International Journal of Epidemiology
©International Epidemiological Association 1996
Vol. 25, No. 6
Printed in Great Britain
Air Pollution and Mortality in
East Berlin during the Winters
of 1981-1989
SIBYLLE I RAHLENBECK* AND HERMANN KAHL*
Rahlenbeck S I (Gondar College of Medical Sciences, PO Box 196, Gondar, Ethiopia), Kahl H. Air pollution and mortality
in East Berlin during the winters of 1981-1989. International Journal of Epidemiology 1996; 25: 1220-1226.
Background. The relationship between air pollution and mortality in East Berlin was examined for the winters of
1981-1989.
Methods. Regression analysis included daily mean levels of sulphur dioxide (SO2) and suspended particulates (SP), and
was controlled for temperature, humidity, week-day, month, and year. Moving averages of previous pollution were also used.
Results. Each pollutant was a significant contributor to excess mortality. The strongest association was found for mortality lagged for 2 days, which depended significantly on the level of SP (p for In SP = 0.876; P = 0.008) and SO2 (p for
In SO2 = 0.635; P = 0.012), when regressed separately. When omitting days with pollutant concentrations above
150 |ig rrr 3 , the pollutant-mortality relationship was linear, and a 100 ng rrr 3 increase was associated with a 6.1% (SP)
and 4.5% (SO2) mortality increase 2 days later, when pollutants were considered separately; this was reduced to 4.6%
(SP) and 2.8% (SO2) increase, when both were considered simultaneously.
Conclusions. The results show that short-term associations between air pollutants and mortality in East Berlin did exist
during the winters 1981-1989. Since the coefficients for SP and SO2 dropped when controlling for the other pollutant
species, a similar strength of association with mortality for both pollutants was found.
Keywords: air pollutants, air pollution, mortality, particulates, sulphur dioxide
Several decades ago high levels of air pollutants were
associated with excess mortality in some populations,
for example in the Meuse Valley in 1930, Belgium,1
and in London 1952.2 Usually, sulphur dioxide (SO2)
and suspended particulates (SP) (US) and black smoke
(BS) (UK) have been used as air pollution indices. The
World Health Organization summarized in 19873 that
short-term mortality excesses might be expected when
ambient levels of particulates (measured as BS) and SO 2
exceed 500 )J.g m~3. However, several recent studies indicate that thresholds, if they exist, are much lower.4"15
Using time series techniques and controlling for
weather variables, epidemiologists have investigated
the classical data sets of London and New York.5-7-16"18
However, no consensus was reached as to which of these
primary pollutants was most responsible for causing
excess mortality. Because of their high collinearity it
is difficult to discriminate their relative contribution.
Furthermore, their physico-chemical behaviour and
toxicity is influenced by weather variables and the presence of other pollutants. Recently, secondary products
such as acid sulphates and sulphuric acid aerosols have
increasingly attracted attention, and it has been argued
that health effects might be related to these compounds
rather than SO2, total suspended particulates (TSP),
SP or BS. 718 " 20 Others, however, have found an independent relationship between SP and mortality,5'8'13'15
whereas SO2 seemed to be an indicator of general air
pollution.
Contrary to the worldwide trend of decreasing air
pollution levels, emissions and air pollution levels remained high over the final decades of the former German
Democratic Republic.13'21 Due to legal restrictions, the
data were previously unavailable for evaluation.
In the present investigation, we examined the association between daily concentrations of SO 2 and SP and
daily mortality in East Berlin, covering eight and a half
winters, from 1 October 1981 to 31 March 1988, as well
as from 1 October 1989 to 31 December 1989. Multivariate procedures were applied, controlling for
temperature, humidity, week-day, month, and year.
MATERIAL AND METHODS
East Berlin
East Berlin was situated at 53.3° North and 13.3° East
in a flat valley of the Spree river, 35 m above sea
level, covering an area of 403 km2; the population was
* Gondar College of Medical Sciences, P O Box 196, Gondar, Ethiopia.
