OUP user menu

The association of daily sulfur dioxide air pollution levels with hospital admissions for cardiovascular diseases in Europe (The Aphea-II study)

Jordi Sunyer , Ferran Ballester , Alain Le Tertre , Richard Atkinson , Jon G Ayres , Francesco Forastiere , Bertil Forsberg , Judith M Vonk , Luigi Bisanti , José M Tenías , Sylvia Medina , Joel Schwartz , Klea Katsouyanni
DOI: http://dx.doi.org/10.1016/S0195-668X(02)00808-4 752-760 First published online: 2 April 2003


The objective of this study is to assess the short-term effect of sulfur dioxide (SO2) air pollution levels on hospital admissions for cardiovascular diseases. Daily mean hospital admissions for cardiovascular diseases, ischemic heart diseases (IHDs), and stroke in seven European areas (the cities of Birmingham, London, Milan, Paris, Rome, and Stockholm, and in The Netherlands) participating in the multicenter European study of air pollution (Aphea-II), were measured. Time series analysis of daily hospital admission counts was performed using poison autoregressive models. A summary regression coefficient for all cities was provided. Daily numbers of all cardiovascular admissions except stroke, and particularly IHDs, rose significantly with an increase of daily SO2levels of the same day and day before. After adjusting for PM10(i.e. particles with size <10μm), the association of SO2with IHD admissions remained significant (i.e. an increase of 0.7%; 95% confidence interval=0.1–1.3, per each 10μg/m3increase of SO2) among subjects younger than 65 years, but not among subjects older than 65. In the older group the increase was only significant for particles (1.3%; CI 0.7–1.8, per each increase in 10μg/m3of PM10). This study provides new evidence for the effects of urban air pollution on cardiac diseases in Europe, and suggests that SO2pollution may play an independent role in triggering ischemic cardiac events. From a Public Health perspective these results suggest that reduction in SO2levels in European cities could imply a reduction of admissions for IHDs.

  • Sulfur dioxide
  • Air pollution
  • Cardiovascular diseases

1 Introduction

In the fog episode in London in 1952 deaths due to heart diseases more than doubled (although less than deaths due to bronchitis).1 Hospital admissions for heart disease were also increased. The London fog was characterized by very high concentrations of black smoke and sulfur dioxide (SO2) in the atmosphere, largely released by domestic coal combustion. In other episodes of air pollution involving the same pollutants, as in the Meuse Valley in 19302 and in Donora, Pennsylvania in 1948,3 many of the people admitted to hospital suffered from chronic cardiac disease. In the last decade, a number of epidemiological studies have reported an acute association between daily counts of hospital deaths and hospital admissions for cardiovascular diseases and levels of air pollution of the preceding days at levels close to or even lower than current national standards, specially in North America,3–13 in Europe,14–20 and in other parts of the world, such as Sydney, Australia,21 and Hong Kong, China.22

In the first multicities European Air Pollution Health Effects Approach (APHEA) study, that covered the period 1985–1990, an association between cardiovascular mortality and both particles (measured as black smoke) and SO2was observed.23 In the second APHEA study, which includes more European cities and covers the period 1990–1996, the association between air pollution and hospital admissions due to cardiovascular causes was measured. Hospital admission is a more sensitive marker than mortality of the environmental effects. An association between particles and admission for ischemic heart diseases (IHDs) in people older than 65 years was reported.24 The objective of this paper is to study the short-term effect of SO2, after adjusting for particles and other air pollutants, on hospital admissions for cardiovascular diseases (i.e. what fraction of daily counts of hospital admissions was associated with SO2pollution levels of the preceding days).

2 Data and methods

Table 1shows daily mean numbers of hospital admissions for cardiovascular diseases (International Classification of Diseases (ICD), 9th revision, 390– 429), IHD (ICD-9, 410–413), and stroke (ICD-9, 430–438) by age groups in seven European areas (all the European cities participating in the hospital admissions project within APHEA-II, with the exception of Barcelona, which was excluded because the detection limit for SO2was set at an arbitrary cut-off of 10μg/m3, a value close to the median). Daily admission data was provided by routine registers. Emergency admissions were used in every city except in Milan, Paris, and Rome where only general admissions data was available. The Netherlands was not able to provide information for all cardiac causes.

