Cerebrovascular Diseases in Helsinki

Low-Level Air Pollution and Hospital Admissions for Cardiac and Cerebrovascular Diseases in Helsinki Antti Pbnkii, MD, and Mikko Virtanen, MSc Introduction Ambient air pollutants are known to increase morbidity and mortality from respiratorv infections and to exacerbate symptoms of chronic obstructive pulmo- -a: nary diseases. Pollutants have also been shown to increase the risk of death from cardiovascular and cerebrovascular diseases, sometimes perhaps even more significantly than the risk of death from.. respiratory causes.'-3 This increased risk of death has been due to high concentra tions of pollutants, as in the London smog episodes of the 1 950s and 1 960s and in the Ruhr area of Germany in 1962 and 1985.1 24' However, recent evidence from pe-, Athens and from various parts of the :....... United States suggests that particulates, X?,',, even at rather low concentrations, may increase the risk of death by exacerbating cardiovascular and cerebrovascular diseases.3'6-2 This observation is important. because the number of people exposed to low-level air pollution is very high, and even a slight increase in the relative risk of death results in a large number of..premature deaths. Although the effects of ambient air pollutants on the risk of death from ischemic cardiac and cerebrovascular disr;9 *?.2-eases are fairly well documented, the effects on the exacerbations of symptoms of these diseases are less well known. We therefore conducted a longitudinal investigation to determine whether low-level ambient air pollutants are associated with.:.' hospital admissions pertaining to ischemic cardiac and cerebrovascular diseases in.38t. -: Helsinki, Finland. The study investigated the relationships between daily counts of hospital admissions and ambient air sulfur dioxide, nitric oxide, nitrogen dioxide, ozonc, and total suspended particulate concentrations; tempcrature; and relative humidity. Materials and Methods Air Pollutants and Meteorologic Variables Air pollution measurements are conducted in Helsinki by district municipal authorities. Sulfur dioxide is measured hourly by coulometric instruments at four automatic monitoring stations, nitric oxide and nitrogen dioxide are measured by chemiluminescence at two stations, and ozone is measured by ultraviolet absorption at one station. Total suspended particulates are collected by high-volume samplers at six stations, four of which collect samples every second day and two of which collect samples every third day. At one weather station, measurements of temperature and relative humidity are recorded hourly. The total number of stations is eight (four in the center of the town and four in the suburbs); these stations range in height from 2 m to 10 m. The sites of these stations have been selected to represent the actual exposure of the population to ambient air pollutants. The daily mean conccntrations of each pollutant at the various stations during the study period (1987 through 1989) were used in the analysis. Helsinki City covers a relatively small area, 185 km2 (the population of Helsinki was 488 604 in 1987, 491 148 in 1988, and 491 777 in 1989). The main sources of air pollutants are energy production by coalfired and oil-fired power plants, road traffic, and, to a small extent, industrialization (Figure 1). In 1987, the total emission The authors are with the Helsinki Citv Center of the Environment. Helsinki. Finland. Requests for reprints should be sent to Antti Ponka. MD. Helsinki City Center of the Environment. Sturenkatu 25. 00510 Helsinki, Finland. This paper was accepted March 22. 1996. American Journal of Public Health 1273

PonkA and Virtanen FIGURE 1-Energy production plants, major roads, air pollution measurement stations, and residential areas in Helsinki. of nitrogen dioxide was 17 600 tons, of which 32% was derived from traffic, 67% from energy production, and 1% from industries. The monitoring stations measuring nitrogen dioxide are situated on the main streets. At street level, 60% to 80% of the nitrogen dioxide level is derived from road traffic. The total emission of sulfur dioxide was 22500 tons, 93% ofwhich derived from energy production, 2% from traffic, and 5% from industries. The stacks of the power plants are 100 to 150 m high, while the exhaust gases from cars spread out at street level. Thus, nitrogen dioxide can be taken as an indicator of pollution due to traffic, and sulfur dioxide can be taken as an indicator of pollution due to energy production. Incidence ofadmissionsforlschemic Cardiac and Cerebrovascular Symptoms The data concerning hospital admissions for exacerbations of symptoms of ischemic cardiac and cerebrovascular diseases were obtained from the register kept of all periods of these illnesses requiring hospitalization. Quality control was achieved by checking diagnoses after every episode of illness, as described later. The register contains information on the dates of hospitalization and on the diagnoses and ages of the patients. The data cover all of the municipal hospitals and Helsinki University Central Hospital, which together treat practically all patients with ischemic cardiac and cerebrovascular