sign in sign up
<<
Home , Confounding

Articles

Assessing Confounding, Effect Modification, and Thresholds in the Association between Ambient Particles and Daily Deaths Joel Schwartz Environmental Program, Department of Environmental Health, Harvard School of Public Health, Boston, Massachusetts, USA

were not available in enough of the cities to I examined the relationship between daily deadt and airborne partides in 10 U.S. cities with vary- allow examination ofthat variable. ing dimatic conditions and seasons in which partide concentraions were high. Airbome partiles Social and economic factors were extract- were associated with significant increases in daily deaths [0.67% increase for a 10 pgWm3 increase ed from the 1990 decennial (7) for in particles; 95% confidence intervl (CI), 0.52-0.81%]. This association was the same in summer use as potential effect modifiers. The variables and winter. To examine potential confounding by other pollutants, I regrsed city- and season- used were the unemployment rate, the per- specific effect sizes against the relationship between airborne particles and other pollutants. centage of the population living below the Controiing for other pollutants did not substantialy (or siicany) chnge the estimated effect poverty level, the percentage of the popula- ofairborne particles. Socioeconomic differences between cities likewise did not modify the effict. tion with a college degree, and the percentage The increase in daily death that occurred out ofhospitals (0.89% per 10 pg/m3; CI, 0.67-1.10%) ofthe population that was nonwhite. was substantially greater than the increase in deaths in hospitals (0.49%; CI, 0.314.68%). This is The assignment of PM1O exposure raised consistent with results previously reported in Philadelphia, Pennsylvania, and suggests that the a number of issues. Many of the locations particle-associated deaths are not just being brought forward by a few days. It is also nt have more than one monitoring location, with recent animal and human studies ofthe mehnisms ofpartide toxcity. Key work airborne but typically only one monitor operates on a particles, air pollution, climate mortality. Environ Helth Pepect 108:563-568 (2000). [Online daily basis, with the others operating every 3 May 20001 third or sixth day. If from all of the btp://ehpnetl.niebs.nib.gpv/doas/2000108p563-568schwarn/abstracth.nm monitors were simply averaged, the daily would change on days when new monitors were included merely because their Studies on four continents have reported in potential effect modifiers for particulate air annual average differs from the monitoring associations between daily concentrations of pollution. Among these are social and eco- station that operates on a daily basis. ambient particles and daily deaths (1,4. The nomic factors that may represent differences The of PM10 measurements also magnitude of the regression coefficients var- in underlying risk. For example, income has can differ from monitoring location to mon- ied, but were remarkably similar compared to been shown to be a potent predictor of life itoring location. Day-to-day changes in epidemiologic studies of other exposures. expectancy. These factors differ among cities which monitors are included in the daily Several arguments have been made to ques- in the United States, and these diffferences average would also result in changes in the tion the relevance ofthese findings for public can be used to explore their role as effect day-to-day variation in the exposure measure health and preventive measures. It has been modifiers for the impact ofairborne particles. that do not represent true changes in expo- argued that the deaths are occurring in per- sure, but only changes in the of sons who were already seriously ill and who Methods monitors. To remove these influences, I used would have died in a few days anyway. It has Data. I selected 10 U.S. cities with approxi- the following algorithm. The annual mean been argued that air pollution is responsible mately daily PM1O (particulate matter < 10 was computed for each monitor for each for the deaths, but that airborne particles are pm) monitoring to provide a reasonable year and subtracted from the daily values of not the responsible agent; rather, other pollu- number of locations for a combined analysis. that monitor. I then standardized these daily tants confound the particle findings. It has The cities were New Haven, Connecticut; deviances from each monitor's annual aver- also been argued that the partide associations Pittsburgh, Pennsylvania; Birmingham, age by dividing by the for only exist at higher concentrations, and Alabama; Detroit, Michigan; Canton, Ohio; that monitor. The daily standardized devia- therefore, most days are below a presumed Chicago, Illinois; Minneapolis-St. Paul, tions for each monitor on each day were threshold for effect; hence public health Minnesota; Colorado Springs, Colorado; and averaged, producing a daily averaged stan- interventions to lower exposure would have Spokane and Seattle, Washington. Daily dardized deviation. I multiplied this by the no impact on most days. deaths in the metropolitan county containing standard deviation of all of the centered Two recent papers addressed the first each city were extracted from National monitor readings for the entire year and argument by showing that the association Center for Health mortality tapes added back in the annual average of all of between daily deaths and airborne particles (6) for the years 1986-1993. I also computed the monitors. This gave a daily average persisted after accounting for any short-term separate daily counts of deaths in the hospi- PMIO concentration for each day in each displacement of (reduced time until) deaths tals and deaths out of hospitals. Minneapolis (3,4). In this paper, I address the latter two and St. Paul were combined and treated as Address correspondence to J. Schwartz, Environ- issues in a multiple-city analysis ofparticulate one city. Daily weather data were obtained mental Epidemiology Program, Department of air pollution and daily deaths. I also indirect- for the same years, from the nearest airport Environmental Health, Harvard School of Public ly address the first issue by an analysis strati- weather station, and daily concentrations of Health, 665 Huntington Avenue, Boston, MA fied by location ofdeath. PM1O, sulfur dioxide, ozone, and carbon 02115 USA Telephone: (617) 432-1245. Fax: (617) et al. that monoxide were obtained for those years from 277-2382. E-mail: [email protected] Recently, Sunyer (5) reported This work was supported in part by NIEHS persons with a previous emergency room visit the U.S. Environmental Protections Agency's grant ES 07410 and by an EPA PM Research for chronic obstructive pulnmonary disease Aerometric Information Retrieval System Center Award. (COPD) had a greater risk of air pollution- (AIRS) monitoring network (Research Received 30 June 1999; accepted 14 January induced mortality. In general, there is interest Triangle Park, NC). Nitrogen dioxide data 2000.

