Initiative, Health Canada. The authors thank the Montreal Urban Community and Environment Canada for providing the air pollution and weather data, and they are grateful to the Ministère de la Santé et des Services sociaux de Québec for providing the health data. They gratefully acknowledge the assistance of David Johnson and Claude Gagnon. Dr. Goldberg gratefully acknowledges receipt of a National Health Scholar Award from the National Health and Research Development Program of Health Canada and support from the Fonds de la Recherche en Santé du Québec. Am J Epidemiol Vol., No., REFERENCES. Lippmann M. Health effects of ozone. A critical review. J Air Pollut Control Assoc 8;:6.. Neuhaus JM, Kalbfleisch JD, Hauck WW. A comparison of cluster-specific and population-averaged approaches for analyzing correlated binary data. Int Stat Rev ;:.. Health effects of outdoor air pollution. Committee of the Environmental and Occupational Health Assembly of the American Thoracic Society. Am J Respir Crit Care Med 6; :.. Strom J, Alfredson L, Malmfors T, et al. Ozone: causation and aggravation of lung diseases. Indoor Environ ;:6 8.. Brunekreef B, Dockery DW, Krzyzanowski M. Epidemiologic studies on short-term effects of low levels of major ambient air pollution components. Environ Health Perspect ; (suppl ):. 6. Adams WC. Effects of ozone exposure at ambient air pollution episode levels on exercise performance. Sports Med 8;:.. Anderson HR, Ponce de Leon A, Bland JM, et al. Air pollution and daily mortality in London: 8. BMJ 6;: Sunyer J, Castellsague J, Saez M, et al. Air pollution and mortality in Barcelona. J Epidemiol Community Health 6; (suppl ):s6 8.. Zmirou D, Barumandzadeh T, Balducci F, et al. Short term effects of air pollution on mortality in the city of Lyon, France, 8. J Epidemiol Community Health 6;(suppl ):s.. Dab W, Medina S, Quenel P, et al. Short term respiratory health effects of ambient air pollution: results of the APHEA project in Paris. J Epidemiol Community Health 6; (suppl ):s 6.. Borja Aburto VH, Loomis DP, Bangdiwala SI, et al. Ozone, suspended particulates, and daily mortality in Mexico City. Am J Epidemiol ;: Kelsall JE, Samet JM, Zeger SL, et al. Air pollution and mortality in Philadelphia, 88. Am J Epidemiol ;6: 6.. Bates DV, Sizto R. Relationship between air pollutant levels and hospital admissions in southern Ontario. Can J Public Health 8;:.. Bates DV, Sizto R. Air pollution and hospital admissions in southern Ontario: the acid summer haze effect. Environ Res 8;:.. Thurston GD, Ito K, Kinney PL, et al. A multi-year study of air pollution and respiratory hospital admissions in three New York State metropolitan areas: results for 88 and 8 summers. J Expo Anal Environ Epidemiol ;:. 6. Burnett RT, Dales RE, Raizenne ME, et al. Effects of low ambient levels of ozone and sulphates on the frequency of respiratory admissions to Ontario hospitals. Environ Res ;6:.. White MC, Etzel RA, Wilcox WD, et al. Exacerbations of childhood asthma and ozone pollution in Atlanta. Environ Res ;6: Schwartz J. PM, ozone, and hospital admissions for the elderly in Minneapolis-St. Paul, Minnesota. Arch Environ Health ;:66.. Schwartz J. Air pollution and hospital admissions for the elderly in Detroit, Michigan. Am J Respir Crit Care Med ;:68.. Schwartz J. Short term fluctuations in air pollution and hospital admissions of the elderly for respiratory disease. Thorax ;: 8.. Schwartz J. Air pollution and hospital admissions for respiratory disease. Epidemiology 6;: 8.. Ponce de Leon A, Anderson HR, Bland JM, et al. Effects of air pollution on daily hospital admissions for respiratory disease in London between 8 88 and. J Epidemiol Community Health 6;(suppl ):s6.. Ponka A, Virtanen M. Low-level air pollution and hospital admissions for cardiac and cerebrovascular diseases in Helsinki. Am J Public Health 6;86: 8. Downloaded from by guest on January

10 86 Goldberg et al.. Schwartz J. Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology ;8:.. Burnett RT, Dales RE, Brook JR, et al. Association between ambient carbon monoxide levels and hospitalization for congestive heart failure in the elderly in Canadian cities. Epidemiology ;8:6. 6. Poloniecki JD, Atkinson RW, Ponce de Leon A, et al. Daily time series for cardiovascular hospital admissions and previous day s air pollution in London, UK. Occup Environ Med ;:.. Burnett RT, Brook JR, Yung WT, et al. Association between ozone and hospitalization for respiratory diseases in 6 Canadian cities. Environ Res ;:. 8. Prescott GJ, Cohen GR, Elton RA, et al. Urban air pollution and cardiopulmonary ill health a. year time series study. Occup Environ Med 8;:6.. Tenias JM, Ballester F, Rivera ML. Association between hospital emergency visits for asthma and air pollution in Valencia, Spain. Occup Environ Med 8;:.. Sheppard L, Levy D, Norris G, et al. Effects of ambient air pollution on nonelderly asthma hospital admissions in Seattle, Washington, 8. Epidemiology ;:.. Schouten JP, Vonk JM, de Graaf A. Short term effects of air pollution on emergency hospital admissions for respiratory disease: results of the APHEA project in two major cities in the Netherlands, 8. J Epidemiol Community Health 6;(suppl ):s.. Frank R, Tankersley CG. The association between airborne particles and daily mortality rate: an explanatory hypothesis. Presented