* Senatsverwaltung fuer Stadtentwicklung und Umweltschutz,
Lentzeallee. Berlin, Germany.
1220
AIR POLLUTION AND MORTALITY IN EAST BERLIN
1 284 553 in 1989. About half of the inhabitants lived
in the centre (Berlin-Mitte) and the adjacent districts of
Prenzlauer Berg, Friedrichshain, and Lichtenberg, that
cover 58 km2 or 14% of the East Berlin area. The
climate was typically continental with an average daily
mean temperature of 5.5°C maximum and 0.0°C minimum for the winter half years. The prevailing wind
direction was west-southwest. Domestic coal burning
constituted an important emission source, and import of
air pollution occurred mainly from southwestern winds
from West-Berlin and the industrial areas southeast and
southwest of Berlin.
Mortality Data
Daily deaths in East Berlin were extracted from the
mortality records of the former East-German registry.
The data set only includes deaths of people aged over
one year. Deaths due to accidents and suicides (ICD-9
numbers > 800) were excluded, as were deaths which
occurred outside East Berlin.
Air Quality and Meteorological Data
Data for SO 2 and SP concentrations were collected
from one station (Parochialstrasse) in the city centre
of East Berlin. SO 2 was measured colometrically on
a 30-minute basis (CM5; Zentrum fuer Umweltgestaltung, Wittenberg). Daily mean values expressed
in |ig nT 3 were taken for the analysis. SP levels
were measured by p-absorption (FH62; Friesicke
and Hoepfner, Schweinfurt). Values taken represent daily means (|ig m"3) that were obtained from
hourly readings from October 1981-November 1988,
and from half-hour readings from December 1988
onwards.
Mean daily temperature (°C) and relative humidity
(%) were taken as weather variables. Data were
obtained from the central station in East Berlin,
Alexanderplatz.22 The measurements were made on a
3-hourly basis.
Data Analysis
Multiple regression models were used to study the
relationship between air pollution indices and daily
mortality, controlling for weather variables. Though
linear regression models assume normally distributed
errors with a constant variance, a criterion not met
when analysing count data, results from linear regression models do not differ from those obtained by Poisson
regression models. 23 ' 24 Models using mortality on the
same day, as well as daily mortality lagged for up to
5 days were specified. Differences of daily mortality
from its 15- and 29-day moving average were also used
as dependent variables. All models included dummy
1221
variables for week-day, month and year. Transformations of the independent pollution variables were used
as indicated by plots, obtained when grouping pollutant
values in increasing order of 50 adjacent values and
plotted against the relevant mortality means (Figure 1);
of these, the ones with the highest R2 and lowest Akaike
Criteria were selected for the model.5
In order to estimate cumulative pollutant effects,
models were also specified with multiple-day moving
averages of pollutant concentrations (up to 6 days, including and excluding the concurrent day). All models
were controlled for covariates; for temperature and
humidity, the moving averages of the same time range
as the pollutants were used.
Data analysis was carried out using the Statistical
Analysis System SAS, Cary, North Carolina, USA.
RESULTS
Daily numbers of deaths ranged from 17 to 64, and mean
number of deaths per day ranged from 36.0 during
winter 1988/1989 to 41.5 during winter 1981/1982
(Table 1). Daily mean concentrations of SO 2 ranged
from 2 to 997 ug m"3, and those of SP from 10 to
566 ng m~3; mean and median levels of SO 2 and SP are
shown in Table 1.
As expected, high positive correlations were observed between daily SO2 and SP (Pearson correlation
coefficient = 0.75), and negative ones for temperature
and SO 2 (Pearson correlation coefficient = -0.47), and
SP (Pearson correlation coefficient = -0.32), respectively.
When SO2 and SP levels were grouped in increasing
order for adjacent values and plotted against mean mortality, lagged for 2 days (Figure 1), the slope obtained
decreased at higher levels for both pollutants, thus
being compatible with a logarithmic transformation.
