Indoor air pollution and vulnerability to
bacterial pneumonia in young children
Lessons from the developing world
Nigel Bruce
Department of Public Health and Policy, University of Liverpool,
UK
Overview
• Indoor (household) air pollution
• Available ‘measures’ of possible bacterial
pneumonia in young children
• Three types of evidence
– Ecological
– Epidemiological studies:
• Systematic review/meta-analysis
• RESPIRE trial
– Mechanistic studies
• Conclusions
• New/ongoing field trials
Household air pollution
• 3 billion use solid fuel as
primary cooking fuel
• 1.2 billion no electricity: use
simple kerosene lamps
• Inefficient stoves/lamps lead
to high emissions of ‘PIC’
• Health-damaging pollutants:
 Small particulates (PM2.5)
 Toxic gases, carcinogens and
irritants
• Typical PM2.5 levels 500
µg/m3, vs. WHO AQGs of 10
• Exposure highest for women
(pregnant) & young children
Solid fuel use for cooking: 2010
Available measures of (possible) bacterial
pneumonia in various types of study
• Mortality (bacterial higher CF):
– ALRI (mix): WHO stats; DHS
– Pneumonia: diagnosed; VA?
• Severe pneumonia (bacterial
more likely to be severe):
– Clinical signs
– Hypoxaemia (pulse oximetry)
• Aetiology:
– Antigen tests (NPA, urine, blood,
lung)
– Lung aspirate or blood culture
• Mechanistic studies:
– In vivo: survival after infection
with S. Pneumoniae (mice)
– In vitro: C-loaded AM killing of S.
pneumoniae.
Ecological association
Death rates from
ALRI in children
under 5 years
(2010)
Source WHO
Ecological association
Death rates from
ALRI in children
under 5 years
(2010)
Source WHO
Percentage of
homes relying on
solid fuels for
cooking (2010)
Source WHO
Systematic review of epidemiological
studies
• Published (Dherani et al 2008)
– Updated
• Eligible studies:
– Cross-sectional, analytic observational, RCT
– Exposure: very few measured HAP or exposure 
fuel, stove-type, behaviour contrast
– Outcome: reported symptoms/signs  community
ALRI  clinical diagnosis  CXR and bacteriology
• Results:
– All non-fatal ALRI (severity not defined); n=21
– Non-fatal, severe ALRI; n=4
– Fatal ALRI; n=4
SRMA: pooled ORs (95% CI)
Outcome
N
I2 (p-value)
Random or
Fixed effect
Publication
bias
(p-value)
All ALRI*
(severity not
defined)
21
61%
(p<0.0001)
Random
Begg’s: 0.56 1.56
P<0.0001
Egger’s: 0.09 (1.33, 1.83)
OR
(95% CI)
P-value
*Includes O’Dempsey (Gambia 1996): Pneumococcal disease on blood culture (79%
pneumonia) OR=2.55 (0.98 – 6.65)
SRMA: pooled ORs (95% CI)
Outcome
N
I2 (p-value)
Random or
Fixed effect
Publication
bias
(p-value)
All ALRI*
(severity not
defined)
21
61%
(p<0.0001)
Random
Begg’s: 0.56 1.56
P<0.0001
Egger’s: 0.09 (1.33, 1.83)
Severe1
4
51% (p=0.10) N/A
Random
2.04
P=0.001
(1.33, 3.14)
Fatal2
4
0% (p=0.64)
Fixed
2.80
P<0.0001
(1.81, 4,34)
N/A
OR
(95% CI)
P-value
1Severe:
includes physician clinical definition (n=3) and low oxygen saturation on pulse
oximetry (n=1)
2Fatal: includes verbal autopsy (n=2), parental recall of signs (n=1) and deaths in hospital
following radiological confirmation of pneumonia (n=1)
*Includes O’Dempsey (Gambia 1996): Pneumococcal disease on blood culture (79%
pneumonia) OR=2.55 (0.98 – 6.65)
RESPIRE Trial
• Objective: impact of HAP reduction on pneumonia
incidence in children <18 months
– Primary: ITT analysis
– Secondary: exposure-response analysis
• Rural, highland communities of Comitancillo and
San Lorenzo, alt. 2200 – 3000 m
• 518 homes (pregnant woman, child <4 months)
randomised to keep open fire or use ‘plancha’
• Children followed to 18 months: ~30,000 child
weeks
• Surveillance for pneumonia cases and all deaths
