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.
In previous decades, numerous time-series studies have demonstrated an association between air pollution and daily mortality figures, 1–4 including data from the Netherlands. 5,6 In these time-series studies, the exposure for the whole study population is estimated by a few air pollution measurement sites, generally background sites. Nevertheless, traffic-related air pollution components are not equally distributed over cities, 7,8 and homes along busy streets can have higher indoor levels of air pollutants than homes along quiet streets. 7 Recent studies showed that the correlation over time between personal exposure and measurements of particles made at background measurement sites is reasonable, 9,10 but also that people living along busy streets had higher levels of personal exposure than people living along quiet streets. 9 Misclassification of exposure may occur if a background site is used to estimate the exposure of people living along busy roads. If this misclassification is unrelated to outcome, the estimated coefficient between exposure and health effects will be attenuated for those people. This attenuation is stronger if the variance of the measurement error is larger. 11 Using traffic-influenced measurement sites might give better exposure estimates for people living along busy roads and reduce the attenuation in effect estimates. At the same time, however, traffic stations generally have a larger range in concentrations than background stations, which will reduce the effect estimate in the same population.
In this study we try to evaluate whether people living along busy roads have higher exposures to air pollution than people living along quiet streets. We estimate the effects of air pollution on daily mortality in the general population and in a subgroup of people living along busy roads using background concentrations. By comparing the size of the effect estimates, we will evaluate whether people living along busy roads have a higher exposure to air pollution. Secondly, we will demonstrate the influence of the type of measurement site on the effect estimates of the total population by using exposure estimates of background and traffic sites.
Subjects and Methods
We obtained counts of total daily deaths within the city of Amsterdam (population 718,000 in 1998) from the Municipal Population Register for the period January 1987 through November 1998. There were separate counts for people living along roads on which more than 10,000 motorized vehicles pass per day (about 10% of the total population) and for people living along roads with fewer vehicles passing per day. For simplicity, when we use “traffic population” or “traffic mortality,” we mean the population or the mortality along roads with more than 10,000 motorized vehicles per day.
We obtained air quality monitoring data from the Amsterdam Environmental Research Institute. We obtained continuous data for sulfur dioxide (SO2), nitrogen dioxide (NO2), nitrogen oxide (NO), carbon monoxide (CO), and ozone (O3) and daily 24-hour averages for black smoke from all background and traffic-influenced sites in Amsterdam. Twenty-four-hour concentrations of particulate matter with an aerodynamic diameter of <10 μm (PM10) were measured once every 3–4 days at background locations. The continuous data except O3 were recalculated to 24-hour averages. Hourly O3 data were recalculated to daily maximum 8-hour means. Missing 24-hour averages were estimated with linear regression using the same component on other sites of the same type if the correlation between the sites was higher than 0.7. 5 Daily 24-hour-averaged concentrations of black smoke and total suspended particulate measured in the city of Rotterdam (located about 80 km from Amsterdam) were used to predict missing black smoke and PM10 values, respectively, on background sites in Amsterdam. 5
Daily city-wide concentrations were estimated for each type of site separately (background or traffic) by taking the average of the available daily measurements for each of the two types.
We obtained meteorological data from the weather station at Amsterdam’s Schiphol Airport of the Royal Dutch Meteorological Institute and the weekly incidence of influenza-type illnesses in the Netherlands from the Dutch Institute of Primary Health Care in Utrecht.
We fitted generalized additive Poisson regression models for the daily death counts in two stages. First, we modeled confounding variables on the total mortality. We filtered out seasonal and long-term temporal patterns. We included smoothed functions of time, influenza and relative humidity, using Akaike’s information criterion to determine the degrees of freedom. 12 We included dummy variables for day of the week. We modeled temperature as two linear variables 13 : “cold” (0 if mean temperature of 3 days before was ≥14°C, 14°C minus mean temperature of 3 days before if temperature was <14°C) and “warm” (0 if temperature of same day was <14°C, temperature of same day minus 14°C at higher values). We checked the model fit by plotting residuals against day of study, temperature, relative humidity, influenza, and fitted values, both for total mortality as well as traffic mortality. In the second stage we included the air pollution variables as continuous variables in the basic models and fitted separate generalized additive Poisson regression models for the daily death counts of the traffic population using concentration of background or street stations. For each component, only days with an estimate for both the background and traffic-influenced sites were used in analysis. Standard errors of the regression coefficients were adjusted for overdispersion. Time-series analyses were done in Splus 2000, and relative risk calculations in Microsoft Excel 97.
Table 1 presents the distribution of mortality counts in the total population as well as the traffic population. Air pollution concentrations at background and traffic-influenced stations are also presented. The levels at the traffic-influenced stations are for most components but especially black smoke, NO, and CO higher than at the background stations. Only the O3 concentration is higher at the background than at the traffic-influenced stations. Black smoke has a low number of observations because before 1993 black smoke was not measured at the traffic-influenced stations. The same is true for CO at the background sites over the years 1995–1998. Spearman correlation coefficients between the traffic-influenced and background concentrations were 0.49 (CO), 0.55 (black smoke), 0.74 (NO), 0.75 (SO2), 0.79 (NO2), and 0.80 (O3).