diseases requiring hospitalization. The study included only the residents of Helsinki. Those patients who were admitted to emergency wards, and thus needed more effective treatment, were also subsequently treated in bed wards but were studied as a separate group. Diagnoses were based on the Intemational Classification of Diseases (ninth edition; ICD-9), published by the World Health Organization. Those included were ischemic cardiac diseases (ICD-9 codes 410 through 414) and cerebrovascular diseases (ICD-9 codes 430 through 438). The following diagnoses were analyzed separately: acute myocardial infarctions (ICD-9 code 410), long-term transient ischemic myocardial attacks (ICD-9 code 411), and short-term transient ischemic myocardial attacks (ICD-9 code 413), as well as cerebral ischemia due to occlusion of extracerebral vessels (ICD-9 code 433) or cerebral vessels (ICD-9 code 434) and transient ischemic cerebral attacks (ICD-9 code 435). Patients with acute appendicitis (ICD-9 codes 540 through 543) were selected as controls. All diagnoses were made by specialists. Case records are compiled for each patient occupying a hospital bed or visiting an outpatient department. The ICD-9 code number for each patient is recorded by the physician after the episode requiring hospitalization when the results of all tests and examinations are available. The case records, as well as the diagnoses, are always seen by a senior specialist who is not directly responsible for treatment but who checks that the diagnostic criteria are valid. If the patient had more than one of the diagnoses just mentioned, only that which primarily led to hospitalization was used in this analysis. Statistical Methods Analytical methodology. The daily numbers of admissions were not normally distributed but skewed, with a low mean and a long tail. Such cases are usually modeled with the Poisson regression,13'14 which was also used in our analysis. The explanatory variables used were the concentrations of nitric oxide, nitrogen dioxide, sulfur dioxide, ozone, and total suspended particulates, along with temperature and relative humidity. For each pollutant, we studied the relation between the daily mean value of the average concentrations of each pollutant at the various stations and the number of admissions on the same day, as well as the number of admissions after lags of 1 to 7 days, because the effects of pollution and weather on admissions may be delayed. All of the concentrations of pollutants were skewed, with some high values. These values seemed to dominate the regression, and logarithmic transformations were used as a means of avoiding this difficulty. Because the number of admission categories was high, the association between any pollutant and a given admission category was tested by comparing models with and without all pollutants and weather variables. Both models included the basic secular variables (e.g., weekday dummies). If the result of such a test was significant, the modeling was repeated after omitting the nonsignificant pollutants and weather variables. In all phases of the modeling, if a pollutant was included in the model, all of its lagged values up to a lag of 7 days were included. The significance of associations between pollutants and admissions was tested by comparing the final model including the pollutant (or pollutants) with a model not including it. The evaluations were made in two different ways. First, all pollutants 1274 American Journal of Public Health September 1996, Vol. 86, No. 9

Pollution and Cardiac Diseases were included in the model simultaneously. Second, the pollutants were included one by one and evaluated separately. Missing values. The study period comprised 1096 days. Information about the concentration of total suspended particulates was lacking for 375 days, owing to the strategy of sampling, and nitric oxide, nitrogen dioxide, and sulfur dioxide values were lacking for 16, 13, and 4 days, respectively. However, the frequencies of admissions during the days with missing observations were similar to those of other days. Thus, the effect of the missing observations was probably small. Testing was done by comparing the numbers of admissions on days with missing values and on days with observed values. The missing values were imputed by means of the standard regression technique. In this method, the independent variables of the analysis in question are divided into two groups, one with missing data and the other without. First, each variable in the first group is used as a dependent variable and the variables in the second group are used as independent variables, and then the missing values are replaced with predictions from this regression. Next, the regression is extended as follows: in each regression, all of the other variables in the first group are added to the list of independent variables. The predictions from this second regression are then used to replace the missing values. In addition to the pollutants and weather variables, the imputation models included year and weekday dummies, as well as trends.15 After the actual modeling, the residuals were studied graphically to determine whether there were any differences