ArticlePDF AvailableLiterature Review

Woodsmoke Health Effects: A Review

Authors:

Abstract and Figures

The sentiment that woodsmoke, being a natural substance, must be benign to humans is still sometimes heard. It is now well established, however, that wood-burning stoves and fireplaces as well as wildland and agricultural fires emit significant quantities of known health-damaging pollutants, including several carcinogenic compounds. Two of the principal gaseous pollutants in woodsmoke, CO and NOx, add to the atmospheric levels of these regulated gases emitted by other combustion sources. Health impacts of exposures to these gases and some of the other woodsmoke constituents (e.g., benzene) are well characterized in thousands of publications. As these gases are indistinguishable no matter where they come from, there is no urgent need to examine their particular health implications in woodsmoke. With this as the backdrop, this review approaches the issue of why woodsmoke may be a special case requiring separate health evaluation through two questions. The first question we address is whether woodsmoke should be regulated and/or managed separately, even though some of its separate constituents are already regulated in many jurisdictions. The second question we address is whether woodsmoke particles pose different levels of risk than other ambient particles of similar size. To address these two key questions, we examine several topics: the chemical and physical nature of woodsmoke; the exposures and epidemiology of smoke from wildland fires and agricultural burning, and related controlled human laboratory exposures to biomass smoke; the epidemiology of outdoor and indoor woodsmoke exposures from residential woodburning in developed countries; and the toxicology of woodsmoke, based on animal exposures and laboratory tests. In addition, a short summary of the exposures and health effects of biomass smoke in developing countries is provided as an additional line of evidence. In the concluding section, we return to the two key issues above to summarize (1) what is currently known about the health effects of inhaled woodsmoke at exposure levels experienced in developed countries, and (2) whether there exists sufficient reason to believe that woodsmoke particles are sufficiently different to warrant separate treatment from other regulated particles. In addition, we provide recommendations for additional woodsmoke research.
Content may be subject to copyright.
Inhalation Toxicology, 19:67–106, 2007
Copyright
c
Informa Healthcare
ISSN: 0895-8378 print / 1091-7691 online
DOI: 10.1080/08958370600985875
Woodsmoke Health Effects: A Review
Luke P. Naeher
Department of Environmental Health Science, College of Public Health, University of Georgia,
Athens, Georgia, USA
Michael Brauer
School of Occupational and Environmental Hygiene, University of British Columbia, Vancouver,
British Columbia, Canada
Michael Lipsett
Department of Epidemiology and Biostatistics, School of Medicine, University of California,
San Francisco, San Francisco, California, USA
Judith T. Zelikoff
Department of Environmental Medicine, New York University School of Medicine, New York,
New York, USA
Christopher D. Simpson and Jane Q. Koenig
Department of Occupational and Environmental Health Sciences, University of Washington, Seattle,
Washington, USA
Kirk R. Smith
Division of Environmental Health Sciences, School of Public Health, University of California, Berkeley,
Berkeley, California, USA
The sentiment that woodsmoke, being a natural substance, must be benign to humans is still
sometimes heard. It is now well established, however, that wood-burning stoves and fireplaces
as well as wildland and agricultural fires emit significant quantities of known health-damaging
pollutants, including several carcinogenic compounds. Two of the principal gaseous pollutants
in woodsmoke, CO and NO
x
, add to the atmospheric levels of these regulated gases emitted by
other combustion sources. Health impacts of exposures to these gases and some of the other
woodsmoke constituents (e.g., benzene) are well characterized in thousands of publications. As
these gases are indistinguishable no matter where they come from, there is no urgent need to
examine their particular health implications in woodsmoke. With this as the backdrop, this
review approaches the issue of why woodsmoke may be a special case requiring separate health
evaluation through two questions. The first question we address is whether woodsmoke should be
regulated and/or managed separately, even though some of its separate constituents are already
regulated in many jurisdictions. The second question we address is whether woodsmoke particles
pose different levels of risk than other ambient particles of similar size. To address these two key
questions, we examine several topics: the chemical and physical nature of woodsmoke; the
exposures and epidemiology of smoke from wildland fires and agricultural burning, and related
Received 22 March 2006; accepted 4 August 2006.
The authors thank the Air Health Effects Division of Health Canada for funding to help prepare the report on which this article is based, but
the views expressed are those of the authors and do not necessarily represent the views of Health Canada. The authors also thank the following
for their assistance: Jamesine Rogers, MPH, and Zohir Chowdhury, PhD, of the University of California at Berkeley; Diana Ceballos, doctoral
candidate at the University of Washington; and Jessica Duffy, MS, doctoral candidate, Shannon Doherty, MS, and Colette Prophete at New York
University School of Medicine.
Address correspondence to Kirk Smith, Division of Enviromental Health Sciences, School of Public Health, University of California, Berkeley,
Berkeley, CA, 94720-7360, USA. E-mail: krksmith@berkeley.edu
67
68 L. P. NAEHER ET AL.
controlled human laboratory exposures to biomass smoke; the epi-
demiology of outdoor and indoor woodsmoke exposures from res-
idential woodburning in developed countries; and the toxicology
of woodsmoke, based on animal exposures and laboratory tests. In
addition, a short summary of the exposures and health effects of
biomass smoke in developing countries is provided as an additional
line of evidence. In the concluding section, we return to the two key
issues above to summarize (1) what is currently known about the
health effects of inhaled woodsmoke at exposure levels experienced
in developed countries, and (2) whether there exists sufficient rea-
son to believe that woodsmoke particles are sufficiently different
to warrant separate treatment from other regulated particles. In
addition, we provide recommendations for additional woodsmoke
research.
As the ability to control fire is often considered the character-
istic distinguishing prehuman and human evolution and wood
is the oldest of human fuels, it is literally true that exposure to
woodsmoke
is as old as humanity itself. Even today, biomass in
the form of wood and agricultural wastes is a significant source
of direct human energy consumption worldwide, representing
about 10% of the total. Of this, about 90% is used in its tradi-
tional forms as household heating and cooking fuels in devel-
oping countries, the rest being modern forms such as power-
plant fuel, principally in developed countries (United Nations
Development Programme [UNDP], 2004). Because household
use dominates total fuel demand in many developing countries,
particularly in rural areas where half of humanity still lives, it is
likely that biomass remains the main source of energy for most
of humanity.
Surprisingly, although the percentage of total fuel demand
constituted by wood declines with economic development, the
absolute amount remains relatively constant. For example, the
average use of biomass fuel per capita in the primarily wealthy
countries participating in the Organization for Economic Coop-
eration and Development (OECD) is quite similar to that in Asia,
which has the world’s largest developing nations (UNDP, 2004).
Of course, per capita use varies substantially with local circum-
stances. Countries with ample wood supplies, such as Finland,
Sweden, and Canada, burn more biomass fuel per capita than
most other countries, while those with low supplies, such as
South Korea and Singapore, burn less (Koopmans, 1999).
Over the past few decades, rising fossil energy costs, the
availability of new technologies, and the desire to use renew-
able sources have led to increases in the use of wood and other
biomass fuels in North America. For example, in Canada, such
fuels increased at about 2.4% annually during the 1990s, more
than half again as fast as overall energy demand (IEA, 2004).
During this same period, the knowledge of, and consequent
concern about, the health effects of air pollution have increased
dramatically around the world, leading to stricter air pollution
regulation and controls. While commercial sources of wood
Here, we use the term “smoke” to refer to the entire mixture of gases, solid
particles, and droplets emitted by combustion.
combustion have been subject to some regulation in North Amer-
ica and Europe, there are still important unregulated sources of
woodsmoke, including household heating stoves and fireplaces.
The latter have been the target of local ordinances in a number of
areas where woodsmoke dominates outdoor air pollution during
some seasons. To attain standards for such important pollutants
as fine particles (PM
2.5
or particulate matter less than 2.5 µm
in diameter), however, additional controls of these household
sources in more areas may be needed.
There are also important nonpoint sources of woodsmoke,
particularly wildland fires and intentional burning of agricul-
tural waste. The apparent increases of accidental wildfires in
some areas may be due to forest management practices, climate
change, and the rise in human population density near fire-prone
areas. In addition, the practice of clearing forested areas through
the use of fire has resulted in several spectacular long-burning
conflagrations in Southeast Asia and elsewhere, which have re-
sulted in a growing concern about the potential health impacts
of such events.
The sentiment that woodsmoke, being a natural substance,
must be benign to humans is still sometimes heard. It is now well
established, however, that wood-burning stoves and fireplaces
as well as wildland and agricultural fires emit significant quan-
tities of known health-damaging pollutants, including several
carcinogenic compounds (e.g., polycyclic aromatic hydrocar-
bons, benzene, aldehydes, respirable particulate matter, carbon
monoxide [CO], nitrogen oxides [NO
x
], and other free radicals)
(Tuthill, 1984; Koenig & Pierson, 1991; Larson and Koenig,
1994; Leonard et al., 2000; Dubick et al., 2002; Smith, 1987;
Traynor et al., 1987). Many of these toxic pollutants present in
woodsmoke are listed in Table 1.
Twoofthe principal gaseous pollutants in woodsmoke, CO
and NO
x
, add to the atmospheric levels of these regulated gases
emitted by other combustion sources. Health impacts of expo-
sures to these gases and some of the other wood smoke con-
stituents (e.g., benzene) are well characterized in thousands
of publications. As these gases are indistinguishable no mat-
ter where they come from, there is no urgent need to examine
their particular health implications in woodsmoke. There are rea-
sons, however, why woodsmoke may be a special case requiring
separate health evaluation.
1. At the point of emissions, woodsmoke contains a vast array
of solid, liquid, and gaseous constituents that change, some-
times rapidly, with time, temperature, sunlight, and interac-
tion with other pollutants, water vapor, and surfaces. Many
constituents are known to be hazardous to human health,
but are not specifically regulated or even fully evaluated.
Current methods of health-effects assessment do poorly in
estimating impacts by summing the effects of separate con-
stituents. The best approach, therefore, is to examine the tox-
icity of the entire mixture, as has been done with the most
well-studied biomass smoke, that from tobacco burning. Al-
though there have been more than 4000 compounds identified
WOODSMOKE HEALTH EFFECTS: A REVIEW 69
TABLE 1
Major health-damaging pollutants from biomass combustion
Compound Examples
a
Source Notes Mode of toxicity
Inorganic
gases
Carbon monoxide
(CO)
Incomplete
combustion
Transported over distances Asphyxiant
Ozone (O
3
) Secondary reaction
product of nitrogen
dioxide and
hydrocarbons
Only present downwind of fire,
transported over long distances
Irritant
Nitrogen dioxide
(NO
2
)
High-temperature
oxidation of
nitrogen in air, some
contribution from
fuel nitrogen
Reactive Irritant
Hydrocarbons Many hundreds Incomplete
combustion
Some transport—also react to form
organic aerosols. Species vary with
biomass and combustion conditions
Unsaturated: 40+,
e.g.,
1,3-butadiene
Irritant, carcinogenic,
mutagenic
Saturated: 25+,
e.g., n-hexane
Irritant, neurotoxicity
Polycyclic aromatic
(PAHs): 20+,
e.g., benzo[a]
pyrene
Mutagenic,
carcinogenic
Monoaromatics:
28+, e.g.,
benzene, styrene
Carcinogenic,
mutagenic
Oxygenatated
organics
Hundreds Incomplete
combustion
Some transport—also react to form
organic aerosols. Species vary with
biomass and combustion conditions
Aldehydes: 20+,
e.g., acrolein,
formaldehyde
Irritant, carcinogenic,
mutagenic
Organic alcohols
and acids: 25+,
e.g., methanol
acetic acid
Irritant, teratogenic
Phenols: 33+, e.g.,
catechol, cresol
(methylphenols)
Irritant, carcinogenic,
mutagenic,
teratogenic
Quinones:
hydroquinone,
fluorenone,
anthraquinone
Irritant, allergenic,
redox active,
oxidative stress and
inflammation,
possibly carcinogenic
Chlorinated
organics
Methylene chloride,
methyl chloride,
dioxin
Requires chlorine in
the biomass
Central nervous system
depressant (methylene
chloride), possible
carcinogens
(Continued on next page)
70 L. P. NAEHER ET AL.
TABLE 1
Major health-damaging pollutants from biomass combustion (Continued)
Compound Examples
a
Source Notes Mode of toxicity
Free
radicals
Semiquinone type
radicals
Little is known about
their formation
Redox active, cause
oxidative stress and
inflammatory
response, possibly
carcinogenic
Particulate
matter
(PM)
Inhalable particles
(PM
10
)
Condensation of
combustion gases;
incomplete
combustion;
entrainment of
vegetation and ash
fragments
Coarse
b
+ fine particles. Coarse
particles are not transported far and
contain mostly soil and ash
Inflammation and
oxidative stress, may
be allergenic
Respirable particles Condensation of
combustion gases;
incomplete
combustion
For biomass smoke, approximately
equal to fine particles
[See below]
Fine particles
(PM
2.5
)
Condensation of
combustion gases;
incomplete
combustion
Transported over long distances;
primary and secondary production
c
Inflammation and
oxidative stress, may
be allergenic
a
Compounds in italics either are criteria air pollutants or are included on the list of hazardous air pollutants specified in Section 112 of the
U.S. Clean Air Act. At least 26 hazardous air pollutants are known to be present in woodsmoke.
b
Coarse particles are defined as those between 2.5 and 10 µminsize.
c
Particles are created directly during the combustion process and also formed later from emitted gases through condensation and atmospheric
chemical reactions.
in tobacco smoke, many dozens of which possess toxic prop-
erties, there are few well-understood links between individ-
ual constituents and many of the health effects known to be
caused by exposure to this mixture.