at the International Symposium on Health Effects of Particulate Matter in Ambient Air, Prague, Czechoslovakia, April,.. Goldberg MS, Bailar JC rd, Burnett R, et al. Identifying subgroups of the general population that may be susceptible to short-term increases in particulate air pollution: a time series study in Montreal, Quebec. Cambridge, MA: Health Effects Institute,.. Goldberg MS, Burnett RT, Bailar JC rd, et al. The association between daily mortality and ambient air particle pollution in Montreal, Quebec.. Nonaccidental mortality. Environ Res ;86:.. Goldberg MS, Burnett RT, Bailar JC rd, et al. The association between daily mortality and ambient air particle pollution in Montreal, Quebec.. Cause-specific mortality. Environ Res ;86: Brook JR, Wiebe AW, Woodhouse SW, et al. Fine particle strong acidity, sulphate, PM, PM., and related gaseous species observed at multiple locations in Canada: concentrations and spatial-temporal relationships. Atmos Environ ; : 6.. Brook JR, Dann TF, Burnett RT. The relationship among TSP, PM, PM., and inorganic constituents of atmospheric particulate matter at multiple Canadian locations. J Air Waste Manag Assoc ;6:. 8. Ozkaynak H, Schatz AD, Thurston GD, et al. Relationships between aerosol extinction coefficients derived from airport visual range observations and alternative measures of airborne particle mass. J Air Pollut Control Assoc 8;:6 8.. Kinney PL, Ozkaynak H. Associations of daily mortality and air pollution in Los Angeles County. Environ Res ;:.. Delfino RJ, Becklake MR, Hanley JA, et al. Estimation of unmeasured particulate air pollution data for an epidemiological study of daily respiratory morbidity. Environ Res ; 6: 8.. Hastie AT, Tibshirani R. Generalized additive models. London, England: Chapman and Hall,.. Priestly MB. Spectral analysis of time series. Burlington, MA: Academic Press, 8.. Verhoeff AP, Hoek G, Schwartz J, et al. Air pollution and daily mortality in Amsterdam. Epidemiology 6;:.. Schwartz J. Air pollution and hospital admissions for the elderly in Birmingham, Alabama. Am J Epidemiol ;: Delfino RJ, Becklake MR, Hanley JA. The relationship of urgent hospital admissions for respiratory illnesses to photochemical air pollution levels in Montreal. Environ Res ; 6:. 6. Delfino RJ, Murphy-Moulton AM, Becklake MR. Emergency room visits for respiratory illnesses among the elderly in Montreal: association with low level ozone exposure. Environ Res 8;6:6.. Delfino RJ, Murphy-Moulton AM, Burnett RT, et al. Effects of air pollution on emergency room visits for respiratory illnesses in Montreal, Quebec. Am J Respir Crit Care Med ;: Spix C, Heinrich J, Dockery D, et al. Air pollution and daily mortality in Erfurt, east Germany, 8 8. Environ Health Perspect ;:8 6.. Vigotti MA, Rossi G, Bisanti L, et al. Short term effects of urban air pollution on respiratory health in Milan, Italy, 8 8. J Epidemiol Community Health 6;(suppl ): s.. Alderson MR, Meade TW. Accuracy of diagnosis on death certificates compared with that in hospital records. Br J Prev Soc Med 6;:.. de Faire U, Friberg L, Lorich U, et al. A validation of cause-ofdeath certification in,6 deaths. Acta Med Scand 6;: 8.. Engel LW, Strauchen JA, Chiazze L Jr, et al. Accuracy of death certification in an autopsied population with specific attention to malignant neoplasms and vascular diseases. Am J Epidemiol 8;:.. Percy C, Stanek E rd, Gloeckler L. Accuracy of cancer death certificates and its effect on cancer mortality statistics. Am J Public Health 8;:.. Van Loveren H, Rmbout PJ, Wagenaar SS, et al. Effects of ozone on the defense to a respiratory Listeria monocytogenes infection in the rat. Suppression of macrophage function and cellular immunity and aggravation of histopathology in lung and liver during infection. Toxicol Appl Pharmacol 88;:.. Jakab GJ, Bassett DJ. Influenza virus infection, ozone exposure, and fibrogenesis. Am Rev Respir Dis ;:. 6. Gilmour MI, Park P, Selgrade MJ. Ozone-enhanced pulmonary infection with Streptococcus zooepidermicus in mice. The role of alveolar macrophage function and capsular virulence factors. Am Rev Respir Dis ;: 6.. Cohen MD, Zelikoff JT, Qu G, et al. Effects of ozone upon macrophage-interferon interactions. Toxicology 6;:. 8. Gilmour MI, Park P, Doerfler DL, et al. Factors that influence the suppression of pulmonary antibacterial defenses in mice exposed to ozone. Exp Lung Res ;:.. Vincent R, Bjarnason SG, Adamson IY, et al. Acute pulmonary toxicity of urban particulate matter and ozone. Am J Pathol ;:6. 6. Bouthillier L, Vincent R, Goegan P, et al. Acute effects of inhaled urban particles and ozone: lung morphology, macrophage activity and plasma endothelin-. Am J Pathol 8;: Adamson IYR, Vincent R, Bjarnason SG. Cell injury and interstitial inflammation in rat lung after inhalation of ozone and urban particulates. Am J Respir Cell Mol Biol ;: Zelikoff JT, Sisco MP, Yang Z, et al. Immunotoxicity of sulfuric acid aerosol: effects on pulmonary macrophage effector and functional activities critical for maintaining host resistance against infectious diseases. Toxicology ;: Becker S, Sokup JM. Exposure to urban air particulate alters the macrophage-mediated inflammatory response to respiratory viral infection. J Toxicol Environ Health ;:. Downloaded from by guest on January Am J Epidemiol Vol., No.,