SHORTTERM EFFECTS OF AIR POLLUTION: RESULTS FROM EPIDEMIOLOGICAL STUDIES

1
SHORT-TERM EFFECTS OF AIR POLLUTION RESULTS
FROM EPIDEMIOLOGICAL STUDIES

• Klea Katsouyanni
• Department of Hygiene and Epidemiology
• University of Athens Medical School
• 2 August 2007

2
Background

• The effects of air pollution on health were
  recognized after severe air pollution episodes in
  Northern Europe and North America between 1900
  and 1965, to which thousands of deaths have been
  attributed.

• The awareness of the consequences led to measures
  which contributed to a substantial decrease of PM
  and other characteristic gaseous pollutants
  concentrations.

3
Background

• From about 1970 to 1990, the prevailing opinion
  among scientists and decision makers was that
  current air pollution levels did not have
  important adverse health effects.

• Since roughly 1990, it became evident that the
  current, relatively lower, air pollution levels
  (mainly ambient particles) had adverse,
  short-term and long-term, health effects
  including an increase in mortality.

4

• THESE RESULTS HAD AN IMPACT ON SETTING GUIDELINES
  AND STANDARDS
• On the U.S. Environmental Protection Agency (EPA)
• The European Union
• The World Health Organisation (WHO)

5
THE EFFECTS OF AIR POLLUTION ON HEALTH ARE OFTEN
CONVENIENTLY CLASSIFIED

• In short-term and long-term effects
• although there is probably a continuum of
  effects in the time scale, which are not yet
  fully understood.

6
What is meant by short-term in this
presentation?

• The effects manifested in the same or the next
  few days (say, up to a week) after a specific
  exposure to an air pollutant or a mixture of air
  pollutants.
• The effects over a short period after exposure,
  say 30-40 days.

7
Usual study designs for the investigation of
short-term effects

• Time series studies (aggregated data usually
  daily based on routinely collected information
  long time series no confounding by individual
  characteristics potential confounding by
  variables which vary on daily basis pollution
  measurements often by fixed monitors)
• Panel studies (cohort followed intensively for
  relatively short time individual data usually
  daily no confounding by individual
  characteristics potential confounding by
  variables which vary on daily basis analysis may
  be done with aggregated or with individual data
  pollution measurements may be individualized)

8
The short-term effects were the first observed
Excess deaths in 52-53 compared with 51-52(From
Bell and Davis, EHP 2001 109 389-394 and London
Smog presentation)
9
Relevant health outcomes (from the WHO ECEH
Report, Quantification of the Health effects of
exposure to air pollution, 2001)
Emergency room visits Visits to
doctor Restricted activity Medication
use Symptoms Impaired pulmonary
function Sub clinical (subtle) effects
Severity of effects
Premature mortality
Hospital admissions
Proportion of population affected
10
HOW DO WE MEASURE PM AND GASES IN THE AIR??

• Black smoke (black particles with mean
  aerodynamic diameter lt4µm) assessed by
  reflectometry. Traditionally used in Europe and
  included in E.C. Directives until 1999. The
  reflectometry units are transformed to µg/m3
  using a calibration curve. BS was used in WHO AQG
  1987
• PM10 (Particles with mean aerodynamic diameter
  lt10µm) measured in µg/m3. Adopted as the main
  particle indicator in the U.S. since 1979 and in
  the E.C. since 1999. Also used in the WHO AQG
  2000 2005.
• PM2.5 (Particles with mean aerodynamic diameter lt
  2.5µm, fine particles). Indicator used for the
  U.S. standards together with ??10.
• Coarse fraction (Particles with mean aerodynamic
  diameter lt10 and gt2.5µm)/ (Ultrafines lt0.1 µm)
• Ozone (usually 1h or 8h in µg/m3 or ppm, ppb )
• NO2 (usually 1h or 24h in µg/m3 or ppm, ppb )
• CO (usually 8h in mg/m3 or ppm, ppb )
• SO2 (usually 24h, also very short-term eg 10 in
  µg/m3 or ppm,ppb)

11
Current limits and guidelines for ambient PM

• ?.C. PM10 (24hour) 50µg/m3 and annual 40µg/m3 and
  20µg/m3 with target years 2005 and 2010
  respectively.
• U.S. E.P.A. PM10 (24hour) 150µg/m3 and annual
  50µg/m3. PM2.5 (24hour) 65µg/m3 and annual
  15µg/m3.