The first question we address, therefore, is whether separate reg-
ulation/management of woodsmoke should be considered,even
though some of its separate constituents are already regulated in
many jurisdictions.
2. Fine particles are thought to be the best single indica-
tor of the health impacts of most combustion sources. Al-
though woodsmoke particles are usually within the size range
thought to be most damaging to human health, their chem-
ical composition is different from those derived from fossil
fuel combustion, on which most health-effects studies have
focused. Because their composition differs from those pro-
duced by fossil fuel combustion, woodsmoke particles may
not produce the same health effects per unit mass as other
combustion particles. Currently, however, except for size,
national regulations and international guidelines do not dis-
tinguish particles by composition.
The second question we address, therefore, is whether
woodsmoke particles pose different levels of risk than other am-
bient particles of similar size.
To address these two key questions, we examine several
topics:
The chemical and physical nature of woodsmoke.
The exposures and epidemiology of smoke from wild-
land fires and agricultural burning, and related con-
trolled human laboratory exposures to biomass smoke.
The epidemiology of outdoor and indoor woodsmoke
exposures from residential woodburning in developed
countries.
The toxicology of woodsmoke, based on animal expo-
sures and laboratory tests.
A short summary of the exposures and health effects of
biomass smoke in developing countries is provided as an addi-
tional line of evidence. At the end we provide recommendations
for additional woodsmoke research.
Although “woodsmoke” is the substance of primary interest in this report,
evidence related to smoke from other biomass (agricultural residues, grass, etc.)
is also examined where relevant.
WOODSMOKE HEALTH EFFECTS: A REVIEW 71
Although cancer-related epidemiology and toxicology are
discussed, we do not attempt a judgment because the Interna-
tional Agency for Research on Cancer (IARC) has just com-
pleted its Monograph #95, which includes an assessment of the
carcinogenicity of household biomass fuel combustion. It was
categorized as Category 2A, probably carcinogenic in humans,
with limited human evidence although supporting animal and
mechanistic evidence (Straif et al., 2006).
BRIEF SUMMARY OF METHODS
The authors searched available biomedical and scientific liter-
ature databases in English for articles dealing with controlled hu-
man exposure, occupational, and epidemiologic health-effects
studies, and toxicologic investigations dealing with woodsmoke;
biomass smoke; forest, vegetation, and wildland fires; agricul-
tural burning; and related terms under developed-country condi-
tions. Because of the scattered nature of the literature, however,
each author also used his or her knowledge of the literature to
identify other papers that did not show up in searches and ma-
terial in the gray literature. We believe that the result is a nearly
complete review of the major relevant publications on these sub-
jects and that there was no bias in selecting papers to review,
although we were not able to apply specific inclusion/exclusion
criteria.
We did not attempt to search for or review all the literature
on the physical and chemical nature of woodsmoke, its envi-
ronmental concentrations and human exposures, or its health
effects in developing-country conditions, such as indoor burn-
ing for cooking. In these arenas, we only try to summarize major
findings by others.
CHEMICAL COMPOSITION OF BIOMASS SMOKE
Wood consists primarily of two polymers: cellulose (50–70%
by weight) and lignin (approximately 30% by weight) (Simoneit
et al., 1998). Other biomass fuels (e.g., grasses, wheat stubble)
also contain these polymers, although their relative proportions
differ. In addition, small amounts of low-molecular-weight or-
ganic compounds (e.g., resins, waxes, sugars) and inorganic
salts are also present in wood. During combustion, pyroly-
sis occurs and the polymers break apart, producing a variety
of smaller molecules. Biomass combustion is typically ineffi-
cient, and a multitude of partially oxidized organic chemicals
are generated in biomass smoke. Biomass smoke contains a
large number of chemicals, many of which have been associ-
ated with adverse health impacts. The major health-damaging
particulate and gaseous chemicals present in biomass smoke are
listed in Table 1, along with some of their main modes of toxic
action.
Tables 2 and 3 summarize the major chemical classes de-
tected in woodsmoke; detailed chemical speciation of the sev-
eral hundred individual compounds that have been detected in
smoke samples is reported in the original references (Rogge
et al., 1998; Schauer et al., 2001; Fine, et al., 2002; McDonald
et al., 2000; Oros & Simoneit, 2001). The studies cited in
TABLE 2
Fine particle emissions and bulk chemical composition in
woodsmoke
Compound class Concentration References
Fine particle
emissions rate
(g/kg of wood
burned)
1.6–9.5 (Schauer et al., 2001;
Fine et al., 2002;
McDonald et al., 2000)
Organic carbon
(wt% of fine
particle mass)
12–101 (Schauer et al., 2001;
Fine et al., 2002;
McDonald et al., 2000)
Elemental carbon
(wt% of fine
particle mass)
0.65–79 (Schauer et al., 2001;
Fine et al., 2002;
McDonald et al., 2000)
Ionic species (wt%
of fine particle
mass)
0.014–1.7 (Schauer et al., 2001;
Fine et al., 2002;
McDonald et al., 2000)
Elemental species
(wt% of fine
particle mass)
a
0.01–4.0 (Schauer et al., 2001;
Fine et al., 2002;
McDonald et al., 2000)
a
Chloride included as an element.
Tables 2 and 3 by Rogge et al., Schauer et al., Fine et al., and
McDonald et al. all attempted to recreate conditions of resi-
dential wood combustion. In contrast, the studies by Oros et
al. aimed at being more representative of wildfire emissions.
More recently, Lee et al. have also described comprehensive
chemical composition of smoke from prescribed burns (Lee
& Baumann, 2005). Although less well characterized, a simi-
lar mixture of chemicals is reported in smoke emissions from
other types of biomass, including grasses, rice straw, sugar-
cane, and ferns (Simoneit et al., 1993, 1998; Rinehart et al.,
2002).
In general, it is difficult to make quantitative comparisons
among emission factors for specific organic compounds re-
ported by different authors. This is because many of the reports
are semiquantitative and the analytical methods used were not
comprehensively validated for each analyte, authentic standards
were frequently not available to calibrate instrument response,
variable combustion conditions (fuel type, moisture content,
combustion device) were used, and emission factors were re-
ported in a variety of units.
It should be noted that most studies have used gas chromatography/mass
spectrometry (GC/MS) to characterize the chemical content of woodsmoke.
GC is a very efficient tool for separating complex mixtures of organic chem-
icals. Combined with MS, the technique allows for highly sensitive, specific
and accurate detection and quantification of a range of organic chemicals in
environmental samples. GC/MS fails to detect compounds that are nonvolatile
or thermally labile, however. The application of novel methods, such as liquid
chromatography–mass spectrometry (LC/MS), that are appropriate for analy-
sis of nonvolatile or thermally labile compounds will further expand the list of
chemicals known to be present in biomass smoke.
72 L. P. NAEHER ET AL.
TABLE 3
Emissions by chemical class for particle and vapor constituents in woodsmoke
Particle-phase Vapor-phase
(mg/kg wood (mg/kg wood
Chemical burned) References burned) References
Carbon monoxide 130,000 (McDonald et al., 2000)
Hydrocarbons
Alkanes (C2–C7) 0.47–570 (Rogge et al., 1998; Fine
et al., 2002)
1.01–300 (Schauer et al., 2001;
McDonald et al., 2000)
Alkenes (C2–C7) 0.58–280 (Rogge et al., 1998; Fine
et al., 2002)
92–1300 (McDonald et al., 2000)
Polycyclic aromatic
hydrocarbons (PAHs)
and substituted PAHs
5.1–32,000 (Oros & Simoneit, 2001;
Fine et al., 2002; Rogge
et al., 1998; McDonald
et al., 2000)
43.4–355 (Schauer et al., 2001;
McDonald et al., 2000)
Methane 4100 (Schauer et al., 2001)
Total nonmethane
hydrocarbons C2–C7
[Included in vapor phase] 390–4000 (Schauer et al., 2001;
McDonald et al., 2000)
Unresolved complex
mixture (UCM)
300–1,130,000 (Oros & Simoneit, 2001;
Fine et al., 2002)
Oxygenated organics
Alkanols 0.24–5400 (Oros & Simoneit, 2001;
Fine et al., 2002)
120–9200 (McDonald et al., 2000)
Carboxylic acids 6200–755,000 (Oros & Simoneit, 2001;
Fine et al., 2002; Rogge
et al., 1998)
2.4 (Schauer et al., 2001)
Aldehydes and ketones [Included in vapor phase] 0.94–4450 (Rogge et al., 1998)
a
(Schauer et al., 2001;
Fine et al., 2002;
McDonald et al., 2000)
Alkyl esters 0.37–4450 (Oros & Simoneit, 2001;
Fine et al., 2002)
Methoxylated phenolic
compounds
28–1000 (Rogge et al., 1998; Fine
et al., 2002; McDonald
et al., 2000)
1200–1500 (Schauer et al., 2001)
Other organics
Other substituted
aromatic compounds
5.0–120,000 (Oros & Simoneit, 2001;
Fine et al., 2002; Rogge
et al., 1998)
110–3600 (Schauer et al., 2001;
McDonald et al., 2000)
Sugar derivatives 1.4–12600 (Oros & Simoneit, 2001;
Fine et al., 2002)
Coumarins and
flavonoids
0.71–12 (Fine et al., 2002)
Phytosteroids 1.7–34.0 (Rogge et al., 1998; Fine
et al., 2002)
Resin acids and
terpenoids
1.7–41,000 (Oros & Simoneit, 2001;
Fine et al., 2002; Rogge
et al., 1998)
21–430 (McDonald et al., 2000)
Unresolved compounds 1.2–120 (Fine et al., 2002) 20–600 (Schauer et al., 2001;
McDonald et al., 2000)
a
Only aldehydes reported.
WOODSMOKE HEALTH EFFECTS: A REVIEW 73
Woodsmoke particles are generally smaller than 1 µm, with a
peak in the size distribution between 0.15 and 0.4 µm (Kleeman
et al., 1999; Hays et al., 2002). As with other combustion mix-
tures, such as diesel and tobacco smoke, fresh woodsmoke con-
tains a large number of ultrafine particles, less than 100 µm,
which condense rapidly as they cool and age. Indeed, most
of the particle mass in aged woodsmoke has been formed
by such condensation processes. Fine particles in this size
range efficiently evade the mucociliary defense system and
are deposited in the peripheral airways, where they may ex-
ert toxic effects. Particles in this size range are not easily re-
moved by gravitational settling and therefore can be transported
over long distances (Echalar et al., 1995). The transport of
biomass combustion particles over hundreds of kilometers has
been extensively documented (Andrae et al., 1988). Haze lay-
ers with elevated concentrations of CO, carbon dioxide (CO
2
),
ozone (O
3
), and nitric oxide (NO) have been observed. During
transport, many of the gaseous species are converted to other
gases or into particles. The “black carbon” from biomass emis-
sions is now thought to contribute to regional and global cli-
mate change as well as adverse health effects in some parts
of the world (Venkataraman et al., 2005; Koch & Hansen,
2005).
Although approximately 5–20% of woodsmoke particulate
mass consists of elemental carbon, the composition of the or-
ganic carbon fraction varies dramatically with the specific fuel
being burned and with the combustion conditions. Detailed anal-
ysis of organic woodsmoke aerosol were conducted by Rogge
et al. (1998), who measured nearly 200 distinct organic com-
pounds, many of them derivatives of wood polymers and resins
(Rogge et al., 1998). Since profiles of specific polycyclic aro-
matic hydrocarbons (PAHs) are likely to vary, many measure-
ments have focused on benzo[a]pyrene (BaP), a probable human
carcinogen.
A number of toxic or carcinogenic compounds are present in
biomass smoke, including free radicals, PAHs, and aldehydes,
as shown in Table 1 (Leonard et al., 2000; Pryor, 1992; Schauer
et al., 2001). Organic extracts of ambient particulate matter
(PM) containing substantial quantities of woodsmoke are 30-
fold more potent than extracts of cigarette smoke condensate in
a mouse skin tumor induction assay (Cupitt et al., 1994), and
are mutagenic in the Salmonella typhimurium microsuspension
and plate incorporation assays (Claxton et al., 2001). Few, if
any, reports exist in which the toxicity of smoke from different
biomass sources was compared and related to differences in the
chemical composition of each smoke type.
Woodsmoke is enriched with several chemicals relative to
pollutant mixtures from other sources of air pollution. Exam-
ples include potassium, methoxyphenols, levoglucosan, retene,
and specific resin acids (e.g., abietic acid) (Khalil & Rasmussen,
2003; Fine et al., 2001, 2002; Schauer et al., 2001; Rogge et al.,
1998; Hawthorne et al., 1992). Many of these chemicals have
been used either individually or in multivariate analyses to quan-
tify woodsmoke emissions (Khalil & Rasmussen, 2003; Schauer
& Cass, 2000; Larsen & Baker, 2003).
