Assessing Confounding, Effect Modification, and Thresholds in the Association between Ambient Particles and Daily Deaths.

I examined the relationship between daily deaths and airborne particles in 10 U.S. cities with varying climatic conditions and seasons in which particle concentrations were high. Airborne particles were associated with significant increases in daily deaths [0.67% increase for a 10 [micro]g/[m.sup.3] increase in particles; 95% confidence interval (CI), 0.52-0.81%]. This association was the same in summer and winter. To examine potential confounding by other pollutants, I regressed city- and season-specific effect sizes against the relationship between airborne particles and other pollutants. Controlling for other pollutants did not substantially (or significantly) change the estimated effect of airborne particles. Socioeconomic differences between cities likewise did not modify the effect. The increase in daily deaths that occurred out of hospitals (0.89% per 10 [micro]g/[m.sup.3]; CI, 0.67-1.10%) was substantially greater than the increase in deaths in hospitals (0.49%; CI, 0.31-0.68%). This is consistent with results previously reported in Philadelphia, Pennsylvania, and suggests that the particle-associated deaths are not just being brought forward by a few days. It is also consistent with recent animal and human studies of the mechanisms of particle toxicity. Key words: airborne particles, air pollution, climate mortality. Environ Health Perspect 108:563-568 (2000). [Online 3 May 2000]

http://ehpnet1.niehs.nih.gov/docs/2000/108p563-568schwartz /abstract.html

Studies on four continents have reported associations between daily concentrations of ambient particles and daily deaths (1,2). The magnitude of the regression coefficients varied, but were remarkably similar compared to epidemiologic studies of other exposures. Several arguments have been made to question the relevance of these findings for public health and preventive measures. It has been argued that the deaths are occurring in persons who were already seriously ill and who would have died in a few days anyway. It has been argued that air pollution is responsible for the deaths, but that airborne particles are not the responsible agent; rather, other pollutants confound the particle findings. It has also been argued that the particle associations only exist at higher concentrations, and therefore, most days are below a presumed threshold for effect; hence public health interventions to lower exposure would have no impact on most days.

Two recent papers addressed the first argument by showing that the association between daily deaths and airborne particles persisted after accounting for any short-term displacement of (reduced time until) deaths (3,4). In this paper, I address the latter two issues in a multiple-city analysis of particulate air pollution and daily deaths. I also indirectly address the first issue by an analysis stratified by location of death.

Recently, Sunyer et al. (5) reported that persons with a previous emergency room visit for chronic obstructive pulmonary disease (COPD) had a greater risk of air pollution-induced mortality. In general, there is interest in potential effect modifiers for particulate air pollution. Among these are social and economic factors that may represent differences in underlying risk. For example, income has been shown to be a potent predictor of life expectancy. These factors differ among cities in the United States, and these differences can be used to explore their role as effect modifiers for the impact of airborne particles.

Methods

Data. I selected 10 U.S. cities with approximately daily [PM.sub.10] (particulate matter [is less than or equal to] 10 [micro]m) monitoring to provide a reasonable number of locations for a combined analysis. The cities were New Haven, Connecticut; Pittsburgh, Pennsylvania; Birmingham, Alabama; Detroit, Michigan; Canton, Ohio; Chicago, Illinois; Minneapolis--St. Paul, Minnesota; Colorado Springs, Colorado; and Spokane and Seattle, Washington. Daily deaths in the metropolitan county containing each city were extracted from National Center for Health Statistics mortality tapes (6) for the years 1986-1993. I also computed separate daily counts of deaths in the hospitals and deaths out of hospitals. Minneapolis and St. Paul were combined and treated as one city. Daily weather data were obtained for the same years, from the nearest airport weather station, and daily concentrations of [PM.sub.10], sulfur dioxide, ozone, and carbon monoxide were obtained for those years from the U.S. Environmental Protections Agency's Aerometric Information Retrieval System (AIRS) monitoring network (Research Triangle Park, NC). Nitrogen dioxide data were not available in enough of the cities to allow examination of that variable.

Social and economic factors were extracted from the 1990 decennial Census (7) for use as potential effect modifiers. The variables used were the unemployment rate, the percentage of the population living below the poverty level, the percentage of the population with a college degree, and the percentage of the population that was nonwhite.

The assignment of [PM.sub.10] exposure raised a number of issues. Many of the locations have more than one monitoring location, but typically only one monitor operates on a daily basis, with the others operating every third or sixth day. If data from all of the monitors were simply averaged, the daily mean would change on days when new monitors were included merely because their annual average differs from the monitoring station that operates on a daily basis.