Environmental Health Perspectives * VOLUME 1081 NUMBER 61 June 2000 563 Articles . Schwartz city. I then computed the mean of the PM1o criterion used was to choose the parameter for simplicity, a Gaussian outcome and imag- concentration on the day of death and the for each variable that minimized Akaike's ine that Xt is the concentration on day t of day preceding death to use as my exposure Information Criterion (14). the pollutant that is causally associated with index. Most studies have found that a 2-day PM1O was treated as having a linear asso- the outcome YI. Hence average is a better predictor ofmortality than ciation with daily mortality in this analysis to Yt = + 131Xt + error.[1 a single day's exposure. Rather than optimiz- facilitate the combination of coefficients PhO ing in each location, I used the same 2-day across cities. was used to Xt is correlated with another pollutant, Z4 average to ensure comparability. reduce sensitivity to outliers in the dependent which is not causally related to Y,. Therefore Analytical methods. For each city, a gen- variable. To reduce sensitivity to outliers in I may write eralized additive was fit the pollution variable, the baseline analysis Xt = 0y + yZt + error. [2] (8,, modeling the logarithm of the expect- was restricted to days when PM1O levels were ed value of daily deaths as a sum of smooth < 150 pg/m3, the currently enforced ambient What happens if Zt is used as the exposure functions of the predictor variables. The standard. This also ensures that the results are variable instead ofXA? Substituting Equation generalized additive model allows regressions unambiguously relevant to questions of revi- 2 into Equation 1 1 have to include nonparametric smooth functions sion ofthose standards. Yt= P0 +P1y0+ + error. [3] to model the potential nonlinear dependence Assessment ofconfounding. Confounding PlyZt ofdaily admissions on weather and season. It is usually assessed by induding the potential I have confounding by the omitted covariate assumes that confounder in the regression. This is a prob- Xp and the coefficient of Zt will be propor- log[E(I1] = PO = S,(XI) + ... +Sp(X), lem for air pollution epidemiology because tional to yi, the slope of the association atmospheric patterns, such as the height of between Xt and Zr This can be illustrated by where Yis the daily count of deaths, E(Y) is the inversion layer, tend to produce parallel some simple simulations. Figure 1 shows the the expected value of that count, the X. are increases and decreases in all air pollutants. results of a simulated example where one the covariates and the Si are the smooth (i.e., This creates considerable collinearity, and variable has a true association with the out- continuously differentiable) functions. For hence instability, in the estimated regression come, and the second variable does not but is the Si I used loess (10, a moving regression coefficients. However, while most pollutants correlated with the first. The slope between smoother. This approach is now standard in tend to go up and down together within each the pollutants varies across different (simulat- air pollution (11). For each city, the increase (in micrograms per cubic ed) cities, which are represented as different covariate, it is necessary to choose a smooth- meter) in one pollutant that accompanies a 1 points in the figure. Figure 1 shows how the ing parameter that determines how smooth pg/m3 increase in another pollutant varies estimated of the noncausal variable the function of that covariate should be. considerably among cities, as this depends on varies with y., the slope between the pollu- Three dasses of predictor variables were used: the source term. For example, some cities tants in each city. The effect size for the non- a smooth function oftime to capture seasonal have very low sulfur fuels, and hence very dif- causal pollutant varies randomnly about a line and other long-term trends in the data, ferent slopes in the association between PM1O with a zero intercept. The zero intercept fol- weather and day-of-the-week variables to cap- and SO2 than in cities with high sulfur fuels. lows from Equation 3, where I see that if^Yl is ture shorter term potential confounding, and In the eastern United States, PM1O peaks in zero, the expected effect size for the non- PM1O. The choice of smoothing parameter the summer, when 03 levels are high and causal variable is zero. If I formalize this by for each set ofvariables is described below. CO levels are low, whereas in many western performing a regression in the second stage, The purpose of the smooth function of U.S. cities, PM1O peaks in the winter. This where, for example, the PM1O effect size (in time is to remove the basic long-term pattern creates considerable variation in the slopes single pollution models) in each town is from the data. Seasonal patterns can vary between PM1O and the other pollutants, par- regressed against the SO2 to PM1O slope in greatly between Birmingham and Spokane, ticularly if analyses are stratified by season. each town, I would expect a zero intercept in for example, and a separate smoothing para- This variation is often larger than the varia- the regression if the effect of PM1O is all due meter