View this table:
Table 1

Daily average of hospital admissions for cardiovascular diseases and air pollutants (in μg/m3) per city

CityCardiovascular (390–429)Ischemic heart disease(410–413)Stroke (430–438)Air pollutants
All ages>65 years<65 years>65 years>65 yearsSO2medianSO290%Black smoke medianBlack smoke 90%PM10medianPM1090%
  • a PM10(particles of size <10μg)=TSP (total suspended particles)×0.75 (originally measured on TSP).

  • b PM13.

The pollutants were those generally measured in European cities and have been described in detail elsewhere.23,24 Pollution levels were assessed at fixed monitoring sites. Particles were measured mostly as PM10(i.e. particles with size <10μm), except in Paris (PM13), and in Milan and Rome (total suspended particulate, TSP). Four areas(Birmingham, London, The Netherlands, and Paris) were also able to provide black smoke data (BS) as a measure of particles. Sulfur dioxide (24-h average), nitrogen dioxide (hourly values and 24-h average), and ozone (hourly value and maximum 8-h average) measurements were available in all cities. Carbon monoxide (8-h average) was also available in all cities except Paris. Stations were chosen to represent background inner-city air quality levels, except for ozone. Thus, stations located in highways were excluded.

Time series analysis was used, the unit of observation being the day and not the individual. The analysis assessed what fraction of the daily variations in hospital admission counts was explained by the daily variations in air pollution of the preceding days, after controlling for other variables that varied in time (such as trend, seasonal patterns, or meteorological factors). Usual risk factors of cardiovascular diseases (such as smoking, diet, or exercise) are not cofactors since they did not vary in the short-term time window analyzed in relation to air pollution daily variations. A Poisson autoregressive model of the hospital admissions time series was constructed which included terms to describe the seasonal patterns in the admissions, theirdependence on temperature and humidity, their association with holiday periods and influenza episodes, and finally air pollution measures.23 These models provided estimates of the effect of air pollution on the mean number of admissions per day. Generalized Additive Models were used to study non-linear relationships between confounders and morbidity,24 thus temperature and humidity were included as a smooth function on the same day of admission or lagged up to 3 preceding days. In every city, the air pollution indicator (i.e. SO2) was introduced as a linear term representing the average of the same and the preceding day's levels. Finally, two pollutant-models were fitted to assess the changes in the SO2estimates after the inclusion of the other pollutants.

In a second step, the heterogeneity of the associations between air pollution levels and daily admission levels among cities was assessed using the chi-square test for heterogeneity. A significant result indicated that the variation in the effect estimates was greater than expected by chance. However, a summary estimate for all the cities that accounted for this additional variation was calculated (random-effects estimate). Where there was significant heterogeneity between estimates further analyses were carried out to investigate possible reasons for this. In this second stage of the analyses, regression models were used to investigate associations between the seven SO2effect estimates (one for each city) and variables describing the health and environmental conditions in each city. In this way factors which may affect the toxicity of the pollutants or the vulnerability of the exposed population (effect modifiers) were explored (such as prevalence of smokers or social class average).

3 Results

Daily counts of cardiovascular admissions, and particularly IHD over 65 years, increased in a statistically significant way in association with an increase of daily SO2levels of the same and the preceding day. Hence, an increase in 10μg/m3of daily average of SO2was associated with an increase of 0.7%; 95% confidence interval=0.1–1.3, of all cardiovascular admissions of the same and the next day. No such pattern was seen for stroke (Table 2). The association with IHD over 65 was homogenous (i.e. positive in all cities except Stockholm, and statistically significant in four of the seven cities) (Fig. 1). The association for all cardiovascular admissions was less homogeneous between cities (it was positive for all cities except for Birmingham, but the magnitude of the positive effect varied widely). Most of this heterogeneity of the association with cardiovascular admissions was explained by the average level of humidity of each city. The association of SO2with cardiovascular admissions (both at all ages and over 65) was stronger at lower levels of humidity (percent increase of all cardiovascular admissions at all ages was 1.67, CI=0.65–2.71, per 10μg/m3of SO2at the lowest levels of humidity and 0.81, CI=0.43–1.19 at the highest).

