1

Traffic- Related Air Pollution:A Critical Review of the Literature on Emissions,

Exposure, and Health Effects

Clean Air Act Advisory CommitteeCrystal City, VA

October 6 – 7, 2009

Dan Greenbaum, PresidentHealth Effects Institute

Goals of the Review

Summarize and synthesize relevant information on air pollution from traffic and its health effects, linking:• Emissions and exposure to traffic air pollution• Exposure to traffic air pollution and health effects• Toxicological data and epidemiologic associations

A preprint of the report was released in May 2009The final Report, following extensive QA/QC, will be published in fall 2009

2

HEI Traffic Review PanelIra Tager—UC Berkeley, ChairKenneth Demerjian—SUNY, AlbanyMark Frampton—U Rochester Michael Jerrett—UC Berkeley Frank Kelly—King’s College Lester Kobzik—Harvard SPHNino Künzli—IMdM, BarcelonaBrian Leaderer—Yale SPHThomas Lumley—U of Washington SPH Frederick Lurmann—Sonoma Tech. Inc Sylvia Richardson—Imperial CollegeJon Samet—Johns HopkinsMichael Walsh—Consultant

Emissions from Motor VehiclesThe Current Context

Significant progress has been made in reduction of pollutant emissions from motor vehicles despite increases in number of vehicles and vehicle miles traveled

Increased urbanization and urban populations have:

• Increased dependence on motor vehicles and traffic congestion

• Changed land use patterns such that more people are near traffic sources of pollution

3

Report Organization

There are many studies (over 400) that have attempted to look at traffic exposure and effects

•However, they are not all of equal quality

4

1. How should we assess Exposure?

• Who is likely to be exposed?• What exposure assessment methods used in

epidemiologic studies?• Pollutant surrogates for traffic exposures (e.g., NO2, EC/BS,

CO, UFPM, benzene, etc.)• Distance from and/or length of roadways• Estimate of traffic density or intensity• Modeling of primary traffic-generated pollutant exposure

VOC (TraceAir) Distance Decay Around Highway 401, Toronto

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

-1000 -800 -600 -400 -200 0 200 400 600

Distance from Expressway (m)

Stan

dard

ized

VO

CLe

vels

Benzene (TraceAir)

THC (TraceAir)

NO2 (Ogawa)

Who is Likely to be Exposed?Highest levels within 300 – 500 meters of a major road

Toronto, Beckerman et al. (2008)

5

Near Roadway Exposure Can Include Large Populations Toronto Example: ~45%

(within 500 meters of an expressway; 100 meters of a major road)

Los Angeles Example: (~44%)

6

Portland, Oregon ~37%

What Markers or Surrogates?

• Pollutant surrogates for traffic exposures (e.g., NO2, EC/BS, CO, UFPM, benzene, etc.)

• Criteria for what is a good surrogate:1. Traffic as the major source2. Emissions vary with other motor vehicle constituents3. Can be measured at low concentrations by reasonably

inexpensive and accurate methods4. Not have independent health effects

7

NO2 as a surrogate

•There is substantial variability in average concentrations by locations.

•NO2 is a potential surrogate for vehicle emissions if it is measured on a fine spatial resolution.

PM2.5 as a Surrogate•Use of PM2.5 as a surrogate is of limited value because many sources contribute to urban PM2.5 and PM2.5concentrations are well mixed within a region

•Current central monitors do not provide sufficient spatial resolution for assessing the contribution of traffic to ambient PM2.5

8

What Markers or Surrogates?

• Pollutant surrogates for traffic exposures (e.g., NO2, EC/BS, CO, UFPM, benzene, etc.)

• Criteria for what is a good surrogate:1. Traffic as the major source2. Emissions vary with other motor vehicle constituents3. Can be measured at low concentrations by reasonably

inexpensive and accurate methods4. Not have independent health effects

• Can provide useful information but none meet all these criteria…

Can We Use Exposure Models?Models used• Proximity models• Geostatistical

interpolation models• Dispersion models• Land-use regression

models• Hybrid models

• Combine a model with time-activity data, or personal/microenvironmental monitoring

• Proximity models are least effective:• Can be confounded by

Socioeconomic Status, Noise, other factors

• Newer models of exposure are better• But should be validated

against some real-world data.