was chosen in each city to reduce the tion in within-city correlations among the to confounding. If both pollutants have a residuals of the regression to "white noise" pollutants, and my approach to confounding (12) (i.e., remove serial correlation). This takes advantage of this fact. It is based on the 1.5 approach was used because each death is an observation that if the PM1O effect is really independent event, and in due to confounding by another pollutant, I residuals indicates there are omitted time- would expect a larger PM1O effect in cities or 1.0 ~ dependent covariates whose variation may seasons where 1 pg/m3 PM10 is representing confound air pollution. Ifthe autocorrelation more ofthat other causal pollutant. ,.5 b | ! t is removed, remaining variation in omitted In this paper I use a hierarchical model- covariates has no systematic temporal pat- ing approach to take advantage of this varia- tern, and hence confounding is less likely. tion to assess confounding. In such an 0.0 This approach has been described previously approach, the first stage consists of standard 0.2 0.4 0.6 0.8 1.0 (12). Sometimes it was necessary to incorpo- regression analyses, producing regression Slope between poliuumt rate autoregressive terms (13) to eliminate coefficients for the exposure or exposures of Figure 1. Scatterplot showing the results of a simu- serial correlation from the residuals. interest. In a second stage, those coefficients lation. Plotted are the effect-size estimate for one The other covariates were temperature, are regressed against explanatory factors. This pollutant as a function of the regression coefficient dew point temperature, and barometric pres- approach has been widely used in the social between it and a confounding pollutant that is sure on the same day, the previous day's sciences (15) and has begun to be applied in causally related to the outcome. The squares show temperature, and day of the week. To allow epidemiology (16). The city-specific Poisson the results when the first pollutant has no causal for association with the outcome. The diamonds show city-specific differences, the smoothing regressions described above are the first stage. the expected results when both pollutants are parameters for these covariates were also The second stage can be used to assess con- causally connected to the outcome, butthe second optimized separately in each location. The founding by cooccuring pollutants. Consider, pollutant confounds the association with the first.

564 VOLUME 1081 NUMBER 61 June 2000 * Environmental Health Perspectives Articles * Ambient particles and daily mortality causal impact on the number of deaths, the the PM0o effect changes for a 1% increase in to be larger than on out-of-hospital deaths. effect size for PM1O in each city may be over- the unemployment rate, for example. Second, the 1952 London smog disaster has stated in a single-pollutant model. In that Assessment oflow-level dose-response rela- been cited as providing biological plausibility case, I would expect a nonzero intercept for tionships. If there is a threshold for the effect to the observed associations at lower concen- PMIO but one that is smaller than the average of PM1O on daily deaths, then the observed trations (17'). If this association is real, one PMIO effect size. This is shown by the dia- slope for PM1O represents an average of the would expect the impact of particulate air monds in Figure 1. These data points are true slope above the threshold and a slope of pollution on deaths in and out of hospitals from a second set of simulations where both zero below the threshold. One unambiguous to show similar patterns to those observed exposures were associated with the outcome. way to determine whether the effect persists at during the London smog disaster. In this case, if I perform a second-stage low PMIO concentrations is to limit the analy- regression, the intercept is an estimate of the sis to days with low concentrations. I chose a Results effect size I would see for PMIO in a city cutoff of 50 gig/m3, well below the current Table 1 shows the populations, mean daily where it is uncorrelated with SO2, which is standard of 150 pg/m3 for PM1O. If a thresh- deaths, and of the environmental vari- to say, the unconfounded PM1O effect size. I old exists above that concentration, I would ables in the 10 study locations. The Census used this approach to examine confounding. expect the mean effect estimate in the 10 data are shown in Table 2. PM1O was only Of course, the actual models fit to mor- cities to fall to zero. If there is a threshold modestly correlated with the weather variables tality data are log-linear. That is, I assume < 50 jg/m3, I would expect the average effect- in most of the 10 locations, and the correla- that size estimate to fall because a larger fraction of tions varied considerably, as shown in Table E(8) = Xoexp(PZ), the days are below the threshold in the 3. There was considerable variation in the restricted analysis than in the analysis that relationship between PM1O and the other air where A0 is the baseline risk before considering included days up to 150 pg/m3. I refit the pollutants across locations and seasons. The pollution. Since the relative