A vast number of papers have shown both the short term^1,2 and long term^3,4 adverse health effects of air pollutants. Public health policy makers have paid attention to the potential deleterious effects of air pollutants, and the focus has now shifted to disease specific issues, such as cardiovascular disease.^5,6 Daily mortality in cardiovascular disease increases when the daily average of air pollutants increases. This acute effect might arise from an increase in the mortality of myocardial infarction. For myocardial infarction, the temporal relation between acute onset and transient increase in the level of air pollutants was examined previously in a case-crossover approach.^7,8 Another approach is to focus on the occurrence of high level concentrations of pollutants and to examine whether there is an increase in mortality during the subsequent hours after such high levels. However, few studies have focused on the short term exposure to high levels of air pollutants and mortality in the subsequent hours.

In Japan, both air pollutant levels and time of death from death certificates are available on an hourly basis. This allowed us to analyse the temporal relation between air pollutants and adverse health events in an hourly scaled measure. Therefore, we examined whether an increase of myocardial infarction deaths occurred after a transient increase in air pollutant levels.

METHODS

Description of data

Tokyo has an area of 2187 km², and about 11.8 million people lived there in the year 1994. Thirty four districts (19 in the urban area and 15 in the suburban area) of the Tokyo Metropolitan area were selected for the study, where 60 ambient air pollution monitoring stations were evenly distributed by the Tokyo Metropolitan government.

The number of myocardial infarction deaths (International Classification of Disease 9th Revision: 410) for each hour was calculated from the Vital Statistics of Japan from 1 January 1990 to 31 December 1994.

In Japan, the criterion used for air pollution is the concentration of suspended particulate matter (SPM)—particles with diameters less than 10 μm, intermediate between and PM_2.5 and PM₁₀. The diameters of SPM are mainly PM₇. The SPM concentrations and temperature are measured every hour at ambient air pollution monitoring stations. These data were obtained from the National Institutes for Environmental Studies and linked with mortality data by district, year, month, day, and hour. When there was more than one ambient air pollution monitoring station in the same district, the average values at these stations were used for the analysis.

Method of examining the effect of SPM on mortality rates

Figure 1 is a hypothetical example that explains the methods we used. This figure shows the SPM concentrations, the critical vale of high SPM levels, exposure windows after SPM concentration reached high levels, deaths attributable to myocardial infarction, and the exposure window used for averaging the temperature values.

Five concentrations (100 μg/m³, 150 μg/m³, 200 μg/m³, 250 μg/m³, and 300 μg/m³) of critical values were set for identifying high concentrations of SPM. Those levels were determined from the distribution of SPM concentrations in Tokyo from 1990 to 1994.

After these levels were established, we selected several window lengths (from one hour to 48 hours) to evaluate the effects of high concentrations of SPM. Each of these windows, so called “exposure windows”, was further categorised into six SPM categories (0–99 μg/m³, 100–149 μg/m³, 150–199 μg/m³, 200–249 μg/m³, 250–299 μg/m³, 300 μg/m³ and over). When different SPM categories overlapped, the higher one was chosen for the analysis.

To adjust for the effect of temperature on myocardial infarction death, an average of hourly temperature before death was used for the analysis. To explore the influence of an exposure window in which hourly temperatures were averaged, several different periods were examined.

In each SPM category, the number of myocardial infarction deaths was counted throughout each period, and the incidence rate per 1000 hours was determined. The reference SPM category was set to 0–99 μg/m³. The rate ratio (incidence rate in any SPM category divided by the reference category) was calculated to compare the different SPM categories.

Statistical analysis

A Poisson regression model was used to adjust for the effects of confounding factors. The average temperature of the period before the myocardial infarction deaths, the hour of day that the deaths occurred, and the region (urban or suburban area) were included in the model as confounders. In modelling the effect of temperature, we used a piecewise linear regression with the disjoint coding method⁹ for treating cold and hot temperature effects separately. In the model, the point where the regression line turned over was set to 25°C, this was the point at which the lowest mortality was seen in the crude analysis. The hour of day that death occurred was included in the model as a dummy variable.

RESULTS

Table 1 shows the summary of SPM concentrations and temperatures during the study period. The median, 90th centile, and maximum concentration of SPM (μg/m³) were 44.0, 115.0, and 1091.3 in the urban areas, and 39.0, 97.0, and 563.0 in the suburban areas, respectively.

Table 2 shows characteristics of the study subjects. There were 14 430 deaths (male: 8,121, female: 6,309) from myocardial infarction during 1990–1994 in Tokyo. The average age at death was 72.1 for men and 78.7 for women. Most of the subjects lived in the urban areas (78.0%) and expired in medical institutions (82.0%).

Table 3 shows the distribution of total hours in each SPM category with six different exposure windows. When the exposure window was short (one to six hours), most of the periods fell in the lowest category (0–99 μg/m³) (one hour: 88.4%, six hours: 77.5%). As the duration of the exposure window increased, the distribution of hours shifted to the higher categories. The proportion of overlap with multiple SPM categories was small when the exposure window was short, and increased with longer exposure windows (24 hours: 18.0%, 48 hours: 27.5%).