Since the natural log transformations gave the highest
R2 and lowest Akaike Criteria for each, SO 2 and SP
concentrations of previous days, they were used for further anaylsis. Since the actual pollutant concentrations
on the concurrent day, and—for SO 2 on the third
previous day—gave a marginally better fit, they were
used (Table 2). For temperature (not shown) the slope
was inverse and nearly linear, though slightly decreasing at both ends. As the majority of the values (>95%)
were in the linear part of the slope, no transformations
were made.
When these variables were regressed on daily
mortality, and mortality lagged for up to 5 days, SP was
a significant predictor of mortality lagged for 1-3 days,
while SO 2 was significant in models using mortality
lagged for I, 2 and 4 days (Table 2). The highest
estimates were observed for the model using mortality
1222
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
TABLE 1 Mean and standard deviation (SD) of daily mortality and daily temperature (° Celsius) and mean and median concentrations
(fig m3) of sulphur dioxide and suspended particulates in East Berlin, winters 1981-1989 (1 October-31 March)
Winter
Mortality
Mean (SD)
Temperature
Mean (SD)
SO2
Mean (median)
SP
Mean (median)
1981-1982
1982-1983
1983-1984
1984-1985
1985-1986
1986-1987
1987-1988
1988-1989
1989a
All
41.5(7.1)
38.7(7.1)
38.5 (6.9)
40.8 (7.2)
40.1 (7.9)
38.0 (6.4)
37.1 (6.3)
36.0 (6.3)
35.6 (6.8)
38.6 (7.0)
3.4 (5.6)
5.9 (4.4)
4.6 (4.8)
3.3 (6.6)
2.8 (6.3)
2.6 (6.7)
5.3(4.1)
6.0 (4.0)
7.1 (4.8)
4.4 (5.6)
184(130)
171 (144)
205 (157)
219(182)
157 (126)
173(125)
130(103)
107 (81)
172 (140)
166 (129)
118(93)
94 (87)
120(100)
116(104)
92 (88)
93 (74)
75(71)
73 (58)
91 (71)
97 (82)
' 1 October-31 December 1989.
44CO
X42-
x
•
•
x
•
JJ
Q
1
<
D
Z 38<
LLJ
•* * * w
•
x
•
•x
*
x"
36*•
340
x
SO2
• SP
!
,—
100 200 300 400 500 600 700
SO2 and SP (/ug/m3)
FIGURE 1 Daily deaths by mean daily
concentrations of sulphur dioxide
and suspended particulates (fig m~3),
grouped for 50 adjacent values of
pollutant concentrations, and corresponding mean daily deaths, 2 days
later, East Berlin, winters 1981-1989
lagged for 2 days. When including both pollutants
simultaneously into those models, where at least one
pollutant was a significant predictor in the singlepollutant model (mortality lagged for 1-4 days), the
estimates of both pollutants were on average reduced
by 40-45%, but there was considerable variation and
no consistent pattern for the various lag times (Table 2).
While SO 2 concentrations were more robust on days /_,
and t_i, SP concentrations were on days t_2 and t_3.
Table 3 presents the models with mortality lagged for
2 days, starting with ln(SO2) or ln(SP), and successively adding other variables. Controlling for week-day
(model 2) did not result in a considerable improvement
of fit, whereas inclusion of dummies for months
(model 3) and years (model 4) did. The addition of
temperature (model 6) caused a significant improvement, and reduced the coefficient of SO 2 and SP by
40%; thus temperature explains much of the mortality
variation attributed to SO 2 or SP in the previous models.
Table 4 shows the effect estimates after omission of
days with SO2 and/or SP levels > 150 (Xg m"3. In this
pollution range the actual concentrations gave a marginally better fit than the log-transformed ones; even in
this low range of pollution, each pollutant was significantly associated with excess mortality, being associated
with a 6.1% mortality increase per 100 (ig m~3 increase
in SP (2.28/37.5), and 4.5% per 100 (ig m"3 increase
in SO 2 (1.72/37.5) levels. After control for the other
pollutant, the SP estimates dropped by 25%, while that
for SO 2 was diminished by 40%; thus, based on the
coefficients in the two-pollutant model, an increase in
SP concentration of 100 u.g rrT3 would result in about
4.6% increase in mortality 2 days later, compared to a
2.8% mortality increase for a 100 fig m~3 increase in
SO 2 (1.71/37.5 and 1.04/37.5).