risks associated individual city analyses with a restriction lim- SO2/PM10 coefficients ranged from a low of with air pollution are generally < 1.1, exp((3Z) iting the analysis to days < 50 gg/m3 to test 0.079 to a high of 1.24. This is more than an - 1 + PZ4 and the results are as before. this hypothesis and combined the results order of magnitude, providing enough power More formally, the two-stage approach using inverse variance weighting. to determine if there is a trend to higher consists of first fitting regressions of daily Location ofdeath analysis. In addition to PM0O slopes in locations where 1 pg/m3 PM10 deaths against PM1O in each location, con- examining all cause mortality, I computed represents more SO2. The same was true for trolling for season, weather, and day of the separate daily counts of deaths occurring in the other pollutants, where the 03/PMIO week. I assume these estimated coefficients and out of hospitals. This is of interest for slopes ranged from -0.22 to 1.07 and the i are normally distributed about some true several reasons. First, it indirectly addresses CO/PM1O slopes ranged from 0.013 to 0.08. city-specific coefficient that is proportional the question ofwhether the time until death Table 4 shows the estimated effect of a to yi, plus possibly an effect of PM10 net of is only being reduced by a few days. One 10 pg/im3 increase in PM1O for all deaths, for confounding, that is, would expect people who are on the brink of deaths out ofhospitals, and for deaths in hos- + death to disproportionately die in hospitals pitals. PMIO was a significant predictor ofall- Oi N(ma &Y,1).° because many are in the hospital already. If cause mortality [0.67% increase for a 10 In the second stage, I estimate a using a air pollution primarily affected those people, pg/m3 increase in PM1O; 95% confidence weighted regression, with inverse variance I would expect its impact on hospital deaths interval (CI), 0.52-0.81%]. The effect size weights. I have added one further refinement to Table 1. Characteristics of the study locations. increase the power of the analysis. In most 1990 PM10 Dew Temperature Pressure cities, 03, CO, and SO2 show greater differ- City Population Deaths (mmHg) ences in their mean level between the indoor- (pg/m3) point (OF) heating season and the warm season than does New Haven 804,219 20.4 28.6 40.1 50.5 29.8 Birmingham 651,525 19.1 34.8 51.7 62.4 29.4 PM1O. This indicates that further variability Pittsburgh 1,336,449 63.3 36.4 41.2 52.1 28.8 in the slope between these pollutants and Detroit 2,111,687 59.7 36.9 40.7 50.9 29.3 PM1O can be obtained by dividing the data in Canton 367,585 9.9 29.31 41 50.4 28.7 each city into the indoor-heating season Chicago 5,105,067 133.4 36.5 39.8 50.3 29.3 (defined as November through April) when Minneapolis-St. Paul 1,518,196 32.3 27.5 35.5 46.3 29.1 CO and SO2 are but is and the Colorado Springs 397,014 6 27.1 28.9 48.9 24.0 high 03 low, Spokane 361,364 8.7 40.6 34.2 47.9 27.5 warm season, when the opposite is true. This Seattle 1,507,319 29.3 32.5 43.9 52.5 29.6 increases our ability to determine whether the PM1O effect size varies with the slope between PM1O and the other pollutants. To accom- Table 2. Demographic characteristics of study locations. plish this, the regressions were fit separately in Percent Percent with Percent below Percent each city in each ofthe two seasons. City unemployed college degree poverty level nonwhite Assessment ofeffect modification. To test New Haven 5.8 24.2 7.9 14 for effect modification, I used social and eco- Birmingham 6.5 19.9 16.0 36 nomic factors in the meta-regression instead Pittsburgh 6.3 22.6 11.5 12 Detroit 12.4 13.7 20.1 43 of the slopes between pairs of air pollutants. Canton 7.2 14.3 11.1 8 This tests for an term, where, for Chicago 8.0 22.8 14.2 37 example, the effect of air pollution increases Minneapolis-St. Paul 4.8 30.7 9.9 11 as the unemployment rate increases. Here Colorado Springs 7.3 25.8 10.4 14 our primary interest is in the coefficient of Spokane 7.3 20.6 13.7 5 the effect modifier, which tells how much Seattle 4.1 32.8 8.0 15

Environmental Health Perspectives * VOLUME 1081 NUMBER 6 1 June 2000 565 Articles * Schwartz was identical in the summer and winter peri- additional 5% of the population nonwhite. be associated with sudden death. A study of ods. However, the effects ofairborne partides These are substantial increases in each of the subjects with implanted cardiac defibrillators were substantially higher for deaths out ofthe postulated effect modifiers, but they are found an increased risk of ventricular tachy- hospital than for deaths in the hospital. associated with no noticeable change in the cardia and ventricular arrhythmia associated Table 4 also shows the results when estimated PM1O effect. with PM2.5 (20). Arrhythmia is one of the restricted to days with PM10 < 50 pg/m3. The major causes of sudden death. Arrhythmia slope of airborne particles was larger when Discussion and sudden death have also been produced in restricted to low air pollution days. These In an analysis of multiple cities across the rats by combustion partides (21) under exper- results are illustrated graphically in Figure 2. United States, PMIO was a significant