View this table:
Table 2

Percentage increase in the daily number of hospital admissions per each increase of 10μg/m3in SO2levels of the same and preceding day

Outcome (ICD-9)SO2
%95% CI
Cardiovascular (390–429)
Cardiovascular over 65 years0.70.31.2
IHD (410–413) below 65 years0.60.21.1
IHD over 65 years1.20.81.6
Stroke (430–438) over 65 years0.0−0.50.5
Fig. 1

Estimated percentage increase in admissions for IHD among subjects aged over 65 years in each city as well as pooled fixed and random-effect estimates associated with a 10μg/m3increase in the average of SO2for 0 and 1 day average lag. Average values of SO2daily mean in μg/m3were 24.3, 23.6, 21.1, 8.5, 17.7, 9.8, and 3.8, respectively, for the cities (in alphabetic order).

None of the other variables examined (such as climatic, socio-economic variables, and others such as prevalence of smoking) helped to explain the heterogeneity between cities.

To analyze if the association with SO2was explained by the levels of other pollutants usually present in urban air such as particles, a second analysis was done fitting two-pollutant models (i.e. adjusting the effect of SO2for a second pollutant). The association with all cardiovascular admissions (both at all ages and above 65 years) weakened and became non-significant after adjusting for CO, NO2, black smoke, and PM10. In contrast, the association of IHD <65 years showed little change after controlling for the other pollutants. For IHD over 65 years of age, the association weakened and became non-significant after adjusting for NO2, black smoke, and PM10(Table 3). Among subjects older than 65 years, the association with IHD was stronger for particles than for SO2, while among subjects younger than 65, the association appeared stronger for SO2(Table 3). A sub-analysis with cities using strictly the same data on PM10(almost the same cities as for the analysis of black smoke) provided the same results.

View this table:
Table 3

Percentage increases (and 95% CI) in hospital admissions for IHDs (ICD-9, 410-413) per 10μg/m3increase in the average of SO2, after adjusting for particles (either PM10or black smoke)

AgeIncluding seven citiesIncluding four cities
<65 years
SO20.6 (0.1–1.1)*0.7 (0.1–1.3)*0.7 (0.1–1.3)*0.9 (0.1–1.7)*
PM100.2 (−0.2–0.5)0.0 (−0.4–0.3)
Black smoke0.1 (−0.3–0.6)−0.3 (0.9–0.3)
>65 years
SO21.2 (0.8–1.6)*−1.4 (−8.0–6.0)1.3 (0.7–1.8)*0.6 (−0.1–1.3)
PM100.7 (0.3–1.1)*1.3 (−1.8–4.5)
Black smoke1.1 (0.7–1.5)*0.8 (0.3–1.4)*
  • a A single pollutant was fitted in each model.

  • *p<0.05.

4 Discussion

This combined analysis summarizes the results from seven European cities concerning the short-term effects of SO2air pollution on hospital admissions for cardiovascular causes. It represents the first multicenter epidemiological study examining the impact of SO2on cardiovascular morbidity in Europe, where the association between air pollution involving SO2and mortality for cardiovascular causes has been consistently shown.23,25 We found a significant effect of SO2on admissions for cardiac causes and IHD for all ages. The percentage increase associated with a 10μg/m3increase in SO2was similar to the effect found for particles (between 0.5 and 1%) in the APHEA study.24 The association was larger for IHD over 65 years old. Given that everyone is exposed to urban air pollutants, for a city such as London with more than 100 admissions per day an increase in 10μg/m3in SO2(its standard deviation was 23μg/m3) would result in an impact of at least one additionaladmission per day due to this pollutant, which if translated to all the European cities would result in a considerable public health impact.

The acute effect of air pollution on cardiovascular disease has already been observed, mainly outside Europe. Table 4 summarizes the results from studies examining the relationship between hospital admissions for cardiovascular diseases and both, SO2and particles, using a systematic review.26 There were 34 reports including particulates, and all except one (97%) found a positive and significant association between particle levels and cardiovascular admissions. For studies including SO2, 17 of the 24 (71%) found positive and significant associations with at least one of the indicators of cardiovascular admissions. Among the 24 reports including SO2, only 12 analyzed both SO2and particles and these are included in Table 4.We have also included a more recent article30 published after completion of the review. An association between cardiovascular hospital admissions and SO2was found in nine out of the 13 studies (Table 4).