Possibly because of the low numbers of daily deaths,
first order autocorrelation was not significant (DurbanWatson statistic). Nevertheless, removal of possibly existing seasonal trends was sought by using differences
of mortality from its 15- and 29-day moving averages.
Of these, differences of mortality from its 29-day moving average decreased autocorrelation to nearly zero
(-0.003). In these models R2 decreased to 3% of its
original value, suggesting that, although some shortterm effects, for example epidemics, are controlled for,
an underestimation of other 'real' effects, for example
high pollution episodes, has also occurred.
AIR POLLUTION AND MORTALITY IN EAST BERLIN
1223
TABLE 2 Coefficients of mean daily concentrations of suphur dioxide and suspended particulates and/or their natural logarithm in
regression models with daily mortality as dependent variable, lagged for 1-5 days; all models are controlled for temperature, humidity,
week-day, month, and year; East Berlin, 1981-1989
Models including
single pollutant
Daily
mortality
P
so2
t
'+i*
'+3a
' • /
P
STE
0.0019
0.0034
0.500*
0.621
0.635*
0.876*
0.0028
0.819*
0.387
0.819*
0.510*
0.566
0.207
0.124
SP
ln(SO 2 )
ln(SP)
ln(SO 2 )
In(SP)
SO 2 C
In(SP)
ln(SO 2 )
ln(SP)
ln(SO 2 )
ln(SP)
ln(SO 2 )
ln(SP)
Models including
both pollutants
0.0015
0.0011
0.335
0.291
0.345
0.590
0.0008
0.694
-0.027
0.841
0.391
0.243
0.246
-0.079
0.0015
0.0030
0.255
0.330
0.254
0.332
0.0015
0.331
0.256
0.331
0.256
0.332
0.257
0.334
%
decrease1"
STE
0.0021
0.0044
0.331
0.429
0.329
0.427
0.0019
0.427
0.331
0.430
0.332
0.431
0.334
0.433
-21%
-68%
-33%
-53%
^16%
-33%
-71%
-15%
5100%
+3%
-23%
-57%
+ 19%
3=100%
a
Daily mortality lagged for 1,2,3 etc. days.
Per cent decrease in the effect estimate in the two-pollutant model compared to the single-pollutant model.
' SO2 gave a slightly better fit than ln(SO2).
* Significant at P < 0.05.
b
TABLE 3 Regression models with the natural logarithm of mean daily concentrations of sulphur dioxide (left side) and suspended
particulates (right side), with daily mortality (2 days lagged) as dependent variable, East Berlin, winters 1981-1989
No. of
predictors
Description
1
2
3
4
5
6
7
one pollutant
model 1+weekdays
model 2+months
model 3+years
model 4+humidity
model 5+temperat.
model 6+ln(SOj/SP)
1
7
12
20
21
22
23
ln(SO 2 )
ln(SP)
P
SEa
R2'b
1.97
2.10
1.54
1.09
1.04
0.64
0.34
0.21
0.22
0.23
0.25
0.25
0.25
0.33
0.058
0.062
0.096
0.135
0.140
0.165
0.165
"CO
Model
no.
SE°
R2'b
2.48
2.61
2.22
1.62
1.54
0.88
0.59
0.31
0.31
0.31
0.31
0.32
0.33
0.43
0.039
0.048
0.101
0.140
0.140
0.162
0.165
* SE = standard error of the coefficient.
b
Adjusted model R2.
c
Plus inclusion of the other pollutant.
