Click on the links below to access all the ArticlePlus for this article.
Please note that ArticlePlus files may launch a viewer application outside of your web browser.
Evidence linking air pollution exposure with adverse health consequences continues to accumulate. Ambient particulate matter is associated with short-term increases in mortality, and effects remain when multiple cities are analyzed with uniform methods.1 In the United States and Europe, multiyear exposure to higher levels of particulate pollution was associated with consistently higher mortality rates, especially for cardiopulmonary causes.2,3
Lower educational attainment has been linked with poor health outcomes, including higher mortality rates.4 The mechanism for these observed associations is not fully understood. People with higher educational status may have improved knowledge and ability to manage personal health and gain access to health care, higher income, better jobs, and stronger social connections that enhance health. Educational attainment may also relate to differences in occupational exposure, living conditions, or baseline health status.
Recently, interest has grown in how air pollution and socioeconomic factors together contribute to disparities in population health. Individual educational attainment has advantages as a socioeconomic indicator for this research theme because, unlike income and occupation, it is not affected by later-life changes in health status. Education has been studied as a modifier of long- and short-term pollution effects on mortality in the United States, Asia, and Europe.2,3,5–12
We evaluated whether education modified associations between pollution and mortality in 3 Latin American cities with substantial inequality, differing patterns of underlying risk factors (eg, smoking behavior), and a larger proportion of less-educated individuals. This topic is timely because the global update of the World Health Organization (WHO) Air Quality Guidelines emphasizes their relevance to developing countries and environmental equity.13 Using data from 3 large Latin American cities (Santiago, Chile; São Paulo, Brazil; and Mexico City, Mexico), we assessed whether education modified the risk of mortality associated with exposure to particulate air pollution during the years 1998–2002.
The 3 Latin American cities are in middle-income countries, represent major population centers, and have temperate climates and high air pollution levels relative to many US and European cities.14 All 3 cities experience substantial inequality as reflected by the Gini coefficient, an index of the inequality of income distribution ranging from zero (perfect equality) to 1 (absolute inequality). The 1999 urban Gini coefficients for Brazil, Chile, and Mexico were 0.63, 0.55, and 0.51, respectively, compared with the typical 0.30 in many industrialized countries.15
Metropolitan Mexico City (altitude 2240 m above sea level) is in a valley with mountains on 3 sides. Major pollution sources include motor vehicles, stationary combustion sources, and a dry lake basin that contributes windblown dust.16 The metropolitan population during the study period was an estimated 18 million.17 We defined the metropolitan area as the 16 delegations of the Federal District and 18 surrounding municipalities of the State of Mexico located in the Valley of Mexico airshed.18
Santiago, Chile is located in a valley to the west of the Andes mountains at about 550 m above sea level. The population was approximately 6 million during the study period.17 Thermal inversions are frequent in the area, especially during the winter season, and because of the topography and lack of cleansing winds, pollutant levels can build to high concentrations. Major pollution sources are industry and motor vehicles, with diesel exhaust estimated to contribute the majority of particulate pollution.14 We defined the metropolitan area as the 32 municipalities of the Province of Santiago that lie within the airshed.19
São Paulo, Brazil (altitude 700 m) is about 60 km from the Atlantic coast, and separated from the coast by a mountain chain. The complex topography of this region and thermal inversions in the winter contribute to numerous air pollution episodes. Motor vehicles contribute substantially to São Paulo's air pollution, but point sources may account for up to half the particulate pollution.14 The use of ethanol in motor vehicle fuel contributes to a unique pollutant mix. The population of São Paulo is estimated at about 10 million.17 Metropolitan São Paulo was defined as the 96 administrative districts of the city.20
Monitors were selected for this study from each city's extensive air pollution and meteorologic monitoring networks based on: (1) adequate daily data availability during 1998–2002; (2) use of the same sampling technology for particulate matter <10 μm in aerodynamic diameter (PM10) within each city; and (3) location within the defined geographic study area.
We evaluated PM10 to facilitate comparison with similar studies examining effect modification by educational attainment, and because particles have been associated with short-term mortality in these cities.20–23 Furthermore, this size fraction has been monitored routinely in all the cities at multiple stations during 1998–2002. Ambient monitor measurements were used as a surrogate for personal exposure. Because all 3 cities have temperate climates with little air conditioning, particulate air pollution penetrates efficiently into indoor spaces.24,25 This and other research26 suggests that ambient PM10 monitors adequately represent longitudinal variation in personal particle exposure.