The variance of [PM.sub.10] measurements also can differ from monitoring location to monitoring location. Day-to-day changes in which monitors are included in the daily average would also result in changes in the day-to-day variation in the exposure measure that do not represent true changes in exposure, but only changes in the sampling of monitors. To remove these influences, I used the following algorithm. The annual mean was computed for each monitor for each year and subtracted from the daily values of that monitor. I then standardized these daily deviances from each monitor's annual average by dividing by the standard deviation for that monitor. The daily standardized deviations for each monitor on each day were averaged, producing a daily averaged standardized deviation. I multiplied this by the standard deviation of all of the centered monitor readings for the entire year and added back in the annual average of all of the monitors. This gave a daily average [PM.sub.10] concentration for each day in each city. I then computed the mean of the [PM.sub.10] concentration on the day of death and the day preceding death to use as my exposure index. Most studies have found that a 2-day average is a better predictor of mortality than a single day's exposure. Rather than optimizing in each location, I used the same 2-day average to ensure comparability.

Analytical methods. For each city, a generalized additive Poisson regression was fit (8,9), modeling the logarithm of the expected value of daily deaths as a sum of smooth functions of the predictor variables. The generalized additive model allows regressions to include nonparametric smooth functions to model the potential nonlinear dependence of daily admissions on weather and season. It assumes that

log[E(Y)] = [[Beta].sub.0] = [S.sub.1] (X.sub.1] + ... + [S.sub.p](X.sub.p]),

where Y is the daily count of deaths, E(Y) is the expected value of that count, the [X.sub.i] are the covariates and the [S.sub.i] are the smooth (i.e., continuously differentiable) functions. For the [S.sub.i] I used loess (10), a moving regression smoother. This approach is now standard in air pollution time series (11). For each covariate, it is necessary to choose a smoothing parameter that determines how smooth the function of that covariate should be. Three classes of predictor variables were used: a smooth function of time to capture seasonal and other long-term trends in the data, weather and day-of-the-week variables to capture shorter term potential confounding, and [PM.sub.10]. The choice of smoothing parameter for each set of variables is described.

The purpose of the smooth function of time is to remove the basic long-term pattern from the data. Seasonal patterns can vary greatly between Birmingham and Spokane, for example, and a separate smoothing parameter was chosen in each city to reduce the residuals of the regression to "white noise" (12) (i.e., remove serial correlation). This approach was used because each death is an independent event, and autocorrelation in residuals indicates there are omitted time-dependent covariates whose variation may confound air pollution. If the autocorrelation is removed, remaining variation in omitted covariates has no systematic temporal pattern, and hence confounding is less likely. This approach has been described previously (12). Sometimes it was necessary to incorporate autoregressive terms (13) to eliminate serial correlation from the residuals.

The other covariates were temperature, dew point temperature, and barometric pressure on the same day, the previous day's temperature, and day of the week. To allow for city-specific differences, the smoothing parameters for these covariates were also optimized separately in each location. The criterion used was to choose the parameter for each variable that minimized Akaike's Information Criterion (14).

[PM.sub.10] was treated as having a linear association with daily mortality in this analysis to facilitate the combination of coefficients across cities. Robust regression was used to reduce sensitivity to outliers in the dependent variable. To reduce sensitivity to outliers in the pollution variable, the baseline analysis was restricted to days when [PM.sub.10] levels were [is less than] 150 [micro]g/[m.sup.3], the currently enforced ambient standard. This also ensures that the results are unambiguously relevant to questions of revision of those standards.

Assessment of confounding. Confounding is usually assessed by including the potential confounder in the regression. This is a problem for air pollution epidemiology because atmospheric patterns, such as the height of the inversion layer, tend to produce parallel increases and decreases in all air pollutants. This creates considerable collinearity, and hence instability, in the estimated regression coefficients. However, while most pollutants tend to go up and down together within each city, the increase (in micrograms per cubic meter) in one pollutant that accompanies a 1 [micro]g/[m.sup.3] increase in another pollutant varies considerably among cities, as this depends on the source term. For example, some cities have very low sulfur fuels, and hence very different slopes in the association between [PM.sub.10] and [SO.sub.2] than in cities with high sulfur fuels. In the eastern United States, [PM.sub.10] peaks in the summer, when [O.sub.3] levels are high and CO levels are low, whereas in many western U.S. cities, [PM.sub.10] peaks in the winter. This creates considerable variation in the slopes between [PM.sub.10] and the other pollutants, particularly if analyses are stratified by season. This variation is often larger than the variation in within-city correlations among the pollutants, and my approach to confounding takes advantage of this fact. It is based on the observation that if the [PM.sub.10] effect is really due to confounding by another pollutant, I would expect a larger [PM.sub.10] effect in cities or seasons where 1 [micro]g/[m.sup.3] [PM.sub.10] is representing more of that other causal pollutant.