12

• During the late 80s and early 90s several
  time-series studies produced evidence of
  short-term effects at relatively low levels of
  pollution
• Their results were put in a broader context and
  were consolidated with the initiation of large
  multi-city studies in Europe and the U.S.

13

• In 1993 the multi-city APHEA AIR POLLUTION AND
  HEALTH a EUROPEAN APPROACH PROJECT was
  initiated
• It included data from up to 30 European cities
  spanning across the continent
• A few key members of the APHEA group Giota
  Touloumi, Evi Samoli, Alain Le Tertre, Richard
  Atkinson, Antonis Analitis, Alexandros Gryparis,
  Ross Anderson

14
Short-term effects of PM on health. Results from
the multi-centre European project Air Pollution
and Health a European Approach (APHEA2)
15
Percent increase in total mortality and 95CIs
associated with an increase of 10µg/m3 in PM10
using loess (upper) and p-splines (lower)
16
Short-term effects of PM10 on health. Results
from the multi-city U.S. (H.E.I funded) project
National Mortality, Morbidity and Air Pollution
Study (NMMAPS)From Samet et al 2000 NEJM 343
1742-9
17
Short-term effects of PM pollution on hospital
respiratory admissions. Results from the
multi-centre European project APHEA based on 8
cities (Atkinson et al, AJRCCM 2002 1641860-6)
18
Short-term effects of PM pollution on hospital
cardiovascular admissions. Results from the
multi-centre European project APHEA based on 8
cities (Le Tertre et al, JECH 2002 56773-9).
19
Short-term effects of PM10 pollution on hospital
admissions. Results from the multi-city U.S.
(H.E.I funded) project National Mortality,
Morbidity and Air Pollution Study (NMMAPS)From
Samet et al, NMMAPS Report, 2000,
www.healtheffects.org
20
APHEA project Shape of the association of total
mortality with PM10 over 6 days (lags 0 to 5)
combined for all cities using a third order
polynomial distributed lag model
Loess
Penalized Splines
21
Hypothetical lag structure corresponding to a
harvesting effect and the estimated shape of the
association of PM10 and daily deaths using a
fourth degree distributed lag in ten APHEA2
cities (Zanobetti et al, Epidemiology 2002
1387-93)
22

• Investigation of effect modification

23
Percent increase in the daily number of deaths
associated with an increase of 10µg/m3 in PM10
concentrations, by levels of important effect
modifiers (?????2, Epidemiology, 2001 12 521-31)
24

• A similar effect modification pattern was
  observed for the effects of BS and PM10 on
  cardiovascular mortality, but not on respiratory
  mortality.
• For respiratory mortality, estimates of PM10 were
  not heterogeneous, whilst those of BS were
  heterogeneous, but not modified by NO2 levels

25
Percent increase in the daily number of deaths
associated with an increase of 10µg/m3 in PM10 or
BS concentrations, by geographical area (APHEA2,
Epidemiology, 2001 2006)
26
(No Transcript)
27
(No Transcript)
28

• The effect estimates for respiratory or cardiac
  admissions were not modified by the NO2 levels.
• There was positive effect modification for
  respiratory admissions among the elderly and
  negative effect modification for IHD admissions
  among the elderly by long term ozone
  concentrations.

29
Sensitive subgroups MI survivorsHospital
cardiac readmissions from 5 cities (HEAPSS
study Circulation 2005, 1123073-9)
30
Short-term effects of ozone on health. Results
from the multi-centre European project Air
Pollution and Health a European Approach
(APHEA2)
31
Shape of the association of respiratory mortality
with O3 8-hour over 6 days (lags 0 to 5) combined
for all cities using a third order polynomial
distributed lag model.
Results from Random Effects
Loess
increase
Results from Random Effects
Day lag
increase
P-Splines
Day lag
32
HEI Special Report, 2003 Revised analysis of
time-series studies of air pollution and health.
Re-analyses of the NMMAPS study

• Dominici et al , NMMAPS mortality, 90 cities (all
  numbers are increases in daily number of events
  per 10µg/m3 increase in PM10)
• GAM-D 0.41 (0.05)
• GAM-S 0.27 (0.05)
• GLM-ns 0.21 (0.06)
• Change of about 50
• Schwartz et al, NMMAPS mortality, 10 cities
  distributed lag model results
• GAM-D 1.3 (1.0-1.5)
• GAM-S 1.1 (0.8-1.4)
• GLM-ns 1.0 (0.7-1.3)
• Penalized splines 1.0 (0.8-1.3)
• Schwartz et al, NMMAPS admissions 14 cities
• Changes of 8-10 for CVD and COPD
• Larger changes for pneumonia admissions