between the imputed and nonimputed values. No clear patterns were detected. Temporalpattems and time series. The number of admissions varied seasonally and also in shorter cycles among the different weekdays. The secular trends and seasonal variation were controlled through the use of dummy variables. The number of admissions during the winter was taken as the baseline because, in every year of the study, admissions were most frequent during that period. The years were divided into four seasons: summer (June through August), autumn (September through November), winter (December through February), and spring (March through May). The number of admissions was highest on Mondays and considerably smaller on weekends, which TABLE 1-Numbers of Admissions for lschemic Cardiac and Cerebrovascular Diseases and the Significance of Their Overall Relations to Pollutants No. Admissions Daily No. Admissions Diagnosis in 3-Year Period Mean Range Deviance P All ischemic cardiac diseases (ICD-9 codes 410-414) Myocardial infarction (ICD-9 code 41 0) Transient long-term ischemic attack (ICD-9 code 41 1) Transient short-term ischemic attack (ICD-9 code 413) Cerebrovascular diseases (ICD-9 codes 430-438) Cerebral ischemia due to occlusion of extracerebral vessels (ICD-9 code 433) Cerebral ischemia due to occlusion of cerebral vessels (ICD-9 code 434) Transient ischemic cerebral attacks (ICD-9 code 435) Appendicitis (ICD-9 codes 540-543) Note. ER = emergency room. 12 664 7 005 4 501 2 677 1 670 1 062 2 134 1 087 7 232 3 737 11.6 1-28 86.7.0053 6.4 0-19 86.0.0061 4.1 0-12 68.6.1201 2.5 0-9 58.9.3702 1.5 0-7 75.4.0432 1.0 0-6 75.7.0411 2.0 0-12 70.7.0895 1.0 0-6 78.4.0257 6.6 0-18 66.5.1684 3.4 0-11 74.8.0478 254 0.2 0-3 49.7.7114 97 0.1 0-2 63.2.237 2 521 1 230 976 563 2 280 is probably explained by delay in seeking treatment. The effect of this day-of-theweek variation was controlled by using the dummy variables, with Monday as a baseline. The cyclic nature of the admissions was also checked up to the fourth order of sine and cosine terms. Because these fluctuations were found to be nonsignificant, they were not included in the final model. Cold weather has been considered to be associated with the symptoms of ischemic cardiac and cerebrovascular diseases.1618 In addition, severe infectious diseases such as influenza are more 2.3 0-8 52.7.600 1.1 0-5 58.2.396 0.9 0-5 49.7.712 0.5 0-4 60.0.475 2.1 0-4 41.3.928 harmful for people with preexisting disease than for healthy persons. Therefore, these two variables had to be controlled. The effect of weather was controlled by using a smoothing technique, and the effect of influenza epidemics was controlled by using multiple dummy variables. Influenza was classified as epidemic when the weekly count of influenza diagnoses in Helsinki was more than 50. The incidence of cardiovascular disease is decreasing in Finland, and this declining trend was seen in the present study. The predicted time series plot, residual time series plots, and periodograms were used to determine the appropriate- September 1996, Vol. 86, No. 9 American Journal of Public Health 1275

Ponkai and Virtanen TABLE 2-Mean Daily Concentrations of Ambient Air Pollutants and Mean Values TABLE 2-Mean Daily Concentrations of Ambient Air Pollutants and Mean Values of Meteorologic Variables in Helsinki: 1987 through 1989 ness of the control of confounding and for improvement of fit (chi-square test) in deciding which terms to include in the model. The raw data showed very strong autocorrelations that were effectively eliminated by using day-of-the-week dummies. The autocorrelations of the standardized residuals were calculated for Mean Range SD Sulfur dioxide,,ug/m3 19 0.2-95 12.6 Nitrogen dioxide,,ug/m3 39 4-170 16.2 Nitric oxide,,ug/m3 91 7-467 61.0 Ozone,,ug/m3 22 0-90 13.1 Total suspended particulates,,ug/m3 76 6-414 51.6 Mean temperature, C 5.4-37.0-26.4 9.3 Relative humidity, % 83 37-100 12.0 TABLE 3-Poisson Regression Models for Associations between Hospital Admissions for Ischemic Cardiac Disease (ICD-9 410-414) and Pollutants in Helsinki All Admissions ER Admissions Parameter Estimate SE P Estimate SE P Intercept 2.308 0.228 <.001 2.200 0.323 <.001 Year: 1988 0.006 0.023.808-0.007 0.027.796 Year: 1989 0.084 0.028.003-1.167 0.046 <.001 Tuesday -0.055 0.038.147-0.039 0.052.452 Wednesday -0.127 0.044.004-0.035 0.059.552 Thursday -0.118 0.044.007-0.092 0.059.122 Friday -0.263 0.044 <.001-0.103 0.059.080 Saturday -0.578 0.049 <.001-0.398 0.064 <.001 Sunday -0.421 0.044 <.001-0.366 0.059 <.001 Spring 0.060 0.027.036 0.088 0.036.016 Summer -0.020 0.028.480-0.004 0.038.918 Autumn 0.005 0.027.854 0.020 0.036.586 Influenza epidemic 0.020 0.040.613 0.037 0.052.478 NO, log, no lag 0.005 0.022.816-0.037 0.028.186 NO, log, lag = 1 day 0.097 0.023 <.001 0.111 0.030 <.001 NO, log, lag = 2 days -0.038 0.022.087-0.020 0.029.487 NO, log, lag = 3 days 0.011 0.023.615 0.030 0.029.304 NO, log, lag = 4 days -0.009 0.023.681-0.006 0.029.828 NO, log, lag = 5 days 0.002 0.023.929-0.006 0.029.825 NO, log, lag = 6 days -0.018 0.022.423-0.023 0.029.419 NO, log, lag = 7 days 0.009 0.022.696-0.008 0.029.794 03, log, no lag -0.074 0.028.007-0.0011 0.027.692 03, log, lag = 