Detailed information about the correlation of the separate monitoring stations is given in Table 2 for black smoke and NO2. The correlation coefficients between the background stations or between the traffic stations lie in the same range. There is no pattern indicating that background stations are better correlated with each other than the traffic stations.
Results of the Poisson regression are presented in Table 3. The first column presents the results of the analysis using mortality in the total population and background concentrations. Most measured components were only weakly associated with daily mortality. Only black smoke and NO2 showed stronger associations. The second column shows the mortality in the traffic population using background stations. The relative risks of black smoke, PM10, CO, and NO2 increase, and those of O3 decrease. The increase in relative risk of black smoke is more than twofold. The last column shows the results of the analysis of the mortality in the total population using traffic sites. Comparing the background and the traffic stations as exposure indicator, the relative risks for all components except O3 are reduced and the confidence intervals narrower. The reduction in relative risk is most pronounced for black smoke and NO2. Confidence intervals of SO2 were only slightly narrower.
Sensitivity analysis that included additional terms for holidays and additional modeling of weeks with high influenza incidence did not change these results (not shown). Separate analysis of summer and winter with reduction of the number of degrees of freedom for time trend to compensate for fewer observations showed that the associations between black smoke and mortality were stronger in summer than in winter. For example, the relative risks using the background air pollution concentrations for the total population were for black smoke lag1 1.638 (95% confidence interval = 1.080–2.485) in summer vs 1.313 (95% confidence interval = 1.069–1.612) in winter. This pattern did not change after exclusion of days with high temperatures or extreme mortality counts (not shown). The other components did not show this difference between summer and winter.
This study showed higher concentrations of all air pollution components except O3 at traffic-influenced sites compared with the background sites. Black smoke and NO2 were associated with daily mortality. Most components, but especially black smoke, are more strongly related with mortality in the traffic population than in the total population. In the total population, using the concentrations at the background stations, the air pollutants showed larger effect estimates than those obtained using traffic stations. The effect estimates of black smoke were stronger in the summer than in the winter.
The higher levels of NO2, NO, CO, and black smoke at the traffic-influenced measurement sites compared with the background sites confirm combustion engines as source of air pollution. The lower O3 levels can be explained by the scavenging effect of NO on O3. Despite the differences in concentration level at the two types of measurement site, the correlation over time between the background and street concentrations was for all components 0.5 or higher. This result indicates the importance of meteorological factors on the variation of daily air pollution concentrations and illustrates our geographically relatively small study area.
The association of black smoke with mortality aligns with the results of an earlier mortality time-series study between 1986 and 1992 in Amsterdam, 5 although the size of the effect estimates of the background concentrations are larger in the present study. A possible explanation might be statistical variation caused by the different time period that we analyzed (1987–1998). Another explanation might be that the different time period resulted in somewhat lower median levels of CO and SO2 than in the study of Verhoeff et al. 5 This finding indicates that the air pollution mixture has changed over the years, which may have had implications for toxicity of particles. Other European studies not only found associations between daily mortality and black smoke levels, 1 but also between daily mortality and NO2 or O3. 6,13–17 Our study also showed an association with NO2 but not with O3.
Differences in effect estimates between summer and winter have been reported by other studies. 1,5,13 Possible causes are, among others, differences in composition of air pollutant mixture between summer and winter, the absence of influenza in the summer as possible confounder, and the higher correlation in the summer between personal and ambient exposure owing to more time outdoors and more frequently open windows. It is not clear why in this study only black smoke shows this difference, but differences in emissions and indoor penetration of particles and gaseous components between summer and winter might be important.
Only 10% of the total Amsterdam population resides along busy roads. Nevertheless, we were still able to show associations between black smoke, NO2, and daily mortality for this subpopulation. These associations were stronger than in the total population. An explanation might be that the traffic population is more highly exposed to the causative components in the air pollution. The higher range in measured concentrations at the traffic stations allows this possibility. The moderate correlation between the daily concentrations at the background and those at the traffic stations is high enough to detect variations in mortality in the same population with both measures of exposure, but leaves enough potential for misclassification. Reflectance measurements of particle filters from measurements made indoors as well as outdoors in Amsterdam showed higher levels in busy than in quiet streets. 7 A personal exposure study in Amsterdam showed that living along a busy road increased average personal exposure to PM10 with 23 μg/m3 compared with background monitor concentrations. The authors expressed doubts, however, about the size of this difference, as the indoor concentrations in the homes of these persons were not elevated. 9 People usually do not spend the whole day in their homes but enter the traffic stream to go to work, shops, family, etc, primarily during daytime when traffic intensity is highest. The places visited might be located in background streets although they live in a street with high traffic density. Personal exposure will be less than the exposure measured at traffic-influenced stations, but above the exposure estimated with the background stations.