9

Criteria for Inclusion of Toxicology and Epidemiology Studies

• Quality of exposure assessment was key…• Studies had to include 1 or more of the following

exposure methods:• Distance from and/or length of roadways• Estimate of traffic density or intensity• Modeling of primary traffic-generated pollutant exposure• Studies of occupations characterized by exposure to traffic• Pollutant surrogates for traffic exposures (e.g., NO2, EC/BS,

CO, benzene, etc.) only if data provided to validate the pollutant as a reasonably specific surrogate for such exposure

2. What Can We Learn from Toxicology?(Example from a somewhat limited database):

Effects of Traffic Exposure on Asthmatics (Zhang HEI 2009)Lung function decline in asthmatics comparing Hyde Park and Oxford Street, London

(although symptoms did not increase…)

10

3. What can we learn from epidemiology?Criteria for Causal Inference

Four categories to test whether traffic causes effects, based on:- how well studies controlled for confounding- consistency of the findings with other studies- quality of the method to estimate exposure

• Sufficient evidence• Suggestive but not sufficient• Inadequate and insufficient evidence• Suggestive of no association

Epidemiology Health Outcomes Evaluated

• Mortality (all cause, cardiopulmonary)• Cardiovascular morbidity• Respiratory disease

• Asthma—childhood/adult• General respiratory symptoms• Lung function-childhood/adult/COPD• Health care utilization

• Non-asthmatic allergy• Birth Outcomes• Cancer

11

1.13 1.161.49

0.87 0.94 1.09 1.09

1.72 1.74

1.13

1.6

2.94

1.58

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

5

per 1

0.3

ug/m

3 in

NO

2 - 2

nd y

ear o

f life

per 1

0.6

ug/m

3 in

NO

2 - 4

th y

ear o

f life

per 1

0.6

ug/m

3 in

NO

2 - 4

th y

ear o

f life

(Alle

rgic

Chi

ldre

n)

per 8

.5 u

g/m

3 N

O2

- 1st

yea

r of l

ife

per 8

.5 u

g/m

3 N

O2

- 2nd

yea

r of l

ife

per I

QR

NO

2 - 1

st y

ear o

f life

per I

QR

NO

2 - 2

nd y

ear o

f life

Gau

derm

an 2

005:

Per

5.7

ppb

NO

2

Jans

sen

2003

: per

18

ug/m

3 N

O2

Hirs

ch 1

999:

per

10u

g/m

3 N

O2

Nor

dlin

g 20

08: p

er 4

4ug/

m3

NO

x

Van

Roo

sebr

oeck

200

8: p

er 1

7ug/

m3

Nic

olai

200

3: >

=57.

4 vs

. < 5

7.4

ug/m

3

Brauer2002

Brauer 2007 Gehring 2002 Morgenstern 2006 Other Authors

OD

DS

RA

TIO

(95%

CI)

Exacerbation of Asthma Symptoms

Increase in Wheeze Per Increment NO2

Synthesis of EvidenceExacerbations withasthma—Sufficientfor causal association

ReasonsLarge number of studies with adequate control for confounding and mostly precise effect estimates

Traffic Exposure and Doctor-Diagnosed Asthma Incidence in Children

NO2 Distance Density Spatial

Synthesis of EvidenceSufficient OR suggestive evidenceReasonsStudies that included both traffic-specific pollutantsand density measures most consistent

12

Long-Term Traffic Exposure and Cardiopulmonary Mortality

1.22

1.38

1.84

1.70

1.05 1.05 1.041.07

1.451.39

1.13

1.95

0.9

1

1.1

1.2

1.3

1.4

1.5

1.6

1.7

1.8

1.9

2

2.1

Jerr

ett 2

009:

Circ

ulat

ory

Fick

elst

ein

2005

:C

ardi

ovas

cula

r

Fick

elst

ein

2005

:St

roke

Geh

ring

2006

:C

ardi

opul

mon

ary

Hoe

k 20

02:

Car

diop

ulm

onar

y

Mah

esw

aran

2003

: Stro

ke

Bele

en 2

008c

:C

ardi

ovas

cula

r

Beel

en 2

008c

:C

ardi

ovas

cula

r

Beel

en 2

008c

:C

ardi

ovas

cula

r

Jerr

ett 2

009:

Circ

ulat

ory

(All)

Jerr

ett 2

009:

Circ

ulat

ory

(pro

xim

ity to

traffi

c)

Jerr

ett 2

005

Residential distance from major road or highway Per10ug/m3

BS

Per10ug/m3

NO2

Per 4 ppb NO2 Per10ug/m3PM2.5

Rel

ativ

e R

isks

(95%

CI)

Proximity BS NO2 PM

Synthesis of EvidenceSuggestive to infer causal association but not yet sufficient

ReasonsToo few studiesRelative imprecision of most estimates

Effects of Traffic Exposure onBirth Outcomes

• Synthesis of Evidence• Insufficient evidence

• Reasons• Only 4 studies met

criteria for inclusions

0.5

0.75

1

1.25

1.5

1.75

2

Qui

ntile

2 v

s. 1

Qui

ntile

3 v

s. 1

Qui

ntile

4 v

s. 1

Qui

ntile

5 v

s. 1

Qui

ntile

2 v

s. 1

Qui

ntile

3 v

s. 1

Qui

ntile

4 v

s. 1

Qui

ntile

5 v

s. 1

Hig

h S

ES

Mid

dle

SE

S

Low

SE

S

Hig

h S

ES

Mid

dle

SE

S

Low

SE

S

Qua

rtile

2 v

s. 1

Qua

rtile

3 v

s. 1

Qua

rtile

4 v

s. 1

NO

2 pe

r 10u

g/m

3

NO

per

10u

g/m

3

PM

2.5

per 1

ug/

m3

<50m

Hig

hway

<150

m H

ighw

ay

<50m

Maj

or ro

ad

NO

2 pe

r 10u

g/m

3

NO

per

10u

g/m

3

PM

2.5

per 1

ug/

m3

Pre-term birth &traffic density

(Wilhelm 2003)

Pre-term birth +LBW & traffic

density (Wilhelm2003)

Pre-term birthin summer &traffic density(Ponce 2005)

Pre-term birthin winter &

traffic density(Ponce 2005)

Birthweight<3000g &

PM2.5exposure

(Slama 2007)

Small for gestational age(Brauer 2008)

LBW at > 37wks (Brauer

2008)

RR

(95%

CI)

Pre-Term PT/LBS Summer Winter LBW SGA LBWDensity Density PT, Density PM2.5 NOx PM2.5 Proximity

13

Conclusions

Exposure

• Traffic-related pollutants impact ambient air quality on a broad spatial scale ranging from roadside, to urban, to regional background

• Based on synthesis of evidence, 300 to 500 meters from major road was identified as the near-source area most impacted by traffic; • variations exist depending on meteorology, background

pollution, and local factors

14

Issues for Exposure Assessment

• None of the pollutant surrogates considered met all criteria foran ideal surrogate• CO, benzene, and NOx [NO2] found in on-road vehicle emissions are

also major components of emissions from all sources• UF PM has not been used in epidemiologic studies so far. It is difficult to

model them because there are no emission inventories • Exposure models are important, but have various degrees of

utility to health studies• The proximity model is the most error-prone • Other models are better:

• Dispersion models (need adequate data)• Land use regression models• Several approaches together (hybrid)

Overall Conclusions

• The data are incomplete on emissions, their transformations, and exposure assessment

• There were, however, enough studies to find• Sufficient evidence for a causal association with

exacerbation of asthma• Suggestive evidence for a number of other health

effects (mortality, lung function, respiratory symptoms, and others)

15

Overall Conclusions II

• Limited evidence of effects but inadequate and insufficient to infer causal associations:• Adult onset asthma• Health care utilization• COPD• Non-asthmatic allergy• Birth outcomes• Cancers

Overall Conclusions III

• A caution: epidemiology studies are based on past estimates of exposure• they may not provide an accurate guide to estimating

health associations in the future• However, given the large number of people living

within 300- 500 meters of a major road, the Panel concluded that exposures to primary traffic generated pollutants are likely to be of public health concern and deserve attention.

16

Thank You!

Dan Greenbaumdgreenbaum@healtheffects.org