pre- imental conditions where the responses Table 4 also shows the estimated effect of dictor of daily deaths. The association was cannot be attributed to cooccurring pollu- PM1O after controlling for potential con- identical in analyses restricted to the indoor- tants. This association is also supported by founding by S02, 03, and CO (i.e., the inter- heating season and the warm months. This is studies of electrocardiogram changes that are cept term in the regression of the baseline consistent with previously published results precursors to arrhythmia. Godleski et al (22 PM1O effect on the coefficient relating PMIO (18). Given the large differences in the con- reported an association between these electro- to each of the other pollutants). For all three centrations of cooccurring pollutants between cardiogram changes and exposure to concen- cooccurring pollutants, the effect size after the summer and winter months, this alone is trated air particles under experimental controlling for confounding was not substan- evidence that the partide associations cannot conditions in animals with preexisting illness- tially (or statistically significantly) different be primarily due to confounding with other es. Similar changes have been reported to be from the baseline result. This is illustrated in pollutants. associated with airborne partides in three epi- Figure 3. These results indicate that there is The association differed by location of demiologic studies using continuous electro- no trend to a higher PMIO coefficient in cities death, with a larger effect on deaths out of cardiogram monitoring in humans (23-25). or seasons with higher slopes between the hospitals. These results are consistent with Increases in heart rate have been associated cooccurring pollutants and PMO1 This is previous reports from Philadelphia (191) and with exposure to airborne partides in studies illustrated in Figure 4, which shows the effect with the experience in the great London in Baltimore, Maryland (25); Germany and size for PMIO in each city and season plotted smog episode of 1952 (17). This suggests Boston, Massachusetts (26); and Utah (27). against the 03/PMIO coefficient. that most of the PM10-associated deaths are Another major cause of sudden death is Figure 5 shows the estimated effect mod- not in people who are desperately ill and thrombotic processes leading to myocardial ification by different measures of social and hence that, in most cases, increased mortality infarctions. Here again, recent animal and economic status. It shows how much more is not a result of time of death simply being human studies indicate that airborne partides of an increase in daily deaths is associated reduced by a few days. may be affecting these processes. Exposure to with a 10 pg/m3 increase in PM1O if the city A higher risk of death out of the hospital combustion partides has been associated with had a 5% higher unemployment rate, an suggests that sudden death is a major compo- increased plasma fibrinogen in rats (28), and additional 5% of the population living nent of the air pollution-associated risk and, an episode of high particulate air pollution under the poverty level, an additional 5% of indeed, "dead on arrival" deaths were most was associated with increased plasma viscosity the population with college degrees, or an strongly associated with air pollution in the in a large epidemiology study (29). The Philadelphia analysis (16). Recently, more findings of the present study are therefore Table 3. Correlations between PM10 and weather mechanistic evidence has been developed that consistent with a growing body of more variables. supports the notion that airborne partides can mechanistic research in humans and animals. Temperature Pressure There was no trend of higher PMIO City (OF) Dew point (mmHg) effect sizes in towns with higher SO2/PM10 New Haven 0.05 -0.11 0.11 slopes, nor in towns with higher 03/PMIO Birmingham 0.26 0.19 0.19 Pittsburgh 0.45 0.44 0.44 Detroit 0.37 0.38 0.38 .ID 1.5 Canton 0.42 0.45 0.45 .E Chicago 0.36 0.32 0.32 Minneapolisa 0.29 0.26 0.26 Colorado Springs -0.34 -0.42 -0.42 2 Spokane -0.01 -0.16 -0.16 U0 a5 Seattle -0.22 -0.29 -0.29 ,E .5*1.16 SID .E aMinneapolis-St.Paul. S accC a Table 4. Estimated effect of a 10-pg/r3 increase in PMIO on daily deaths in the meta-analysis. Percent increase 0. CL Model in deaths 95% Cl c E 0 -i Overall 0.67 0.52-0.81 0 yv Summer only 0.67 0.48-0.86 Cab"Ory ° Winter only 0.66 0.45-0.87 0.0 Figure 2. Percent increase in daily deaths associ- Basic au2 Cu U3 In hospitals 0.49 0.31-0.68 ated with a 10-pg/m3 increase in PM10 from six Out of hospitals 0.89 0.67-1.10 separate analyses. Hosp, hospital. Results are Confounder Days < 50 pg/m3 0.87 0.62-1.12 shown for all deaths for summer and winter com- Figure 3. Effect of a 10-pg/m3 increase in all Confounding by bined (Sum/Win), summer only, winter only, deaths in the basic analysis and analysis using the SO2 0.57 0.25-0.90 deaths in hospitals, deaths out of hospitals, and intercept term from the meta-regression of the CO 0.90 0.42-0.97 all deaths, but restricted to days when PM10 was PM10 effect size in each city against the relation- 03 0.69 0.53-1.26 < 50 pg/mi3. ship between PM10 and S02' 03, and CO.

