Health Effects of Aerosols - Aerosols: Measurement, Dosimetry, and Health Effects - page 609

TABLE 23.14 (continued)
Representative Sample of Time Series Studies that Relate Daily Changes in Measures of
PM Aerosol to Daily Changes in Non-Accidental Mortality

General Population-Based Time Series Studies
Study
Population

Years Pollutants a Methods Results References

Time series studies of association between exposure to mobile sources and daily mortality
Amsterdam,
the
Netherlands

1987-1998 Trafic counts,

black smoke (BS),
PM 10 , SO 2 , CO,
NO 2 , NO, O 3

PR, GAM Higher pollutant levels at

trafic-inluenced monitoring
sites. Effects greatest when
trafic-inluenced site data are
applied to persons who lived on
roads with >10 4 vehicles/day;
most consistent for BS
RR/100 μg/m 3 BS (lag 1), 1.89
(1.20, 2.95)

[287]

TS, time series; PR, Poisson regression; GEE, generalized estimating equations; GAM, generalized additive models; OLS,
ordinary least-squares regression; CLR, conditional logistic regression; COH, coeficient of haze; RR, relative risk;
IQR, interquartile (25th-75th) range; COPD, chronic obstructive lung disease; CVD, cardiovascular diseases.
a All studies include some variables to control for meteorological effects, season, and, for some, day of the week effects.
SO 2 , CO, and NO 2 are noted since they often are surrogates for ixed (SO 2 ) or mobile (CO, NO 2 ) source PM aerosol. O 3
is included, since in some areas (e.g., eastern United States) photochemical smog is highly correlated with ambient PM
aerosol levels.
b This is a follow-up study to that in Ref. [301] and uses a longer time series and adds data on PM 2.5 to those of PM 10 .
c Ninety-ive percent CI.
d This study is included since pollutant levels are quite low. Maximum PM 10 = 33.9 μg/m 3 , 90th percentile = 22.8 μg/m 3 .
e Case-crossover designs are matched designs in which each subject serves as his or her own control. Unlike the population-
based time series studies, case-crossover designs provide control for individual covariates and, with proper sampling,
control of temporal confounding [222,304].

pollutants in the statistical models (Figure 23.24, bottom panel). There was evidence for geographic
heterogeneity, with effect estimates being greatest in the northeast and least in the northwest (Figure
23.25a). City-speciic estimates for all-cause and cardiorespiratory deaths showed heterogeneity;
however, within regions, there seems to be relative homogeneity (Figure 23.25b—note circled areas
for southern California and the northeast). This heterogeneity likely represents a combination of
differences in the ambient pollutant mixtures between regions as well as demographic differences.
In the original analyses, the percentage of adults without a high school diploma in each city had an
independent effect on daily mortality. This was not included in the reanalysis. The APHEA project
reported percentage increases in total daily mortality of approximately 0.7% per 10 μg/m 3 increase
in PM 10 (average of lag 0, 1—see footnote a in Table 23.16) for 21 cities where such data were avail-
able (Table 23.16) [228]. The results were relatively robust to the types of modeling strategy chosen.
These results are somewhat lower than those expected with a similar average of lags for NMMAPS
(Figure 23.24, top panel).
A number of studies have attempted to identify speciic components of mass or particle number
as the source of increased risk for daily mortality (Table 23.17). Two separate reports from the
Harvard Six Cities study found that only ine PM was associated with increases in daily mortality;
no association was observed for coarse PM or a “crustal” component based on source attribution