Levoglucosan is sugar anhydride derived from the pyrolysis
of the major wood polymer cellulose. Levoglucosan is one of
the most abundant organic compounds associated with particles
in woodsmoke (Fine et al., 2001, 2002). It is stable in the envi-
ronment and has been used extensively to estimate woodsmoke
levels in ambient PM samples (Schauer & Cass, 2000; Katz
et al., 2004; Larson et al., 2004). Levoglucosan is present in
other biomass smoke samples, including smoke from tobacco,
grasses, and rice straw (Sakuma & Ohsumi, 1980; Simoneit,
et al., 1993). Under conditions in which woodsmoke dominates
the biomass smoke contribution to ambient aerosol, however,
levoglucosan can be considered a unique tracer for woodsmoke
(Schauer & Cass, 2000).
Methoxyphenols are a class of chemicals derived from the py-
rolysis of the wood polymer lignin. This class of chemicals spans
a range of volatilities from relatively volatile (e.g., guaiacol)
to exclusively particle-associated (e.g., sinapinaldehyde). These
chemicals are relatively abundant in woodsmoke, although the
most abundant compounds are predominantly in the vapor phase
(Hawthorne et al., 1989; Schauer et al., 2001). Accurate chemi-
cal analysis of the methoxyphenols, however, has proved to be
an analytical challenge, and many of the methoxyphenols were
found to be chemically reactive—a property that would under-
mine their suitability as tracers for biomass smoke (Simpson
et al., 2005). Methoxyphenols have been used as woodsmoke
tracers in multivariate source apportionment models to deter-
mine the proportion of urban fine PM derived from woodburning
(Schauer & Cass, 2000).
The organic chemical composition has been used to distin-
guish smokes from different biomass fuels. Smoke from hard-
wood versus softwood burning can be distinguished by the rela-
tive proportions of substituted guaiacols compared to syringols
(Hawthorne et al., 1989; Oros & Simoneit, 2001; Schauer &
Cass, 2000). Mono- and dimethoxyphenols are also present in
small amounts in grass and grain smokes, but the major phe-
nolic compounds in grass smoke are p-coumaryl derivatives
(Simoneit, et al., 1993). Diterpenoids (e.g., dehydroabietic acid)
are abundant in smoke from gymnospems (conifers) compared to
angiosperms (Schauer et al., 2001). Certain chemicals may even
be unique to smoke from specific tree species (e.g., juvabione
from balsam fir), although the atmospheric stability of such com-
pounds and hence their utility as source-specific exposure mark-
ers has not been established (Fine et al., 2001; Oros & Simoneit,
2001).
Emission factors for fine particles are highly dependent on
the fuel characteristics and burn conditions (smoldering vs. flam-
ing). Similarly, emission factors for specific organic chemicals
are influenced by fuel moisture content and burn conditions,
although the relationships may not parallel those observed for
fine particles (Khalil & Rasmussen, 2003; Guillen & Ibargoitia,
1999).
74 L. P. NAEHER ET AL.
FOREST FIRE AND AGRICULTURAL BURNING:
EXPOSURE AND HEALTH STUDIES
In contrast to the large amount of information relating urban
PM to human health impacts, there is only a limited number of
studies directly evaluating the community health impacts of air
pollution resulting from the burning of biomass. Several reviews
have discussed the health impacts and pollutants associated with
woodsmoke air pollution (Larson & Koenig, 1994; Pierson et al.,
1989; Vedal, 1993; Boman et al., 2003, 2006). Although the em-
phasis of these reviews was on community exposures resulting
from burning of wood in fireplaces and wood stoves, many of
the conclusions are relevant to the broader understanding of veg-
etation fire air pollution. The World Heath Organization (WHO)
has published a document describing Health Guidelines for Veg-
etation Fire Events,
which also contains a review of evidence
linking air pollution from vegetation fires with human health ef-
fects. Specific information relating agricultural and forest/brush
burning with human health effects is summarized next and pre-
sented in Table 4.
On a regional basis, during vegetation fire episodes PM is the
air pollutant most consistently elevated in locations impacted by
fire smoke (Sapkota et al., 2005). For example, during fires in
southern California, PM
10
concentrations were 3–4 times higher
than during nonfire periods, while particle number, and CO and
NO concentrations were increased by a factor of 2. The con-
centrations of NO
2
and O
3
were essentially unchanged or even
lower (Phuleria et al., 2005). Further, measurements indicate
that that biomass combustion emissions can be transported over
hundreds of kilometers such that local air quality is degraded
even at great distances from fire locations (Sapkota et al., 2005).
Smoke from African and Brazilian savanna fires has been shown
to contain substantial quantities of fine particles (Artaxo et al.,
1991; Echalar et al., 1995). Mass concentrations ranged from
30 µg/m
3
in areas not affected by biomass burning to 300 µg/m
3
in large areas (2 million km
2
) with intense burning. Additional
studies of fine particle (<2 µm) composition associated with
biomass burning in the Amazon Basin was reported by Artaxo
et al. (1994), who found 24-h average PM
10
and PM
2.5
mass con-
centrations as high as 700 and 400 µg/m
3
, respectively (Artazo
et al., 1994). In one of the few measurements of rural community
air pollution associated with large tropical forest fires, Reinhardt
and Ottmar measured formaldehyde, acrolein, benzene, CO, and
respirable PM (PM
3.5
)inRondonia, Brazil, during the peak of
the 1996 biomass burning season (Reinhardt et al., 2001). Of
the species measured, respirable particle levels were elevated
5-10 times above background, with mean levels of 190 µg/m
3
and levels as high as 250 µg/m
3
measured during several of the
12-h sampling periods. The mean CO level was 4 ppm, which
is similar to levels measured in moderately polluted urban ar-
eas, but below the level expected to be associated with acute
health impacts. Benzene levels (11 µg/m
3
average) were higher
www.who.int/docstore/peh/Vegetation fires/Health Guidelines final 3.pdf
than those measured in other rural areas and were comparable
to those measured in cities.
Measurements from Southeast Asia also indicate that parti-
cles are the main air pollutant elevated during periods of vege-
tation fire-related air pollution (Radojevic & Hassan, 1999). For
example, during a 2- to 3-mo period in 1994, 24-h PM
10
levels
up to 409 µg/m
3
were recorded in Kuala Lumpur (Hassan et al.,
1995), and levels ranged from 36 to 285 µg/m
3
(unspecified
average time) in Singapore (Nichol, 1997). In a 1997 vegeta-
tion fire episode, PM
10
levels as high as 930 and 421 µg/m
3
were measured in Sarawak (Malaysia) and Kuala Lumpur, re-
spectively, while 24-h levels in Singapore and southern Thailand
were somewhat lower (Brauer, 1998). Closer to the fire source
in Indonesia, PM
10
concentrations as high as 1800 µg/m
3
were
measured over an unspecified period (Kunii et al., 2002). In
February–May 1998 a more limited vegetation fire episode af-
fected regions of Borneo. In Brunei, 24-h PM
10
levels as high
as 440 µg/m
3
were measured during this period (Radojevic &
Hassan, 1999).
Wildland Firefighters
In general, wildland firefighters experience greater exposure
from forest fire smoke than members of the general public. Pat-
terns of exposure can be intense in initial fire-suppression efforts
or in situations involving thermal inversions. Workshifts are fre-
quently 12 to 18 h and can last for more than 24 h. In large
fires, prolonged work shifts can last for many days. In wildland
firefighting, it is not feasible to use a self-contained breathing
apparatus; often the only respiratory protection used is a cot-
ton bandana tied over the nose and mouth. Moreover, many of
the tasks in wildland firefighting are physically demanding and
require elevated pulmonary ventilation rates, which can result
in substantial doses of smoke to the respiratory tract. Off-shift
smoke exposures may occur as well, depending on the location
of the base camp (where firefighters eat and sleep) in relation
to the fire and the prevailing meteorology. With the intensity of
smoke exposures, it is not surprising that respiratory problems
accounted for about 40% of all medical visits made by wild-
land firefighters during the Yellowstone firestorm of 1988 (U.S.
Department of Agriculture, 1989).
There have been several investigations of both exposures and
health impacts of smoke exposure among wildland firefighters.
Exposure assessment can represent a major logistical challenge,
considering that the work often takes place on steep terrain in
remote locations and may involve extreme physical exertion. In
addition, exposure assessment must of necessity be limited to
relatively few of the thousands of substances in biomass smoke.
By extension, the few health studies that have been undertaken
have not involved concurrent exposure assessment, but have fo-
cused on cross-shift or cross-seasonal respiratory effects.
Reinhardt and Ottmar (2000) undertook an exposure as-
sessment of breathing-zone levels of acrolein, benzene, carbon
dioxide, CO, formaldehyde, and PM
3.5
among firefighters at 21
wildfires in California between 1992 and 1995. Interestingly,
WOODSMOKE HEALTH EFFECTS: A REVIEW 75
TABLE 4
Summary of selected epidemiologic studies of large-scale vegetation fires
Population Endpoints measured Results Reference
All ages Emergency room visits Increased respiratory visits in communities exposed to fire smoke (Duclos et al.,
1990)
All ages Emergency room visits,
hospital admissions
Increased emergency-room visits and hospital admissions for
asthma and bronchitis during fire period relative to same
period in previous year
(Sorenson et al.,
1999)
All ages Acute respiratory
distress hospital visits
Increase in acute respiratory distress inhalation therapy visits
associated with indirect measure (sedimentation) of air
pollution during sugar-cane burning season in Brazil
(Arbex et al.,
2000)
All ages Outpatient visits Increased visits for asthma, upper respiratory tract symptoms,
and rhinitis during vegetation fire episode periods of elevated
PM
10
in Malaysia
(Brauer, 1998)
All Ages Outpatient visits,
hospital admissions,
mortality
Increase in PM
10
from 50 to 150 µg/m
3
during vegetation fire
episode periods associated with increase in outpatient visits in
Singapore for upper respiratory tract symptoms (12%), asthma
(37%), and rhinitis (26%). No increase in hospital admissions
or mortality
(Emmanuel,
2000)
All Ages Emergency room visits Increased asthma visits with PM
10
during episode of exposure to
biomass burning emissions in Singapore
(Chew et al.,
1995)
All Ages Emergency room visits No increase in asthma visits with PM
10
during episode of
exposure to bushfire emissions in Australia
(Copper et al.,
1994)
All Ages Emergency room visits No increase in asthma visits with PM
10
During episode of
exposure to bushfire emissions in Australia
(Smith et al.,
1996)
All Ages Emergency room visits Increased asthma visits associated with PM
10
, especially for
concentrations exceeding 40 µg/m
3
(Johnston et al.,
2002)
All Ages Physician visits for
respiratory,
cardiovascular, and
mental illness
A46to78% increase in physician visits for respiratory illness
during a 3-wk forest fire period in Kelowna, British Columbia
(Moore et al.,
2006)
All Ages Hospital admission for
respiratory illness
Daily hospital emission rates for respiratory illness increased
with levels of PM
10
for bushfire and nonbushfire periods
(Chen et al.,
2006)
All ages,
>65 yr
Mortality 0.7% (all ages) and 1.8% (ages 65–74) increases in adjusted
relative risk of nontrauma mortality per 10-µg/m
3
increase in
PM
10
Kuala Lumpur, Malaysia, for 1996-1997, including
vegetation fore episode period
(Sastry, 2002)
Adults with
COPD
Symptoms Significant increase in symptom index (dyspnea, cough, chest
tightness, wheezing, sputum production) on two days of
elevated PM2.5 (65 µg/m
3
) relative to control days (14
µg/m
3
). Days of elevated PM attributed to fire smoke by
satellite imaging
(Sutherland,
2005)
Adults Asthma medication,
lung function,
asthmatic and other
respiratory symptoms
Increased prevalence of respiratory symptoms and various
asthma indicators, decreased lung function post-rice stubble
burning period relative to period prior to burning in three
communities in Iran
(Golshan et al.,
2002)
Adult mili-
tary recruits
Blood markers of
inflammation
Bone marrow stimulated to release immature polymorphonuclear
leukocytes into blood during period of exposure to forest fire
smoke relative to period following smoke exposure
(Tan et al., 2000)
Children Respiratory hospital
admissions
Increased pediatric respiratory hospital admissions associated
with increased biomass smoke markers (potassium and black
carbon) during sugar-cane burning season in Brazil
(Cancado et al.,
2002)
Children Lung function Decreased lung function in children during vegetation fire
episode compared to preepisode measurements
(Hisham-Hashim
et al., 1998)
76 L. P. NAEHER ET AL.
exposures to the gases were generally well below time-weighted
average occupational health standards. However, some of the
fires resulted in high-level peak exposures to heavy smoke. Res-
pirable particle (PM
3.5
)exposures on multiday fires averaged
0.72 mg/m
3
on the fireline, and 0.5 mg/m
3
over the work shift,
with peak concentrations of 2.3 and 2.93 mg/m
3
. The corre-
sponding exposures to CO were 4.0 and 2.8 ppm, with peak
(2-h time-weighted average [TWA]) exposures of 38.8 ppm and
30.5 ppm. The particle concentrations are about 10 to 30 times
higher than 24-h average ambient air quality standards for PM
2.5
(currently 65 µg/m
3
in the United States).