When estimating cumulative effects of pollutants, the
of the pollutants was significantly associated with an
pollutant
increase in mortality. When including moving averages
concentrations (not transformed) of the previous 4 days
of both pollutants, the effect estimates for SP were
(not including the concurrent day) gave the best fit
reduced considerably to about 40% of their original
(Table 5). In models including only one pollutant, each
values, while those for SO 2 were reduced to only 70%.
model
using
the
moving
average of the
1224
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
TABLE 4 Coefficients of mean daily concentrations of SO2 and SP after exclusion of days with mean concentrations of sulphur dioxide
and/or suspended particulates exceeding 150 fig m3 (n = 866) in models with daily mortality, lagged for two.days, as dependent variable
Mean
concentration
(Mg m3)
SO2/_.
SPr,
85.2
66.3
Model including
both pollutants
Model including
one pollutant
0.0172*
0.0228*
STE
J
STE
0.0071
0.0083
0.0104
0.0171
0.0080
0.0094
* Siginificant at P < 0.05.
TABLE 5 Coefficients of moving averages of mean daily concentrations
mortality as dependent
variable
Moving
average 3
of sulphur dioxide and suspended particulates
Models including
single pollutant
SO 2
SP
SO 2
SP
SO 2
SP
in models with daily
Models including
both pollutants
P
STE
P
STE
0.0039*
0.0070*
0.0040*
0.0110*
0.0044*
0.0078*
0.0016
0.0033
0.0018
0.0035
0.0018
0.0037
0.0030
0.0025
0.0025
0.0041
0.0032
0.0032
0.0025
0.0049
0.0027
0.0054
0.0029
0.0056
a
Moving average of pollutant concentration of previous / days.
* Significant at P < 0.05.
DISCUSSION
A significant association was found between ambient
SO 2 and SP concentrations and daily number of deaths
on the ensuing 4 days in East Berlin during the winters
of 1981-1989. In single pollutant models, controlled for
covariates, the mortality effect was most pronounced on
the second day. Graphic analysis of the relationship
(Figure 1) revealed curvilinear functions with steeper
slopes at lower mean concentrations that level off with
increasing pollutant concentrations. This has been
reported previously, as has been the lack of an indication for the existence of a threshold. 5 " 91314 - 24
Previous studies have found, varying with the degree
of pollution observed and species measured, mortality
increases of about 4-8% per 100 (ig m~3 increase of
particulates, 5 ' 913 " 15 - 23 ' 24 and between 0.5 and 5% per
100 fig m"3 increase in SO 2 levels. 6 1 0 " 1 3 This is in
agreement with our results. The majority of previous
studies, however, which used the same or similar pollution variables, found one of the two to be the major
predictor of excess mortality, rendering the other one
insignificant when both were included simultaneously.
Most recent studies have reported particulates or BS as
having a stronger association with excess mortality than
SO 2 , for example in Detroit, Philadelphia, Steubenville,
Ohio, London, and Erfurt. 5 ' 6 ' 8 ' 913 Spix et al.u have
recently examined the association of daily air pollution
and mortality in Erfurt, which is located about 200 km
southwest of Berlin. SO 2 levels, available from 1980
until 1989, were in general, much higher than in Berlin,
with a winter median of 364 u.g m"3. Daily mortality in
Erfurt, lagged for 2 days, increased by 10% for an increase in SO2 from 23 to 929 ng m"3 (5th to 95 lh percentile); this is comparable to our results. Only for the
years 1989 and 1990 were SP data available in Erfurt
and the SP levels were comparable to those in Berlin.
During these 2 years, the effect of SP was stronger than
that of SO 2 , with an increase from 15 to 331 |ig m -3
being associated with a 22% increase in mortality on
the same day. Though SO 2 levels were high during these
AIR POLLUTION AND MORTALITY IN EAST BERLIN
two years (winter medians of 275 and 208 jig m 3 ), their
mortality effect was very small then, and reduced to
nearly zero, when SP was considered simultaneously,
while the effect of SP was only slightly decreased in the
two-pollutant model. Therefore, the authors conclude,
as have others who found similar results in other
locations,5>6>8>9 that the mortality effect is caused by
particluates, while SO2 acts as a proxy of another pollution agent or as a mere indicator for general air pollution. While our results agree with those from Erfurt and
other sites regarding the logarithmic nature of the relationship, the magnitude of the effects, and the higher
effect estimates for SP than for SO 2 , there are remarkable differences. Firstly, SP mortality effects were
lagged in Berlin, as has also been reported from other
locations,5'14'15'24 while they occurred on the concurrent
day in Erfurt. Secondly and more importantly, when
considering SO 2 and SP simultaneously in Berlin, SP
effect estimates were—on average—destabilized by the
same amount as were SO 2 estimates. Though, when
considering the day with the maximum effect, that is
after 2 (and 3) days, the SP mortality association was
robuster than the SO 2 one, the former was severely
diminished after control for SO2, when mortality was
lagged for 1 or 4 days. In models using moving averages
of pollutant concentrations over the previous 2-4 days,
when controlling for the other pollutant, mortality
associations were stronger with SO 2 than with SP.