View this table:
Table 4

Results for SO2and particulates from time series studies analyzing the association between cardiovascular hospital pollution and air pollution

Author/s (year of publication)Site/period of studyOutcomeAge groupEstimates of the associationa
Schwartz and Morris (1995)4Detroit, USA/1986–1989Ischemic heart disease (ICD-9, 410–414)>65 years0.30.6
Schwartz (1997)8Tucson, USA/1988–1990Cardiovascular disease (ICD-9, 390–429)>65 yearsNS1.2
Burnett et al. (1997)9Toronto, Canada/1992–1994 (summer)Cardiac disease (ICD-9, 410–414,427,428)>65 yearsSO2-lh: 2.31.4
Poloniecki et al. (1997)16London, UK/1987–1994All circulatory diseases (ICD-9, 390–459)All ages0.5BS: 1.6
Acute myocardial infarction (ICD-9, 410)0.6BS: 2.0
Prescott et al. (1998)18Edinburgh, UK/1992–1995Cardiovascular diseases (ICD-9, 410–414,427,428,434–440)>65 yearsNS4.8
Wong et al. (1999)22Hong Kong, China/1994–1995Cardiovascular diseases (ICD-9, 410–417,420–438,440–444)All ages1.60.6
>65 years2.10.8
Ischemic heart disease (ICD-9, 410–414)All agesNSNS
Atkinson et al. (1999)27London, UK/1992–1994All circulatory diseases (ICD-9, 390–459)All ages0.90.8
0–64 yearsNS1.3
Ischemic heart disease (ICD-9, 410–414)>65 years1.81.0
Burnett et al. (1999)7Toronto, Canada/1980–1994Ischemic heart disease (ICD-9, 410–414)All ages0.6NS
Moolgavkar (2000)28Three counties: Los Angeles (LA), Cook, Maricopa, USA/1987–1995Cardiovascular disease (ICD-9, 390–429)>65 yearsLA, 4.8; Cook, 1.4; Maricopa, 2.5LA,0.6; Cook, 0.8; Maricopa, NS
Lipmann et al. (2000)13Detroit, USAIschemic heart disease (ICD-9, 410–414)>65 yearsNS1.2
Ballester et al. (2001)20Valencia, Spain/1994–1996All circulatory diseases (ICD-9, 390–459)All ages3.0BS: 1.5 (only significant in summer)
Heart diseases (ICD-9, 410–414,427,428)3.6BS: 1.5 (only significant in summer)
Biggeri et al. (2001)29Six Italian cities (Turin, Milano, Bologna, Firenze, Rome, and Palermo)/1995–1999Cardiovascular disease (ICD-9, 390–429)All ages2.80.8
Anderson et al. (2001)30West Midlands (Birminghan), UK/1994–1996Cardiovascular disease (ICD-9, 390–429)All agesNSNS
Ischemic heart disease (ICD-9, 410–414)>65 yearsNSNS
Sunyer et al. (this paper)APHEA2: Birminghan, London, Milan, Netherlands, Paris, Rome, and StockholmCardiovascular disease (ICD-9, 390–429)All ages0.70.4
<65 years0.6NS
Ischemic heart disease (ICD-9, 410–413)>65 years1.2NS
  • a Results expressed as the increase (%) in hospital admissions for an increase in the pollutant levels of 10μg/m3.

  • b If PM10was not available we chose BS estimates.

  • NS=non-significant results. Results for pollutants other than SO2and PM10/BS are not shown here. >65 years=persons aged 65 years and older; 0–64 years=persons aged under 65 years.

The present study refers to the assessment of air pollution as a trigger of cardiovascular disease admissions, and mainly of IHD. The underlying hypothesis was that frail individuals were at higher risk of having an attack on days with higher air pollution levels. This study, then, did not refer to the classical risk factors that mainly ascertain cumulative or chronic relationships. While epidemiological findings, such as the present study, are very consistent, the issue of the biological plausibility is weak since the underlying mechanisms of an acute heart effect of air pollution are unknown. Several physiopathological pathways have been proposed for the relationship between particulate air pollution and cardiovascular health.31–33 One of the major hypotheses is that particles induce activation of some mediators/indicators of alterations in blood coagulability.33 Another kind of mechanism examined includes those indicative of cardiac autonomic control, such as heart rate and various indices of its variability.34 The lack of an association with stroke is an intriguing finding since effects of air pollution on hemostatic profile would be expected to increase the number of stroke events.