We defined exposure as daily air pollution levels averaged from selected monitors across the metropolitan area. Previous use of this approach in Mexico City yielded estimates of association similar to those derived by assigning exposure based on proximity of the decedent's municipality of residence to an individual monitor.27 Furthermore, using city-wide averages may reduce exposure measurement error because many people move throughout the city and any given ambient monitor may not be representative of personal exposures. We averaged data from 7 monitors in São Paulo, 8 in Santiago, and 4 in Mexico City. We excluded the Xalostoc station in Mexico City. This monitor is near industrial sources (Retama, personal communication, January 2005), as confirmed by systematically high concentration readings and relatively lower correlations with other monitor sites,28 and is thus not likely to be representative of wider population exposure. Pearson correlations among the selected monitors within each city were all at least 0.50. In Mexico City and Santiago, PM10 is measured hourly with the Rupprecht & Patashnick tapered element oscillating microbalance (TEOM) sampler. In São Paulo, PM10 is measured continuously using a β-radiation method. Daily 24-hour averages of PM10 in micrograms per cubic meter at local temperature and pressure were calculated from these readings.
Daily temperature and relative humidity are measured by automatic recording instruments in the city monitoring networks. We averaged all values from the monitors reporting on any given day to calculate the mean, minimum, and maximum of both parameters. From these values, we calculated apparent temperature (AT) in degrees Celsius, intended to reflect an individual's perceived air temperature, with the following formula:
where Ta is air temperature and Td is dew point temperature (calculated from relative humidity), all in degrees Celsius.29
Daily deaths occurring from 1998 to 2002 were obtained in electronic records from the vital statistics authorities in each country: the Brazilian Municipal Mortality Information Improvement Program; the Chilean National Institute of Statistics; and the Mexican National Institute of Statistics, Geography, and Information. We included only individuals who both lived and died within the metropolitan area. Causes of deaths in all 3 cities were coded using the International Classification of Diseases 10th revision. We eliminated deaths from external causes (accidents, poisoning, and violence), codes S and above. We further classified the deaths into cardiovascular (I codes <800) and respiratory (J 100–118, 120–189, 209–499, and 690–700) causes. Final analysis datasets for each city included deaths occurring from 1 January 1998 to 30 December 2002. The last day of 2002 was excluded due to administrative irregularities resulting in a dip in death counts in Mexico City on that day.
In each city, educational attainment is recorded on the death certificates in a categorical manner, although Chile has an additional variable, years-within-category. Because the categories differed slightly by city and because some coding conventions changed from year to year within cities, we created a uniform stratification to provide comparability across educational categories, as follows: none, some primary, some secondary, secondary or more, and missing or unknown. Years of education corresponding to primary and secondary categories, respectively, were 1–6 and 7–11 years (Mexico City and Santiago) and 1–7 and 8–11 years (São Paulo). In Chile, the years-within-category variable also allowed a sensitivity analysis considering separately those who had completed a high school degree (12 years) compared with those with postsecondary education.
We aimed to examine effects of air pollution exposure on mortality and whether effects were modified by educational attainment, including cumulative effects up to 5 days before the day of death. To achieve this goal, we used time series regression, in which the aim is to evaluate whether daily death counts (the outcome) are associated with increases in daily pollution levels, controlling for other factors that are associated temporally with pollution and mortality (eg, weather and season). Because the unit of analysis is the day, individual-level confounders (eg, smoking habits, obesity) that are unlikely to vary in time with pollution are not of general concern. For the cumulative 6-day effect analysis, we used a distributed lag model as applied in a multicity study,30 using a harmonized approach across all cities, as has been recently recommended.31
First, we fit city-specific robust Poisson regression models to all daily death counts. We controlled for potential temporal confounders of the association between PM10 and daily mortality by using indicator variables for day of week and a natural cubic spline for time to address seasonality and long-term trend. We chose the number of degrees of freedom (df) for the temporal trend spline separately for each city by minimizing the sum of autocorrelation in the residuals over 30 days.32 PM10 exposure was modeled for several different timescales. Single-day concentrations were included in models for lag 0 (day of death) and lag 1 (the day previous to the day of death).
