Cohort studies for air pollution health effects - UNECE队列研究空气污染对健康的影响-联合国欧洲经济委员会

《Cohort studies for air pollution health effects - UNECE队列研究空气污染对健康的影响-联合国欧洲经济委员会》由会员分享，可在线阅读，更多相关《Cohort studies for air pollution health effects - UNECE队列研究空气污染对健康的影响-联合国欧洲经济委员会（5页珍藏版）》请在装配图网上搜索。

1、Design of cohort studies for air pollution health effectsBert Brunekreef, PhDProfessor of Environmental HealthInstitute for Risk Assessment Sciences (IRAS)University of UtrechtPO Box 80176 3508 TD Utrecht NetherlandsPhone 31 30 2535400 FAX 31 30 2535077 EMAIL B.Brunekreefvet.uu.nl1. Whats a cohort s

2、tudy?“In epidemiology, the word cohort is often used to designate a group of people who share a common experience or condition (). Often, there are two cohorts in the study, one of them is described as the exposed cohort, the other is thought of as the unexposed, or reference, cohort; if there are m

3、ore than two cohorts, each may be characterized by a different level or type of exposure.” (1). In these words, cohort studies are being introduced in one of the most authoritative textbooks, Modern Epidemiology by Rothman and Greenland. In a cohort study, the incidence of disease and/or death is st

4、udied in the respective cohorts over a period of time. They can be either retrospective (looking back in time) or prospective (following cohorts over a certain period of time). Retrospective cohort studies often use study populations that were defined in the past in, for example, a twin registry or

5、in employment records, and that can still be traced today. Prospective cohort studies have the advantage of optimal control over the design to answer the principal study hypothesis. Prospective cohort studies are in principle attractive to study health effects of air pollution because pitfalls of ot

6、her designs can to some extent be avoided. Pitfalls of ecologic studies relating air pollution to public health indicator data (such as mortality statistics) include absence of information at the individual level about potential confounders such as smoking, dietary habits and occupational exposures.

7、 Pitfalls of cross-sectional and case-control population studies include absence of historic exposure data and hence, limited possibilities to establish the temporal order of cause and effect. Prospective cohort studies in which a suitably large number of subjects are being followed over time, howev

8、er, are expensive; also, long periods of follow-up are usually needed to come to firm conclusions. In this contribution, some design issues related to cohort studies of air pollution and health will be discussed.2. Examples of recent cohort studies of air pollution and health2.1. MortalityThree coho

9、rt studies, all conducted in the U.S., have drawn a large amount of attention in the last decade (2-4). The first, the Harvard Six Cities Study, was published in 1993 (2). It documented a relationship between fine particulate matter and survival in a cohort of over 8,000 adults, living in six U.S. c

10、ommunities, that was followed for about 15 years, from 1974-1989. Exposure was classified as the average of concentrations of a number of different air pollutants including PM2.5, sulfate, ozone, SO2 and PM10 over a period of some eight years, covering the late 1970s and the first half of the 1980s.

11、 For each city and pollutant, each individual within that city was assumed to have had the same exposure. Comparing most to least polluted city, a relative risk of 1.26 was found for total mortality. The second study, the ACS-II study, was published in 1995 (3). It documented a relationship between

12、sulfate and fine particulate matter and survival in a cohort of over 500,000 subjects, living in about 150 communities, that was followed for about 8 years, from 1981-1989. Exposure was classified using existing data on sulfate and PM2.5 measured for one year in 1980/1981. Again, each individual wit

13、hin a city was assigned the same exposure value. Comparing most to least polluted city, a relative risk of was found or sulfate, and of for fine particulate matter. The third study, the California-based AHSMOG study (4), found a relationship between PM10 and lung cancer and non-cardiac respiratory d

14、eaths in a cohort of about 7,000 Seventh-Day Adventists that were followed over a period of 10 years. Exposure in this study was classified as a weighted average of concentrations of ozone and PM10 measured at nearby monitoring stations, thus allowing for more within-city variation of exposure class

15、ification than in the other two studies. The analysis did not permit expression of the findings in terms of most vs. least polluted community. All three studies found relationships between air pollution and mortality at levels of pollution that were generally below existing air quality standards. Th