Materna et al. (1992) also found extremely high particle ex-
posures among California wildland firefighters during the 1987–
1989 fire seasons. Table 5 presents their data on PM exposures.
These investigators also sampled for 12 PAHs and found all
below 1 µg/m
3
. The highest CO levels were associated with
tending gasoline-powered pumping engines rather than from
smoke exposure per se. An aldehyde screen detected formalde-
hyde, acrolein, furfural, and acetaldehyde. Most levels were
well below occupational exposure limits; however, formalde-
hyde (which was detected in all samples) in several instances ex-
ceeded such limits (maximum TWA [226 min] = 0.42 mg/m
3
).
In general, these studies demonstrate that of the various mea-
sured constituents of smoke, PM tends to be the most consis-
tently elevated during wildland firefighting in relation to health-
based exposure standards.
In the first report of cross-seasonal changes in respiratory
symptoms and lung function in wildland firefighters, Rothman
et al. (1991) examined 69 Northern California firefighters who
were nonsmokers or former smokers who had not smoked in
at least 6 mo. There were significant cross-seasonal increases
in reported cough, phlegm, wheeze, and eye and nasal irrita-
tion. Only eye irritation, however, was significantly associated
with firefighting activity (r = .48, p < .001), while the associ-
ation of wheeze with firefighting in the last 2 wk of the study
wasofborderline significance (r = .25, p = .07). There were
small, but statistically significant, declines in several measures
of pulmonary function across the season, with the strongest
relationships for the highest exposure category in the final
week preceding the follow-up spirometry. The associations be-
TABLE 5
Personal TWA particle exposures among California wildland
firefighters
Particle Mean
metric Site/activity (mg/m
3
) Range (mg/m
3
)
TSP Base camp/waiting 3.3 1.8–4.4
in staging area
TSP Fireline/mop-up 9.5 2.7–37.4
Respirable Fireline/mop-up 1.8 0.3–5.1
Respirable Prescribed burn 1.2 0.2–2.7
Note. Modified from Materna et al. (1992).
came weaker and less significant with the progressive inclu-
sion of additional weeks prior to the spirometry. Across the
8-wk study, several lung function metrics exhibited significant
declines, including FEV1 (1.2%, confidence interval [CI]
0.5, 2.0%), FEV1/FVC (0.006, CI 0.001, 0.01), and
although FVC also declined, this change was not significant.
Those in the highest category for hours worked in the week pre-
ceding spirometry experienced larger decrements in lung func-
tion (FEV1 =−2.9% [130 ml] and FVC =−1.9% [101 ml]).
These changes were not affected by adjustment for potential
confounders (not specified). The use of a cotton bandana for
respiratory protection was not associated with any measurable
protection.
Liu et al. (1992) examined cross-season changes in pul-
monary function and airway hyperresponsiveness in 63 wildland
firefighters in northern California and Montana in 1989. They
were tested before the start of the fire season and within 2 wk of
discharge from service. Though pre- and post-season spirometric
measurements were within the normal range for all participants,
there were significant cross-seasonal declines in FVC, FEV1,
and FEF25–75 of 0.09 L, 0.15 L, and 0.44 L/s, respectively.
There was no significant relationship with any of the covariates
measured, including smoking status, history of allergy, asthma,
or upper/lower respiratory symptoms, specific firefighting crew
membership, or seasonal versus full-time employment. Airway
responsiveness to methacholine increased significantly across
the fire season, which was not affected by gender, history of
smoking, allergy, full-time versus seasonal employment, or crew
membership. This study suggests that, in addition to persistent
cross-seasonal changes in lung function, firefighting may also be
associated with increased airway hyperresponsiveness, although
the effect was not significant.
Letts et al. (1991) conducted a health survey of 78 wildland
firefighters in Southern California. There were no changes in
symptom prevalence cross-seasonally, nor were there any sig-
nificant associations with exposure (defined as low, medium, and
high, based on hours of work and weighted by visual estimates
of smoke intensity). There were small, nonsignificant changes in
FEV1 and FVC. The decrements in FEF25–75 and FEV1/FVC,
however, were both significant (2.3%, CI 4.2, 0.5% and
0.5%, CI 1.0, 0.1%). The changes in FEF25–75 showed
a nonsignificant exposure-response trend ( p = 0.08) of: 0.5%,
1.9%, and 4.7% for the low-, medium-, and high-exposure
groups, respectively. Interestingly, however, there were no asso-
ciations with the number of seasons of firefighting, days since
the last fire, or age. Although these investigators concluded that
there was limited evidence of cross-seasonal effects of firefight-
ing on lung function, they indicated that the season in which
their survey was conducted involved an atypically low number
of firefighting hours. Moreover, the baseline was established in
June, reportedly “before significant smoke exposure occurred,
All confidence intervals reported here are at the 95% level.
WOODSMOKE HEALTH EFFECTS: A REVIEW 77
though the extent of firefighting preceding the initial measure-
ment was not documented.
In addition to examining cross-seasonal lung function
changes, Betchley et al. (1997) also examined cross-shift
changes among forest firefighters in the Cascade Mountains
of Oregon and Washington (Betchley et al., 1997). Among
76 workers examined at the beginning and immediately after
prescribed burns, mean declines in FVC, FEV1, and FEF25–
75 were 0.065 L, 0.150 L, and 0.496 L/s, respectively. These
changes were significant even after adjusting for respiratory in-
fections in the preceding 4 wk, smoking status, any “lung con-
dition, and allergy. Examining cross-seasonal changes in 53
firefighters, the values for these same measures were 0.033 L,
0.104 L, and 0.275 L/s, respectively. The changes for FEV1
and FEF25–75 were significant, and remained so even af-
ter adjustment for the same potential confounders and effect
modifiers. There were no significant cross-seasonal changes
in respiratory symptoms. The cross-seasonal lung function
measurements and symptom reports were taken, on average,
78 days after the last occupational firefighting activities of the
season. In a subsequent analysis of a subset of these work-
ers (n = 65) who had been working when several combustion
products were measured, the lung function decrements ob-
served were not found to be specifically associated with PM
3.5
,
acrolein, carbon monoxide, or formaldehyde (Slaughter et al.,
2004).
Investigators in Sardinia compared lung function among 92
wildland firefighters with a “control” group of policemen (Serra
et al., 1996). The testing was undertaken in late spring, just
prior to the onset of the principal fire season. The two groups
had identical mean values for FVC and TLC,
and showed no
significant differences for FRC,
DLCO, or DLCO
/TLC. The
firefighters, however, demonstrated modestly lower lung func-
tion test results for FEV1, FEV1/FVC, FEF50, FEF25, and RV.
∗∗
Although there were significant differences in age and height
between the two groups (the firefighters were older and shorter,
both of which would favor lower mean lung function), the sig-
nificant differences in lung function remained after multivariate
control for age, height, smoking status, and pack-year history
for current smokers. The investigators found no relationship of
pulmonary function with years of service or with the number of
fires extinguished over their careers. Cough and expectoration
were more common among firefighters, but these differences
were not significant.
Total lung capacity (TLC) is the volume of air contained in the lungs after
maximal inhalation.
Functional residual capacity (FRC) measures the amount of air remaining
in the lungs after a normal tidal expiration.
Carbon monoxide diffusing capacity (DLCO) provides an assessment of
the ability of gases to diffuse across the blood–gas barrier, that is, from the
alveoli into the blood.
∗∗
Residual volume (RV) is the amount of air remaining in the lungs after a
maximal exhalation.
Wildland firefighting can involve intermittent prolonged ex-
posures to high concentrations of respirable particles, which
consist of mixtures unique to each situation. Exposures to ele-
vated levels of CO and respiratory irritants such as formaldehyde
also occur, but respirable particles probably represent the prin-
cipal exposure of concern. The few health studies conducted on
such workers have documented cross-seasonal decrements in
lung function, increased airway hyperresponsiveness, and in-
creased prevalence of respiratory symptoms. Rothman et al.
(1991) demonstrated that recent cumulative exposures were
more strongly associated with greater changes in lung function
than were more remote exposures. At least one study has also
shown acute cross-shift spirometric changes as well (Liu et al.,
1992). There has been no long-term follow-up of the respira-
tory health of wildland firefighters, however. Among municipal
firefighters, chronic pulmonary dysfunction may result from re-
peated smoke exposure, particularly among those who do not use
respiratory protective devices (Tepper et al., 1991; Sparrow et al.,
1982). It is unknown whether cessation of exposure among wild-
land firefighters during the off-season may allow for recovery
and reversibility of effects, in contrast to municipal firefighters,
who can be exposed year-round. In any case, the relatively small
effects demonstrated among firefighters cannot be quantitatively
extrapolated to nonoccupational exposures, as the demands of
the job require a degree of physical fitness and resilience far
beyond that found in most of the general population.
Forest and Brush Fires
Several studies in North America have evaluated the health
impacts associated with forest and brush fires. In the first study
examining the effect of wildfire smoke on the general popula-
tion, Duclos and colleagues evaluated the impact of a numerous
large forest fires on emergency room (ER) visits to 15 hospi-
tals in 6 counties in California (Duclos et al., 1990). The au-
thors calculated observed-to-expected ratios of ER visits, based
on the numbers of visits during two reference periods. Dur-
ing the approximately 2
1
2
-wk period of observation, ER vis-
its for asthma and chronic obstructive pulmonary disease in-
creased by 40% ( p < .001) and 30% ( p = .02), respectively.
Significant increases were also observed for bronchitis (ob-
served [O]/expected [E] = 1.2, p = .03), laryngitis (O/E = 1.6,
p = .02), sinusitis (O/E = 1.3, p = .05), and other upper respira-
tory infections (O/E = 1.5, p < .001). Exposure assessment was
problematic, however, as few PM
10
or other monitors were lo-
cated downwind of the fires. The highest PM
10
concentration
measured was 237 µg/m
3
.Incontrast, several measurements
of total suspended particles (TSP) exceeded 1000 µg/m
3
; the
highest recorded value was 4158 µg/m
3
. Exposure to forest fire
smoke can be unpredictable, changing with wind direction, in-
tensity of the fire, precipitation, and other variables. The few air
quality measurements available to these investigators could not
serve to reliably characterize population exposures, which is a
general limitation of all wildfire studies. In addition, this study
was subject to other typical limitations of ER analyses related
78 L. P. NAEHER ET AL.
to behavioral and economic factors (e.g., perceptions of illness
severity, access to other health care providers, and availability
of health insurance, with the latter more problematic in the U.S.
than elsewhere).
Although no air pollutant concentrations were reported, the
impact of wildfires in Florida on ER visits to eight hospitals in
1998 were compared to visits during the same 5-wk period in
the previous year. From 1997 to 1998, ER visits increased sub-
stantially for asthma (91%), bronchitis with acute exacerbation
(132%), and chest pain (37%), while visits decreased for painful
respiration (27%) and acute bronchitis (20%). Though based on
smaller numbers, there were modest changes in the number of
hospital admissions (increases of 46% for asthma and 24% for
chest pain) (Sorenson et al., 1999). Although this study suggests
that wildfire smoke exposure resulted in increased ER visits for
respiratory disease and symptoms, no firm conclusions are pos-
sible. There was only one reference period selected, which might
not provide a stable basis for comparison, and no statistical test-
ing was undertaken.
In a retrospective evaluation of the health impacts of a large
wildfire in a northern California Native American reservation,
visits to the local medical clinic for respiratory illness increased
by 52% over the same period the prior year (Mott et al., 2002).
During the ten weeks that the fire lasted, PM10 levels exceeded
150 µg/m
3
(24-h average) 15 times, and on 2 days the lev-
els exceeded 500 µg/m
3
.Weekly concentrations of PM
10
were
strongly correlated with weekly visits for respiratory illness dur-
ing the fire year (r = .74), but not in the prior year (r =−.63). In a
community survey of 289 respondents, more than 60% reported
respiratory symptoms during the smoke episode; 20% reported
symptoms persisting at least 2 wk after the smoke cleared. In-
dividuals with preexisting cardiopulmonary diseases reported
significantly more symptoms before, during, and after the fire
than those without such illnesses. The investigators also retro-
spectively evaluated the efficacy of several public health inter-
ventions in symptom reduction: (1) filtered and unfiltered masks
distributed free of charge; (2) vouchers for free hotel accom-
modations in towns away from the smoke to assist evacuation
efforts; (3) high-efficiency particulate air (HEPA) cleaners dis-
tributed for residential use; and (4) public service announce-
ments (PSAs) about exposure reduction strategies. Mott and
colleagues found that increased duration of use of a residen-
tial HEPA air cleaner was associated with decreased odds of
reporting increased symptoms (odds ratio [OR] 0.54, CI 0.32,
0.89), with an inverse trend of symptom reporting with increas-
ing duration of use. Similarly, ability to accurately recall a PSA
was also associated with reduced odds for respiratory symp-
toms. In contrast, there was no detectable beneficial effect of
evacuation from smoky areas or of the use of masks. However,
the timing and duration of evacuation were not optimal. On the
days with the highest recorded smoke concentrations, over 80%
of the subjects had not evacuated. That mask use was not pro-
tective is not surprising; the masks were distributed without fit
testing and had variable filtration efficiencies. Moreover, none
of the interventions was randomized, and in fact individuals with
smoke-related health effects or a prior diagnosis of respiratory
or cardiovascular disease were given priority to receive hotel
vouchers and HEPA air cleaners. Finally, due to the retrospec-
tive nature of the investigation, recall bias may have affected the
results based on the survey.