In this paper I use a hierarchical modeling approach to take advantage of this variation to assess confounding. In such an approach, the first stage consists of standard regression analyses, producing regression coefficients for the exposure or exposures of interest. In a second stage, those coefficients are regressed against explanatory factors. This approach has been widely used in the social sciences (15) and has begun to be applied in epidemiology (16). The city-specific Poisson regressions described above are the first stage. The second stage can be used to assess confounding by cooccuring pollutants. Consider, for simplicity, a Gaussian outcome and imagine that [X.sub.t] is the concentration on day t of the pollutant that is causally associated with the outcome [Y.sub.t]. Hence

[1] [Y.sub.t] = [[Beta].sub.0] + [[Beta].sub.1] [X.sub.t] + error.

[X.sub.t] is correlated with another pollutant, [Z.sub.t] which is not causally related to [Y.sub.t]. Therefore I may write

[2] [X.sub.t] = [[Gamma].sub.0] + [[Gamma].sub.1] [Z.sub.t] + error.

What happens if [Z.sub.t] is used as the exposure variable instead of [X.sub.t]? Substituting Equation 2 into Equation 1 I have

[3] [Y.sub.t] = [[Beta].sub.0] + [[Beta].sub.1] [[Gamma].sub.0] + [[Beta].sub.1] [[Gamma].sub.1] [Z.sub.t] + error.

I have confounding by the omitted covariate [X.sub.t], and the coefficient of [Z.sub.t] will be proportional to [[Gamma].sub.i] the slope of the association between [X.sub.t] and [Z.sub.t]. This can be illustrated by some simple simulations. Figure 1 shows the results of a simulated example where one variable has a true association with the outcome, and the second variable does not but is correlated with the first. The slope between the pollutants varies across different (simulated) cities, which are represented as different points in the figure. Figure 1 shows how the estimated effect size of the noncausal variable varies with [[Gamma].sub.i], the slope between the pollutants in each city. The effect size for the noncausal pollutant varies randomly about a line with a zero intercept. The zero intercept follows from Equation 3, where I see that if [[Gamma].sub.1] is zero, the expected effect size for the noncausal variable is zero. If I formalize this by performing a regression in the second stage, where, for example, the [PM.sub.10] effect size (in single pollution models) in each town is regressed against the [SO.sub.2] to [PM.sub.10] slope in each town, I would expect a zero intercept in the regression if the effect of [PM.sub.10] is all due to confounding. If both pollutants have a causal impact on the number of deaths, the effect size for [PM.sub.10] in each city may be overstated in a single-pollutant model. In that case, I would expect a nonzero intercept for [PM.sub.10] but one that is smaller than the average [PM.sub.10] effect size. This is shown by the diamonds in Figure 1. These data points are from a second set of simulations where both exposures were associated with the outcome. In this case, if I perform a second-stage regression, the intercept is an estimate of the effect size I would see for [PM.sub.10] in a city where it is uncorrelated with [SO.sub.2], which is to say, the unconfounded [PM.sub.10] effect size. I used this approach to examine confounding.

[Figure 1 ILLUSTRATION OMITTED]

Of course, the actual models fit to mortality data are log-linear. That is, I assume that

E([Y.sub.t]) = [[Lambda].sub.0]exp([[Beta][Z.sub.t],

where [[Lambda].sub.0] is the baseline risk before considering pollution. Since the relative risks associated with air pollution are generally [is less than] 1.1, ex([[Beta][Z.sub.t]) ~ 1 + [Beta][Z.sub.t], and the results are as before.

More formally, the two-stage approach consists of first fitting regressions of daily deaths against [PM.sub.10] in each location, controlling for season, weather, and day of the week. I assume these estimated coefficients [[Beta].sub.i] are normally distributed about some true city-specific coefficient that is proportional to [[Gamma].sub.i], plus possibly an effect of [PM.sub.10] net of confounding, that is,

[[Beta].sub.i] ~ N([Alpha] + [Delta][Gamma].sub.i], [Sigma]).

In the second stage, I estimate [Alpha] using a weighted regression, with inverse variance weights.

I have added one further refinement to increase the power of the analysis. In most cities, [O.sub.3], CO, and [SO.sub.2] show greater differences in their mean level between the indoor-heating season and the warm season than does [PM.sub.10]. This indicates that further variability in the slope between these pollutants and [PM.sub.10] can be obtained by dividing the data in each city into the indoor-heating season (defined as November through April) when CO and [SO.sub.2] are high but [O.sub.3] is low, and the warm season, when the opposite is true. This increases our ability to determine whether the [PM.sub.10] effect size varies with the slope between [PM.sub.10] and the other pollutants. To accomplish this, the regressions were fit separately in each city in each of the two seasons.