1 day 0.097 0.032.003-0.145 0.038 <.001 03, log, lag = 2 days -0.034 0.033.291 0.130 0.045.004 03, log, lag = 3 days 0.011 0.033.746-0.007 0.048.874 03, log, lag = 4 days 0.032 0.032.033-0.003 0.045.947 03, log, lag = 5 days 0.004 0.032.913 0.004 0.043.931 03, log, lag = 6 days -0.008 0.032.793-0.045 0.043.295 03, log, lag = 7 days -0.008 0.028.764-0.015 0.038.689 Deviance 1173.5 1287.6 Pearson X2 1151.0 1213.7 df 1060 1060 Note. ER = emergency room; NO = nitric oxide; 03 = ozone. each model.19 Because patients seldom seek treatment within a week from the previous admission, the risk groups (i.e., the groups producing the admissions) did not include the same subjects on successive days. Therefore, the dependence on previous observations involved two different sources: the effect of the shrinking of the risk group and the delayed or cumulative effect of the pollutants. Because the true size of the risk group was unknown, the former effect was assumed to be nonsignificant and was omitted from the models. Such an omission could have caused additional problems if a significant pollutant had been omitted from the model. Therefore, the autocorrelation pattern of the residuals was examined closely to detect such a possibility. Results Number ofadmissions There were 12 664 periods of hospitalization due to ischemic cardiac diseases and 7232 periods due to cerebrovascular episodes during the 3-year period (daily averages of 11.6 and 6.6 admissions, respectively). Every day, at least one such patient was admitted. More than half of the patients were admitted through emergency wards (55.3% of those with cardiovascular disease and 51.7% of those with cerebrovascular disease). The numbers of patients, by diagnosis, mode of admission, and age group, are presented in Table 1. Pollutants and Meteorologic Variables Concentrations of most pollutants in the ambient air were relatively low. The mean concentrations of sulfur dioxide, nitric oxide, nitrogen dioxide, and ozone during the 3-year period (averages of the means at the respective stations) were 19,ug/m3, 92 p,g/m3, 39,ug/m3, and 22,ug/m3, respectively. The mean total suspended particulate concentration was 76 p,g/m3, a high value resulting from the meteorologic conditions and from erosion of street surfaces caused by the studded tires used during the winter, as well as from the use of sand on the streets to treat icy surfaces. Mean daily values and ranges for the different variables are presented in Table 2. The mean 8-hour concentration of ozone was 16.1,ug/m3, and the highest 8-hour concentration was 50.1 ug/im3. When the different stations were compared, the mean long-term concentrations of sulfur dioxide had a 1.4- to 2.3-fold variation, those of total suspended particulates had a 2.6- to 2.8-fold variation, and those of nitrogen dioxide had a 1.3- to 1.6-fold variation. Sulfur dioxide, nitric oxide, and nitrogen dioxide had highly significant negative correlations with temperature, whereas ozone correlated positively with temperature. Sulfur dioxide correlated at highly significant levels with nitrogen dioxide and total suspended particulates and correlated negatively with ozone. 1276 American Journal of Public Health September 1996, Vol. 86, No. 9

Pollution and Cardiac Diseases TABLE 4-The Results of Poisson Regression Models Showing Significant Associations between Hospital Admissions for Transient lschemic Attacks (ICD-9 411, 413) and Cerebrovascular Diseases (ICD-9 430-438) and Pollutants in Helsinki Transient Long-Term lschemic Attack Transient Short-Term Ischemic Attack Cerebrovascular Diseases All Admissions ER Admissions (ER Admissions) (ER Admissions) Parameter Estimate SE P Estimate SE P Estimate SE P Estimate SE P Pollutant, log, no lag 0.121 0.069.080 0.161 0.089.071 0.155 0.101.128-0.039 0.054.467 Pollutant, log, lag = 1 day 0.043 0.068.526 0.033 0.089.715-0.056 0.105.591-0.025 0.057.657 Pollutant, log, lag = 2 days -0.072 0.070.299-0.080 0.089.371 0.128 0.109.591 0.083 0.058.156 Pollutant, log, lag = 3 days 0.182 0.069.008** 0.204 0.089.023* -0.016 0.107.884 0.032 0.058.580 Pollutant, log, lag = 4 days 0.021 0.070.759 0.061 0.089.496 0.097 0.107.362-0.030 0.057.596 Pollutant, log, lag = 5 days 0.088 0.069.201 0.015 0.088.869-0.096 0.105.362-0.016 0.059.781 Pollutant, log, lag = 6 days -0.029 0.069.679 0.045 0.089.609 0.234 0.110.034* 0.162 0.061.008** Pollutant, log, lag = 7 days -0.017 0.068.804 0.005 0.088.951-0.259 0.094.006** -0.072 0.054.182 Pollutant TSP TSP NO2 NO2 Deviance 1343.8 1224.2 1251.1 1376.1 Pearson X2 1164.8 1136.4 1178.9 1244.6 Note. Estimates for time variables are not presented. ER = emergency room; TSP = total suspended particulates. *P <.05; **P <.01 (P values for the estimates). Relation ofair Pollutants and Weather to Admissions Admissions due to all ischemic cardiac diseases showed a statistically significant association (P =.005 for the whole model) with pollutants in a model including all five pollutants. Admissions increased significantly with a 1-day lag after high nitric oxide (relative risk [RR] = 1.10 for a 2.7-fold increase, 95% confidence interval [CI] = 1.05, 1.15; P <.001) and ozone (RR = 1.10, 95% CI = 1.03, 1.18; P =.003; Table 