Traffic related air pollution and acute hospital admission for respiratory diseases in Drammen, Norway

Transcription

1 European Journal of Epidemiology 18: , Ó 2003 Kluwer Academic Publishers. Printed in the Netherlands. RESPIRATORY DISEASES Traffic related air pollution and acute hospital admission for respiratory diseases in Drammen, Norway Bente Oftedal 1, Per Nafstad 1,2, Per Magnus 1, Sonja Bjørkly 3 & Anders Skrondal 1 1 Division of Epidemiology, Norwegian Institute of Public Health, Oslo; 2 Buskerud University College, Drammen; 3 Drammen Kommune, Drammen, Norway Accepted in revised form 28 November 2002 Abstract. The aim of this study was to estimate the associations between seven ambient air pollutants [particulate matter (PM 10 ), nitrous dioxide (NO 2 ), sulfur dioxide (SO 2 ), ozone (O 3 ), benzene, formaldehyde and toluene] and acute hospital admissions for respiratory diseases in Drammen, Norway Time-series analysis of counts was performed by means of generalized additive models with log link and Poisson distribution. The results showed that benzene was the pollutant having the strongest association with respiratory diseases for the total study period, the relative risk of an interquartile increase of benzene was with 95% confidence interval: The corresponding results were ( ) for formaldehyde, ( ) for toluene, ( ) for NO 2, ( ) for SO 2, ( ) for O 3 and ( ) for PM 10. Dividing the total study period into two 3-year periods, there was a substantial reduction in the exposure levels of the volatile organic compounds (benzene, formaldehyde and toluene) from the first to the second period. Separate analyses for the second time period showed weaker association between these pollutants and the health outcome. This study provides further evidence for short-term respiratory health effects of traffic related air pollution. Key words: Air pollution, Benzene, Hospital admissions, NO 2,PM 10, Respiratory disease Background Time-series analyses have demonstrated associations between daily concentrations of several air pollutants and daily number of deaths and hospital admissions for citizens in several cities around the world [1 14]. Most often measures of particulate air pollution have been found to be associated with adverse health effects [1 12]. However, a study from Drammen in Norway from 1994 to 1997 found that benzene, among eight ambient air pollutants, was the pollutant most highly associated with acute hospital admissions for respiratory diseases [15]. We carried out a followup of the study from Drammen by prolonging the study period with three additional years. The aim of the study was to further explore the associations between ambient air pollutants and acute hospital admissions for respiratory diseases using a larger data set. Methods The study from Drammen is described in detail elsewhere [15]. Briefly, the study was a timeseries analysis including number of acute hospital admissions for respiratory diseases and daily concentration levels of eight different pollutants [particulate matter 10 lm in diameter (PM 10 ), nitrous oxide (NO), nitrous dioxide (NO 2 ), sulfur dioxide (SO 2 ), ozone (O 3 ), benzene, formaldehyde and toluene]. The study period was from November 1994 to December We prolonged the study period with three additional years, from January 1998 to December The study population was the same [15] while we confined the outcome to acute hospital admissions for respiratory diseases classified according to the International Classification of Diseases, 9th revision codes and 10th revision codes J00 J99. Data collection on environmental exposure (temperature, humidity, influenza epidemics and air pollution) was conducted identically, except that the authorities had ceased the registration of NO [15]. The other pollutants are listed in Table 1. PM 10 was measured by the tapered element oscillating microbalance technique [16] and the other pollutants were measured by differential optical absorption spectroscopy [17]. Benzene concentrations were lost due to technical problems for 22 consecutive days in May and June The other variables had missing values for some days. We chose to include only days with information on all the variables listed in Table 1, including temperature and humidity.