More recently, Sutherland and colleagues reported an in-
crease in an index of respiratory symptoms (dyspnea, cough,
chest tightness, wheezing, and sputum production) among a
panel of 21 subjects with COPD associated with 2 days of el-
evated ambient particle levels resulting from a forest fire near
Denver, CO. On the 2 days in which symptom scores were in-
creased, average PM
2.5
concentrations increased to 63 µg/m
3
relative to an average of 14 µg/m
3
on control days (Sutherland
et al., 2005). During this same fire as well as several other fires
in Colorado, the indoor infiltration of particulate matter was
measured and the effectiveness of room HEPA-filter air clean-
ers was assessed in a total of eight homes. A decrease in PM
2.5
concentrations of 63–88% was measured in homes in which air
cleaners were operated, relative to homes without air cleaners.
In the homes without the air cleaners, measured indoor PM
2.5
concentrations were 58–100% of the concentrations measured
outdoors (Henderson et al., 2005).
Moore and colleagues assessed the impact of elevated con-
centrations of PM
2.5
associated with forest fires on outpatient
physician visits for respiratory disease. Two large fires burning
adjacent to urban areas in British Columbia, Canada, resulted in
intermittent elevations (140–200 µg/m
3
)indaily average PM
2.5
concentrations over a 5-wk period in August and September
2003. In the city with the highest levels of PM and that was
closest to a fire, weekly physician visits for respiratory disease
were increased approximately 45–80% relative to average rates
corresponding to those weeks during the previous 10 yr. No
statistically significant increases were observed in the city with
lower fire-related PM increases and neither city experienced el-
evated physician visits for cardiovascular diseases (Moore et al.,
2006). However, as many patients experiencing symptoms char-
acteristic of acute cardiovascular events go directly to a hospital
emergency department, it is possible that the health database
used in this investigation may not have been capable of identi-
fying circulatory outcomes of interest during the study period.
During 1994, bush fires near Sydney, Australia led to ele-
vated PM
10
levels (maximum hourly values of approximately
250 µg/m
3
) for a 7-day period. Two studies of asthma emer-
gency room visits during the bushfire smoke episode failed to
detect any association with air pollution (Copper et al., 1994;
Smith et al., 1996). The report by Copper et al. (1994) was in
the form of a letter to The Lancet, with few details provided.
The investigators examined only three inner-city hospitals, pre-
ferring to avoid the influence of “patients who presented with
direct effects of smoke inhalation,” which might have occurred
had they included hospitals with catchment areas closer to the
fires. They compared the numbers of asthma ER visits for the
week before the bushfires (January 1–8), the fire period (January
WOODSMOKE HEALTH EFFECTS: A REVIEW 79
9–20), and afterward (January 21–31), and found no difference
among the 3 periods. These comparisons were based on rela-
tively small numbers, however, with fewer than 100 visits for
asthma during the entire month for all 3 hospitals. The report
by Smith et al. (1996) involved a comparison of the proportions
of asthma to total ER visits to seven hospitals during the week
of high smoke levels compared to the same week the prior year.
There was no difference in these proportions, nor was there a
relationship between the maximum daily nephelometric particle
measurement and the number of asthma ER visits in multiweek
regression models. Although it appears that the bushfire smoke
did not have an impact on asthma ER visits, this study is limited
by the use of a single reference period. In addition, the regres-
sion analysis is likely to have had very limited statistical power,
with relatively few days of observation.
A recent analysis of these same fires and lung function (mea-
sured as peak expiratory flow rate [PEFR]) did not detect any
association between either PM
10
levels or an indicator variable
representing the fire period and evening PEFR in 25 asthmatic
children, although 20 children without airway hyperreactivity
showed a significant decrease in PEFR with increasing same-
day PM
10
concentrations (Jalaludin et al., 2000). Whether this
represents a true lack of association or an artifact of experimental
design is difficult to ascertain. Thirty-two children in this analy-
sis were recruited during the week of the fire. There did not seem
to be any examination of whether there was a learning period
for these children (during which the initial PEFR measurements
might have been more variable), nor was there any discussion of
the quality control for recording the measurements, or even what
the PEFR protocol was. Of the 32 children (mean age = 9.2 yr),
25 had a physician’s diagnosis of asthma; however, only 12
of the 32 had evidence of airway hyperresponsiveness, which is
considered a hallmark of asthma. Although the regression model
included indicator variables for use of asthma medications, there
could nonetheless still have been residual confounding by med-
ication use. In other words, the use of asthma medications might
still have had enough of an effect on lung function to obscure
a relationship between PEFR and smoke exposure, despite the
attempt to control for this influence statistically. Finally, due to
the timing of subject recruitment, it is not clear how many child-
days of observation during the fires actually contributed to the
analysis. The reported data suggest that this study is likely to
have had very limited statistical power.
The results from these studies appear to conflict with those
conducted in North America. As noted earlier, however, all have
significant limitations that suggest caution in generalizing the re-
sults. It is also possible that there is less respiratory toxicity from
bushfire smoke than from forest fire smoke due to chemical and
physical differences between the two. Two more recent stud-
ies from Australia have reported associations between bushfire
smoke and health impacts. For example, a study undertaken at
the only hospital in Darwin (northwestern Australia) evaluated
the association between daily asthma ER visits (adjusted for
influenza and day-of-week effects) and measured PM
10
over a
7-mo period, which included 2 bushfire smoke episodes. Bush-
fires represent the principal regional source of significant levels
of air pollution in Darwin during the dry season in which this in-
vestigation took place. Increased asthma visits were associated
with PM
10
concentrations, especially for days on which PM
10
concentrations exceeded 40 µg/m
3
(Johnston et al., 2002). The
adjusted rate ratio per 10-µg/m
3
increase in PM
10
was 1.20 (CI
1.09, 1.34). The largest association was observed for a 5-day lag,
comparing days when PM
10
exceeded 40 µg/m
3
with those on
which PM
10
was less than 10 µg/m
3
(adjusted rate ratio = 2.56
[CI 1.60, 4.09]). Unlike the prior studies of biomass smoke con-
ducted in Australia, this investigation clearly had adequate sta-
tistical power to detect an association between PM and asthma
visits. Though the time-series analysis did not control for pollen
or mold, which are not routinely monitored in Darwin, the inves-
tigators considered it “extremely unlikely” that either of these
would vary systematically with bushfire smoke. This assessment
by the authors is probably true, but without analyzing the smoke
for these bioaerosols, it is not possible to state definitively that
they did not confound the results.
Chen et al. (2006) evaluated the relationship between respira-
tory hospital admissions in Brisbane, Australia, and particulate
matter (PM
10
)fora3
1
2
-yr period that included 452 days (35%
of the study period) categorized as days with bushfires (>1ha
burned) in the study region, based on review of fire records. Dur-
ing the bushfire periods, the median of daily respiratory hospital
admissions in Brisbane was 34 (range: 9–76) and the daily mean
PM
10
was 18.3 µg/m
3
(range: 7.5–60.6 µg/m
3
), compared to a
median of 32 respiratory hospital admissions per day (range: 7–
91) and daily mean PM
10
of 14.9 µg/m
3
(range: 4.9–58.1 µg/m
3
)
during non-bushfire days. The authors categorized PM
10
values
into low (<15), medium (15–20) and high (>20), rather than
using a continuous variable for PM
10
. This may have resulted
in a loss of information about the potential impacts of extreme
values and possibly a bias toward (or away from) the null hy-
pothesis of no effect (Dosemeci et al., 1990). In addition, the
authors noted that the single PM
10
monitor used in this study
was upwind of many of the fires, indicating that the populations
affected were exposed to higher PM
10
and smoke concentrations
than those reported, which could have resulted in an overesti-
mate of the magnitude of effect. Nonetheless, for both bushfire
and nonbushfire periods, increased PM
10
concentrations were
associated with increased relative risks for respiratory hospital
admissions, with some suggestion of slightly stronger associa-
tions on the days with the highest daily PM
10
concentrations (i.e.,
>20 µg/m
3
)onbushfire (RR = 1.19, CI 1.09, 1.30, for same-day
PM
10
concentrations) versus nonbushfire (RR = 1.13, CI 1.06,
1.23) days. The results of this study are consistent with many
other time-series investigations of PM and, at a minimum, indi-
cate that the associations between PM
10
and respiratory health
admissions on bushfire days were at least as great as those on
days when other sources of PM
10
predominated.
Major regional episodes of air pollution from vegetation fires
in Southeast Asia have been the subject of several investigations
80 L. P. NAEHER ET AL.
and surveillance programs. An analysis of emergency room vis-
its for asthma in Singapore during a 1994 episode of regional
pollution resulting from forest and plantation fires reported an
association between PM
10
and emergency room visits for child-
hood asthma. During the “haze” period, mean PM
10
levels were
20% higher than the annual average. Although a time-series anal-
ysis was not conducted, the authors suggested that the associa-
tion remained significant for all concentrations above 158 µg/m
3
(Chew et al., 1995).
Reports from surveillance monitoring activities conducted
during major Southeast Asian episodes in 1997 and 1998 also
indicated effects on health care utilization. In Singapore, for
example, there was a 30% increase in hospital attendance for
“haze-related” illnesses: A time-series analysis indicated that a
PM
10
increase of 100 µg/m
3
was associated with 12%, 19%,
and 26% increases in cases of upper respiratory tract illness,
asthma, and rhinitis, respectively. It is not clear why rhinitis
constituted a separate diagnostic category in this investigation,
rather than being included with upper respiratory tract illness.
This analysis did not observe any significant increases in hospi-
tal admissions or mortality (Emmanuel, 2000). Similar findings
were also observed in Malaysia (Brauer, 1998; Leech et al.,
1998).
Preliminary results from a study of 107 Kuala Lumpur
schoolchildren found statistically significant decreases in lung
function between preepisode measurements in June–July 1996
and measurements conducted during the haze episode in
September 1997 (Hisham-Hashim et al., 1998). A convenience
sample questionnaire survey conducted in Indonesia during the
1997 haze episode also suggested acute impacts on respiratory
and cardiovascular symptoms (Kunii et al., 2002). Of 539 inter-
viewees, 91% reported respiratory symptoms (cough, sneezing,
runny nose, sputum production, or sore throat), 44% reported
shortness of breath on walking, 33% reported chest discom-
fort, and 23% reported palpitations. Although the numbers were
small, respondents with asthma or heart disease tended to ex-
perience a greater proportion of moderate and severe symptoms
relative to those without preexisting disease. Despite these find-
ings, however, the cross-sectional nature of the sampling and
reporting and the absence of an unexposed reference popula-
tion weaken any inference of a causal relationship between the
smoke and these symptoms.
In another study of the 1997 Southeast Asia haze episode,
Tan and colleagues (2000) obtained blood samples at weekly
intervals from 30 Singaporean military recruits who followed
standardized outdoor routines during the episode. The mean
24-h PM
10
level during the episode was 125.4 µg/m
3
. Ana-
lyzing the numbers of immature inflammatory cells (polymor-
phonuclear cells or PMNs) in the subjects’ blood in relation to
daily measures of several pollutants, these investigators found
the strongest relationship with same-day PM
10
, though a 1-day
lag of this metric was also statistically significant. Although
these results are insufficient to establish a causal relationship,
they suggest that smoke inhalation stimulated the bone marrow
to eject immature PMNs into the circulation.
Recently, Mott et al. reported several related examinations of
the Indonesian fires on hospitalizations and survival (Mott et al.,
2005). In analyses of the fire period (August through October
1997) compared with a 31-mo baseline period (January 1995
through July 1997), they reported fire-related increases of 50%
and 83% for admissions due to COPD and asthma among indi-
viduals aged 40 to 64, and an increase of 42% for COPD among
individuals aged 65 and older. In a time-series analysis in which
the baseline period was used to generate predicted numbers of
hospitalizations by age group for the fire period, the observed
admissions were significantly elevated for several respiratory
categories (asthma and COPD), principally among the 40–64 yr
age stratum. There was no significant elevation of admissions for
total circulatory diseases, though observed ischemic heart dis-
ease (IHD) admissions (n = 6) for the 18–39 yr age stratum were
slightly above the 95% upper limit predicted (n = 5.7). However,
the small numbers involved, coupled with the absence of a sig-
nificant elevation of IHD admissions in older age groups, sug-
gest caution in interpreting this relationship. Finally, Mott and
colleagues examined repeat hospitalizations and survival during
the fire period compared with the corresponding periods in 1995
and 1996. Individuals over age 65 with prior hospitalizations for
any cardiorespiratory disease, any respiratory disease, or COPD
in particular were more likely to be re-hospitalized during the
fire period, especially for respiratory causes, compared with the
corresponding periods in 1995 and 1996. In particular, individ-
uals with a prior history of hospitalization for COPD were more
likely to be rehospitalized for COPD or die from any cause
during the fire period (an approximately 44% increase for both
outcomes combined); this phenomenon was only manifest when
smoke levels exceeded approximately 150 µg/m
3
.