566 VOLUME 1081 NUMBER 61 June 2000 * Environmental Health Perspectives Articles * Ambient particles and daily mortality or CO/PM10 slopes. This indicates that the measure personal exposure to partides of all that was seen in an air pollution episode PM1O effects are not likely to be caused by sources, including resuspended house dust, where of the pollution effect is well confounding by other pollutants. These environmental tobacco smoke, and cooking accepted. That pattern, moreover, is consis- results address the issue ofwhether the PM10 aerosols. Hence, personal exposure to particles tent with recent animal and human data on effect is due to the other pollutants: they do of outdoor origin are more dosely related to the effects ofparticles on risk factors for sud- not address the question of whether those outdoor concentrations than some interpreta- den death. Finally, the public health benefit other pollutants have significant associations tions of personal monitoring data suggest. of each incremental reduction of 1 pg/m3 with daily deaths as well. This will be This has been confirmed by Janssen et al. appears to be higher at the lower air pollu- addressed in a later study. (36), who found correlations between tion levels that prevail on most days. This Recent animal studies, in which exposure personal partide monitors in adults and out- suggests that intervention strategies that can be controlled and limited to airborne par- door monitors were much higher after exclud- lower average levels, rather that those that tides, support the finding of an independent ing environmental tobacco smoke (ETS) address the few peak days, are the most particle effect. For example, Zelikoff et al. exposure. Janssen et al. (36) also highlighted appropriate. This is an important considera- (30) reported that exposure to concentrated another key issue. Most of the difference tion, as a number of cities (e.g., Mexico City, air partides after infection with streptococcal between personal PM exposure and outdoor Mexico; Athens, Greece) have adapted strate- pneumonia was associated with a doubling of concentrations reflects cross-sectional varia- gies that limit driving or industrial activity on the area of lung involvement and a doubling tions among persons. For time-series studies, peak pollution days. Such approaches do of the bacterial burden of rats within 48 hr. it is the longitudinal correlation that matters, lower average levels, but are costly and disrup- Effects of particle exposure on influenza mor- and Janssen et al. (36) reported considerably tive, and the same effort put into reducing tality have also been noted (31). higher longitudinal correlations between per- everyday emissions appears likely to produce The PM1O effect was not substantially sonal PM exposure and outdoor concentra- greater public health benefit. modified by socioeconomic status measured tions, with a median of 0.70 for PMIO in the at the city level, but when the analysis was absence of ETS exposure. Finally, two recent REFERENCES AND NOTES restricted to days with PM1O concentrations articles examined the statistical implications of < 50 pg/m3, the effect was greater. These the measurement error. Schwartz and Levin 1. Pope CA, Dockery DW, Schwartz J. Review of epidemio- logic evidence of health effects of particulate air pollu- results are inconsistent with a threshold for (37) pointed out that most of the difference tion. Inhal Toxicol 7:1-18 (1995). PM1O at any concentrations except those between personal and central measurements 2. Katsouyanni K, Touloumi G, Spix C, Schwartz J, Balducci substantially < 50 Indeed, they sug- of exposure in the time-series context are F, Medina S, Rossi G, Wojtyniak D, Sunyer J, Bacharova pg/m3. L, et al. Short term effects of ambient sulphur dioxide and gest that the PM1O slope increases at lower Berkson error, and hence do not bias the esti- particulate mater on mortality in 12 European cities: concentrations, rather than approaching mates. Zeger et al. (38) have explored the results from time series data from the APHEA project. Br zero. This tendency for a lower slope at high issue in more detail and have shown that the Med J 314:1658-1663 (1997). 