Next, we used an unconstrained distributed lag approach, in which PM10 concentrations on the day of death and up to 5 previous days were entered as separate terms, and the coefficients of these terms were summed for an overall unbiased estimate of effect.33 We fit separate models that estimated and then summed PM10 coefficients for lag period 0–5 days. To control for potential confounding effects of weather, we included mean apparent temperature for the corresponding PM10 lag. For example, the lag 1 effect of PM10 includes control for lag 1 temperature, and models with 6 PM10 terms included 6 terms for temperature, consistent with recent research.34
The models were fit using S-Plus software (Insightful Corp., Seattle, WA) using a strict convergence criterion. The city-specific models for single-day exposure were of the following form:
where E(Ytc) is the expected daily death count on day t for city c; Xt−lc is the PM10 concentration in city c on day t at a lag l;ATt−lc is mean daily apparent temperature in city c on day t at lag l; βc and γc reflect the relationship between PM10 and AT and the log of expected mortality rate within city c, respectively; ns is natural cubic spline, dftime is the df for temporal trends; DOWt is an indicator variable for the day of the week on day t, θc is its coefficient, and αc is a constant. For the distributed lag model, the exposure term βcXt−lc is replaced by j=0toJβjcXt−jQc representing the PM10 concentrations over J days (J = 5 in our case), and the corresponding lags forj=0toJγjcATt−jc.
We calculated mortality rate ratios and standard errors from the particle parameter estimate(s) and their variances and covariances (when multiple PM10 terms were included). Standard errors of the regression coefficients were adjusted for any over- or under-dispersion. From the rate ratios, we derived the percent change in mortality per 10-μg/m3 increase in PM10 concentration and the corresponding 95% confidence interval (CI). This same approach was then applied in each city to death counts stratified by age (all adults; adults 65 and older); educational level (among adults); and then educational level further stratified by sex. Effect modification was evaluated by considering whether CIs for differing strata of education enclosed the point estimates of the other strata.
Apparent temperature and PM10 were not highly correlated; Pearson correlation coefficients were 0.13, −0.05, and −0.15 for Mexico City, São Paulo, and Santiago, respectively. In all cities, PM10 exceeded the World Health Organization recommended annual guideline, with Santiago experiencing the highest pollution levels (Table 1). The PM10 levels in Santiago are somewhat lower than those reported in other recent studies,35 probably because we used concentrations measured by TEOM monitors as opposed to the gravimetric measurements evaluated in other research. The TEOM monitors burn off the volatile fraction, a substantial portion of the mass in diesel-generated particles.
In time-series plots, all cities showed substantial winter peaks in both pollution and mortality, as well as relatively moderate temperatures year-round. In all cities, slightly more men than women died during the study period (Table 2). Only a small proportion of those dying had more than a high school education, and in all cities, the percentage of men with higher education was double that of women. Relatively few death certificates lacked education information in Mexico City and Santiago, but in São Paulo, 25% did. Individuals with missing education data were on average 6 years younger and more likely to be female (5% higher) compared with those with education recorded.
Point estimates of associations between PM10 and all-cause adult mortality were positive for all 3 cities (Table 3). Air pollution effects on mortality differed little by sex, and point estimates of effect were not consistently different among the older adults compared with all adults.
We did not see consistent gradients of increasing effect sizes with lower levels of education among all adults (Table 4). In fact, in São Paulo, the opposite effect was observed, with monotonically increasing effects with increasing education for several lags considered among all adults and among men. In several instances, especially in Santiago, we saw a bimodal pattern of stronger effects in the lowest and highest educational categories. In Mexico City, associations with PM10 were highest among the most educated women for all lags considered.
Formal effect-measure modification, as evaluated by considering whether CIs for a given stratum enclosed the corresponding point estimates in other strata, was seen principally in São Paulo for the total population and among men for some of the lags, especially when comparing those with the most education and those with the least. The bimodal pattern of stronger associations at the extremes also fit this formal criterion for effect modification among the total population in Santiago. Mexico City associations were more homogeneous across educational strata and most CIs enclosed the point estimates of other strata.