3) concentrations; however, on days with high ozone concentrations, the relative risk of admission decreased (RR = 0.93,95% CI = 0.88,0.98; P =.007). The risk of emergency admissions due to all ischemic cardiac diseases showed similar relations to nitric oxide and ozone. Relative risks were 1.14 (95% CI = 1.06, 1.23;P <.001) 1 day after high nitric oxide concentrations, 0.87 (95% CI = 0.80, 0.94; P <.001) 1 day after high ozone concentrations, and 1.13 (95% CI = 1.03, 1.24; P =.004) 2 days after high ozone concentrations. The relative risk denotes the relative change in risk when the pollutant changes one unit on the log scale (i.e., an approximately 2.7-fold change). Overall admissions due to cerebrovascular diseases did not show a significant association with pollution; however, admissions via the emergency room did show a significant association. On any given day, the relative risks associated with nitric oxide and nitrogen dioxide were usually opposite in the model including both of these compounds. When nitric oxide was excluded from the model, the association of nitrogen dioxide with admissions due to cerebrovascular diseases was still significant with a 6-day lag (RR = 1.18, 95% CI = 1.04, 1.33; P =.008; Table 4). Significant associations were observed among all patients with long-term and short-term transient ischemic myocardial attacks (ICD-9 codes 411 and 413). The risk for all admissions due to longterm transient ischemic attacks was high 3 days after high concentrations of total suspended particulates (RR = 1.20, 95% CI 1.05, 1.38; P.008). When nitric = = oxide was excluded from the model, the risk for emergency admissions due to short- term ischemic myocardial attacks increased on the sixth day after a high concentration of nitrogen dioxide (RR = 1.26, 95% CI = 1.01, 1.57; P =.034) and decreased again on the seventh day (RR = 0.77, 95% CI = 0.64, 0.93; P =.006). Admissions due to myocardial infarction were not related to the pollutants. When the patients were analyzed separately in two age groups (those less than 65 years of age and those 65 years of age or older), the overall and emergency admissions due to all ischemic cardiac diseases were significantly associated with pollutants only among the younger patients. For the subgroups, in contrast, the association was significant only among older persons suffering short-term transient ischemic myocardial attacks. No age dependence was observed among patients with long-term transient ischemic myocardial attacks or cerebrovascular diseases. When the pollutants were included in the model one by one, the associations described earlier remained significant, although the relative risks were slightly lower than in the corresponding models including all pollutants simultaneously. The only exceptions were overall and emergency admissions due to ischemic cardiac diseases, which were no longer significantly associated with ozone after a 1-day lag. Temperature, relative humidity, and the concentration of sulfur dioxide had no significant effects on the daily number of admissions. However, the numbers of emergency admissions due to all ischemic cardiac diseases were higher (RR 1.09, = 95% CI 1.02, 1.18; P.013) during the = = spring and those due to cerebrovascular diseases were lower (RR = 0.88, 95% CI 0.79, 0.97; P.013) during the = = summer than during other seasons, irrespective of temperature. Admissions due to appendicitis (n 2280) bore = no relationship to the concentrations of pollutants or to the season. Discussion Several studies have revealed associations between ambient air pollutants and daily or weekly overall mortality and cause-specific mortality due to respiratory, cardiac, and cerebrovascular diseases1-39-11,2b-2; however, some studies September 1996, Vol. 86, No. 9 American Journal of Public Health 1277

Ponkfi and Virtanen have not found these associations at a significant level.629 During the London smog episode in December 1952, when the concentrations of sulfur dioxide and black smoke reached values of 2000,ug/m3, the increase in total mortality was 2.6-fold in the second week of the episode.1'3'20 Increases were seen in all age groups, but especially among the elderly. The largest relative increases were seen in deaths due to bronchitis and emphysema (9.5-fold), tuberculosis (5.5-fold), pneumonia and influenza (4.1-fold), and myocardial degeneration (2.8-fold). According to coroners' records from parts of the London area, there was a substantial increase in reports of respiratory factors as contributors to the underlying nonrespiratory causes of death.1'3 Thus, it is obvious that the pollutants increased the mortality rate by exacerbating preexisting chronic cardiovascular and respiratory diseases or precipitating acute conditions, such as heart attacks or pneumonia.3 Schwartz3'7 studied the associations between daily mortality rates and total suspended particulate concentrations in Philadelphia in 1973 through 1980 by comparing the number of deaths on the 5% of days with the highest particulate air pollution and on the 5% of days with the lowest pollution. The mean particulate concentrations were 141 and 47,ug/m3, respectively. The overall relative risk of dying on