3. Zeger SL, Dominici F, Samet J. Harvesting-resistant esti- concentrations has been noted in London remaining bias is negative-that is, an under- mates of air pollution effects on mortality. Epidemiology (32) and in the APHEA study (Air Pollution estimation of the effect. Hence, measurement 10:171-175 (1999). and Health: a European Approach) (33). A error in exposure is an unlikely cause of these 4. Schwartz J. Harvesting and long-term exposure effects in the relationship between air pollution and mortality. study of six U.S. cities recently reported a associations. Am J Epidemiol 151:440-448(2000). higher slope for PM2 5 when the analysis was In sum, this study provides evidence that 5. Sunyer J, Schwartz J, Tobias A, Macfarlane D, Garcia J, restricted to days < 30 pg/m3 (34). airborne particles influence the number of Anto JM. Patients with chronic obstructive pulmonary One limitation of studies such as these is deaths that these effects disease are at increased risk of death associated with daily and are not urban particle air pollution: a case-crossover analysis. the use of outdoor monitoring stations rather primarily attributable to other air pollutants. Am J Epidemiol 151:50-56(2000). than personal exposure monitors. Because the The data show the same pattern of higher 6. National Center for Health Statistics. Public Use Data difference between these measurements can relative effect on deaths out of the hospital Tape Documentation. Mortality Detail. Hyattsville, MD:National Center for Health Statistics, 1986-1993. be large, some have questioned whether the 7. U.S. Department of Commerce, Economics and Statistics associations reported in daily time-series stud- Administration, Bureau of the Census. 1990 Census of the ies could be causal. Several recent papers have Population and Housing. Washington:U.S. Government Printing Office, 1992. Available: http://www.census.gov/ addressed parts of this issue. Wilson and Suh main/www/cen1990.html [cited May 1999]. (35) pointed out that outdoor monitors are 005 8. Hastie T, Tibshirani R. Generalized Additive Models. surrogates for personal exposure to partides of 0.10 l London:Chapman and Hall, 1990. 9. Schwartz J. Air pollution and daily mortality in outdoor origin, such as motor vehicle exhaust A Birmingham, Alabama. Am J Epidemiol 137:1136-1147 and sulfates. Current personal monitors (1993). 10. Cleveland WS, Deviin SJ. Robust locally-weighted regression and smoothing scatterplots. J Am Stat Assoc 74:829-836 (1988). 11. Schwartz J. Generalized additive models in epidemiology. 1 1.0 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 -0.10 heart disease in eight U.S. counties. Epidemiology 10:17-22 11999). Poverty Unemployment College Nonwhite 13. Brumback BA, Ryan LM, Schwartz JD, Neas LM, Stark Effect modifier PC, Burge HA. Transitional regression models with appli- cation to environmental time series. J Am Stat Assoc 00.2 .0 0.2 0.4 0.6 0.8 1.0 Figure 5. Incremental effect (above the baseline) 449:16-28(2000). of 10 pg/m3 PM10 on all deaths associated with a 14. Akaike H. Information theory and an extension of the O?PM1o coeffleient 5% increase in the poverty rate, the unemploy- maximum likelihood principal. In: 2nd International Figure 4. Plot of the PM10 effect size in each sea- ment rate, the proportion of the population with a Symposium on Information Theory (Petrov BN, Csaki F, son in each city versus the regression coefficient college degree, and the proportion of the popula- eds). Budapest:Akademiai Kiado, 1973;267-281. relating 03 to PM10 in that season and city. tion that is nonwhite. 15. Bryk AS, Raudenbush SW. Hierarchical linear models:

Environmental Health Perspectives * VOLUME 1081 NUMBER 61 June 2000 567 Articles * Schwartz