Control and intervention stoves
Traditional open 3-stone fire:
kitchen 48-hour PM2.5 levels of
500 - 1000 μg/m3
The plancha chimney wood stove,
locally made and popular with
households
Overview of child health outcomes assessment
Follow-up at weekly visit
Home
Weekly visit
• Well
Community
centre
Hospital
Study doctor
examines
Assessed by
duty doctor
• Mild illness
•Pulse oximetry
• Referral to
study doctor
•If pneumonia,
RSV* test and
refer for CXR
Study team
obtain CXR
and inpatient
data and
diagnosis
•Refer if very ill
Child dies
Verbal
autopsy
Child dies
Health outcome
definitions
Verbal
autopsy
* Respiratory syncytial virus
Home IAP and exposure assessment
methods
• All homes:
– 48 hr CO tube child
(3 monthly)
– mother (6 monthly)
• Random sub-sample
(n=40+40):
–
–
–
–
–
3-monthly
CO (tube, Hobo)
PM (filter, pump)
Continuous PM
Mother breath CO
(COHb)
Effect of intervention stove on (i) kitchen IAP
and (ii) personal exposure
8
Geometric mean CO (ppm)
7
6
↓90%
5
↓61%
4
↓52%
3
2
1
0
Kitchen
Child
Open fire
Mother
Plancha
Smith et al, J Exp Sci Env Epidemiol 2009
Physician-assessed outcomes (ITT)
Case finding
Outcome
Physician
diagnosed
pneumonia
Investigations:
- Pulse
oximetry
- RSV direct
antigen test
- Chest X-ray
RR (95% CI)
P-value
All
0.78 (0.59, 1.06)
0.095
- Severe (hypoxic)
0.67 (0.45, 0.98)
0.042
CXR +ve
0.74 (0.42, 1.15)
0.231
- CXR +ve & hypoxic
0.68 (0.36, 1.33)
0.234
RSV +ve
0.76 (0.42, 1.16)
0.275
- RSV +ve & hypoxic
0.87 (0.46, 1.51)
0.633
RSV -ve
0.79 (0.53, 1.07)
0.192
- RSV –ve & hypoxic
0.54 (0.31, 0.91)
0.026
Plancha
Open fire
Exposureresponse analysis
• Mean PM2.5 exposure
equivalent (µg/m3):
• OF: 250
• Plancha:125
• Lowest exposure
decile ~50 µg/m3
• Statistically significant
E-R relationships
• Implications: low
exposure (<30-50
µg/m3) needed to
prevent most cases
Mechanisms: focus on HAP
Pollutants
• Carbonaceous PM (<10
microns; <5 into alveoli)
• Gases (irritant, toxic):
– NO2, CO
• Hydrocarbons (cancer):
– Benzene
• Polyaromatic HC (cancer):
– benzo [A] pyrene
• Aldehydes (irritant):
– Formaldehyde
– Acrolein
Mechanisms: focus on HAP
Pollutants
Defence mechanisms
• Carbonaceous PM (<10
microns; <5 into alveoli)
• Gases (irritant, toxic):
Filtering
Muco-ciliary
clearance
– NO2, CO
• Hydrocarbons (cancer):
– Benzene
• Polyaromatic HC (cancer):
Physical
barrier of
epithelium
– benzo [A] pyrene
• Aldehydes (irritant):
– Formaldehyde
– Acrolein
Immune response including:
alveolar macrophages (AM),
opsonisation, IgA, IgG, surfactant,
plasma, etc.
AM function: carbon loading
• Biomass fuel users
show higher carbon
loading in AMs
• Human (BAL) study
(Malawi)*
– Wood fuel users
higher AM (p<0.01)
– Also for kerosene
lighting (P<0.001)
*Fullerton et al 2009
Impaired AM phagocytic function
• Human AM*:
– UF-CB; DEP
– 4 tests (silica, microorganisms)
– All ↓phagocytosis
• Rat AM** (see graph):
– Carbon-loaded AM
– reduced Strep
pneumoniae killing
• Mice AM***:
– CAP particles
– S. pneumoniae
– Increased adherence,
but reduced killing
– Iron chelation reversed
*Lundborg et al 2006
**Lundborg et al 2007 (graph)
***Zhou et al 2007
Oxidative stress
Mudway et al
2005
• Human respiratory tract lining
fluid model
• PM obtained from dung fuel
(DC PM-sample 1)
• Antioxidant (Ascorbate)
depleted by PM
Oxidative stress
Mudway et al
2005
• Human respiratory tract lining
fluid model
• PM obtained from dung fuel
(DC PM-sample 1)
• Antioxidant (Ascorbate)
depleted by PM
• Metal chelating agent (DPTA)
inhibits effect
• Conclude that redox active
metals in PM are important
In vivo survival following infection
Studies of mice
infected with S.