Because of the possibility of an age-cohort effect (ie, older generations may be less likely to have attained higher education, and thus older people disproportionately represented in the lower-educated strata), we conducted the same sex-stratified analysis among those aged 65 and older (Table 5). A pattern of increasing effects with increasing education was observed in Mexico City for 6-day cumulative estimates, and this same pattern, seen for all adult mortality, persisted in Brazil. Otherwise no systematic changes were seen across the cities when evaluating just the older population.
A sensitivity analysis for Santiago including only those with postsecondary education found reductions in cumulative (6-day) PM10 effects from a 2.00% increase in mortality per 10-μg/m3 increment in PM10 (95% CI = 0.93%–3.07%) to 0.86% (0.48%–1.23%). Among those in Brazil with missing education data, the effect estimates were not substantially different from those in the total population (data not shown).
To evaluate possible systematic differences in residence by educational attainment, which could be relevant if particle levels or composition differs across the cities, we created maps with quintiles of education for the highest and lowest educational categories (eFigure). In both Mexico City and São Paulo, more highly educated people tended to cluster in the center city; residence patterns by education were more heterogeneous in Santiago.
In these Latin American cities, estimated PM10 exposure in the short term was positively associated with daily mortality, consistent with evidence from other parts of the world.1 However, no consistent gradients in effect modification by educational status were observed across these 3 cities, in contrast to results in other regions. In North America and the Netherlands, relative risks of mortality associated with long-term exposure to particulate pollution have been found to increase monotonically by decreasing levels of educational attainment.2,3,8 In Poland, this gradient was evident only in women.9 Similar patterns of increasing risk among the less educated for short-term effects of ambient particles on mortality have also been seen in the United States and Hong Kong,10,11 although in 1 US study, effect modification was only modest.12 In French cities, however, both long- and short-term associations between particles and mortality showed higher associations among the more highly educated, when considering 3 educational attainment categories.5,6
The heterogeneous patterns we observed by educational attainment may relate to the way education interacts with other aspects of the social context to affect health and underlying susceptibility. A survey of elderly adults in these cities examined equity of access to health care by socioeconomic indicators, including education.15 Health care access is one tangible route by which socioeconomic circumstances may operate to affect susceptibility to air pollution. For example, use of certain medications and vitamin supplementation, more likely in those receiving adequate health care, can mitigate the effects of air pollution on cardiac and respiratory health.36,37 In that survey, only in Santiago did access to health care decrease with decreasing educational level. In São Paulo, the opposite effect was observed, and in Mexico City, access to health care was more associated with levels of health insurance than with educational attainment.15
This survey suggests that, although all of these socioeconomic variables are related, education interacts in complex ways with other socioeconomic circumstances in influencing health and quality of life among elderly people in these cities. Thus, heterogeneity would be expected when evaluating how environmental quality may affect health among people with differing educational levels. The previously cited survey relied on questionnaire reports on educational level. To our knowledge, nothing has been published on the quality of information on education in death certificates in these countries, which is a limitation of our study. Assuming that the educational strata have some validity, it is possible that people who are less educated and who survive to older ages, despite infection, poor health care, inadequate nutrition, or other disadvantages throughout the life course, represent highly resilient “healthy survivors,” whereas the more highly educated groups include individuals with a larger range of frailty, thus explaining the results seen in our cities and in France.5,6
Although evidence suggests that the health of men and women can be patterned differently by educational status (eg, life expectancy reductions with diminishing years of schooling were even greater among Chilean men compared with women38), and lower education modified air pollution risk only among women in Poland,9 we did not see consistently different effects by sex in these Latin American cities.
The gradients of air pollution risk by educational status observed in previous studies may be due to varying exposures, differing prevalence of comorbid conditions, or other systematic differences related to degree of education39 processes not possible to elucidate fully with the study designs used to date, including ours. We report relative effect estimates (percent change in mortality); equivalent effect estimates combined with higher baseline mortality rates among the less educated could translate into a higher absolute burden among the lower strata. Similarly, if the city-wide average exposure assignment scheme systematically underestimates the exposure or relative toxicity of exposure among less educated people (who may, for example, live closer to roadways) the observed effect could be biased downward. Several other factors may contribute to the patterns we observed.