Assessment of effect modification. To test for effect modification, I used social and economic factors in the meta-regression instead of the slopes between pairs of air pollutants. This tests for an interaction term, where, for example, the effect of air pollution increases as the unemployment rate increases. Here our primary interest is in the coefficient of the effect modifier, which tells how much the [PM.sub.10] effect changes for a 1% increase in the unemployment rate, for example.

Assessment of low-level dose--response relationships. If there is a threshold for the effect of [PM.sub.10] on daily deaths, then the observed slope for [PM.sub.10] represents an average of the true slope above the threshold and a slope of zero below the threshold. One unambiguous way to determine whether the effect persists at low [PM.sub.10] concentrations is to limit the analysis to days with low concentrations. I chose a cutoff of 50 [micro]g/[m.sup.3], well below the current standard of 150 [micro]g/[m.sup.3] for [PM.sub.10]. If a threshold exists above that concentration, I would expect the mean effect estimate in the 10 cities to fall to zero. If there is a threshold [is less than] 50 [micro]g/[m.sup.3], I would expect the average effect-size estimate to fall because a larger fraction of the days are below the threshold in the restricted analysis than in the analysis that included days up to 150 [micro]g/[m.sup.3]. I refit the individual city analyses with a restriction limiting the analysis to days [is less than] 50 [micro]g/[m.sup.3] to test this hypothesis and combined the results using inverse variance weighting.

Location of death analysis. In addition to examining all cause mortality, I computed separate daily counts of deaths occurring in and out of hospitals. This is of interest for several reasons. First, it indirectly addresses the question of whether the time until death is only being reduced by a few days. One would expect people who are on the brink of death to disproportionately die in hospitals because many are in the hospital already. If air pollution primarily affected those people, I would expect its impact on hospital deaths to be larger than on out-of-hospital deaths. Second, the 1952 London smog disaster has been cited as providing biological plausibility to the observed associations at lower concentrations (17). If this association is real, one would expect the impact of particulate air pollution on deaths in and out of hospitals to show similar patterns to those observed during the London smog disaster.

Results

Table 1 shows the populations, mean daily deaths, and means of the environmental variables in the 10 study locations. The Census data are shown in Table 2. [PM.sub.10] was only modestly correlated with the weather variables in most of the 10 locations, and the correlations varied considerably, as shown in Table 3. There was considerable variation in the relationship between [PM.sub.10] and the other air pollutants across locations and seasons. The [SO.sub.2]/[PM.sub.10] coefficients ranged from a low of 0.079 to a high of 1.24. This is more than an order of magnitude, providing enough power to determine if there is a trend to higher [PM.sub.10] slopes in locations where 1 [micro]g/[m.sup.3] [PM.sub.10] represents more [SO.sub.2]. The same was true for the other pollutants, where the [O.sub.3]/[PM.sub.10] slopes ranged from -0.22 to 1.07 and the CO/[PM.sub.10] slopes ranged from 0.013 to 0.08.

Table 1. Characteristics of the study locations.

                           1990               [PM.sub.10]
City                    Population   Deaths   ([micro]g/
                                              [m.sup.3])

New Haven                  804,219     20.4          28.6
Birmingham                 651,525     19.1          34.8
Pittsburgh               1,336,449     63.3          36.4
Detroit                  2,111,687     59.7          36.9
Canton                     367,585      9.9         29.31
Chicago                  5,105,067    133.4          36.5
Minneapolis--St. Paul    1,518,196     32.3          27.5
Colorado Springs           397,014        6          27.1
Spokane                    361,364      8.7          40.6
Seattle                  1,507,319     29.3          32.5

                        Dew     Temperature     Pressure
City                    point   ([degrees] F)   (mmHg)

New Haven                40.1      50.5           29.8
Birmingham               51.7      62.4           29.4
Pittsburgh               41.2      52.1           28.8
Detroit                  40.7      50.9           29.3
Canton                   41        50.4           28.7
Chicago                  39.8      50.3           29.3
Minneapolis--St. Paul    35.5      46.3           29.1
Colorado Springs         28.9      48.9           24.0
Spokane                  34.2      47.9           27.5
Seattle                  43.9      52.5           29.6
Table 2. Demographic characteristics of study locations.

                        Percent      Percent   Percent   Percent
City                    unemployed   with      below     nonwhite
                                     college   poverty
                                     degree    level

New Haven                   5.8       24.2        7.9       14
Birmingham                  6.5       19.9       16.0       36
Pittsburgh                  6.3       22.6       11.5       12
Detroit                    12.4       13.7       20.1       43
Canton                      7.2       14.3       11.1        8
Chicago                     8.0       22.8       14.2       37
Minneapolis--St. Paul       4.8       30.7        9.9       11
Colorado Springs            7.3       25.8       10.4       14
Spokane                     7.3       20.6       13.7        5
Seattle                     4.1       32.8        8.0       15
Table 3. Correlations between [PM.sub.10] and weather
variables.