Only one other study has evaluated the impacts of air pol-
lution from vegetation fires on mortality. Sastry (2002) evalu-
ated the population health effects in Malaysia of air pollution
generated by a widespread series of fires that occurred mainly
in Indonesia between April and November 1997. The results
showed that the haze from these fires was associated with dele-
terious effects on population health in Malaysia and were in
general agreement with the mortality impacts associated with
particles in urban air (Sastry, 2002). A 10-µg/m
3
increase in
PM
10
measured in Kuala Lumpur was associated with 0.7 % (all
ages) and 1.8% (ages 65–74) increases in adjusted relative risks
of nontraumatic mortality. Visibility-based estimates of PM con-
centrations in Kuching, a city closer to the fire sources, were also
associated with increased mortality.
In a subsequent toxicological examination involving rabbits, these same
investigators found that repeated PM
10
instillations into the respiratory tract
resulted in increased production of PMNs in the bone marrow and an acceleration
of their release into the blood, both of which were associated with the numbers
of particles ingested by the animals’ alveolar macrophages (Mukae et al., 2001).
WOODSMOKE HEALTH EFFECTS: A REVIEW 81
With the exception of three of the Australian bushfire in-
vestigations, all of which have significant structural limitations,
the epidemiologic studies of indoor and community exposure to
biomass smoke indicate a generally consistent relationship be-
tween exposure and increased respiratory symptoms, increased
risk of respiratory illness, including hospital admissions and
emergency room visits, and decreased lung function. Several
studies suggest that asthmatics are a particularly susceptible sub-
population with respect to smoke exposure, which is consistent
with the results of many studies of the impacts of ambient air
pollution. The effects of community exposure to biomass air
pollution from wildfires on mortality have not been sufficiently
studied to support general conclusions.
Agricultural Burning
There have been few studies of the impacts of agricultural
burning, despite growing concern about its potential impact on
human health (Tenenbaum, 2000). In one Canadian study, 428
middle-aged subjects with slight-to-moderate airway obstruc-
tion were surveyed about respiratory symptoms during a 2-wk
period of exposure to straw and stubble combustion products.
During the exposure period, 24-h average PM
10
levels increased
from 15–40 µg/m
3
to 80–200 µg/m
3
. One-hour levels of CO and
nitrogen dioxide reached 11 ppm and 110 ppb, respectively. To-
tal volatile organic compound levels increased from preepisode
levels of 30–100 µg/m
3
to 100–460 µg/m
3
during the episode.
Although 37% of subjects were not bothered by smoke at all,
42% reported that several respiratory symptoms (cough, wheez-
ing, chest tightness, shortness of breath) developed or became
worse due to the air pollution episode and 20% reported that
they had breathing trouble. Subjects with asthma and chronic
bronchitis were more likely to be affected, and women appeared
to be more susceptible than men for several symptoms (cough,
shortness of breath, nocturnal awakening) (Long et al., 1998). In
contrast, current cigarette smokers reported significantly fewer
symptoms than the former smokers constituting the rest of the
study population. This study indicates that, besides woodsmoke,
biomass air pollution from agricultural burning is associated
with increased respiratory symptoms among a susceptible pop-
ulation with preexisting lung disease.
A time-series study in California suggested that agricultural
burn smoke was associated with serious exacerbations of asthma.
The association between asthma hospital admissions and the
burning of rice field stubble and waste rice straw was examined
in Butte County, California, over a 10-yr period (Jacobs et al.,
1997). Although burning was not associated with any measure-
ments of major air pollutants (probably because monitors were
not sited to provide optimal measurement of burn smoke), burn
acreage was significantly associated with an increased risk of
asthma hospitalization and showed an exposure-response trend.
The greatest risk of hospitalization was observed on days when
500 or more acres were burned (relative risk [RR] 1.23, CI 1.09,
1.39).
A recent cross-sectional study in three rural villages in
Iran also evaluated the relationship between rice stubble burn-
ing and respiratory morbidity, especially asthmatic symptoms
(Golshan et al., 2002). During a burning period lasting several
weeks, PM
10
concentrations doubled. Based on responses to a
physician-administered survey before and after this episode, the
investigators reported significant increases in the prevalence of
asthma attacks, use of asthma medications, the occurrence of
nocturnal sleep disturbances, and other respiratory symptoms
among 994 residents of an agricultural region. Several measures
of pulmonary function also decreased significantly.
The relationship of rice stubble burning with asthma was also
studied in Niigata prefecture, Japan (Torigoe et al., 2000). In
this study, measured PM
10
concentrations were associated with
monthly asthma hospital admissions and ER visits in a region
where rice straw burning emissions led to high particle concen-
trations during the September–October burning season. During
the period 1994–1998, both asthma ER visits and hospitaliza-
tions were significantly higher in September than in almost all
other months of the year except October and November (for ER
visits; hospitalizations in the month of December were also not
significantly different from September). Although PM
10
levels
were not associated with monthly ER visits for asthma, the in-
vestigators reported a significantly higher number of asthma ER
visits on days when rice straw burning occurred and the follow-
ing 2 days (7.1 ± 3.9) versus other days (4.5 ± 3.3). The latter
comparison would have better time resolution than an analysis
of monthly average of asthma exacerbations, and should proba-
bly be accorded greater weight. Although this investigation also
involved a parental questionnaire suggesting more asthma ex-
acerbations in children during the rice burning season than at
other times of the year, an autumn peak in asthma flares is also
common in other parts of the world where rice burning does
not occur. In general, multiple findings in this investigation are
suggestive of a rice smoke effect on asthma, but several limita-
tions of the study design constrain both causal inference and the
generalizability of the findings.
A metric commonly used as a surrogate of exposure to
biomass smoke is the amount of agricultural land burned, as
in the Butte County, California, study mentioned earlier (Jacobs
et al., 1997). Norris (1998) evaluated the association between
acres of grass seed residues burned around Spokane, WA, and
visits to local emergency departments for asthma. (Norris, 1998).
During one burning event, peak PM
10
concentrations in Spokane
reached 100 µg/m
3
(Figure 1). Using a bivariate indicator
(20 days with >499 acres burned) for the exposure surrogate,
an association with increased emergency department visits for
children was observed (RR 1.30, CI 1.08, 1.58).
Afew studies have specifically examined air pollution and
health effects associated with the burning of sugar cane. In
Brazil, daily indirect measurements (sedimentation of particle
mass) of air pollution during the sugar cane burning season in
1995 were associated with the number of patients visiting hospi-
tals for inhalation therapy for acute respiratory distress (Arbex
82 L. P. NAEHER ET AL.
FIG. 1. PM
10
measured downwind of a grass-burning event in Spokane, WA, at the Rockwood residential monitoring site (Septem-
ber 1994). Note: Modified from Norris (1998).
et al., 2000). The relative risk of such a hospital visit associated
with an increase of 10 mg in the sediment was 1.09 (1–1.19);
this association displayed an exposure-response relationship as
well. Boopathy and colleagues presented a descriptive analysis
of asthma hospital visits to a medical center in Houma, LA, dur-
ing 1998–1999 (Boopathy et al., 2002). The area served by this
medical center accounted for approximately 27% of Louisiana’s
sugar-cane cultivation during this period. Although no air pol-
lution measurements were available, asthma hospital visits in-
creased dramatically during the October–December sugar-cane
burning season. As noted earlier, however, an autumn peak in
asthma exacerbations is common, and respiratory infections (the
main precipitating factor for severe asthma attacks) also typi-
cally increase in frequency during this time. Therefore, it would
be inappropriate to infer a causal relationship between sugar-
cane burning and asthma hospital visits based on this descriptive
study. Boeniger and coworkers (1991) conducted an exposure
assessment of smoke during sugar-cane harvesting in Hawaii
in 1987 (Boeniger et al., 1991). They collected both area and
personal samples. The concentration of PM increased by at least
20 and up to 70 times the measured background levels at the sam-
pling sites chosen, but were highly variable, making exposure
assessment difficult. A subsequent study of Hawaiian sugar-cane
workers, however, reported no elevated morbidity or mortality
rates or decreased lung function (Miller et al., 1993).
Together, these epidemiologic studies suggest that exposure
to products of agricultural burning, specifically the burning
of rice stubble/straw, may be associated with exacerbation of
asthma. In a chamber study of smoke generated by controlled
burning of rice stubble straw, Solomon and colleagues exposed
13 adults with allergic rhinitis (age range 24–55) at rest to fil-
tered air, rice-straw smoke (RSS) at 200 µg/m
3
or at 600 µg/m
3
for 30 min, or RSS at 200 µg/m
3
on 3 consecutive days. Bron-
choalveolar lavage (BAL) was conducted at 6 h postexposure.
Of a variety of cell types and cytokines measured in BAL fluid,
the investigators found a near doubling of epithelial cells only
after the 3-day exposure, but no difference from filtered air ex-
posures in total white blood cells, macrophages, PMNs, lym-
phocytes, eosinophils, or interleukin-8 under any of the RSS
exposure conditions. Interestingly, this effect was not observed
at a higher concentration (600 µg/m
3
) delivered over a shorter
time interval, suggesting that repeated exposures may be neces-
sary, at least among individuals with allergic rhinitis (Solomon,
2003).
Several studies have also reported an increased risk (odds ra-
tios of 1.5–2.5) of lung cancer and mesothelioma among sugar-
cane workers, although specific job activities were not evalu-
ated and exposure measurements were not made (Rothschild &
Mulvey, 1982; Brooks et al., 1992). A case-control study (118
histologically confirmed lung cancer cases and 128 controls with
other cancers matched by age, sex, district of residence, and tim-
ing of diagnosis) in India found an increased risk of lung can-
cer in sugar-cane workers associated with postharvest burning
(odds ratio = 1.8, 95% CI 1.03.3) (Amre et al., 1999). It has
been suggested that this association may be due to the liberation
of asbestos-like biogenic silica fibers in sugar cane smoke.
RESIDENTIAL WOODSMOKE IN DEVELOPED
COUNTRIES: EXPOSURE AND HEALTH STUDIES
During winter in areas where wood is available, woodburning
is common in essentially every part of the developed world for
household heating. It is also popular for recreational use in fire-
places. This has implications for area-wide ambient levels and
indoor pollution as well as what can be called “neighborhood”
pollution, outdoors but sometimes localized in neighborhoods
where woodstoves are in use. Here we do not attempt to summa-
rize the evidence on the contribution of woodsmoke to ambient
pollution in the developed world, but provide typical examples
in different regions.
WOODSMOKE HEALTH EFFECTS: A REVIEW 83
FIG. 2. Source apportionment results for wintertime PM
10
in
San Jose, CA (1993–1994). Note: Modified from Fairley (1990).
Ambient and Neighborhood Levels
Source apportionment studies indicate that woodsmoke is a
major source of ambient PM during winter months in several
parts of the United States. Figure 2 shows data from San Jose,
CA, that indicate that 42% of the PM
10
during winter months
could be attributed to wood burning (Fairley, 1990). Chemical
mass balance receptor modeling of fine particles in Fresno and
Bakersfield, CA during wintertime identified both hardwood and
softwood as sources of PM and organic compounds (Schauer &
Cass, 2000), which were likely to have been due to residential
woodburning.
Outdoor PM levels in Seattle, WA; are also heavily influ-
enced by residential woodstoves. Data from 3 years of sampling
in Seattle were analyzed for sources using positive matrix fac-
torization (PMF) (Maykut et al., 2003). The PMF analysis found
that vegetative burning contributed 34% to the total sources of
PM in Seattle over 3 yr (Figure 3).
Another study utilized a large data set from a 2-year expo-
sure assessment and health effects panel study in Seattle dur-
ing September 2000–May 2001. Indoor, outdoor, personal, and
fixed-site PM monitoring data were available. The samples were
analyzed for elements using XRF, and positive matrix factoriza-
tion (PMF) was used to apportion sources (Larson et al., 2004).
Five sources contributed to indoor and outdoor samples: veg-
etative burning, mobile emissions, secondary sulfate, a chlo-
rine source, and a crustal-derived source. Vegetative burning
contributed the largest fraction of PM mass in all the samples
(35%, 49%, and 62% in indoor, outdoor, and personal mass,
respectively).
The distribution of particle-phase organic compounds has
been measured in communities with children participating
in the Southern California Children’s Health Study (CHS)
(Manchester-Neesvig et al., 2003). Concentrations of levoglu-
cosan, a good tracer for woodsmoke aerosol, were seen in all
12 CHS communities (Figure 4). The average concentration
FIG. 3. Source apportionment results for PM
10
in Seattle, WA
(1996–1999). Note: Modified from Maykut et al. (2003).
increased substantially in the winter, as would be expected
for woodsmoke emissions. The concentrations of levoglucosan
were highest at the Atascadero site, which is about 15 miles in-
land. Earlier, these investigators identified two additional sugar
anhydride tracers of woodsmoke (galactosan and mannosan) in a
study of urban sites in the San Joaquin Valley, California (Nolte
et al., 2001). These data may allow a separate estimation of the
effects of woodsmoke exposure on health outcomes.