We saw consistently stronger effects of PM10 on mortality among the most educated. In Santiago and São Paulo, previous analyses showed higher effects of PM10 on daily mortality in districts where the more educated reside,20,40 although the opposite trend was seen in another study of São Paulo when using smaller geographic units.22 Clustering of residence by education may lead to systematic differences in exposure to particles of differing origin and composition. The maps in the eFigure suggest that the centers of Mexico City and São Paulo have a higher proportion of more educated people, but people move across the city and so it is difficult to attribute our results fully to these residential differences.
The stronger effect among the educated may also relate to patterns of prevalence of smoking in these societies. In Mexico, smoking was more prevalent among higher income and more highly educated households41,42; however, in Chile and Brazil, the opposite pattern was observed, with less educated households smoking more.43,44 Smoking may lead to enhanced susceptibility to air pollution effects, and the strong trend toward higher associations with mortality among the more educated men in Mexico City is consistent with population smoking prevalence patterns. Nutritional status may also affect susceptibility; in both Brazil and Chile, the more highly educated consume more fruits and vegetables,43,45 which may increase resilience to adverse effects of pollution exposure. To our knowledge, no information on nutritional and dietary patterns by socioeconomic level in Mexico has been published. Underlying disease may also be distributed differently by education. Although we did not analyze secondary causes, we did not observe important differences in mortality cause by educational level, other than cardiovascular cause mortality, which was reduced proportionally among the most highly educated in Santiago.
The study of effect modification of short-term pollution by education may be sensitive to analytic method. Although we used stratified analyses as has been commonly done in time series studies, models using interaction terms could be explored in the future. Sparse numbers in the extreme educational categories, especially when further stratified by age and sex, may have limited our power to see contrasting effects, and differing modeling approaches could potentially enhance power. An analysis of 4 US cities used a time-series approach similar to ours, that is, stratifying death counts by educational status.12 A more recent study of 20 US cities in which the case-crossover approach was used11 (and including the same 4 US cities evaluated in the first report) showed a much stronger gradient by education than the earlier study. The more recent study presents only pooled results, precluding direct comparisons of the influence of method on the effects for the 4 cities. Future comparative analyses in those cities and these Latin American cities could illustrate the influence of method.
We used a consistent set of decision rules for exposure assignment across the cities. However, PM10 measurement methods other than those used in these cities, such as gravimetric methods, may have different associations with mortality.
We did not evaluate other socioeconomic indicators in this study. Although death certificates in the 3 cities include an occupation variable, both the quality and the comparability of these indicators across the 3 cities are limited, and large numbers of observation are missing. Mexican death certificates also have an additional variable for social security or health care insurance, which has proven to be a good predictor for some health outcomes in surveys conducted in Mexico. However, this variable has not been used based on death certificates because it is incomplete and unreliable. Area-level indicators of socioeconomic position could be constructed using Census data. However, inference using such indicators is strongly limited by the coarse resolution of the geographic units recorded on death certificates, as we saw in previous work in Mexico City and São Paulo,16,20 so we confined our analysis to individual education.
This analysis evaluated air pollution and mortality across several timescales in 3 Latin American cities, using a uniform framework, and examined effect modification by individual educational status. Unlike several studies in other countries, we did not observe gradients of increasing associations between particles and mortality with decreasing educational levels. These differing patterns in Latin America may be due to how pollution exposures, underlying susceptibilities, or lifestyle and behavioral factors correlate with educational status in these societies.
We thank Roberto Muñoz, Marcela Bravo, and Jesuíno Romano for data assistance; James Escobar, Mateus Habermann, and Tenaya Sunbury for help with figures; Antonella Zanobetti for programming advice; and the reviewers and editor whose suggestions greatly improved the manuscript.
1. Bell ML, Samet JM, Dominici F. Time-series studies of particulate matter. Annu Rev Publ Health
2. Pope CA III, Burnett RT, Thun MJ, et al. Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. JAMA
3. Hoek G, Brunekreef B, Goldbohm S, et al. Association between mortality and indicators of traffic-related air pollution in the Netherlands: a cohort study. Lancet
4. Lynch J, Kaplan G. Socioeconomic position. In: Berkman L, Kawachi I, eds. Social Epidemiology
. New York, NY: Oxford University Press; 2000:13–35.