the days with high pollution was 1.08 (P <.0001). The relative risk of dying from chronic lung disease (ICD-9 codes 490 through 496) was 1.25 (P =.11); from pneumonia (ICD-9 codes 480 through 487), 1.13 (P =.18); from cardiovascular diseases (ICD-9 codes 390 through 429), 1.09 (P <.0001); from cerebrovascular diseases (ICD-9 code 436), 1.15 (P =.046); and from lung cancer, 1.19 (P =.002). Qualitatively, these results parallel those observed in London. Other studies in the United States have also found associations between deaths due to cardiovascular causes and particulates810 or primary motor-vehicle-related pollutants together (e.g., carbon monoxide, nitrogen dioxide, and particulates).ii Although many studies have investigated the association between pollutants and daily mortality, few have explored the effect of pollutants on the daily number of hospital admissions due to cardiovascular and cerebrovascular diseases.23-032 Recently, Schwartz and Morris3O reported the results of the first longitudinal study concerning the association between air pollution and ischemic cardiac disease (ICD-9 codes 410 through 414) with 0- to 2-day lags in Detroit. The concentration of respirable particulates (i.e., particulates with an aerodynamic diameter less than 10,um) was a significant predictor of admissions, with a relative risk of 1.018 for a 32-,ug/m3 increase in respirable particulates. Sulfur dioxide and carbon monoxide, but not ozone, were also associated with admissions, but only in a singlepollutant model. The relative risks were 1.014 for a 51-,ug/m3 increase in sulfur dioxide and 0.018 for a 1.5-mg/m3 increase in carbon monoxide. Mean values were as follows: respirable particulates, 48,g/m3; ozone, 82,ug/m3; sulfur dioxide, 72.6,ug/m3; and carbon monoxide, 2.7 mg/m3. Measurements were not made for nitrogen dioxide. Levels of pollutants were higher than in the present study in Helsinki, but the results are in accord, although the effects of lags of more than 2 days were not investigated in Detroit. According to an earlier study in Allegheny County, Pennsylvania, the daily numbers of urgent and emergency hospital admissions due to heart diseases (ICD-8 codes 393 through 398, 410 through 427) and other circulatory diseases (ICD-8 codes 390 through 392, 400 through 404, 430 through 458) increased significantly when concentrations of sulfur dioxide and particulates were high.23 A recent time series study from Ontario revealed an association between ambient air sulfate and cardiac admissions (ICD-9 codes 410,413, 427, 428),31 and another study conducted in seven large US cities showed an association between carbon monoxide and admissions due to congestive heart failure.32 Effects of air pollution episodes on hospital admissions due to cardiac causes have also been reported in the Ruhr area of Germany and in London.212 Wichmann et al.2 studied the effects of a smog episode in the Ruhr area in 1985, when the maximal average 24-hour levels were 440,ug/m3 for total suspended particulates, 620 P,g/m3 for sulfur dioxide, and 170,ug/m3 for nitrogen dioxide. During the 1-week episode, there were significant increases in hospital admissions for cerebral circulatory failure, heart arrhythmia, coronary insufficiency, and chronic bronchitis. In London during a 1-week smog episode in 1991, the maximum 1-hour nitrogen dioxide level was 795,ug/m3, the maximum 1-hour sulfur dioxide level was 349 Rg/m3, and the maximum daily average of black smoke was 148,ug/m3. During the episode, the mortality from cardiovascular diseases (RR = 1.14) and from all causes (RR = 1.10) increased significantly, and the number of hospital admissions for ischemic heart disease increased nonsignificantly (RR = 1.04). Increases in mortality and morbidity due to pollution have been reported to be associated with increases in airborne particulates and, less often, with sulfur dioxide.28 Particulates are derived mainly from the combustion of fossil fuels, but their detailed composition may not be critical for the association with increased mortality during air pollution episodes. Coal combustion for household energy production, car traffic, and a steel mill have been reported as sources of airborne particulates increasing the risk of death.1'`11 In Helsinki, however, most of the particulates (up to 80%) are derived from the street surfaces due to the effect of studded tires on the asphalt and from sand spread on the streets to prevent slipperiness. Thus, the particulates in Helsinki differ in composition from those found in cities with warmer climates and, as a result, probably comprise more chemically stable and less toxic compounds. The importance of nitrogen dioxide as a cause of increased mortality is insufficiently understood. Most previous studies have included comprehensive measurements of sulfur dioxide and particulates but little or no information about nitrogen dioxide and ozone. Some observations suggest that the associations between ozone and nitrogen dioxide concentrations and daily mortality counts are not significant. 