Table 4 shows the rate ratios of myocardial infarction deaths adjusted for temperature, the hour of day when deaths occurred, and region. In every SPM category, the rate ratios were higher than 1.00. The adjusted rate ratios in each SPM category ranged from 1.07 to 1.13 in the 100–149 μg/m³ category, from 1.09 to 1.24 in the intermediate categories (150–199, 200–249, and 250–299 μg/m³), and from 1.14 to 1.40 in the highest category (300 μg/m³ and over). When the exposure window was one hour, the rate ratio was 1.40 (CI: 1.00 to 1.97) in the highest category. Although not significant, the rate ratios in the highest category still remained large when the exposure windows were short (two to six hours). As the duration of the exposure window increased, the adjusted rate ratios of the intermediate and highest categories showed similar values. For every exposure window duration, a gradual increase in the rate ratio was seen in the 100–149 and 150–199 μg/m³ categories. When exposure window was one to six hours, an increase in the rate ratio was mainly seen as the category increased, although a slight decrease was sometimes observed in the 200–249 μg/m³ category. There was little influence of temperature changes on the adjusted rate ratio.

DISCUSSION

This study showed an increase in the rate ratio of myocardial infarction deaths occurred once the SPM level exceeded 100 μg/m³. A remarkably high rate ratio was seen within one hour after the SPM level exceeded 300 μg/m³. A gradual increase in the rate ratio was found as the category of SPM concentration level increased only when the exposure window was below six hours.

With environmental quality standards in Japan, the daily average of SPM normally stays below 100 μg/m³. This shows that the transient effects we saw were in situations with highly concentrated air pollutants. We assumed that a concentration of air pollutants during a certain hour (point exposure) caused the events in the following hours (exposure window), not merely events occurring during the same hour (point events). When using hourly scaled measures, our approach is straightforward for examining transient effects on events after exposure to high levels of air pollutants. A prior study using a case-crossover approach showed that a transient increase of air pollutants caused myocardial infarction onset during the reference period of sampling. Our approach was to examine the effect of transient rise in air pollutants using the entire reference period.

The exact duration between an exposure to air pollutants and death from myocardial infarction is difficult to estimate. In experimental studies in animals and humans, potential biological mechanisms linking air pollution and cardiovascular diseases have been investigated.^10,11 These studies show an acute response within a few hours after exposure to air pollutants, such as reduced heart rate variability, raised ST-segment elevation, and vasoconstriction. In the study using a case-crossover approach,⁷ an increased rate ratio of myocardial infarction was associated with an increase in pollutants during the two hour period before the infarction onset (1.48 with an increase of 25 μg/m³ PM_2.5), and in the 24 hour period one day before the infarction onset (1.69 with an increase of 20 μg/m³ PM_2.5). Another study showed that an increase in myocardial infarction onset was seen within one hour after exposure to traffic.¹² To investigate the effect of exposure duration on myocardial infarction deaths, we studied several different exposure windows after high level exposure to SPM, from an extremely short interval (one hour) to a prolonged period (48 hours).

We selected the higher category when multiple SPM categories overlapped. From previous studies, it is obvious that the effect of air pollutants in a higher category is larger than in a lower category.^3,4 These previous results suggest that the effect of a higher SPM category is much stronger than that of a lower category during a period of overlap. In the same SPM category, the crude incidence declined slightly with an increase in the exposure window. This was presumably because the incidence was diluted when the incidence of a lower SPM category moved to a higher category. This dilution might have been less in a shorter period (one to six hours) in which the proportion of overlapping SPM categories was below 10%.

In the study with the case-crossover approach,⁷ a high rate ratio of myocardial infarction onset was seen two hours and one day after there was an increased concentration of particular matter. In our study, remarkably high rate ratios in the highest SPM category were noted within one hour after reaching a high SPM concentration. Although not significant, high rate ratios were also seen in the short exposure windows (two to six hours). The adjusted rate ratios had wide confidence intervals because there were extremely few hours within the highest SPM category when the exposure window was short. For example, total hours in the highest SPM category contained only about 7000 hours (0.5%) when the exposure window was six hours. The high rate ratio in the highest SPM category during several short exposure windows suggested that deaths from myocardial infarction are likely to occur within a few hours of increased SPM concentrations. In 1994, 70% of people died within four hours after the onset of myocardial infarction in Japan.¹³ The increased mortality within one to six hours after high SPM concentrations suggests that the onset of myocardial infarction occurs more frequently just after reaching a high concentration of SPM.

We searched the medical literature to identify confounding factors that strongly affected myocardial infarction deaths. We treated temperature^14–16 and the hour that myocardial infarction occurred^17,18 as confounding factors in the analysis. As there were apparent differences between urban and suburban areas, such as lifestyle and environment, we also included a regional effect as a confounding factor. As the onset of myocardial infarction has a circadian variation, it was necessary to include the hour of the day that myocardial infarctions occurred in the model. As death certificates in Japan do not include information about the time that myocardial infarction occurred, we used the hour of day when death occurred. Because most deaths from myocardial infarction occur within a few hours of myocardial infarction onset,¹³ we considered it appropriate to use the hour of death instead of the hour of myocardial infarction onset. To adjust for the effect of temperature in the model, we averaged the temperature over a period of time before death. As we were not certain of the exact time period to use for the temperature averaging, we used various time periods (from three hours to 24 hours) and found that the time period had little influence on the incidence rate ratio.

We used the rate of myocardial infarction deaths per 1000 hours instead of person hours in this study, because information on person hours was not available. The range of population in the study area from 1990 to 1994 was between 11.68 million (1991) and 11.87 million (1992). We assumed that the population in the study area was stable and used the rate of myocardial infarction deaths per 1000 hours to compare the different SPM categories.

In conclusion, we examined the transient effect of SPM on myocardial infarction deaths and found an increased incidence rate ratio within a few hours after reaching high concentrations of SPM. When we focused on the effect only a few hours after SPM reached high levels, there was a gradual increase in the incidence rate ratio as the SPM concentration increased.