Environmental Engineering Reference
In-Depth Information
TABLE 23.16
Association between a 10  μ g/m 3  Increment in PM 10  (Average of Lag 0.1) a
and Percentage Increase in Total Daily Mortality
GAM b  with Default 
Convergence Criterion
GAM with Strict 
Convergence Criterion
GLM with 
Natural Spline
GLM with 
Penalized Spline
0.7% (0.6%-0.8%) c
0.7% (0.5%-0.7%)
0.4% (0.3%-0.6%)
0.6% A(0.4%-0.7%)
Source: Katsouyanni, K. et al., Sensitivity analysis of various models of short-term effects of ambient
particles on total mortality in 29 cities in APHEA2, in Special Report: Revised Analyses of
Time-Series Studies of Air Pollution and Health , Health Effects Institute, Ed, Health Effects
Institute, Boston, MA, pp. 157-164, 2003.
GAM, generalized additive model; GLM, generalized linear model.
a Not directly stated but surmised from Table 2 in Ref. [314].
b All results based on ixed-effects models. Results similar to random-effects models.
c Author's calculations from coeficients in Table 1 of Ref. [228].
(e.g., diabetes [study of hospital admissions; [230] ] congestive heart failure [231]). As noted previ-
ously, persons with underlying respiratory diseases seem to have an increased risk of cardiovas-
cular deaths.
The form of the exposure-response relationship between daily changes in PM aerosol and daily
mortality has been investigated extensively (e.g., see Refs [232-237]). In general, the data have
been more consistent with a no-threshold model than models with thresholds, although alternative
models with thresholds still remain a consideration (Figure 13.6 from Ref. [28]). The NMMAPS
study evaluated data from the 20 largest U.S. cities to assess the most likely exposure-response
relationships between daily changes in PM aerosol and daily mortality [238]. A summary analysis
for total mortality suggests that a no-threshold model is most consistent with the data for PM 10
(Figure 23.27). A more detailed analysis, based on the distribution of posterior probability of a
threshold, suggested that a threshold was not likely for cardiovascular mortality, but was consistent
for the data for “other” causes of death (Figure 23.28). In the case of all-cause and cardiovascular
mortality, any likely threshold was below the federal annual PM 10 annual 24 h standard of 50 μg/m 3
in force at the time over which the PM data were evaluated (Figure 23.28). In a commentary that
accompanied the publication of the NMMAPS analysis, Pope [237] reviewed the various factors
that could inluence the detection of threshold (statistical methods, publication bias, measurement
error) and presented a graphical summary of some of the more important time series studies that
have addressed exposure-response relationships (Figure 23.29). He concluded that the weight of
evidence “further indicates that assumptions or scientiic priors of no-effects threshold levels for
PM are not well supported by the empiric evidence” [237]. Other investigators, although in the
minority, raise the argument that measurement errors preclude accurate speciication of the expo-
sure-response relationship [239].
In the end, the overall health impact of the associations between daily changes in PM aerosol and
daily mortality depends upon the extent to which life expectancy is shortened by exposure [240]. If
the people who are dying were those whose deaths were advanced by only a few days (a phenom-
enon that has been termed “harvesting” or mortality displacement [241]), there is general agreement
(see Ref. [240] for an alternative view) that the daily increases in mortality with increases in daily
PM are not due simply to harvesting [241-245]. Schwartz carried out analyses based on smooth-
ing windows of 15, 30, 45, and 60 days and observed that the percentage increases in mortality
(1979-1986) per 10 μg/m 3 increase in PM 2.5 in Boston peaked at 15, 60, 60 days for COPD, pneu-
monia, and ischemic heart disease (IHD) deaths, respectively (Figure 23.30) [242]. COPD showed
 
Search WWH ::




Custom Search