                   Temperature                 Pressure
City               ([degrees] F)   Dew point   (mmHg)

New Haven              0.05          -0.11       0.11
Birmingham             0.26           0.19       0.19
Pittsburgh             0.45           0.44       0.44
Detroit                0.37           0.38       0.38
Canton                 0.42           0.45       0.45
Chicago                0.36           0.32       0.32
Minneapolis(a)         0.29           0.26       0.26
Colorado Springs      -0.34          -0.42      -0.42
Spokane               -0.01          -0.16      -0.16
Seattle               -0.22          -0.29      -0.29

(a) Minneapolis--St. Paul.

Table 4 shows the estimated effect of a 10 [micro]g/[m.sup.3] increase in [PM.sub.10] for all deaths, for deaths out of hospitals, and for deaths in hospitals. [PM.sub.10] was a significant predictor of all-cause mortality [0.67% increase for a 10 [micro]g/[m.sup.3] increase in [PM.sub.10]; 95% confidence interval (CI), 0.52-0.81%]. The effect size was identical in the summer and winter periods. However, the effects of airborne particles were substantially higher for deaths out of the hospital than for deaths in the hospital.

Table 4. Estimated effect of a 10-[micro]g/[m.sup.3] increase in [PM.sub.10] on daily deaths in the meta-analysis.

                      Percent increase
Model                 in deaths          95% CI

Overall                   0.67           0.52-0.81
Summer only               0.67           0.48-0.86
Winter only               0.66           0.45-0.87
In hospitals              0.49           0.31-0.68
Out of hospitals          0.89           0.67-1.10
Days < 50 [micro]g/       0.87           0.62-1.12
[m.sup.3]

Confounding by
  [SO.sub.2]              0.57           0.25-0.90
  CO                      0.90           0.42-0.97
  [O.sub.3]               0.69           0.53-1.26

Table 4 also shows the results when restricted to days with [PM.sub.10] [is less than] 50 [micro]g/[m.sup.3]. The slope of airborne particles was larger when restricted to low air pollution days. These results are illustrated graphically in Figure 2.

[Figure 2 ILLUSTRATION OMITTED]

Table 4 also shows the estimated effect of [PM.sub.10] after controlling for potential confounding by [SO.sub.2], [O.sub.3], and CO (i.e., the intercept term in the regression of the baseline [PM.sub.10] effect on the coefficient relating [PM.sub.10] to each of the other pollutants). For all three cooccurring pollutants, the effect size after controlling for confounding was not substantially (or statistically significantly) different from the baseline result. This is illustrated in Figure 3. These results indicate that there is no trend to a higher [PM.sub.10] coefficient in cities or seasons with higher slopes between the cooccurring pollutants and [PM.sub.10]. This is illustrated in Figure 4, which shows the effect size for [PM.sub.10] in each city and season plotted against the [O.sub.3]/[PM.sub.10] coefficient.

[Figures 3-4 ILLUSTRATION OMITTED]

Figure 5 shows the estimated effect modification by different measures of social and economic status. It shows how much more of an increase in daily deaths is associated with a 10 [micro]g/[m.sup.3] increase in [PM.sub.10] if the city had a 5% higher unemployment rate, an additional 5% of the population living under the poverty level, an additional 5% of the population with college degrees, or an additional 5% of the population nonwhite. These are substantial increases in each of the postulated effect modifiers, but they are associated with no noticeable change in the estimated [PM.sub.10] effect.

[Figure 5 ILLUSTRATION OMITTED]

Discussion

In an analysis of multiple cities across the United States, [PM.sub.10] was a significant predictor of daily deaths. The association was identical in analyses restricted to the indoor-heating season and the warm months. This is consistent with previously published results (18). Given the large differences in the concentrations of cooccurring pollutants between the summer and winter months, this alone is evidence that the particle associations cannot be primarily due to confounding with other pollutants.

The association differed by location of death, with a larger effect on deaths out of hospitals. These results are consistent with previous reports from Philadelphia (19) and with the experience in the great London smog episode of 1952 (17). This suggests that most of the [PM.sub.10]-associated deaths are not in people who are desperately ill and hence that, in most cases, increased mortality is not a result of time of death simply being reduced by a few days.