Therefore, the strong and superior effect of SP over SO2
found in Erfurt and elsewhere, 5 ' 14 ' 15 was not seen in
Berlin during the winters 1981-1989. There is certainly
some collinearity problem here that makes it difficult to
discriminate their relative contribution. Furthermore,
the physico-chemical behaviour as well as toxicity at
a given site, are modified by temperature, humidity,
distribution of existing particle sizes, and the presence
of other pollutants.3'19 Several recent reports have emphasized the importance of acidic aerosols and acidic
particulates 16,18-20 as more plausible candidates for causing health effects than the pollutants measured routinely.
The notion that the observed associations in East
Berlin are probably confounded by other co-existing
pollutants cannot be excluded from the data. However,
causality of the relationships—at least in part—is
supported by studies relating these air pollutants to
cause-specific-mortality11 and morbidity. At ambient
concentrations of < 100 |ig m"3 for SO 2 and BS, significant associations between pollutant levels and emergency room admissions for chronic pulmonary diseases
have been reported from Barcelona, Spain.12 Similar
TSP levels were associated with increased incidence of
croup cases in German cities, 25 and hospital admissions
from respiratory conditions were associated with PMIO
1225
levels in Utah, USA.26 The relative contribution of
these, and the importance of other existing pollutants,
remains to be investigated in further studies.
Constraints of the present study include the imprecise exposure data, derived from only one sampling
point (the only one existing over the whole period),
the lack of information on other pollutants existing, and
finally, the use of a rather crude measure of health
effects. Despite these constraints however, significant
associations were found between SO2 and SP levels,
and mortality in East Berlin. These associations were
significant at concentrations well below air pollution
standards.
In summary, the association between air pollutants
and daily mortality in East Berlin is quantitatively
similar to results of other recent studies, though the
often reported stronger mortality association of particulates compared to SO 2 was not found here. It would
be of great value to investigate these issues in greater
detail, covering more pollutants, and their possible
interactions.
ACKNOWLEDGEMENTS
The authors thank Dr D Obrikat, Institut fuer
Strahlenschutz, Berlin, Dr Casper, and Dr W Mehnert,
Robert-Koch-Institut, Berlin for co-operation and Dr
J Stolwijk, Yale University for advice. Thanks are
extended to Dr J Hanley, McGill University and
Dr Eshetu, Dept. of Statistics, Addis Ababa University
for computational help, Dr R Waller, Dept. of Health,
London, Dr K Katsouyanni, Athens, K Salmon, Belfast
and Dr Prinz, Marburg for valuable comments.
REFERENCES
' Firket J. Fog along the Meuse Valley. Trans Faraday Soc 1936;
32: 1192-97.
2
Ministry of Health, UK. Mortality and Morbidity during
the London Fog of December 1952. London: HMSO, 1954,
Reports on Public Health and Medical Subjects No. 95.
3
World Health Organization. Air Quality Guidelines for Europe.
Copenhagen: WHO Regional Office for Europe, 1987.
WHO Regional Publications, European Series No.23.
4
Hatzakis A, Katsouyanni K, Kalandidi A, Day N, Trichopoulos
D. Short-term effects of air pollution on mortality in
Athens. Int J Epidemiol 1986; IS: 73-81.