5. Filleul L, Rondeau V, Cantagrel A, et al. Do subject characteristics modify the effects of particulate air pollution on daily mortality among the elderly? J Occup Environ Med
6. Filleul L, Rondeau V, Vandentorren S, et al. Twenty five year mortality and air pollution: results from the French PAARC survey. Occup Environ Med
7. Laurent O, Bard D, Filleul L, et al. Effect of socioeconomic status on the relationship between atmospheric pollution and mortality. J Epidemiol Community Health
8. Health Effects Institute. Reanalysis of the Harvard Six Cities Study and the American Cancer Society Study of Particulate Air Pollution and Mortality (A special report of the Institute's Particle Epidemiology Reanalysis Project). Cambridge, MA: Health Effects Institute, 2000; Available at: http://pubs.healtheffects.org/view.php?id=6
. Accessed April 10, 2007.
9. Krzyzanowski M, Wojtyniak B. Ten-year mortality in a sample of an adult population in relation to air pollution. J Epidemiol Community Health
10. Ou CQ, Chung RY, Thach TQ, et al. The modifying effects of individual socio-economic status on mortality risk associated with air pollution [abstract]. Epidemiology
11. Zeka A, Zanobetti A, Schwartz J. Individual-level modifiers of the effects of particulate matter on daily mortality. Am J Epidemiol
. 2006; 163:849–859.
12. Zanobetti A, Schwartz J. Race, gender, and social status as modifiers of the effects of PM10 on mortality. J Occup Environ Med
. 2000;42:469– 474.
13. World Health Organization (WHO). WHO air quality guidelines global update: Report on a working group meeting, Bonn Germany, 18–20 October, 2005. Available at: http://www.euro.who.int/Document/E87950.pdf
. Accessed December 6, 2006.
14. Maisonet M, Gouveia N. An assessment of health effects of urban air pollution in Latin America and the Caribbean. Pan American Health Organization.
2005; ISBN 92 75 12598 8.
15. Wallace SP, Gutierrez VF. Equity of access to health care for older adults in four major Latin American cities. Pan Am J Pub Health
16. O’Neill MS, Loomis D, Borja-Aburto VH. Ozone, area social conditions, and mortality in Mexico City. Environ Res
. 2004;94:234 –242.
17. Brinkhoff T. The population of the major cities and agglomerations of countries in America. Available at: http://www.citypopulation.de/America.html
. Accessed December 7, 2006.
18. Comisión Ambiental Metropolitana. Map of metropolitan Mexico City under jurisdiction of CAM-the Metropolitan Environment Commission. Available at: http://www.sma.df.gob.mx/sma/modules.php?name=News&file=article&sid=60
. Accessed December 12, 2006.
19. Cifuentes LA, Krupnick AJ, O'Ryan R, et al. Urban Air Quality and Human Health in Latin America and the Caribbean. Inter American Development Bank.
2005 Available at: http://www.iadb.org/sds/ENV/publication/publication_2492_4240_e.htm
. Accessed January 18, 2007.
20. Gouveia N, Fletcher T. Time series analysis of air pollution and mortality: effects by cause, age and socioeconomic status. J Epidemiol Community Health
. 2000;54:750 –755.
21. Cifuentes LA, Vega J, Kopfer K, et al. Effect of the fine fraction of particulate matter versus the coarse mass and other pollutants on daily mortality in Santiago, Chile. J Air Waste Manage
. 2000;50: 1287–1298.
22. Martins MC, Fatigati FL, Vespoli TC, et al. Influence of socioeconomic conditions on air pollution adverse health effects in elderly people: an analysis of six regions in Sao Paulo, Brazil. J Epidemiol Community Health
23. Tellez-Rojo MM, Romieu I, Ruiz-Velasco S, et al. Daily respiratory mortality and PM10 pollution in Mexico City: importance of considering place of death. Eur Resp J
24. Rojas-Bracho L, Suh HH, Oyola P, et al. Measurements of children’s exposures to particles and nitrogen dioxide in Santiago, Chile. Sci Total Environ
25. Rojas-Bracho L. Evaluación del grado de exposición a aeroparticulas en los habitantes de la zona centro de la Ciudad de México [Masters thesis]. Universidad Nacional Autonoma de México.