A growing number of studies35–37 support the hypothesis that composition of ultrafine particles38 (those with size less than 100nm in diameter) and transition metals39 could explain, in a substantial part, these harmful effects of air pollution on the cardiovascular system. However, an additive role of SO2, or other gaseous pollutants such as CO, could not be discarded.7,40 Particles and SO2may act via a different mechanism rather than interacting with each other through the same mechanistic route. In a recent study, a change in heart rate variability in humans on exposure to SO2(200ppb for 1h) was attributed to stimulation of receptors in the upper respiratory tract.31 Such a mechanism of SO2may be operating in patients with IHD. Among the studies on cardiovascular admissions (Table 4), after adjusting for particles, the association of SO2remained significant only in three studies,7,20,28 while the association of particles is less affected by the adjustment for SO2. However, in a recent study in Hong Kong, abatement of SO2levels had a notable impact in reducing adverse health effects, although particle levels remained stable.41 The fact that PM10, black smoke, CO, and NO2reduce the association of SO2with cardiovascular admissions in the present study may be because these pollutants probably come from the same source (i.e. diesel exhaust or fossil fuel combustion) or may increase together due to the meteorological conditions. To disentangle the specific effect of each of the pollutants in the urban atmospheres is not possible in the present study. Only the association of SO2with IHD in individuals aged under 65 was not confounded by particles or CO, although it was slightly reduced after adjusting for NO2levels. This suggests that SO2may play an independent role in triggering ischemic cardiac events. A different question is why SO2is associated with IHD in the younger but not in the older age group, after adjusting for particles. One explanation could be that the daily activity pattern of the young group could result in a lower measurement error, using ambient measurements as surrogates for personal measurements, than in the older group.

A problem of the present study refers to lack of checking of diagnosis labels. This problem could affect present results if variation in coding occurs within the same city on a daily basis. However, given that this error, if it occurs, is unlikely to be related with time variations in air pollution levels, the consequent bias would reduce the magnitude of the associations estimated. One problem encountered with data comparability on health outcome was that Paris and Milan were not able to provide emergency admissions only. However, a sub-analysis excluding these cities yielded very similar results, suggesting that the error produced by including both hospital admissions and emergency admissions was small. In a sub-analysis of the London data, emergency admissions accounted for the majority of hospital admissions for cardiovascular causes. On the other hand, the potential problem of using routinely collected data on air pollution has been extensively evaluated and, if anything, reduced the estimatedcoefficients.4,12,30

The present study did not incorporate the classical risk factors for cardiovascular diseases even though these factors may change by season. However, the statistical procedure carried out here adjusted for all the seasonal effects, and a residual confounding of these factors (such as seasonal variations on diet or exercise) is unlikely. A different question refers to the modification of the effect of air pollution by certain factors (such as age, sex, or smoking). Thus, the effect of air pollution could be more important in smokers than in non-smokers.37 Time series analysis is not suitable for assessing susceptibility factors, which is beyond the present study. However, we performed a hierarchical analysis trying to ascertain which city variables could explain differences in the coefficients of the association between air pollution and cardiac admissions. Neither age and sex structure nor smoking prevalence explain these variations, but this analysis was only conducted with seven areas, and is of limited value.

It should be noted that certain frequently encountered problems of meta-analysis do not apply in our study. There was no selection bias in the sense that the participating cities were not selected by the results of the short-term analysis, but by their ability to provide data. All the city specific analyses were centralized in one place, giving more homogeneity in the application of the methodology defined within the APHEA-2 statistical group. We also set rules for exposures and confounding factor measurements. Similarly, problems due to the usual epidemiological design such as the lack of inclusion of classical risk factors of cardiovascular diseases do not apply to the present study since they did not vary from one day to the next.

Overall, this study provides new evidence for the effects of urban air pollution on cardiac diseases in Europe, suggesting that air pollutants could trigger a MI in subjects with vulnerable arteries. Mechanistically, most of the evidence refers to the hemodynamic and hemostatic effects of air pollutants, but solid evidence for the role of these possible physio-pathological mechanisms is lacking. The present analysis shows that from a public health perspective, urban atmospheres in Europe are associated with admissions for IHD, and that SO2is a surrogate of these mixtures, in some cases (such as in IHD in younger than 65 years) in a stronger way than particles. Therefore, these results suggest that reduction in SO2levels in European cities could imply a reduction of admissions for IHDs.


The APHEA 2 study is supported by the European Commission (EC) Environment and Climate 1994–98 Programme (Contract number ENV4-CT97-0534). The Swedish group did not receive funding from the EC.


View Abstract