5
Schwartz J, Marcus A. Mortality and air pollution in London:
a time series analysis. Am J Epidemiol 1990; 131:
185-94.
6
Schwartz J, Dockery D. Increased mortality in Philadelphia
associated with daily pollution concentrations. Am Rev
Resp Dis 1992; 145: 600-04.
7
Ostro B. A search for a threshold in the relationship of air pollution to mortality: a reanalysis of data on London winters.
Environ Health Perspect 1984; 8: 397-99.
1226
8
INTERNATIONAL JOURNAL OF EPIDEMIOLOGY
Schwartz J. Paniculate air pollution and daily mortality in
Detroit. Environ Res 1991; 56: 204-13.
9
Schwartz J, Dockery D W. Particluate air pollution and daily
mortality in Steubenville, Ohio. Am J Epidemiol 1992;
135: 12-22.
10
Derriennic F, Richardson S, Moule A, Lelouch J. Short-term
effects of sulphur dioxide pollution on mortality in two
French cities. Inl J Epidemiol 1989; 18: 186-97.
" Katsouyanni K, Karakatsani A, Messari I, Toulomi D, Hatzakis
A, Kalandidi A, Trichopoulos D. Air pollution and causespecific mortality in Athens. J Epidemiol Community
Health 1990; 44: 321-24.
12
Sunyer J, Saez M, Murillo C, Castellsague J, Martinez F, Anto J.
Air pollution and emergency room admissions for chronic
obstructive pulmonary disease: a 5-year study. Am J
Epidemiol 1993; 137: 701-05.
13
Spix C, Heinrich J, Dockery D el al. Air pollution and daily
mortality in Erfurt, East Germany, 1980-1989. Environ
Health Perspect 1993; 101: 518-26.
14
Schwartz J. Air pollution and daily mortality in Birmingham,
Alabama. Am J Epidemiol 1993; 137: 1136-47.
15
Pope A C, Schwartz J, Ransom M R. Daily mortality and PMIO
pollution in Utah Valley. Arch Environ Health 1992; 47:
211-17.
16
Ito K, Thurston G D, Hayes C, Lippmann M. Associations of
London, England, daily mortality with paniculate matter,
sulphur dioxide and acidic aerosol pollution. Arch Environ
Health 1993; 48: 213-20.
17
Thurston G D, Ito K, Lippmann M, Hayes C. Reexamination of
London, England mortality in relation to exposure to acidic
aerosols during 1963-1972 winters. Environ Health
Perspect 1989; 79: 73-89.
18
Ozkaynak H, Spengler J D. Analysis of health effects resulting
from population exposures to acidic precipitation precursors. Environ Health Perspect 1985; 63: 45-55.
"Larson T V. The influence of chemical and physical forms
of ambient air acids on airway doses. Environ Health
Perspect 1989; 79: 7-13.
20
Speizer F E. Studies on acidic aerosols in six cities in a new
multi-city investigation. Environ Health Perspect 1989;
79: 61-67.
21
Thriene B. Die lufthygienische Belastung in der DDR—
Situation und Erfordernisse (Ambient air pollution burden
in the GDR—situation and needs). Off Cesundheitswes
1991; 51: 54-60.
22
Meteorologischer Dienst der DDR. Taglicher Wetterbericht des
Meteorologischen Dienstes der DDR (Daily weather report
of the meteorologic agency of the GDR). Potsdam
1981-1990.
23
Kinney P L, Ito K, Thurston G D. A sensitivity analysis of
mortality/PM|0 associations in Los Angeles, lnhal Toxicol
1995; 7: 59-69.
24
Schwartz J. Air pollution and daily mortality: a review and
meta-analysis. Environ Res 1994; 64: 36-52.
25
Schwartz J, Spix C, Wichmann H E, Malin E. Air pollution and
acute respiratory illness in five German communities.
Environ Res 1991; 56: 1-14.
26
Pope C A. Respiratory disease associated with PM |0 pollution in
Utah, Salt Lake, and Cache Counties. Arch Environ Health
1991; 46: 90-97.
(Revised version received May 1996)