In Canada, with cold winters and abundant forests,
woodsmoke is a major source of particle emissions. Figure 5
shows that household woodsmoke is responsible for more than
30% of annual PM emissions in 8 provinces and more than 10%
in the remaining 4. It is also more responsible for a significant
fraction of VOC emissions.
Christchurch, New Zealand, is another city impacted by
woodsmoke. It is estimated that more than 90% of wintertime
ambient PM comes from heating stoves and open fires burning
wood (McGowan et al., 2002). Frequent periods of air stag-
nation compound the problem by trapping PM near the ground,
and local meteorologists estimate that the relatively even mixing
results in fairly homogeneous PM exposure to the population.
Emissions inventories in Launceston, Australia, indicate that
household woodburning accounted for 85% of annual PM
10
emissions in 2000 and that a 50% reduction would be needed in
order the city to meet air quality standards.
Source apportionment studies in Denmark show that house-
hold woodburning was responsible for 47% of national PM
2.5
emissions in 2002. In addition, household woodburning grew by
about 50% during the 1990s, as compared to only 7% for total
energy use.
A recent phenomenon in the United States has been the use of
backyard wood-fired boilers for heating homes, which have not
http://www.ec.gc.ca/science/sandejan99/article1 e.html
84 L. P. NAEHER ET AL.
FIG. 4. Spatial distribution of winter time levoglucosan in Southern California (1995–1996). Note: Modified from Manchester-
Neesvig et al. (2003).
been regulated and often produce substantial pollution locally
(Johnson, 2006).
Indoor Levels
Relatively few measurements seem to have been reported
of indoor concentrations of woodsmoke in developed-country
households. A case-control study of woodstoves and health in
Navajo children in Arizona did include measurements of indoor
concentrations of respirable particles (PM
10
)in90households
FIG. 5. Importance of woodsmoke emissions in Canada by province. Note: Data from Health Canada. See footnote on page 86.
(Robin et al., 1996). Cases were children from birth to 24 months
of age hospitalized with acute respiratory illnesses and controls
that were not hospitalized. Sixty-three percent of the cases had
wood stoves in their homes, compared with 51% of the controls.
TWA concentrations (15-h) ranged from 22.2 µg/m
3
in houses
that used gas or electricity to 100 µg/m
3
in homes that heated
with wood alone.
Early studies of woodsmoke health effects often used the
presence or absence of a wood stove in the home as the
WOODSMOKE HEALTH EFFECTS: A REVIEW 85
indicator of exposure (see next section). Due to penetration
of woodsmoke particles indoors, these exposures may not be
due exclusively to indoor sources of woodsmoke. It has been
shown in a woodsmoke-impacted community that particles read-
ily penetrate inside residences (Anuszewski et al., 1998). The
contribution from outdoor-generated particles to indoor and per-
sonal exposure in Seattle, WA, residences has been estimated
using a recursive model (Allen et al., 2003, 2004). Nonlinear
regression was used to estimate particle penetration, particle
decay rate, and particle infiltration. Estimates of particle infil-
tration agree well with those derived from sulfur-tracer meth-
ods (R
2
= .78) (Sarnat et al., 2002). In a sample of 44 resi-
dences, outdoor-generated particles accounted for an average
of 79 ± 17% of the indoor PM concentration. These data sug-
gest that in epidemiologic studies of associations between health
outcomes and outdoor PM, much of the exposure to outdoor par-
ticles can occur inside the home. Other factors, such as the age of
the house, opening of windows, and air conditioning, can affect
penetration. In one study, home air conditioning was associated
with lower penetration of outdoor particles; moreover, the asso-
ciations between PM
10
and hospital admissions were lower in
cities with a higher prevalence of air conditioning (Janssen et al.,
2002). These findings imply that even woodstoves and fireplaces
operating well that vent most smoke outside may produce sub-
stantial exposures through penetration back into the house, a
characteristic of “neighborhood pollution.
Health Effects of Residential Woodburning
To date, only a single controlled exposure study of human
exposure to woodsmoke itself seems to have been published
(Barregard et al., 2006; Sallsten et al., 2006). Thirteen sub-
jects were exposed to realistic concentrations of woodsmoke
(200–300 µg/m
3
PM
2.5
) generated under controlled conditions
for two 4-h sessions, spaced 1 wk apart. In this study, expo-
sure to woodsmoke resulted in small exposure-related changes
in levels of inflammatory mediators and coagulation factors.
In addition, evidence of increased free radical-mediated lipid
peroxidation was observed in 9 of the 13 subjects. Although
this is the only controlled study of woodsmoke exposure pub-
lished to date and it observed a small number of subjects, it
is suggestive of woodsmoke-associated systemic inflammatory
effects.
The majority of information regarding direct human health ef-
fects associated with woodsmoke exposure is derived from a rel-
atively large number of epidemiologic studies have documented
respiratory effects of residential woodburning, especially in chil-
dren. One of the earliest studies was conducted in Michigan
by Honicky et al., who compared respiratory symptoms in
31 children who lived in homes with wood stoves with 31 chil-
dren who lived in homes without wood stoves (Honicky et al.,
A thorough summary of emissions from woodsmoke was published several
years ago (Larson & Koenig, 1994).
1985). Symptoms were categorized as mild, moderate, and se-
vere. The two groups did not differ with respect to mild symp-
toms, but differed significantly for severe symptoms ( p < .001).
A similar study was conducted in Boise, ID, by Butterfield et al.,
where respiratory symptoms were tracked in 59 children under
the age of 5
1
2
years during a winter season (Butterfield et al.,
1989). Symptoms such as wheeze, cough, and nocturnal awak-
ening were associated with presence of a woodstove.
Morris et al. (1990) evaluated the impact of indoor
woodsmoke child health on a Navajo reservation in Arizona by
assessing use of a well-child clinic (Morris et al., 1990). For 58
case-control pairs, the odds ratio (OR) for a serious acute lower
respiratory infection (ALRI: bronchiolitis or pneumonia) asso-
ciated with the presence of a wood stove was 4.2 ( p < .0012).
A more recent case-control study among slightly younger (1–
24 mo) Navajo children reached similar, but nonsignificant con-
clusions (OR 5.0, CI 0.6, 42.8) (Robin et al., 1996). Measured 15-
hPM
10
levels above 65 µg/m
3
were more frequent in households
with wood cookstoves (OR 7.0, CI 0.9 to 56.9). Adjustment for
potential confounders (including the number of children living
in the house, lack of running water or electricity, difficulty with
transportation to the clinic, type of home, and the temperature
on the PM
10
sampling day) had relatively little effect on the
magnitude of the associations. The low number of cases (45)
likely affected the precision of the estimates, reducing the in-
vestigators’ ability to detect significant associations between use
of wood-burning devices and respiratory infections. It is note-
worthy, however, that the magnitude of effect exceeds those gen-
erally found in developing-country studies of ALRI in children
(discussed later).
A questionnaire study of respiratory symptoms compared
residents of 600 homes in a high woodsmoke area of Seattle,
WA , with 600 homes (questionnaires completed for one parent
and two children in each residence) of a low woodsmoke area
(Browning et al., 1990). PM
10
concentrations averaged 55 and
33 µg/m
3
, respectively. When all age groups were combined,
no significant differences were observed between the high- and
low-exposure areas. There were, however, statistically signifi-
cantly higher levels of congestion and wheezing in 1- to 5-year-
olds between the 2 areas for all three questionnaires (1 baseline
questionnaire and 2 follow-up questionnaires which asked about
acute symptoms). This study supports findings from the other
investigations suggesting that young children are particularly
susceptible to adverse effects of woodsmoke.
In Seattle, WA, 326 elementary school children were stud-
ied during the heating seasons of 1988–1989 and 1989–1990
(Koenig et al., 1993). Monthly or bimonthly spirometry values
were collected during the school year. PM exposure was mea-
sured by light scattering using nephelometers. The exposure
metric used was the 12-h nighttime average (7 p.m. to 7 a.m.) to
reflect the hours when woodsmoke is most elevated. A random-
effects statistical model compared changes in FEV
1
and FVC
with changes in the light-scattering coefficient. The 26 chil-
dren with asthma showed a significant decrement (18 ml/µg/m
3
86 L. P. NAEHER ET AL.
PM
2.5
) for both measures of lung function. Children without
asthma showed no significant changes in lung function associ-
ated with PM values.
A companion study evaluated the impact of particulate mat-
ter on emergency room visits for asthma in Seattle (Schwartz
et al., 1993). A significant association was observed between
PM
10
particle levels and emergency room visits for asthma. The
mean PM
10
level during the 1-yr study period was 30 µg/m
3
.
At this concentration, PM
10
appeared to be responsible for 125
of the asthma emergency room visits. An exposure response re-
lationship was also observed down to very low levels of PM
10
,
with no evidence for a threshold at concentrations as low as
15 µg/m
3
. The authors indicate that on an annual basis 60%
of the fine particle mass in Seattle residential neighborhoods is
from woodburning.
Overall, health effects research in Seattle shows associa-
tions between PM
2.5
and lung function decrements in children
(Koenig et al., 1993), visits to emergency departments for asthma
(Norris et al., 1999), hospitalizations for asthma (Sheppard
et al., 1999), and increases in asthma symptoms in children
(Yu et al., 2000), as well as increases in exhaled nitric oxide
(Koenig et al., 2003, 2005). Since woodburning is the primary
source of fine particles in the Seattle airshed, the health effects
studies suggest a causal relationship.
Lung function in 410 schoolchildren in Klamath Falls, OR,
was studied during winter in high- and low-exposure areas were
studied where it has been estimated that woodsmoke accounts
for as much as 80% of winter period PM
10
(Heumann et al.,
1991). Winter PM
10
levels in the high exposure area ranged
from approximately 50 to 250 µg/m
3
, while levels in the low
exposure area ranged from 20 to 75 µg/m
3
. Lung function de-
creased during the wood-burning season for the children in the
high-exposure area, but not in the low-exposure area.
Two studies were conducted in Montana to evaluate acute
changes in lung function in children within a single community
at different levels of air pollution, and also to evaluate cross-
sectional differences in lung function between communities with
different air quality levels, as an indication of chronic impacts
(Johnson et al., 1990). Acute lung function decrements mea-
sured in 375 children were associated with increased levels of
particulates. The 24-h averages ranged from 43 to 80 µg/m
3
and from 14 to 38 µg/m
3
for PM
10
and PM
2.5
, respectively. The
chronic impact study also associated small decrements in lung
function with residence in communities with higher levels of air
pollution. Although particle composition was not measured di-
rectly in this study, measurements conducted in the acute study
community during the same period attributed 68% of the PM
3.5
to woodsmoke (Larson & Koenig, 1994).
Another study examined the relationship of woodstoves to
otitis media and asthma in a case-control study of home en-
vironmental air pollutants in Springville, NY (Daigler et al.,
1991). That study found use of woodstoves was more likely
to be present in homes of children with otitis media (OR 1.7,
CI = 1.03, 2.89).
In contrast, in a larger, prospective study of 904 infants in
Connecticut and Virginia, Pettigrew et al. found no relation-
ship between either woodstove or fireplace use and either single
episodes of otitis media or recurrent otitis media, which was de-
fined as 4 or more episodes during 1 yr (Pettigrew et al., 2004).
Data on infant respiratory symptoms (in this case, a physician’s
diagnosis of an ear infection) and hours of use of secondary
heating sources were collected in telephone interviews with the
mothers every 2 wk for 1 yr. Although both woodstove and fire-
place use were significantly associated with the outcomes in
bivariate models, these associations were absent in multivariate
models that adjusted for gender, race, day care, number of chil-
dren in the household, duration of breast-feeding, winter heating
season, use of gas appliances, season of birth, maternal educa-
tion, maternal history of asthma and allergy, and pets. On the
other hand, in the same study, woodstove but not fireplace use
was associated with total days of cough in these infants (RR
1.08, CI 1.00,1.16) (Triche et al., 2002).
In a panel study of adults (ages 18–70) in Denver, CO (Ostro
et al., 1991), the use of a fireplace or woodstove was associated
with an increase in daily moderate or severe shortness of breath
(OR 1.3, CI 1.1, 1.4). Use of woodstoves or fireplaces was second
only to the presence of smokers in the home, and more strongly
associated with shortness of breath than use of gas stoves or
occupational exposures. As this study included only subjects
with moderate to severe asthma, however, the findings may not
be generalizable across the entire clinical spectrum of asthma.