1'25 In a 1-year follow-up in St. Louis and eastern Tennessee, however, it was possible to study the associations of nitrogen dioxide and ozone with daily mortality when the mean and highest nitrogen dioxide concentrations were 24 to 39,ug/m3 and 96 pg/m3, respectively, and the mean and highest ozone concentrations were 23 to 45,ug/m3 and 128,ug/m3, respectively. Neither nitrogen dioxide nor ozone was significantly associated with total mortality.25 Our findings suggest that nitrogen dioxide is associated with exacerbation of the symptoms of ischemic cardiac and cerebrovascular diseases. The same is true for nitric oxide, which as such is not harmful to health; this finding is possibly explained by the various chemical reactions in the ambient air. The nitrous emissions due to combustion consist mostly of nitric oxide, which oxidizes to nitrogen dioxide in the air. The concentrations of pollutants measured in Helsinki in 1987 through 1989 were lower than in the studies cited 1278 American Journal of Public Health September 1996, Vol. 86, No. 9

earlier, although the levels of particulates in the most polluted parts of Helsinki were close to the average values reported in recent studies in the United States.3,8,9,2426,30 The mean daily nitrogen dioxide and sulfur dioxide values we observed in Helsinki were lower than the guideline values issued by the World Health Organization (i.e., 150 and 125,ug/m3, respectively).33 The guideline for nitrogen dioxide was exceeded during only 1 day, the mean daily concentrations of nitrogen dioxide and sulfur dioxide being 39 and 19 plg/m3, respectively. The mean 8-hour levels of ozone were also always below 100 to 120,ug/m3, the guideline value provided by the World Health Organization. However, significant associations were seen between daily nitrogen dioxide and ozone concentrations and the daily counts of emergency room admissions. In contrast, total suspended particulate concentrations often exceeded the World Health Organization guidelines. The results suggest that, among patients with ischemic cardiac or cerebrovascular diseases, ambient air pollutants may increase the symptoms at lower concentrations than previously believed or than listed in the guidelines given in many countries and by the World Health Organization. Usually, the effects of pollutants on health are rapid and may be observed during the same day or after a lag of I to 3 days. Spix et al.27 have summarized the lag in the effect of pollutants on total mortality, stating that insofar as nobody has seen lagged effects of more than 2 days, a 1- to 2-day lag has been observed by several investigators. However, several studies have not looked for lags of longer than 2 days. In a similar study, we observed earlier that sulfur dioxide and nitrogen dioxide, but not total suspended particulates or ozone, were associated with hospital admissions due to chronic bronchitis and emphysema in Helsinki.34 When sulfur dioxide was high, a significant peak of admissions was observed during the same day, but the strongest effect of nitrogen dioxide was delayed; after a small initial peak with a 1- to 3-day lag, it reached a second, statistically significant peak after 6 days. This may reflect a dual mechanism of the effect of pollutants. The rapid effect may be due to direct irritation, and the later effect may be due to secondary infections. Similarly, in the present study, the number of admissions increased first with mainly a 1- to 2-day lag and increased a second time with a 6-day lag. Our results suggest that even lowlevel air pollution may significantly exacerbate symptoms of ischemic cardiac and cerebrovascular diseases, as measured by frequency of hospitalization. This observation is in agreement with a number of previous studies that have shown an association between mortality from these diseases and ambient air pollutionl-3'7-9,11,23 and with recent studies suggesting that hospital admissions due to heart diseases are also associated with air pollution.'2'3(032 Our findings further suggest that these exacerbations may be associated with nitrogen dioxide, which is derived mainly from exhaust gases of car traffic, and not merely with particulates alone or with sulfur dioxide. Although the results of the present study are in good accord with those of other investigations of associations between pollutants and mortality due to cardiovascular diseases,"3,7'35 it is still possible that the relations observed were due to chance. There are also possible sources of bias, such as the slight difference in the basic population from day to day due to the absence of those patients who were hospitalized due to ischemic cardiac or cerebrovascular disease. Confirmatory findings are needed, because there always may be biases in hospital data for population-based studies of air pollution. The inconsistencies now observed between the effects of various pollutants and hospital admissions due to ischemic cardiac and cerebrovascular diseases also emphasize the need for further studies. However, it is possible that any pollutant principally targeting the respiratory tract-whether nitrogen dioxide, sulfur dioxide, total suspended particulates, or ozone-causes cardiovascular stress and increases the likelihood of clinical symptoms. O References 1. Mortality and Morbidity during the London Fog of December 1952. London, England: Her Majesty's Public Health Service; 1954. 