2 672 Table 1. Means and standard deviations (SD) of daily number of hospital admissions and daily concentrations of air pollutants (lg/m 3 ) in Drammen, Norway Total period Mean SD Mean SD Mean SD Hospital admissions Respiratory diseases (ICD and ICD-10 J00-J99) Air pollutants Benzene Formaldehyde Toluene PM NO SO O Dataset includes days (n = 974 ( ), n = 973 ( ) and n = 1938 (Total Period)) with valid measures for all the listed variables including temperature and humidity. ICD-9/ICD-10 = International Classification of Diseases, 9th revision/10th revision. Statistical methods Time-series analysis of counts [18] was performed by means of generalized additive models (GAM) [19] with log link and Poisson distribution as described for the study [15]. The convergence criteria for GAM in S-plus were changed from the default value 10 )3 to 10 )15 to assure that the convergence of its iterative estimation procedure reached the maximum likelihood value [20]. To be able to compare the results from the different study periods, we used the interquartile range for the total period in the relative risk estimation. The stagewise modeling strategy based on the Akaike Information Criterium (AIC) was retained in this study too [15]. Adhering to the Health Effects Institute (HEI) report of 2000 [21], the smoothing parameter for the cubic spline smoother of time trends was now chosen as 7 degrees of freedom per year, instead of 12 degrees of freedom per year as used before [15]. The results from both approaches were consistent, in accordance with the sensitivity results in the HEI report of 2000 [21]. Otherwise, we adhered to the previous modeling strategy [15]. Results The daily mean concentration of benzene was 14.9 lg/m 3 for the first 3-year period, while it decreased to 5.9 lg/m 3 for the last (Table 1). The daily mean benzene level dropped gradually during the total study period, from 16.8 lg/m 3 in 1994 to 4.0 lg/m 3 in The same decreasing trend occurred for formaldehyde and toluene, while PM 10 and NO 2 levels were rather stable. The level for SO 2 was very low in both periods. The correlation between the pollutants was of the same order of magnitude as for the study (from )0.47 to 0.78) [15]. The exception was a reduced correlation between benzene and the other volatile organic compounds (VOCs) formaldehyde and toluene [0.51/ 0.04 (for / ) and 0.69/0.16, respectively]. The results from the stagewise modeling strategy suggest that the background model for the previous period [15] with the addition of summer vacation is the preferred. The model for the total period substituted a 3-day lag of temperature and summer vacation with a current lag of humidity and Christmas vacation. For the new period the model is the same as for the total period, with a smooth spline function of a 2-day lag of temperature taking the place of humidity. Furthermore, there was no sign of autocorrelation of the residuals. In the total time period benzene was the pollutant showing the strongest association to respiratory diseases (Table 2). We also found associations with formaldehyde, toluene, NO 2 and SO 2, but a lower association with PM 10. The smoothed log relative risk of hospital admissions for respiratory diseases for the pollutants with the strongest effects, benzene (Figure 1) and NO 2 (Figure 2), showed linear relationships, thereby supporting the inclusion of these variables as linear components in the statistical model. In two-component models including PM 10 and one other pollutant, the associations were in general weaker. Benzene still had the strongest association, the relative risk for the total period was (with 95% confidence interval: ) for benzene and ( ) for PM 10. Corresponding figures for a model with NO 2 and PM 10 were ( ) and ( ). Including benzene and NO 2 in the same model, the relative risk was ( ) for benzene and ( ) for NO 2.

3 Table 2. Relative risks (RR) with 95% confidence intervals (CI) of hospital admissions for respiratory diseases by an interquartile (IQ) increase in daily concentration for the total period Total period IQ RR 95% CI RR 95% CI RR 95% CI Benzene Formaldehyde Toluene PM NO SO O Models with one pollutant only. 673 Comparing the results from the total period with the study, the associations with the VOCs were reduced. Therefore we also modelled the last time period only and found no associations between hospital admissions for respiratory diseases and VOCs for this period. The association with PM 10 was also lower than in the study, but this was not the case for NO 2 and SO 2. The association with NO 2 was positive and of the same magnitude for both time periods. Discussion Figure 1. Log relative risk of hospital admissions for respiratory diseases vs. the concentration of benzene for the total period, using a single-pollutant model adjusted for background variables. The dashed lines give 95% confidence limits, and the vertical lines along the x-axis indicate the number of observations. Figure 2. Log relative risk of hospital admissions for respiratory diseases vs. the concentration of NO 2 for the total period, using a single-pollutant model adjusted for background variables. The dashed lines give 95% confidence limits, and the vertical lines along the x-axis indicate the number of observations. We found positive associations between daily number of hospital admissions for acute respiratory diseases and concentrations of several air pollutants. For NO 2 and SO 2 the associations did not change substantially from the first to the second 3-year period, while it was reduced for the VOCs. Contrary to many other time-series studies, information was available enabling us to estimate associations between health outcome and seven different air pollutants. As earlier reported benzene was the air pollutant with the strongest association with the health outcome during the study. The study period has now been extended with three new years. The results of the analyses of the 6-year period confirmed many of the earlier findings and strengthen the evidence of short-term health effects of air pollution exposure in Drammen. Traffic related air pollutants including NO 2 and the VOCs seemed to be of particular importance. We have no explanation for why PM 10 concentrations were so weakly associated with the health outcome. However, urban air is always a mixture of pollutants and daily level of any urban air pollutant could be considered only as an indicator of air pollution exposure. The effect of extremely low SO 2 concentrations supports such a view. In all time series studies of associations between daily levels of air pollution and short term health effects, exposure is measured at the population level and is based on measurements from one or a few monitoring stations. The rationale for this approach