In a study of 888 women living in nonsmoking households
in Connecticut and Virginia, Triche and colleagues analyzed
daily respiratory symptom data collected during the winter heat-
ing seasons of 1994–1996 (Triche et al., 2005). Using Poisson
regression and controlling for age, race, allergic status, num-
ber of children, education, type of dwelling (single-family vs.
multi-unit), and state of residence, these investigators found that
each hour-per-day use of a fireplace was associated with several
reported respiratory symptoms, including cough (RR 1.05, CI
1.01, 1.09), sore throat (RR 1.04, CI 1.00, 1.08), chest tightness
(RR 1.05, CI 0.99, 1.12), and phlegm (RR 1.04, CI 0.99, 1.09).
These results suggest that use of a fireplace for4hwould increase
the risk of such symptoms by about 16–20%. No such associ-
ations were found for woodstove use, which the investigators
suggested may have been due to greater indoor emissions from
fireplaces.
Several large time-series studies have been conducted in com-
munities with known woodsmoke sources. The first was con-
ducted in Seattle over a 1-yr period (September 1989 Septem-
ber 1990) (Schwartz et al., 1993), during which there were 2955
emergency department visits for asthma to 8 hospitals. PM
10
TWA over 24 h ranged from 6 to 103 µg/m
3
, with a mean
of 29.6. In Poisson regressions controlling for weather, season,
time trends, age, hospital, and day of the week, the daily counts
of emergency room visits for persons under age 65 were sig-
nificantly associated with PM
10
exposure on the previous day.
The relative risk for a 30 µg/m
3
increase in PM
10
was 1.12
WOODSMOKE HEALTH EFFECTS: A REVIEW 87
TABLE 6
Woodsmoke in developed countries: A sample of studies
Location Woodsmoke concentration Source
Outdoors
Santa Clara County, CA 42% of CMB (Fairley, 1990)
Seattle, WA 49% of total PM
2.5
mass (Larson et al., 2004)
Atascadero, CA Levoglucosan (Manchester-Neesvig et al., 2003)
Atlanta, GA 11% of total PM
2.5
mass (Polissar et al., 2001)
Vermont 10–18% of PM
2.5
(Polissar et al., 2001)
Christchurch, New Zealand 90% of PM
2.5
in winter (McGowan et al., 2002)
Indoor/personal
Seattle, WA; personal 62% of total PM
2.5
mass (Larson et al., 2004)
Seattle, WA; indoor 35% of total PM
2.5
mass (Larson et al., 2004)
Fort Defiance, AZ Indoor PM
10
dominated by (Robin et al., 1996)
woodstove smoke
(CI 1.04, 1.2). A significant exposure-response trend was found
up to nearly 60 µg/m
3
.Woodsmoke contributed approximately
85% of the wintertime PM in residential areas during the study
period.
Two time-series studies have been conducted in Santa Clara
County, California, an area in which woodsmoke is the single
largest contributor to winter PM
10
(see Table 6). Particulate lev-
els are highest during the winter in this area. The first study was
one of the initial mortality time-series studies which indicated
an association between relatively low PM
10
levels and increased
daily mortality (Fairley, 1990). A study of asthma emergency
room visits in Santa Clara County and winter PM
10
found a
relative risk for an emergency visit, adjusted to a 60-µg/m
3
in-
crease in PM
10
,tobe1.4 (CI 1.2, 1.7) at 20
F (Lipsett et al.,
1997).
A study in Christchurch, New Zealand, examined the asso-
ciation between hospital admissions and PM
10
for the period
1988–1998. Ambient PM
10
levels during the study period av-
eraged 25 µg/m
3
, with a maximum of 283 µg/m
3
. The results
were stratified into total cardiac and total respiratory admissions.
The estimated percentage increases per interquartile increase in
PM
10
(approximately 15 µg/m
3
) for all age groups was 3.37 (CI:
2.3–4.4) for respiratory admissions and 1.26 (CI: 0.3–2.2) for
cardiac admissions, but with no increase for ischaemic heart dis-
ease (McGowan et al., 2002). As noted in Table 6, woodsmoke
makes up 90% of wintertime PM
10
. One interpretation of these
data is that fine particles from wood burning are more closely as-
sociated with adverse respiratory effects than adverse cardiovas-
cular effects. Data from Seattle, WA; support this interpretation,
as studies show PM
2.5
in Seattle associated with asthma aggra-
vation (Koenig et al., 1993, 2003) but do not find similar associa-
tions with cardiac events such as myocardial infarction (Sullivan
et al., 2005) or sudden cardiac address (Levy et al., 2001).
On the other hand, several studies have failed to find asso-
ciations between woodstove use and respiratory health (Tuthill,
1984; Eisner et al., 2002). The Tuthill study evaluated health
outcomes associated with woodsmoke or formaldehyde expo-
sures in children. An association was seen between respiratory
symptoms and prevalence of respiratory disease and estimated
exposure to formaldehyde but not seen between these endpoints
and estimated exposure to wood smoke. Eisner et al. (2002) stud-
ied asthma outcomes in adult subjects exposed to combustion
sources indoors that included woodsmoke and environmental to-
bacco smoke. Although higher use of woodstoves and fireplaces
was associated with more severe asthma at baseline, there was
no association between use of wood burning devices and asthma
aggravation after the 18-mo follow-up.
Summary of Residential Woodsmoke Epidemiology
The studies discussed in this section do have some limita-
tions. For instance, in common with most nonoccupational air
pollution epidemiologic studies, few had personal exposure in-
formation. These studies, however, do encompass a gradient
of health impacts associated with woodsmoke and PM. The
indicators of adverse effects run from increases in respiratory
symptoms to lung function decreases to visits to emergency de-
partments and finally hospitalizations. It is highly unlikely that
this pyramid of adverse effects could be built if the associations
reported were not real.
In assessing the ”strength” of air pollution health effects data,
Bates (1992) concluded that the question of coherence is crucial.
He went on to state that such coherence may exist at different lev-
els: within epidemiologic data, and between epidemiologic and
toxicological data, and between epidemiologic data and con-
trolled studies. Woodsmoke exposure of residents inside their
homes is supported by the infiltration data discussed earlier in
this section. Therefore, it is reasonable to conclude that exposure
to the concentrations and durations of woodsmoke associated
with residential woodburning is likely to cause a variety of ad-
verse respiratory health effects. The biological plausibility for
this conclusion is supported both by the toxicology literature,
limited controlled exposure studies, and the wealth of data on
88 L. P. NAEHER ET AL.
health effects of biomass burning in developing countries re-
viewed below.
Other reviewers have come to similar conclusions (McGowan
et al., 2002). Boman et al reviewed the literature relating to ad-
verse health effects from ambient exposure to woodsmoke and,
comparing the results of studies of acute exposure to those done
in areas without much woodsmoke, concluded that there was no
reason to think that the adverse impacts of acute woodsmoke
exposure would be less than those associated with other sources
of ambient PM (Boman et al., 2003).
Statisticians are attempting to derive models that will allow
source apportionment data to be added to health endpoint analy-
sis without creating undue bias (Lumley & Liu, 2003). Creation
of such models will help apportion specific health outcomes to
specific sources such as woodburning.
This short summary of published studies shows that signif-
icant exposures to ambient woodsmoke do occur in developed
countries and that important health effects have been demon-
strated to result.
BIOMASS USE IN DEVELOPING COUNTRIES
Indoor Air Pollution From Household Fuels
Throughout human history, the largest exposures to particle
air pollution probably occurred in households through use of
wood and other forms of biomass as sources of cooking, dry-
ing, and space-heating energy. Even today, such uses probably
account for the majority of human exposure to respirable PM
worldwide because of the continued high dependence on such
household fuels (Smith, 1993). As shown in Figure 6, for ex-
ample, about half of the world’s households are still thought
to cook with solid fuels on a daily basis (Smith et al., 2004).
FIG. 6. Map of solid fuel use. Source: Smith et al. (2004).
FIG. 7. Emissions and energy characteristics of typical Indian
cookstoves. Note improvement in combustion and total effi-
ciency moving from solid to liquid and gaseous fuels and great
reduction in emissions per unit energy delivered. Source: Data
from Smith et al. (2000).
Of this, about 95% consists of wood and agricultural residues.
Household use of mineral coal for cooking, which makes up the
remainder, is mainly confined to China.
In simple devices, like the household stoves commonly used
in developing countries, biomass fuel does not combust cleanly.
Systematic emissions studies in India and China, for example,
have generally validated the so-called “energy ladder” concept
with regard to the emissions from combustion of household fuels
(Smith et al., 1994). As shown in Figure 7, the energy ladder
WOODSMOKE HEALTH EFFECTS: A REVIEW 89
starts at the bottom with low-quality biomass fuels, such as cow
dung, moves up through crop residues, to wood. Further up the
ladder lie liquid and then gaseous fuels (kerosene and liquefied
propane gas, LPG), with electricity being at the top (i.e., with
the lowest emissions). Nominal combustion efficiency (percent
of fuel carbon emitted as CO
2
)isaslow as 80% for the poorer
fuels and reaches more than 99% with gaseous fuels (Smith
et al., 2000). In combination with the low thermal efficiency of
solid fuel stoves, the result is differences in emissions per meal
of nearly two orders of magnitude between gaseous and solid
fuels. In addition to cleanliness, the cost, complexity, and ease
of household use generally increase as one moves up the ladder
(Office of Technology Assessment, 1992). Broadly, as average
household income increases in societies, usage tends to move up
the ladder, although not always to the last rung (electricity). This
is shown by econometric studies at the national level (Mehta,
2003). In individual communities, however, the situation is often
more complicated, particularly during transition phases, when
households may straddle several rungs of the ladder at once by
using multiple fuels depending on prices, seasons, availability,
and so forth (Sinton et al., 2004).
As noted elsewhere in this report, poor combustion efficiency
creates high emission factors for wood and other biomass across
a wide range of health-damaging pollutants. High emissions,
however, do not necessarily lead to high exposures unless they
reach human breathing zones. Unfortunately, however, condi-
tions in hundreds of millions of Third World households are
nearly ideal to maximize exposures from emissions. A large, but
unknown, fraction of daily cooking is done in unvented stoves,
that is, stoves in which the emissions are released directly into the
living area and not vented through a chimney or hood. Although
there are not systematic surveys in developing-country settings,
about 200 studies of indoor air quality (IAQ) measurements in
households using solid fuels have been published, more than
half from China.
Table 7 shows a summary of these studies for
the two most widely measured pollutants, PM and CO. These
studies have been published between 1968 and 2003. The stud-
ies in South Asia were mainly conducted in Nepal and in India,
with only one reported from Bangladesh. The studies in Africa
come mainly from Kenya, Gambia, and South Africa. Most of
the Latin American studies have been conducted in Guatemala
and Mexico.
Most of these studies were conducted in rural settings and
attempted to characterize the distribution of concentration lev-
els in the kitchens, with the earlier studies reported from the
highlands in different parts of the world. Also, there is little
information available on seasonal effects or differences across
the various meals cooked in a day. Meal cooking time varied
from study to study, generally between 30 min and 3 h, with
one study reporting up to 8 h. Although several studies made
See the Chinese IAQ database (Sinton et al., 1996) and the non-
Chinese IAQ database (Saksena et al., 2003) both available at http://ehs.
sph.berkeley.edu/krsmith.
comparisons between the traditional and the improved stoves,
in this summary table that distinction has not been made.
A highly polluting source releasing pollution indoors at times
and places when people are always present (household cook-
ing) has a potential to produce high exposure. Put another way,
the associated intake fraction (fraction of material released that
is actually inhaled by someone) is orders of magnitude higher
for indoor than for outdoor sources of air pollution (Bennett
et al., 2002). Although the uncertainties are large, the available
evidence would indicate that the total exposure to combustion-
derived fine particles from indoor solid fuel use is larger than that
from all outdoor sources of pollution in the world (Smith, 1993).
Even in communities where most households use chimneys,
however, the intake fraction can be substantially higher than
for typical outdoor sources since the smoke may sit in the area
among the houses in what is called “neighborhood pollution.
Such pollution may not be fully reflected by ambient moni-
toring data, but may nevertheless substantially influence local
exposures (Smith et al., 1994). This same phenomenon exists
in developed countries as well, for example, from household
fireplaces, as discussed earlier. Because of their almost universal
role as household cooks, the highest exposures from household
use of solid fuels, however, seem usually to occur to women
and their youngest children who are with them during cooking,
although significant exposures can accrue to other household
members as well (Balakrishnan et al., 2004).
Although few studies have linked linked IAQ measurements
to ill health, a growing number of epidemiologic studies have
found significant risks of various exposure indicators and ill
health in developing-country biomass-using households. Such
exposure indicators include use of solid or “dirty” fuel versus
liquid/gas “clean” fuel; using a stove with a flue or without; years
cooking with solid fuel, and, for infants, being carried on their
mother’s back while cooking or not. Taking advantage of the
increasing number of such studies, the recent global Compar-
ative Risk Assessment (CRA) managed by the WHO included
indoor as well as outdoor air pollution among the 26 risk factors
examined (Ezzati et al., 2002).
The available epidemiologic evidence was divided into three
categories, as shown in Table 8. Considered sufficient in quan-
tity and quality to justify inclusion in the global CRA was evi-
dence only in the top category: acute lower respiratory infections
(ALRI: pneumonia) in young children, chronic obstructive pul-
monary disease (COPD) in adults, and lung cancer in adults (for
coal smoke only). The odds ratios shown in the table are the