2. Wichmann HE, Mueller W, Allhoff P, et al. Health effects during a smog episode in West Germany in 1985. Environ Health Perspect. 1989;79:89-99. 3. Schwartz J. What are people dying of on high air pollution days? Environ Res. 1994;64:26-35. 4. Schwartz J, Marcus A. Mortality and air pollution in London: a time series analysis. Am JEpidemiol. 1990;131:185-194. 5. Steiger H, Brockhaus A. Studies on the connection between air pollution and mortality in the Ruhr area [in German]. Die Naturwissenschaften. 1966;53:498. 6. Katsouyanni K, Karakatsani A, Messari I, et al. Air pollution and cause specific Pollution and Cardiac Diseases mortality in Athens. J Epidemiol Community Health. 1990;44:321-324. 7. Schwartz J, Dockery DW. Increased mortality in Philadelphia associated with daily air pollution concentrations. Am Rev Respir Dis. 1992;145:600-604. 8. Schwartz J. Air pollution and daily mortality in Birmingham, Alabama. Am J Epidemio. 1993;137:1136-1147. 9. Pope CA, Schwartz J, Ransom MR. Daily mortality and PM1o pollution in Utah Valley. Arch Environ Health. 1992;47:21 1-217. 10. Fairley D. The relationship of daily mortality to suspended particulates in Santa Clara County, 1980-1986. Environ Health Perspect. 1990;89:159-168. 11. Kinney PL, Ozkaynak H. Associations of daily mortality and air pollution in Los Angeles County. Environ Res. 1991;54:99-120. 12. Andersson HR, Limb ES, Bland JM, Ponce de Leon A, Strachan DP, Bower JS. Health effects of an air pollution episode in London, December 1991. Thorax. 1995;50: 1188-1193. 13. McCullagh P, Nelder JA. Generalized Linear Models. 2nd ed. London, England: Chapman & Hall; 1989. 14. Breslow NE, Day NE. Statistical Methods in CancerResearch II. The Design andanalysis of Cohort Studies. Lyon, France: International Agency for Research on Cancer; 1982. 15. Little RJA, Rubin DB. Statistical Analysis with Missing Data. New York, NY: John Wiley & Sons Inc; 1987. 16. Rose G. Cold weather and ischaemic heart disease. BrJPrev Soc Med. 1966;20:97-100. 17. Rogot E, Padgett SJ. Associations of coronary and stroke mortality with temperature and snowfall in selected areas of the United States. Am J Epidemiol. 1976;103: 565-575. 18. Kunst AE, Looman CWN, Mackenbach JP. Outdoor air temperature and mortality in the Netherlands: a time-series analysis. Am JEpidemiol. 1993;137:331-341. 19. Schwartz J, Wypij D, Dockery D, et al. Daily diaries of respiratory symptoms and air pollution: methodological issues and results. Environ Health Perspect. 1991;90: 181-187. 20. Logan WPD. Mortality in the London fog incident, 1952. Lancet. 1953;264:336-338. 21. Buechley RW, Riggan WB, Hasselblad V, VanBruggen JB. SO2 levels and perturbations in mortality. Arch Environ Health. 1973;27:134-137. 22. Mazumdar S, Schimmel H, Higgins ITT. Relation of daily mortality to air pollution: an analysis of 14 London winters, 1958/59-1971/72.Arch Environ Health. 1982;37:213-220. 23. Mazumdar S, Sussman N. Relationships of air pollution to health: results from the Pittsburgh study. Arch Environ Health. 1983;38:17-24. 24. Schwartz J, Dockery DW. Particulate air pollution and daily mortality in Steubenville, Ohio. Am J Epidemiol. 1992;135:12-19. 25. Dockery DW, Schwartz J, Spengler JD. Air pollution and daily mortality: associations with particulates and acid aerosols. Environ Res. 1992;59:362-373. September 1996, Vol. 86, No. 9 American Journal of Public Health 1279

Ponkil and Virtanen 26. Schwartz J. Particulate air pollution and daily mortality in Detroit. Environ Res. 1991;56:204-213. 27. Spix C, Heinrich J, Dockery D, et al. Air pollution and daily mortality in Erfurt, East Germany, 1980-1989. Environ Health Perspect. 1993;101:518-526. 28. Derriennic F, Richardson S, Mollie A, Lellouch J. Short-term effects of sulphur dioxide pollution on mortality in two French cities.intjepidemiol. 1989;18:186-197. 29. Steiger H. An investigation on the mortality during a smog situation in western Ruhrgebiet on the 17th January 1979. Zentralblatt fur Baktenologie, Mikrobiologie und Hygiene. I Abteilung Originale B; Umweltshygiene, Krankenhaushygiene, Arbeitshygiene, Praventive Medizin. 1980;171:445-447. 30. Schwartz J, Morris RD. Air pollution and hospital admissions for cardiovascular disease in Detroit, Michigan.Am JEpidemiol. 1995;142:23-35. 31. Burnett RT, Dales R, Krewski D, Vincent R, Dann T, Brook JR. Associations between ambient particulate sulfate and admissions to Ontario hospitals for cardiac and respiratory diseases. Am J Epidemiol. 1995;142:15-22. 32. Morris RD, Naumova EN, Munasinghe RL. Ambient air pollution and hospitalization for congestive heart failure among elderly people in seven large US cities. Am JPublic Health. 1995;85:1361-1365. 33. Air Quality GuidelinesforEurope. Copenhagen, Denmark: World Health Organization; 1984. 34. Ponka A, Virtanen M. Chronic bronchitis, emphysema, and low-level air pollution in Helsinki, 1987-1989. Environ Res. 1994;65: 207-217. 35. Xu X, Dockery DW, Gao J, Chen Y. Air pollution and daily mortality in residential areas of Beijing, China. Arch Environ Health. 1994;49:216-222. 1280 American Journal of Public Health September 1996, Vol. 86, No. 9