A higher risk of death out of the hospital suggests that sudden death is a major component of the air pollution-associated risk and, indeed, "dead on arrival" deaths were most strongly associated with air pollution in the Philadelphia analysis (16). Recently, more mechanistic evidence has been developed that supports the notion that airborne particles can be associated with sudden death. A study of subjects with implanted cardiac defibrillators found an increased risk of ventricular tachycardia and ventricular arrhythmia associated with [PM.sub.2.5] (20). Arrhythmia is one of the major causes of sudden death. Arrhythmia and sudden death have also been produced in rats by combustion particles (21) under experimental conditions where the responses cannot be attributed to cooccurring pollutants. This association is also supported by studies of electrocardiogram changes that are precursors to arrhythmia. Godleski et al (22) reported an association between these electrocardiogram changes and exposure to concentrated air particles under experimental conditions in animals with preexisting illnesses. Similar changes have been reported to be associated with airborne particles in three epidemiologic studies using continuous electrocardiogram monitoring in humans (23-25). Increases in heart rate have been associated with exposure to airborne particles in studies in Baltimore, Maryland (25); Germany and Boston, Massachusetts (26); and Utah (27).

Another major cause of sudden death is thrombotic processes leading to myocardial infarctions. Here again, recent animal and human studies indicate that airborne particles may be affecting these processes. Exposure to combustion particles has been associated with increased plasma fibrinogen in rats (28), and an episode of high particulate air pollution was associated with increased plasma viscosity in a large epidemiology study (29). The findings of the present study are therefore consistent with a growing body of more mechanistic research in humans and animals.

There was no trend of higher [PM.sub.10] effect sizes in towns with higher [SO.sub.2]/[PM.sub.10] slopes, nor in towns with higher [O.sub.3]/[PM.sub.10] or CO/[PM.sub.10] slopes. This indicates that the [PM.sub.10] effects are not likely to be caused by confounding by other pollutants. These results address the issue of whether the [PM.sub.10] effect is due to the other pollutants: they do not address the question of whether those other pollutants have significant associations with daily deaths as well. This will be addressed in a later study.

Recent animal studies, in which exposure can be controlled and limited to airborne particles, support the finding of an independent particle effect. For example, Zelikoff et al. (30) reported that exposure to concentrated air particles after infection with streptococcal pneumonia was associated with a doubling of the area of lung involvement and a doubling of the bacterial burden of rats within 48 hr. Effects of particle exposure on influenza mortality have also been noted (31).

The [PM.sub.10] effect was not substantially modified by socioeconomic status measured at the city level, but when the analysis was restricted to days with [PM.sub.10] concentrations [is less than] 50 [micro]g/[m.sup.3], the effect was greater. These results are inconsistent with a threshold for [PM.sub.10] at any concentrations except those substantially [is less than] 50 [micro]g/[m.sup.3. Indeed, they suggest that the [PM.sub.10] slope increases at lower concentrations, rather than approaching zero. This tendency for a lower slope at high concentrations has been noted in London (32) and in the APHEA study (Air Pollution and Health: a European Approach) (33). A study of six U.S. cities recently reported a higher slope for [PM.sub.2.5] when the analysis was restricted to days [is less than] 30 [micro]g/[m.sup.3] (34).

One limitation of studies such as these is the use of outdoor monitoring stations rather than personal exposure monitors. Because the difference between these measurements can be large, some have questioned whether the associations reported in daily time--series studies could be causal. Several recent papers have addressed parts of this issue. Wilson and Suh (35) pointed out that outdoor monitors are surrogates for personal exposure to particles of outdoor origin, such as motor vehicle exhaust and sulfates. Current personal monitors measure personal exposure to particles of all sources, including resuspended house dust, environmental tobacco smoke, and cooking aerosols. Hence, personal exposure to particles of outdoor origin are more closely related to outdoor concentrations than some interpretations of personal monitoring data suggest. This has been confirmed by Janssen et al. (36), who found median correlations between personal particle monitors in adults and outdoor monitors were much higher after excluding environmental tobacco smoke (ETS) exposure. Janssen et al. (36) also highlighted another key issue. Most of the difference between personal PM exposure and outdoor concentrations reflects cross-sectional variations among persons. For time--series studies, it is the longitudinal correlation that matters, and Janssen et al. (36) reported considerably higher longitudinal correlations between personal PM exposure and outdoor concentrations, with a median of 0.70 for [PM.sub.10] in the absence of ETS exposure. Finally, two recent articles examined the statistical implications of the measurement error. Schwartz and Levin (37) pointed out that most of the difference between personal and central measurements of exposure in the time--series context are Berkson error, and hence do not bias the estimates. Zeger et al. (38) have explored the issue in more detail and have shown that the remaining bias is negative--that is, an underestimation of the effect. Hence, measurement error in exposure is an unlikely cause of these associations.