The effect estimates on total population mortality were larger using the concentrations at the background stations than when using concentrations at the traffic stations. The lower variance in daily air pollution concentrations at the background stations compared with the traffic stations may have caused this. With a smaller range in air pollutant concentrations, we tried to explain the same range in mortality counts, which resulted in a larger effect estimate. A given increase of pollutant concentrations at the background site is associated with a larger increase at the traffic site. The larger confidence intervals of the effect estimates when using background concentrations are probably also caused by the smaller range in concentrations.
From the results of this study, it is not possible to conclude which of the two exposure estimates is the most appropriate for the traffic population. The higher relative risks of the population living along busy roads (traffic population) compared with those of the background population using background concentrations as exposure estimate suggest that the exposure of the traffic population is higher than the estimate of the background stations. Nevertheless, the use of exposure measurements obtained from traffic stations for the total population showed that using traffic stations to measure exposure will lead to underestimates of the effect when the traffic sites overestimate the exposure range.
Part of the heterogeneity found in the effect estimates of different time-series studies might be caused by differing definitions of what is regarded as a background station and the differing assessments of the proportion of the total study population that is living along busy roads.
1. Schwartz J. Air pollution
and daily mortality
: a review and a meta-analysis. Environ Res 1994; 64: 36–52.
2. Katsouyanni K, Touloumi G, Spix C, Schwartz J, Balducci F, Medina S, Rossi G, Wojtyniak B, Sunyer J, Bacharova L, Schouten JP, Ponka A, Anderson HR. Short term effects of ambient sulphur dioxide and particulate matter on mortality
in 12 European cities: results from time series data from the APHEA project. BMJ 1997; 314: 1658–1663.
3. Vedal S. Ambient particles and health: lines that divide. J Air Waste Manage Assoc 1997; 47: 551–581.
4. Pope CA III. Mortality
and air pollution
: associations persist with continued advances in research methodology. Environ Health Perspect 1999; 107: 613–614.
5. Verhoeff AP, Hoek G, Schwartz J, van Wijnen J. Air pollution
and daily mortality
in Amsterdam. Epidemiology 1996; 7: 225–230.
6. Hoek G, Schwartz JD, Groot B, Eilers P. Effects of ambient particulate matter and ozone on daily mortality
in Rotterdam, the Netherlands. Arch Environ Health 1997; 52: 455–463.
7. Fischer PH, Hoek G, van Reeuwijk H, Briggs DJ, Lebret E, van Wijnen JH, Kingham S, Elliott PE. Traffic
-related differences in outdoor and indoor concentrations of particles and volatile organic compounds in Amsterdam. Atm Environ 2000; 34: 3713–3722.
8. van Vliet P, Knape M, de Hartog J, Janssen N, Harssema H, Brunekreef B. Motor vehicle exhaust and chronic respiratory symptoms in children living near freeways. Environ Res 1997; 74: 122–132.
9. Janssen NAH, Hoek G, Brunekreef B, Harssema H, Mensink I, Zuidhof A. Personal sampling of PM10
in adults: relation between personal, indoor and outdoor concentrations. Am J Epidemiol 1998; 147: 537–547.
10. Wallace L. Indoor particles: a review. J Air Waste Manage 1996; 46: 98–126.
11. Zeger SL, Thomas D, Dominici F, Samet JM, Schwartz J, Dockery D, Cohen A. Exposure measurement error in time-series studies of air pollution
: concepts and consequences. Environ Health Perspect 2000; 108: 419–426.
12. Hastie TJ, Tibishirani RJ. Generalized additive models. London: Chapman and Hall, 1990.
13. Hoek G, Brunekreef B, Verhoeff A, van Wijnen J, Fischer P. Daily mortality
and air pollution
in the Netherlands. J Air Waste Manage Assoc 2000; 50: 1380–1389.
14. Touloumi G, Katsouyanni K, Zmirou D, Schwartz J, Spix C, Ponce de Leon A, Tobias A, Quennel P, Rabczenko D, Bacharova L, Bisanti L, Vonk JM, Ponka A. Short-term effects of ambient oxidant exposure on mortality
: a combined analysis within the APHEA project. Am J Epidemiol 1997; 146: 177–185.
15. Touloumi G, Katsouyanni K, Zmirou D, Schwartz J, Spix C, Ponce de Leon A, Tobias A, Quennel P, Rabczenko D, Bacharova L, Bisanti L, Vonk JM, Ponka A. Short-term effects of ambient oxidant exposure on mortality
: a combined analysis within the APHEA project. Am J Epidemiol 1997; 146: 177–185.
16. Bremner SA, Anderson HR, Atkinson RW, McMichael AJ, Strachan DP, Bland JM, Bower JS. Short-term associations between outdoor air pollution
in London 1992–4. Occup Environ Med 1999; 56: 237–244.
17. Health effects of outdoor air pollution
: state of the art. Part 2. Committee of the Environmental and Occupational Health Assembly of the American Thoracic Society. Am J Respir Crit Care Med 1996; 153: 477–498.