16、e potential effects on life expectancy have been estimated to be in the order of 1 year at realistic exposure contrasts in the industrial developed world (5,6). The public health impact of such an effect on life expectancy is large (7).Especially the first two studies have been extensively scrutiniz

17、ed, in view of the potential policy implications. If pollution levels have to be appreciably decreased, the associated costs of abatement measures are very high. An extensive re-analysis of the first two studies has essentially corroborate their findings (8) although questions have been raised with

18、respect to the role of non-causal spatial correlation of the exposure and health data in explaining the associations, and with respect to the role of co-pollutants, especially SO2. As exposure classification was based on central-site monitoring in the first two studies, the role of small-scale spati

19、al heterogeneity in exposure could not be addressed in these studies. The AHSMOG study used weighted averages of data derived from a number of nearby monitoring sites as a basis for exposure classification, improving upon the use of single central site monitoring, but still ignoring local source con

20、tributions.Similar cohort studies from other parts of the world have not been reported to date. A study from France, presented in abstract form, suggests that there is a relationship between SO2, measured in the 1970s, and mortality upon follow-up of subjects who were studied at the time (9). A rece

21、nt cohort study from the Netherlands has suggested that exposure to traffic-related air pollution (soot and NO2) may be associated with reduced survival (10). Design issues related to cohort studies of air pollution and mortality will be discussed further in section2.2. MorbidityA few cohort studies

22、 have looked at morbidity endpoints. A recent example is the Southern California Air Pollution study (11) in which the development of lung function and respiratory disease in school children is being studied in relation to the air pollution components ozone and fine particulate matter. The study is

23、being conducted in 12 communities chosen to represent the range in air pollution concentrations found in the greater Los Angeles area. First results suggest that exposure to fine particulate matter (but not ozone) is associated with reduced lung function growth in children. Exposure assessment in th

24、is study was based on dedicated monitoring of ozone, nitrogen dioxide, PM10, PM2.5 and acid vapor in all areas. An example of an ongoing study on effects of traffic-related air pollution on morbidity is the EU funded TRAPCA study, in which three birth cohorts (studies in which children are being fol

25、lowed from birth) in Holland, Sweden and Germany are being utilized to study effects of air pollution on the incidence of asthma and allergy in young children (12). Exposure assessment in this study is based on dedicated monitoring of PM2.5, Black Smoke (as proxy for diesel exhaust) and nitrogen dio

26、xide on 40 locations in each of the three study areas. The data are then being used to develop a stochastic model relating air pollution concentrations to measures of population and traffic density which are available in Geographic Information Systems (GIS). The GIS data are then used to estimate lo

27、ng-term exposure to air pollution at the address of each study subject.3. Design issuesIn this section, a number of design issues will be discussed, related to cohort studies of air pollution health effects. These are:a. study hypothesisb. exposure assessmentc. confounder assessmentd. effect modific

28、ation3a.Study hypothesisIn most of the quoted studies on effects of long-term exposure to air pollution on survival, there has been no explicit hypothesis with respect to the time sequence of exposure and effect. In comparison, the (vast) literature on the effects of active smoking on survival, much

29、 consideration is given to issues such as age when smoking started, number of years smoked and other cumulative measures of exposure such as pack-years, as well as to latency (the minimum time that needs to elapse between the start of exposure and the effect. Although the discussion section of publi

30、shed papers usually does address the issue of historic exposures to some extent, lack of data is then often used as an argument to ignore the exposure history. It remains unclear, therefore, to what extent the associations seen in recent cohort studies were produced by historic exposure contrasts, w

31、hich may have been much larger than current or recent contrasts. It is not entirely clear that historic exposures cannot be estimated with some reliability. A recent case-control study from Sweden has used historic emission data to model exposure of lung cancer cases and controls over a period of mo

32、re than 30 years; a relationship with nitrogen dioxide as a measure of traffic-related air pollution was found, but not with sulfur dioxide as a measure of energy-generation related pollution (13).Another issue related to study hypotheses is that originally, air pollution was thought to mainly affec

33、t the respiratory system, so that not much attention was paid to collecting data on other (notably cardiovascular) endpoint and their determinants (see also below, confounding). This is clear for instance in the Six Cities Study, which was designed more than 25 years ago. Very few cohort studies hav

34、e been started with air pollution as the primary hypothesis to begin with. Several of the quoted examples (3, 10, 12) are studies started for other reasons than air pollution. As a result, these studies are not optimal in terms of exposure contrasts, endpoint information, confounder information etc.