In sum, this study provides evidence that airborne particles influence the number of daily deaths and that these effects are not primarily attributable to other air pollutants. The data show the same pattern of higher relative effect on deaths out of the hospital that was seen in an air pollution episode where causality of the pollution effect is well accepted. That pattern, moreover, is consistent with recent animal and human data on the effects of particles on risk factors for sudden death. Finally, the public health benefit of each incremental reduction of 1 [micro]g/[m.sup.3] appears to be higher at the lower air pollution levels that prevail on most days. This suggests that intervention strategies that lower average levels, rather that those that address the few peak days, are the most appropriate. This is an important consideration, as a number of cities (e.g., Mexico City, Mexico; Athens, Greece) have adapted strategies that limit driving or industrial activity on peak pollution days. Such approaches do lower average levels, but are costly and disruptive, and the same effort put into reducing everyday emissions appears likely to produce greater public health benefit.

REFERENCES AND NOTES

(1.) Pope CA, Dockery DW, Schwartz J. Review of epidemiologic evidence of health effects of particulate air pollution. Inhal Toxicol 7:1-18 (1995).

(2.) Katsouyanni K, Touloumi G, Spix C, Schwartz J, Balducci F, Medina S, Rossi G, Wojtyniak D, Sunyer J, Bacharova L, et al. Short term effects of ambient sulphur dioxide and particulate mater on mortality in 12 European cities: results from time series data from the APHEA project. Br Med J 314:1658-1663 (1997).

(3.) Zeger SL, Dominici F, Samet J. Harvesting-resistant estimates of air pollution effects on mortality. Epidemiology 10:171-175 (1999).

(4.) Schwartz J. Harvesting and long-term exposure effects in the relationship between air pollution and mortality. Am J Epidemiol 151:440-448 (2000).

(5.) Sunyer J, Schwartz J, Tobias A, Macfarlane D, Garcia J, Anto JM. Patients with chronic obstructive pulmonary disease are at increased risk of death associated with urban particle air pollution: a case-crossover analysis. Am J Epidemiol 151:50-56 (2000).

(6.) National Center for Health Statistics. Public Use Data Tape Documentation. Mortality Detail. Hyattsville, MD:National Center for Health Statistics, 1986-1993.

(7.) U.S. Department of Commerce, Economics and Statistics Administration, Bureau of the Census. 1990 Census of the Population and Housing. Washington:U.S. Government Printing Office, 1992. Available: http://www.census.gov/ main/www/cen1990.html [cited May 1999].

(8.) Hastie T, Tibshirani R. Generalized Additive Models. London:Chapman and Hall, 1990.

(9.) Schwartz J. Air pollution and daily mortality in Birmingham, Alabama. Am J Epidemiol 137:1136-1147 (1993).

(10.) Cleveland WS, Devlin SJ. Robust locally-weighted regression and smoothing scatterplots. J Am Stat Assoc 74:829-836 (1988).

(11.) Schwartz J. Generalized additive models in epidemiology. In: International Biometric Society, Invited Papers. 17th International Biometric Conference. Hamilton, Ontario, Canada:International Biometric Society, 1994;55-80.

(12.) Schwartz J. Air pollution and hospital admissions for heart disease in eight U.S. counties. Epidemiology 10:17-22 (1999).

(13.) Brumback BA, Ryan LM, Schwartz JD, Neas LM, Stark PC, Burge HA. Transitional regression models with application to environmental time series. J Am Stat Assoc 449:16-28 (2000).

(14.) Akaike H. Information theory and an extension of the maximum likelihood principal. In: 2nd International Symposium on Information Theory (Petrov BN, Csaki F, eds). Budapest:Akademiai Kiado, 1973;267-281.

(15.) Bryk AS, Raudenbush SW. Hierarchical linear models: applications and data analysis methods. Advanced Quantitative Techniques in the Social Sciences Series. Vol 1. Newbury Park, CA:Sage Publications, 1992.

(16.) Witte JS, Greenland S, Bird CL, Haile RW. Hierarchical regression analysis applied to a study of multiple dietary exposures and breast cancer. Epidemiology 5:612-621 (1994).

(17.) Mortality and Morbidity during the London Fog of December 1952. Rpt No 95 on Public Health and Medical Subjects. London:Her Majesty's Stationary Office, 1954.

(18.) Kelsall JE, Samet JM, Zeger SL, Xu J. Air pollution and mortality in Philadelphia: 1974-1988. Am J Epidemiol 146:750-762 (1997).

(19.) Schwartz J. What are people dying of on high air pollution days? Environ Res 64:26-35 (1994).

(20.) Peters A, Liu E, Verrier RL, Schwartz J, Gold DR, Mittleman M, Baliff J, Oh A, Allen G, Monahan K, et al. Air pollution and incidences of cardiac arrhythmia. Epidemiology 11:11-17 (2000).

