Health Effects of Aerosols - Aerosols: Measurement, Dosimetry, and Health Effects - page 610

TABLE 23.15
Multicity Time Series Studies of the Effect of Daily Changes in PM Aerosol
on Daily Mortality

Study and References Locations and Pollutants Methods

The National Morbidity, Morbidity
and Air Pollution Study (NMMAPS)

Ninety largest cities in the
United States for 1987-1994

Three-stage regional model

Initial analyses : [307-309] Subanalysis of 20 of the

largest in this group [309]
PM 10 , O 3 , CO, SO 2 , NO

1. Within-city variability estimated with
log-linear semi-parametric model

Reanalyzes after discovery of the
GAM problem : a [238,310]

Meteorological data 2. Within-region variability of the true

regression coeficient with weighted
second-stage regression
3. Between-region variability in the true
regional regression coeficients
Heterogeneity across cities and regions
explored

Air Pollution and Health: a European
Approach (APHEA)
[228, pp. 311-315]

European cities Individual city analysis

APHEA1—15 cities, black
smoke, PM 10 , SO 2 , NO 2 , O 3

1. Standard procedures for Poisson
regression

APHEA2—29 cities, black
smoke, PM 10 , SO 2 , NO 2 , O 3

Between-city analyses
2. Assume city-speciic means normally
distributed around overall mean
3. Evaluate effect modiiers

a The initial analyses used the default convergence criteria in the GAM algorithm in S-Plus. These default criteria did not
lead to convergence and resulted in standard errors of parameter estimates that were too small. Subsequent reanalyzes were
carried out with more strict convergence criteria in S-Plus and with a generalized linear model approach with natural and
penalized splines. These issues are explored in detail in the references in the table and are summarized in Ref. [226].

[121,128]. A similar conclusion was reported in a Spokane, WA, study in which the contribution
of coarse PM was evaluated with an indicator for days with dust storms [126]. A meta-analysis
based on 19 U.S.-based studies came to the same conclusions [229]. In contrast, studies from
California, Mexico City, and Santiago, Chile, all reported that daily changes in coarse PM aerosol
had as large or larger effects on daily mortality than did ine PM aerosol (Table 23.17). The extent
of differences in the composition of coarse PM (concentrations of iron, bioaerosol components,
and a “tail” of combustion product PM that extends into the coarse range) were not evaluated by
any of these studies. Based on the available data, it does not appear appropriate to extrapolate the
relative contributions of ine and coarse PM to the associations with mortality from one location
to another.
A large study was undertaken in Germany to evaluate the relative strength of associations
between changes in daily mortality and changes in particle number in the ultraine range and mass
fractions [138]. The authors concluded that the associations were driven by particle number effects.
However, inspection of the data (Figure 23.26) does not provide a difference in the exposure-
response relationships between particle number, ine mass, and total PM 10 mass. A commentary by
the reviewers of the report concluded that the evidence did not favor one component over another
(begins on p. 93 of Ref. [138]).
Among the studies that have reported associations between daily changes in PM and daily
mortality, there has been a general consistency that the effects are greater in the elderly (usually
deined as age 65 years and older) and among individuals with selected underlying disease