26. Sarnat JA, Brown KW, Schwartz J, et al. Ambient gas concentrations and personal particulate matter exposures: implications for studying the health effects of particles. Epidemiology
27. O’Neill MS, Loomis D, Borja Aburto VH, et al. Do associations between airborne particles and daily mortality in Mexico City differ by measurement method, region, or modeling strategy? J Expo Anal Environ Epidemiol
28. O’Neill MS, Loomis D, Torres-Meza VM, et al. Estimating particle exposure in the Mexico City metropolitan area. J Expo Anal Environ Epidemiol
29. Steadman RG. The assessment of sultriness. Part II: Effects of wind, extra radiation and barometric pressure on apparent temperature. J Appl Meteorol
30. Zanobetti A, Schwartz J, Samoli E, et al. The temporal pattern of mortality responses to air pollution: a multicity assessment of mortality displacement. Epidemiology
31. Samet JM, White RH. Urban air pollution, health, and equity. J Epidemiol Community Health
32. Touloumi G, Samoli E, Pipikou M, et al. Seasonal confounding in air pollution and health time-series studies: effect on air pollution effect estimates. Stat Med
33. Zanobetti A, Schwartz J, Samoli E, et al. The temporal pattern of respiratory and heart disease mortality in response to air pollution. Environ Health Perspect
. 2003;111:1188 –1193.
34. Goodman PG, Dockery DW, Clancy L. Cause-specific mortality and the extended effects of particulate pollution and temperature exposure. Environ Health Perspect
. 2004;112:179 –185.
35. Koutrakis P, Sax SN, Sarnat JA, et al. Analysis of PM10, PM2.5, and PM2.5–10 concentrations in Santiago, Chile, from 1989 to 2001. J Air Waste Manage
36. Romieu I, Sienra-Monge JJ, Ramirez-Aguilar M, et al. Antioxidant supplementation and lung functions among children with asthma exposed to high levels of air pollutants. Am J Respir Crit Care Med
37. Schwartz J, Park SK, O’Neill MS, et al. Glutathione-S-transferase M1, obesity, statins, and autonomic effects of particles: gene-by-drug-byenvironment interaction. Am J Respir Crit Care Med
. 2005;172:1529– 1533.
38. Vega J, Hollstein R, Delgado I Socioeconomic differentials and mortality in middle-income nations. In: Evans T, Whitehead M, Diderichsen F, et al, eds. Challenging Inequities inHealth: From Ethics to Action.
New York: Oxford University Press; 2001:123–137.
39. Pope CA III, Burnett RT, Thurston GD. Pollution-related mortality and educational level [author reply]. JAMA
40. Cifuentes LA, Bravo M. Differential mortality effects of PM10 by educational level and residence in Santiago, Chile _poster_. Health Effects Institute Annual Meeting, Boston, MA, 2004.
41. Campuzano I. Comportamiento de los fumadores en Mexico segun las Encuestas Nacionales de Adicciones 1988 a 1998. In: Valdes-Salgado R, Lazcano-Paonce E, Hernandez-Avila M, eds. Primer Informe Sobre el Combate al Tabaquismo. Mexico ante el Convenio Marco para el Control del Tabaco Mexico
. Cuernavaca, Mexico: Instituto Nacional de Salud Publica; 2005.
42. Vazquez-Segovia LA, Sesma-Vazquez S, Hernandez-Avila M. Tobacco consumption in Mexican households: results from the National Household Income and Expenditure Survey, 1984-2000. Salud Publica de Mexico
. 2002;44(suppl 1):S76 –S81.
43. Chile. II Encuesta Nacional de Calidad de Vida y Salud. 2006. Available at: http://epi.minsal.cl/epi/html/sdesalud/calidaddevida2006/index.htm
. Accessed October 17, 2007.
44. Monteiro CA, Cavalcante TM, Moura EC, et al. Population-based evidence of a strong decline in the prevalence of smokers in Brazil (1989 –2003). Bull World Health Organ
45. Jaime PC, Monteiro CA. Fruit and vegetable intake by Brazilian adults, 2003. Cadernos de Saude Publica
. 2005;21:S19 –S24.