applications and data analysis methods. Advanced Heart rate variability associated with particulate air pol- 31. Clarke RW, Hemenway DR, Frank R, Kleebarger SR, Quantitative Techniques in the Social Sciences Series. lution. Am Heart J 138:890-899 (1999). Longphre MV, Jakab GJ. Particle associated sulfate Vol 1. Newbury Park, CA:Sage Publications, 1992. 24. Liao D, Creason J, Shy C, Williams R, Watts R, Zweidinger exposure enhances murine influenza mortality 16. Witte JS, Greenland S, Bird CL, Haile RW. Hierarchical R. Daily variation of particulate air pollution and poor car- [Abstract]. Am J Respir Crit Care Med 155:A245 (1997). applied to a study of multiple dietary diac autonomic control in the elderly. Environ Health 32. Schwartz J, Marcus A. Mortality and air pollution in exposures and breast cancer. Epidemiology 5:612-621 Perspect 107:521-525 11999). London: a time series analysis. Am J Epidemiol 131:185-194 (1994). 25. Gold DR, Litonjua A, Schwartz J, Verrier M, Milstein R, (1990). 17. Mortality and Morbidity during the London Fog of Larson A, Lovett E, Verrier R. Ambient pollution and heart 33. Katsouyanni K, Touloumi G, Spix C, Schwartz J, Balducci December 1952. Rpt No 95 on Public Health and Medical rate variability. Circulation 101:1267-1273 (2000). F, Medina S, Rossi G, Wojtyniak B, Sunyer J, Bacharova Subjects. London:Her Majesty's Stationary Office, 1954. 26. Peters A, Perz S, Doring A, Stieber J, Koenig W, L, et al. Short term effects of ambient sulphur dioxide and 18. Kelsall JE, Samet JM, Zeger SL, Xu J. Air pollution and Wichmann HE. Increases in heart rate during an air pol- particulate matter on mortality in 12 European cities: mortality in Philadelphia: 1974-1988. Am J Epidemiol lution episode. Am J Epidemiol 150(10):1094-109811999). results from time series data from the APHEA project. Br 146:750-762 (1997). 27. Pope CA, Dockery DW, Kanner RE, Villegas GM, Schwartz Med J 314:1658-1663(1997). 19. Schwartz J. What are people dying of on high air pollu- J. Oxygen saturation, pulse rate, and particulate air pollu- 34. Schwartz J, Dockery DW, Neas LM. Is daily mortality tion days? Environ Res 64:26-35 (1994). tion: a daily time series panel study. Am J Respir Crit Care associated specifically with fine particles? J Air Waste 20. Peters A, Liu E, Verrier RL, Schwartz J, Gold DR, Med 159:365-372(1999). Manag Assoc 46:2-14 (1996). Mittleman M, Baliff J, Oh A, Allen G, Monahan K, et al. 28. Gardner SY, Costa DL. Particle-induced elevations in 35. Wilson WE, Suh H. Fine particles and coarse particles: Air pollution and incidences of cardiac arrhythmia. white blood cell count and plasma fibrinogen levels in rats concentration relationships relevant to epidemiologic Epidemiology 11:11-17 (2000). [Abstract]. Am J Respir Crit Care Med 157:A152 (1998). studies. J Air Waste Manag Assoc 47:1238-1249 (1997). 21. Watkinson WP, Campen MJ, Costa DL. Cardiac arrhyth- 29. Peters A, Doring A, Wichmann HE, Koenig W. Increased 36. Janssen NA, Hoek G, Brunekreef B, Harsseman H, mia induction after exposure to residual oil fly ash parti- plasma viscosity during an air pollution episode: a link to Mensink I, Zuidhof A. Personal sampling of PM10 in cles in a rodent model of pulmonary hypertension. mortality? Lancet 349:1582-1587 (1997). adults: relation between personal, indoor, and outdoor Toxicol Sci 41(2):209-216 (1998). 30. Zelikoff JT, Nadziejko C, Fang T, Gordon C, Premdass C, concentrations. Am J Epidemiol 147:537-547 (1998). 22. Godleski JJ, Lovett EG, Sioutas C, Killingsworth CR, Cohen MD. Short term, low-dose inhalation of ambient par- 37. Schwartz J, Levin R. Drinking water turbidity and health. Krishnamurthi GG, Hatch V, Wolfsom M, Ferguson ST, ticulate matter exacerbates ongoing pneumonococcal Epidemiology 10:86-90 (1999). Koutrakis P, Verrier RL. Impact of inhaled concentrated infections in Streptoccus pneumoniae-infected rats. In: 38. Zeger SL, Thomas D, Dominici F, Samet J, Schwartz J, ambient air particles on canine electrocardiographic pat- Proceedings of the Third Colloquium on Particulate Air Dockery D, Cohen A. Exposure measurement error in terns [Abstract]. Am J Respir Crit Care Med 157:A260 (1998). Pollution and Human Health (Phalen RF, Bell YM, eds). time-series studies of air pollution: concepts and conse- 23. Pope CA Ill, Verrier RL, Lovett EG, Larson AC, Raizenne Irvine, CA:University of California, Air Pollution Health quences. Environ Health Perspect 108:419-426 (2000). ME, Kanner RE, Schwartz J, Villegas GM, Dockery DW. Effects Laboratory, 1999;8-94to 8-101.

Maximize your exposure!

Place your position announcement in EHP and receive a free listing on the Environmental Health information Service at * o -visited nearly 8oo,ooo times a month! For moire infoirmration, call 919-541-5257.

568 VOLUME 1081 NUMBER 61 June 2000 * Environmental Health Perspectives