35、 The ACS study (3) was primarily designed to study lifestyle factors in relation to cancer incidence; in the participating communities, study populations were not representative of the community population at large. Information on occupational exposures was poor; subjects were spread out over large

36、metropolitan areas, in which air pollution exposure was characterized by measurements conducted at a single site, for a limited period of time, with uncertain quality (8). 3bExposure assessmentAs mentioned, exposure assessment has been very different in the different cohort studies. Especially the c

37、ohort studies that try to link air pollution to survival have, of necessity, had to work with incomplete data, either in terms of exposure history, or air pollution components, or both. As the Swedish case-control study (13) has shown, there is perhaps room for improvement by utilizing historic data

38、 on emissions. Another example is a study on college freshman in California, in which detailed retrospective data on residential history and time-activity patterns were collected (14, 15).It should be noted that the cohort studies on children (11,12) provide a much better opportunity to do concurren

39、t and state-of-the-art exposure assessment, simply because there is not much of an exposure history to deal with. Such studies, however, deal with morbidity rather than mortality, and the major questions about the effect of current levels of air pollution on survival will have to be answered by stud

40、ies on adult populations in which the problems surrounding historic exposures cannot be so easily avoided. 3cConfoundingWhen cohort studies were designed to investigate respiratory health endpoints primarily, there is a risk that there is insufficient information on confounding factors related to ot

41、her (notably cardiovascular) health endpoints. In the Six Cities Study (2), there was a lack of dietary information for this reason, and body mass index was used as a surrogate. When cohort studies were designed to study another association than the association between air pollution and survival to

42、begin with, important confounder information may be missing. Even when occupational histories have been collected, this information may be of limited use in establishing confounding of associations between air pollution and cardiovascular deaths (16). Again, the cohort studies on children have bette

43、r opportunities to investigate confounding, as new questions can more easily be added to a next round of questionnaires.3dEffect modificationIn the main cohort studies on air pollution and survival, there was no apparent effect modification by age or smoking. The AHSMOG study (4) found effect modifi

44、cation by gender, and also by baseline anti-oxidant vitamin intake, with smaller air pollution effects found in subjects with high anti-oxidant vitamin intake at baseline. The re-analysis of the Six Cities and ACS studies found significant effect modification by educational status, the effect of air

45、 pollution being restricted mainly to subjects with low and medium educational status (8). It is not clear whether this represents differences in exposure to air pollution or other factors related to education such as nutritional status. It would be of considerable research and public health interes

46、t to investigate in other cohort studies also in more detail the extent to which effects of air pollution are stronger in some sub-groups of the study population than in others.4 Concluding remarksOne could argue that the long-term effects of air pollution on survival are important enough to warrant

47、 some new cohort studies on air pollution and survival (17). Such studies could utilize sophisticated new methods for measuring air pollution exposures, and they could take on board information on all factors that are now seen as potentially important confounders. The main disadvantages of starting

48、new studies are their expense, and, perhaps more important, the long period needed to produce meaningful results. However, when we realize that the Harvard Six Cities Study, which was designed almost three decades ago, is still among the most important studies today, and when we realize that the cos

49、t of reducing air pollution further is enormous, it would seem foolishly shortsighted not to invest in one or two new Six Cities Studies so that the unresolved questions we have today can be answered with confidence in the years ahead.In the meantime, a number of approaches can be tried to maximize

50、the use of existing cohort studies designed to answer other questions. As has been suggested elegantly by Kunzli and Tager (18), many cohort studies already exist in the world that could potentially be utilized to answer questions on air pollution effects. One of their suggestions is to actively exp

51、lore the use of Geographic Information Systems to impute air pollution exposures at the address rather than the community level, and the quoted examples from Sweden (12, 13), the Netherlands (10, 12) and Germany (12) can be seen as steps in this direction. The Swedish example also suggests that we s