4 674 is that daily level of exposure in an urban area is assumed to vary similarly for most of the population even though the absolute levels could be different. With the current approach only conditions that show a daily variation pattern similar to the variation in air pollution could confound the association. This excludes conditions like tobacco smoking, which is unlikely to vary with daily levels of ambient air pollution. For the VOCs, and especially benzene, the association was substantially lower in the second than the first 3-year period. The rather dramatic reduction in ambient benzene (and other VOCs) exposure during the study period suggested that these results might be reconciled by factors related to the exposure. The main sources of benzene in ambient air in Norway are combustion of fuel by vehicles and burning of wood for heating. The reduction in ambient benzene concentrations can be explained by a gradual replacement of vehicles without catalysts during the 1990s, reinforced by new regulations. Introduction of new regulations has also lead to reduced benzene concentrations in petrol [22]. Potential explanations for the weaker association in the second compared to the first time period include (1) the low levels of benzene in the last period, (2) the small variation in benzene in the last period and (3) more measurement error in the last period. Reduced concentration would be an explanation if benzene only has health effects when exposure exceeds a threshold. However, as far as we know there is no evidence of such a threshold. Small variation in ambient benzene levels reduces the precision of estimated health effects of benzene exposure, making it harder to detect associations in the last time period. In this type of studies exposure is measured at one or a few stations at fixed locations, which is likely to introduce measurement error for individual exposure. Measurement error could be higher for low compared to high ambient levels of VOCs and hence produce attenuated associations in the second period. This study provides further evidence for short-term respiratory health effects of urban air pollution. Traffic related air pollution seems to be of particular importance. Benzene was the pollutant with the strongest association with the health outcome, but the association changed from the first to the second 3- year period. The substantial reduction in benzene exposure during the period might have contributed to this, but further studies of the associations between health effects and benzene are needed. References 1. Dockery DW, Pope III CA. Acute respiratory effects of particulate air pollution. Ann Rev Public Health 1994; 15: Lebowitz MD. Epidemiological studies of the respiratory effects of air pollution. Eur Respir J 1996; 9: Schwartz J, Marcus A. Mortality and air pollution in London: A time series analysis. Am J Epidemiol 1990; 131: Schwartz J, Morris R. Air pollution and hospital admissions for cardiovascular disease in Detroit, Michigan. Am J Epidemiol 1995; 142: Schwartz J, Slater D, Larson TV, Pierson WE, Koenig JQ. Particulate air pollution and hospital emergency room visits for asthma in Seattle. Am Rev Respir Dis 1993; 147: Schwartz J. Air pollution and hospital admissions for respiratory disease. Epidemiology 1996; 7: Schwartz J. Air pollution and hospital admissions for cardiovascular disease in Tucson. Epidemiology 1997; 8: Anderson HR, Spix C, Medina S, et al. Air pollution and daily admissions for chronic obstructive pulmonary disease in 6 European cities: Results of the APHEA project. Eur Respir J 1997; 10: Katsouyanni K, Touloumi G, Spix C, et al. Short term effects of ambient sulfur dioxide and particular matter on mortality in 12 European cities: Results from time series data from the APHEA project. Br Med J 1997; 314: Zimirou D, Schwartz J, Saez M, et al. Time series analysis of air pollution and cause-specific mortality. Epidemiology 1998; 9: Kelsall JE, Samet JM, Zeger SL, Xu J. Air pollution and mortality in Philadelphia, Am J Epidemiol 1997; 146: Moolgavkar SH, Luebeck EG, Anderson EL. Air pollution and hospital admissions for respiratory causes in Minneapolis St. Paul and Birmingham. Epidemiology 1997; 8: Burnett RT, Dales RE, Brook JR, Raizenne ME, Krewski D. Association between ambient carbon monoxide levels and hospitalizations for congestive heart failure in the elderly in 10 Canadian cities. Epidemiology 1997; 8: Fusco D, Forastiere F, Michelozzi P, et al. Air pollution and hospital admissions for respiratory conditionsin Rome, Italy. Eur Respir J 2001; 17: Hagen JA, Nafstad P, Skrondal A, Bjørkly S, Magnus P. Associations between outdoor air pollutants and hospitalization for respiratory diseases. Epidemiology 2000; 11: Patashnick H, Rupprecht EG. Continuous PM-10 measurements using the tapered element oscillating microbalance. J Air Waste Manag Assoc 1991; 41: Platt U. Differential optical absorption spectroscopy (DOAS). In: Sigrist MW (ed.), Air Monitoring by Spectroscopic Techniques. New York: John Wiley and Sons, Zeger SL. A regression model for time series of counts. Biometrika 1988; 75: Hastie TJ, Tibshirani RJ. Generalized Additive Models. London: Chapman & Hall, 1990.

5 Kaiser J. Air pollution risks. Software glitch threw off mortality estimates. Science 2002; 296: Health Effects Institute. The National Morbidity, Mortality, and Air Pollution Study. Part I: Methods and Methodologic Issues. Health Effects Institute, Number 94, Part I, June Larssen S, Hagen LO. Luftkvaliteten i norske byer (Quality of air in Norwegian cities). NILU-OR 69. Kjeller, Norway: Norwegian Institute for Air Research, Address for correspondence: Bente Oftedal, Division of Epidemiology, Norwegian Institute of Public Health, P.O. Box 4404 Nydalen, 0403 Oslo, Norway Phone: ; Fax: bente.oftedal@fhi.no