33
GAM problems Sensitivity of APHEA2 results for
PM10 and COPD and asthma admissions in adults
over the age of 65 years. GAM-D, GAM-S, NS models
(from the HEI Special Report, 2003).
34
HEI Special Report Revised analysis of
time-series studies of air pollution and
health.Overview of results

• Comparison of GAM-D and GAM-S
• Change of gt40 2 studies (Canada and USA)
• Change of 10-40 6 studies (Canada and USA)
• Change of lt10 16 studies (7 European)
• Comparison of GAM-S and GLM-ns
• Change of gt40 5 studies (1 European)
• Change of 10-40 9 studies (3 European)
• Change of lt10 12 studies (4 European)
• Some studies reported models with different df
  and concluded that effect estimates are sensitive

35
Europe Canada
U.S.A.
36
Exposure misclassificationIn most
epidemiological studies measurements from fixed
site monitors have been used. How do these
represent the population or individual exposure?

• There is some evidence that personal exposure to
  PM is correlated over time with measurements from
  ambient monitors (Janssen et al Am J Epidemiol
  1998 147 537-47).
• Zeger at al (Env Health Perspect 2000 108
  419-26) have provided a conceptual framework for
  exposure misclassification in air pollution
  studies. Limited application indicated that the
  use of measurements from fixed site monitors led
  to underestimation of the effects of PM10.
• Sarnat et al (EHP 2001 109 1053-61) in a study
  in Baltimore showed that ambient gaseous
  pollutant measurements were not correlated with
  personal exposures for the same pollutant, but
  were correlated with personal PM2.5 exposures. If
  this is right, then ambient gaseous pollutant
  measurements are only a surrogate to PM
  exposures!
• Georgoulis et al (Atm Env 2002 36963-74) in the
  EXPOLIS study, has shown that the most consistent
  and significant determinant of personal CO
  exposure is the ambient level.
• There is clearly a need for further research.

37

• Results from Panel studies

38
Adjusted ORs between daily source-specific PM2.5
concentrations and occurrence of ST-segment
depression (ULTRA study, Lanki et al 2006EHP
114655-60)
39
Adjusted (also mutually) ORs between indicator
elements of PM2.5 sources and occurrence of
ST-segment depression (ULTRA study, Lanki et al
2006 EHP 114655-60)
40
Sensitive subgroups patients with pre-existing
heart or pulmonary disease?

• In a panel of 29 patients with COPD or asthma or
  IHD (Lagorio et al, Environ Health 2006, 511) ,
  lung function decrements in relation to various
  pollutants were investigated. Among COPD
  patients, lung function decreased with increasing
  PM2.5 and NO2. In asthma patients there were
  effects of NO2. No effects of the coarse fraction
  were observed. No association was observed in IHD
  patients.
• In a panel of 30 patients with COPD or recent MI
  (Wheeler et al, EHP 2006, 114560), increasing
  PM2.5 concentrations had an effect on heart rate
  variability (HRV) of the COPD patients whilst NO2
  concentrations affected the HRV in both groups of
  patients.

41
Conclusions (1)

• The current levels of ambient particles in Europe
  and North America have short term health effects
  which include an increase in daily mortality and
  hospital admissions for specific causes.
• The effects are not due to harvesting.
• Effect modifiers have been identified.
• Sensitive subgroups appear to be those with
  pre-existing cardiac or pulmonary disease and the
  elderly.

42
Conclusions (2)

• The exposure response curve at current pollutant
  concentrations, appears linear and compatible
  with a no-threshold model.
• The results are very consistent across method,
  space and time, but concern small relative risks.
  
• However, if the ubiquity of exposure and the
  existence of sensitive sub-groups are taken into
  account, it is shown that they constitute an
  important public health problem.

43
Conclusions (3)

• Indirect evidence from large studies shows that
  air pollution originating from traffic plays the
  most important role in the observed short-term
  health effects.
• Efforts to estimate or measure indicators of
  pollution by source re-enforce the above
  evidence.
• The appears to be heterogeneity in the responses
  of individuals according to their health status,
  with reference to the most harmful pollutants and
  to the relevant outcomes. However, the role of
  traffic pollution appears as a common
  denominator.