Pneumoniae
• Hatch (1985):
– Poorer survival
with PM
– For CB and AAP
derived PM
• Tellabati (2010)
Tellabati et al 2010
– Increased
survival with PM
(p<0.001)
– Used UF-CB
Summary: evidence for causality
Bradford Hill viewpoints
#
1
Viewpoint
Strength of association
Summary of evidence
2
Consistency across
populations/study designs
Majority of studies find report increased risk with
exposure (not all significant)
3
Specificity
N/A
4
Temporality (exposure
precedes outcome)
Exposure has preceded infection in all studies;
longitudinal studies available
5
Biological gradient
Statistically significant gradients in two studies
6
Biological plausibility
Studies show range of mechanisms are affected
(Ciliatoxic; ↓AM function; ↑oxidative stress, &c)
7
Coherence with natural
history, animal studies
HAP exposure consistent with mortality;
Some animal evidence available
8
Experiment
RESPIRE; adult cohort study from China
9
Analogy
Other main sources (AAP, smoking) increase risk
OR>2 for severe/fatal pneumonia
Conclusions and next steps
• 2.8 billion people exposed to high levels of HAP; >1
billion children through pregnancy and post-natally
• Does this cause bacterial pneumonia?
– Good evidence for ‘ALRI’
– Most ALRI in developing countries is bacterial pneumonia
– Evidence for severe, fatal, non-RSV, pneumococcal
disease
– Mechanistic studies show plausible pathways and effects
• What is needed to confirm?
– New RCTs (... Ghana, Nepal, Malawi, India)
– Include: exposure assessment, aetiology and severity
– Further mechanistic studies (in vitro and in vivo)
• Vaccine world?
– Reducing HAP may reduce risk via LBW, PTB, and in first
few months of life before vaccine has full effect
New and ongoing RCTs:
Birth outcomes and ALRI
Country Investigator
group
Intervention
Investigations
Status
Malawi
Liverpool;
Wellcome Trust
R/Centre
Fan stove
• Severity
• Aetiology
• Exposure
• Mechanisms
Preparation
phase
Nepal
Johns Hopkins
Rocket stove
LPG
• Severity
• Aetiology
Ongoing
Ghana
Columbia
University;
Kintampo
R/Centre
Fan stove
LPG
• Severity
• Aetiology
• Exposure
• Mechanisms
Recruiting
India
UC Berkeley;
INCLEN
TBC: Fan
stove and/or
LPG
• TBC
Pilot studies
Thank you!
Trends in SFU: 1980 - 2010
(b)
Open fire
Variance=0.36
0.5
1.0
1.5
Plancha stove
Variance=0.31
0.0
Probability Density
2.0
Exposure distributions in plancha and
open fire groups
-2
-1
0
1
Child Mean CO (ln(ppm))
2
3
Impact of 50% increase in exposure
Pneumonia
classification
Cases/child OR (95% CI; p-value) with doubling of exposure
weeks
A: Unadjusted
B: Adjusted for
C: As for B
confounders
plus stove type
All
263/
30270
1.22
1.05, 1.41)
P=0.011
1.25
(1.06, 1.48)
P=0.010
1.28
(1.05, 1.56)
P=0.015
Hypoxaemic
136/
30317
1.35
(1.12, 1.61)
P=0.001
1.38
(1.12, 1.69)
P=0.002
1.39
(1.07, 1.81)
P=0.014
Radiological
85/
30317
1.45
(1.11, 1.90)
P=0.006
1.45
(1.09, 1.93)
P=0.011
1.66
(1.15, 2.40)
P=0.007
Hypoxaemic
and
radiological
53/
30323
1.71
(1.25, 2.32)
P=0.001
1.71
(1.20, 2.44)
P=0.003
2.09
(1.29, 3.38)
P=0.003
The average exposure reduction for the intervention group was 50%
In vivo survival following infection
Studies of mice
infected with S.
Pneumoniae
• Hatch (1985):
Tellabati et
al 2010
RSV infection (Lambert 2003):
• Mice treated with CB, then infected with
RSV
• No increased replication of RSV
• Later increase in neutrophils and TNF
• 2o bacterial infection only seen for CB+RSV
– Poorer survival
with PM
– For CB and AAP
derived PM
• Tellabati (2010)
– Increased
survival with PM
(p<0.001)
– Used UF-CB
Integrated exposure-response function:
child ALRI incidence
Household Air Pollution
SHS
AAP
All ALRI:
mixed viral
and bacterial
Integrated exposure-response function:
child ALRI incidence
Household Air Pollution
SHS
AAP
Average
RESPIRE
plancha
Average
LMIC
exposure
Estimate
for
SRMA
All ALRI:
mixed viral
and bacterial
Integrated exposure-response function:
child ALRI incidence
Household Air Pollution
SHS
AAP
0.78
0.60
2.8
2.2
1.7
Average
RESPIRE
plancha
Estimate
for
SRMA
Average
LMIC
exposure