Bates, David V.
This issue of Epidemiology contains 3 meta-analyses of the extensive data relating ambient ozone levels to daily mortality.1–3 When I was a student at Cambridge, my tutor used to throw things across the table at us if we did not always mention the most important fact first; so for those whose attention span is short, all 3 studies report a small but substantial association between ozone levels and total mortality.
These studies were commissioned by the same agency, but the authors were free to carry out the analysis as they saw fit, and all 3 differ. One set of authors1 used data from 14 U.S. cities, 13 Canadian cities, and 21 European cities, and excluded data from the National Morbidity and Mortality Air Pollution Study (NMMAPS) and from Mexico City. The second2 used the data from the NMMAPS study of 95 cities together with European studies for a total input of 144 datasets. (These authors had already published an analysis of the NMMAPS data alone.4) The third3 was more restricted, using data from 7 U.S. cities plus other worldwide data for various parts of the analysis. Another difference is that one paper1 used data on the prevalence of air conditioning in both the United States and Canada.
Bayesian hierarchical models were used in the analyses. PM interaction with ozone was generally found to be unimportant. All 3 studies noted that the response function was higher in summer (when ozone levels are higher) than in winter, which means that if the data are not stratified by season, the overall response outcome is likely to be diminished. Other notable factors were that the prevalence of air conditioning affected the outcome1; the NMMAPS data alone yielded lower response outcomes than most other analyses, and there was generally satisfactory concordance between U.S. and European data.
Levy et al1 found that total mortality increased by 0.86% per 10 ppb in summer. Bell et al2 found an increase of 0.83% per 10 ppb in total mortality overall, and agreed that U.S. and non-U.S. data were similar. Ito et al3 provided a detailed seasonal breakdown and showed that the main effect occurred in the warm season. In their analysis for a single pollutant model, these authors plotted data from 8 U.S. regions, 8 European cities, 2 Australian cities, plus Mexico City, Sao Paulo, Santiago, and 2 regions of South Korea (See Fig. 1 of Ito et al3). Negative data points were noted for 5 cities, and all the rest were positive. The highest was for Brisbane, Australia, at approximately a 3.5% increase in mortality per 10 ppb for the 24-hour average ozone. Reviewing all the data, I would regard the value of 0.86% change in mortality per 10 ppb as a minimal figure, because inclusion of data from Brisbane and Mexico City would increase this significantly.
A recent report of European data5 summarized data from 23 regions with mortality data over a 3-year period. The authors reported no association between ozone and mortality during the winter months but an association in summer, with a mean increase of 0.33% in total mortality, 0.45% in cardiovascular deaths, and 1.13% in respiratory deaths per 10-μg/m3 ozone. Because this is equivalent to 5 ppb, the percentage increases should all be doubled for a 10-ppb change. Particulate matter (PM10) was not a confounder, although there was some interaction with nitrogen dioxide (NO2) and carbon monoxide (CO).
We have known, since Haagen-Schmidt's work in 1952, that tropospheric ozone is formed from nitrogen dioxide in the presence of hydrocarbons and sunlight in a complex series of reactions. It is a difficult pollutant to control, owing to the complex nature of its formation, but there are other aspects that make epidemiologic studies difficult.
For example, Northern Hemisphere cities, including Los Angeles, have a distinct “ozone season.” The correlations between temperature and ozone levels are high, making analyses complex because heat waves are themselves responsible for an increase in mortality. In a recent severe heatwave in The Netherlands, over 400 deaths attributed to heat were probably due to the accompanying high ozone levels.6 The same has been reported in Britain.7
The interrelationship between heat and tropospheric ozone is not straightforward. Atmospheric scientists in Toronto have recently shown that not only are surface temperature- and ambient ozone-related, but elevated ozone levels have the effect of increasing surface temperature.8 In Brisbane, Australia, and Mexico City, levels of ozone vary little throughout the year; hence, there is no “seasonality” factor. It may therefore be significant that these 2 cities yielded the highest response outcome, because they are least likely to be confounded by other seasonality factors. If this is the case, excluding them from a meta-analysis will necessarily have the effect of lowering the dose–response metric.
The question of personal exposures is also complex. Many factors contribute to the formation of ozone, whereas nitric oxide (NO) emitted from vehicles actually “quenches” ozone to form nitrogen dioxide (NO2); hence, values in the center of heavily urbanized cities will be lower than in the suburbs. It may be that the most vulnerable members of a population, living in the inner city, are personally exposed to lower ozone levels than wealthier people in the suburbs. One of the 3 studies reported here1 took the prevalence of air conditioning into account; homes with such units will have lower indoor ozone levels than those without.
Another factor in personal exposure is the time course of tropospheric ozone formation, which usually reaches a maximum between 12 noon and 3 pm. Children in air-conditioned schools are exposed to indoor ozone levels that are only approximately 15% of ambient outdoor levels, whereas children going outside to play in the early afternoon will encounter the highest ozone of the day. It is possible to list the factors that may account for individual differences in ozone exposure, but it is not easy to incorporate them in any meaningful way into studies of outcomes. A recent unique study of ozone levels in South Carolina during one summer month showed 3-fold differences in ozone levels across regions of the state.9
Yet another challenge in the study of ozone effects is that in northern latitudes, respiratory illnesses (and hence outcomes such as hospital admissions and respiratory mortality) are at their highest during winter months when ambient ozone is at its lowest. If summer and winter data are combined in one analysis, it may be difficult to see any ozone effect. It was for this reason that in a time-series study of southern Ontario in 1987,10 I stratified the data into summer and winter seasons. This uncovered a clear relationship between ambient ozone and hospital respiratory admissions—but only in summer.