52、hould perhaps contemplate the use of case-control studies more than we have done in recent years.References1. Rothman K & Greenland S. Modern Epidemiology, 2nd edition. Lippincott-Raven, 19982. Dockery DW, Pope CA III, Xu X, Spengler JD, Ware JH, Fay ME, Ferris BG, Speizer FE. An association between

53、 air pollution and mortality in six US cities. N Engl J Medicine 1993: 329: 1753-1759 3. Pope CA III, Thun MJ, Namboodiri MM et al. Particulate air pollution as a predictor of mortality in a prospective study of U.S. adults. Am J Resp Crit Care Med 1995; 151: 669-6744. Abbey DE, Nishino N, McDonnell

54、 WF, Burchette RJ, Knutsen SF, Beeson WL, Yang JX. Long-term inhalable particles and other air pollutants related to mortality in nonsmokers. Am J Resp Crit Care Med 1999; 159: 373-3825. Brunekreef B. Air pollution and life expectancy: is there a relationship? Occup Environ Med 1997; 54: 781-7846. N

55、evalainen J, Pekkanen J. The effect of particulate air pollution on life expectancy. Sci Total Environ 1998 Jun 30;217(1-2):137-1417. Kunzli N, Kaiser R, Medina S, Studnicka M, Chanel O, Filliger P, Herry M, Horak F Jr, Puybonnieux-Texier V, Quenel P, Schneider J, Seethaler R, Vergnaud JC, Sommer H.

56、 Public-health impact of outdoor and traffic-related air pollution: a European assessment. Lancet 2000; 356: 795-8018. Reanalysis of the Harvard Six Cities Study and the American Cancer Society Study of Particulate Air Pollution and Mortality, 2000. A Special Report of the Institutes Particle Epidem

57、iology Reanalysis Project - Final version. Health Effects Insitute, Cambridge, MA, 20009. Baldi I, Roussillon C, Filleul L, Tessier JF, Kauffmann F, Brochard P, Nejjari C. Effect of air pollution on long-term mortality: description of mortality rates in relation to pollutant levels in the French PAA

58、RC study. Eur Resp J 1999; 14 (suppl 30): 392S10. Hoek G, Brunekreef B, Goldbohm S, Fischer P, van den Brandt P. Long-term exposure to traffic related air pollution is associated with mortality in a Dutch cohort study. Submitted for publication.11. James Gauderman W, McConnell R, Gilliland F, London

59、 S, Thomas D, Avol E, Vora H, Berhane K, Rappaport EB, Lurmann F, Margolis HG, Peters J. Association between air pollution and lung function growth in southern california children. Am J Respir Crit Care Med 2000; 162: 1383-139012. http:/www.slm.wau.nl/eoh/research/609660.htm13. Nyberg F, Gustavsson

60、P, Jarup L, Bellander T, Berglind N, Jakobsson R, Pershagen G. Urban air pollution and lung cancer in Stockholm. Epidemiology 2000; 11: 487-49514. Kunzli N, Lurmann F, Segal M, Ngo L, Balmes J, Tager IB. Association between lifetime ambient ozone exposure and pulmonary function in college freshmen-r

61、esults of a pilot study. Environ Res 1997; 72: 8-23 15. Kunzli N, Lurman F, Segal M, Ngo L, Balmes J, Tager IB. Reliability of lifetime residential history and activity measures as elements of cumulative ambient ozone exposure assessment. J Expo Anal Environ Epidemiol 1996; 6: 289-310 16. Steenland

62、K, Fine L, Belkic K, Landsbergis P, Schnall P, Baker D, Theorell T, Siegrist J, Peter R, Karasek R, Marmot M, Brisson C, Tuchsen F. Research findings linking workplace factors to CVD outcomes. Occup Med 2000; 15: 7-6817. Brunekreef B. All but quiet on the particulate front. Am J Resp Crit Care Med 1999; 159: 354-35618. Kunzli N, Tager IB. Long-term health effects of particulate and other ambient air pollution: research can progress faster If We want It To. Environ Health Perspect 2000; 108: 915-918