(21.) Watkinson WP, Campen MJ, Costa DL. Cardiac arrhythmia induction after exposure to residual oil fly ash particles in a rodent model of pulmonary hypertension. Toxicol Sci 41(2):209-216 (1998).

(22.) Godleski JJ, Lovett EG, Sioutas C, Killingsworth CR, Krishnamurthi GG, Hatch V, Wolfsom M, Ferguson ST, Koutrakis P, Verrier RL Impact of inhaled concentrated ambient air particles on canine electrocardiographic patterns [Abstract]. Am J Respir Crit Care Med 157:A260 (1998).

(23.) Pope CA III, Verrier RL, Lovett EG, Larson AC, Raizenne ME, Kanner RE, Schwartz J, Villegas GM, Dockery DW. Heart rate variability associated with particulate air pollution. Am Heart J 138:890-899 (1999).

(24.) Liao D, Creason J, Shy C, Williams R, Watts R, Zweidinger R. Daily variation of particulate air pollution and poor cardiac autonomic control in the elderly. Environ Health Perspect 107:521-525 (1999).

(25.) Gold DR, Litonjua A, Schwartz J, Verrier M, Milstein R, Larson A, Lovett E, Verrier R. Ambient pollution and heart rate variability. Circulation 101:1267-1273 (2000).

(26.) Peters A, Perz S, Doring A, Stieber J, Koenig W, Wichmann HE. Increases in heart rate during an air pollution episode. Am J Epidemiol 150(10):1094-1098 (1999).

(27.) Pope CA, Dockery DW, Kanner RE, Villegas GM, Schwartz J. Oxygen saturation, pulse rate, and particulate air pollution: a daily time series panel study. Am J Respir Crit Care Med 159:365-372 (1999).

(28.) Gardner SY, Costa DL. Particle-induced elevations in white blood cell count and plasma fibrinogen levels in rats [Abstract]. Am J Respir Crit Care Meal 157:A152 (1998).

(29.) Peters A, Boring A, Wichmann HE, Koenig W. Increased plasma viscosity during an air pollution episode: a link to mortality? Lancet 349:1582-1587 (1997).

(30.) Zelikoff JT, Nadziejko C, Fang T, Gordon C, Premdass C, Cohen MD. Short term, low-dose inhalation of ambient particulate matter exacerbates ongoing pneumonococcal infections in Streptoccus pneumoniae-infected rats. In: Proceedings of the Third Colloquium on Particulate Air Pollution and Human Health (Phalen RF, Bell YM, eds). Irvine, CA:University of California, Air Pollution Health Effects Laboratory, 1999;8-94 to 8-101.

(31.) Clarke RW, Hemenway DR, Frank R, Kleeberger SR, Longphre MV, Jakab GJ. Particle associated sulfate exposure enhances murine influenza mortality [Abstract]. Am J Respir Crit Care Med 155:A245 (1997).

(32.) Schwartz J, Marcus A. Mortality and air pollution in London: a time series analysis. Am J Epidemiol 131:185-194 (1990).

(33.) Katsouyanni K, Touloumi G, Spix C, Schwartz J, Balducci F, Medina S, Rossi G, Wojtyniak B, Sunyer J, Bacharova L, et al. Short term effects of ambient sulphur dioxide and particulate matter on mortality in 12 European cities: results from time series data from the APHEA project. Br Med J 314:1658-1663 (1997).

(34.) Schwartz J, Dockery DW, Neas LM. Is daily mortality associated specifically with fine particles? J Air Waste Manag Assoc 46:2-14 (1996).

(35.) Wilson WE, Suh H. Fine particles and coarse particles: concentration relationships relevant to epidemiologic studies. J Air Waste Manag Assoc 47:1236-1249 (1997).

(36.) Janssen NA, Hoek G, Brunekreef B, Harsseman H, Mensink I, Zuidhof A. Personal sampling of PM10 in adults: relation between personal, indoor, and outdoor concentrations. Am J Epidemiol 147:537-547 (1998).

(37.) Schwartz J, Levin R. Drinking water turbidity and health. Epidemiology 10:86-90 (1999).

(38.) Zeger SL, Thomas D, Dominici F, Samet J, Schwartz J, Dockery D, Cohen A. Exposure measurement error in time--series studies of air pollution: concepts and consequences. Environ Health Perspect 108:419-426 (2000).

Joel Schwartz Environmental Epidemiology Program, Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts, USA

Address correspondence to J. Schwartz, Environmental Epidemiology Program, Department of Environmental Health, Harvard School of Public Health, 665 Huntington Avenue, Boston, MA 02115 USA. Telephone: [phone]. Fax: [phone]. E-mail: [e-mail]

This work was supported in part by NIEHS grant ES 07410 and by an EPA PM Research Center Award.

Received 30 June 1999; accepted 14 January 2000.