The relation between ozone and respiratory illness is so well established that emergency admissions have been suggested as a surrogate measure of ozone. A research group in New Jersey11 recently concluded that ambient ozone levels in the summer can be reliably predicted from hospital emergency and admission data. They reported, “Sufficient databases exist for ER (Emergency Room) visits by asthmatics in Northern and Central New Jersey, and throughout the state for hospital admissions, for these health outcomes to be used as health-based indicators, complementing air-monitoring data in assessing whether improvements in public health are occurring because of reduction in emissions of precursors of ozone.”
The 3 new meta-analyses appearing here,1–3 along with the recent European study,5 each have unique features and appear to resolve the question of whether ambient ozone levels are associated with increased mortality. It seems unlikely that PM2.5 is an important confounder, and the effect of ozone appears to be independent of temperature. A final question—that of biologic plausibility—is in some ways the easiest to answer. Ozone is capable of causing inflammation in the lung at lower concentrations than any other gas. Such an effect would be a hazard to anyone with heart failure and pulmonary congestion, and would worsen the function of anyone with advanced lung disease.
“Background” levels of ozone are steadily rising in both hemispheres. Ozone is no respecter of frontiers.12 The increasing emissions of precursors, particularly NO2, in Asia13 are predicted to raise the “background” ozone level in Western America and Canada by between 5 and 10 ppb. Levels of ozone over the Atlantic Ocean have also been rising.14 In addition, global warming is expected to increase tropospheric ozone levels, although the magnitude of this effect is uncertain.
Canada and the World Health Organization have both proposed a standard or guideline of 0.06 ppm for an 8-hour exposure. The United Kingdom is aiming at 0.05 ppm for 8 hours as a maximum by the year 2005. In the United States, the Environmental Protection Agency will be reviewing the federal ozone standard during 2005. All these jurisdictions recognize that what is involved is a delicate balancing act with no margin between present exposure levels and adverse effects on health. In Los Angeles, ozone exposures in schoolchildren have been shown to be associated with school absences for respiratory illnesses,15 and economists have been busy calculating the economic burden.16 When one adds together such effects on a national basis, plus the influence of ozone on hospital admissions and emergency visits, and now the impact of premature mortality, one does not have to be an economist to see that the overall economic burden of this pollutant must be enormous.
In such a context, no one could maintain that these 3 meta-analyses (which support the European data) are of academic interest only. What they point to is the urgent need to reduce public exposures to ambient ozone by all possible means.
ABOUT THE AUTHOR
DAVID BATES trained in lung disease at St. Bartholomew's Hospital during London's smog disaster of 1952, leading to a lifelong research interest in the health effects of ozone. He migrated to Canada in 1972, where he served as Dean of the Medical School at University of British Columbia. He continues his research and writing there as an Emeritus Member of the faculty.
1. Levy JI, Chemerynski SM, Sarnat JA. Ozone exposure and mortality: an empirical Bayes meta-regression analysis. Epidemiology. 2005;16:458–468.
2. Bell ML, Dominici F, Samet JM. A meta-analysis of time-series studies of ozone and mortality with comparison to the National Morbidity, Mortality, and Air Pollution Study. Epidemiology. 2005;16:436–445.
3. Ito K, DeLeon SF, Lippmann M. Associations between ozone and daily mortality: analysis and meta-analysis. Epidemiology. 2005;16:446–457.
4. Bell ML, McDermott A, Seger SL, et al. Ozone and short-term mortality in 95 US urban communities, 1987–2000. JAMA. 2004;292:2372–2378.
5. Gryparis A, Forsberg B, Katsouyanni K, et al. Acute effects of ozone on mortality from the ‘Air Pollution and Health: A European Approach’ Project. Am J Respir Crit Care Med. 2004;170:1080–1087.
6. Fischer PH, Brunekreef B, Lebret E. Air pollution related deaths during the 2003 heat wave in The Netherlands. Atmospheric Environment. 2004;38:1083–1085.
7. Stedman JR. The predicted number of air pollution related deaths in the UK during the August 2003 heatwave. Atmospheric Environment. 2004;38:1087–1090.
8. Beaney G, Gough WA. The influence of tropospheric ozone on the air temperature of the city of Toronto, Ontario, Canada. Atmospheric Environment. 2002;36:2319–2325.
9. Franklin AC, Salmon LG, Wolfson JM, et al. Ozone measurements in South Carolina using passive samplers. J Air Waste Manage Assoc. 2004;54:1312–1320.
10. Bates DV, Sizto R. Hospital admissions and air pollutants in Southern Ontario: the acid summer haze effect. Environ Res. 1987;43:317–331.
11. Weisel CP, Cody RP, Georgopoulos PG, et al. Concepts in developing health-based indicators for ozone. Int Arch Occup Environ Health. 2002;75:415–422.
12. Holloway T, Fiore A, Hastings MG. Intercontinental transport of air pollution: will emerging science lead to a new hemispheric treaty? Environ Sci Technol. 2003;37:4535–4542.
13. Akimoto H. Global air quality and pollution. Science. 2003;302:1716–1719.
14. Lelieveld J, Van Aardenne J, Fischer H, et al. Increasing ozone over the Atlantic Ocean. Science. 2004;304:1483–1487.
15. Gilliland FD, Berhane K, Rappaport E, et al. The effects of ambient air pollution on school absenteeism due to respiratory illness. Epidemiology. 2001;12:43–54.
16. Hall JV, Brajer V, Lurmann FW. Economic valuation of ozone-related school absences in the South Coast Air Basin of California. Contemporary Economic Policy. 2003;21:407–417.
© 2005 Lippincott Williams & Wilkins, Inc.