The Guatemala Randomized Stove Trial

1
The Guatemala Randomized Stove Trial

• A Brief History and Introduction
• Kirk R. Smith, June 26, 2005

2
(No Transcript)
3
First person in human history to have her
exposure measured doing one of the oldest tasks
in human history
Kheda District, Gujarat, India 1981
Filter
Pump
4
Improved Stoves in Gujarat??
Mean pollution in traditional stoves 6400 ug/m3
/- 4600 ug/m3
Mean pollution in smokeless chulas 4600 ug/m3
/- 2900 ug/m3
5
Diseases for which we have health effects data
Chronic obstructive lung disease
Pneumonia ALRI
Lung cancer
Asthma
Blindness
Low birth weight
Tuberculosis
Heart disease?
Early infant death
6
How important is ARI?

• ALRI is responsible for 2 million deaths in
  children lt5 annually, 95 in developing
  countries. Largest single cause of death.
• Morbidity from ARI is the leading cause of
  illness in children in developed countries as
  well.

7
Randomized trials

• The Gold standard in epidemiologic research.
• By randomization one obtains two completely equal
  groups, on both known and unknown confounders.
• Intervene in one group and follow changes over
  time.
• Since it is known they were equal at start, one
  can say with confidence that the change in one
  group is caused by the intervention.
• Required in drug trials, but often not possible
  with environmental risk factors (how would one
  randomize cities for an outdoor air pollution
  study?)

8
The history of a randomized field trial17 years!

• Based on 1981 study, international workshop in
  1984 identified ALRI as best single target for a
  RCT on IAP
• Proposals to BOSTID in 1987,89 by Kirk, MR
  Pandey, John Steinhoff
• Original site developed in Jumla Nepal (2500 m,
  Sample size 300)

9
Five-day walk from nearest road
10
The history of a randomized field trial17 years!

• Nepal study not funded.
• WHO-Geneva created committee in 1990 to find best
  site worldwide (Nigel and Kirk meet)
• Searched for field sites in Mexico, Nepal, India,
  Zimbabwe, Kenya, Turkey, South Africa, Ethiopia,
  Peru, and Guatemala.
• Guatemala chosen as best site

11
The history of a randomized field trial17 years!

• Starting in 1992, 11 small pilot studies in
  Guatemala with Institute for Nutrition of Central
  America and Panama (INCAP), including field
  surveys, testing of intervention stove and
  monitoring equipment, and ethnographic studies.
• Proposals not funded by MRC and EU in 1990s
• First NIEHS proposal with Emory and INCAP in 1999
• Second with MERTU in 2000
• Funded by NIEHS after additional revisions June
  2001

12
Current evidence linking ALRI and Air Pollution

• Tobacco active and passive
• Outdoor air pollution
• Ten observational epi-studies on biomass smoke
  ARI in developing countries.
• Animal studies

13
IAP and Childhood ALRIStudies included in
Meta-Analysis

• 9 Case-control South Africa, Zimbabwe, Nigeria,
  Tanzania, Gambia, Brazil, India, Argentina 6
  adjusted for confounders n 4311 Odds Ratios
  2.2-9.9
• 3 Cohort Nepal, Gambia 2 adjusted for
  confounders n 910 Odds Ratios 2.2-6.0
• 1 Case-fatality Nigeria Hospitalized patients
  n 103 Odds Ratio 8.2
• 2 US Case-control n 206 Adjusted for
  confounders. Odds Ratios 4.8

14
The risk of ALRI associated with use of solid
fuels, in children aged lt5 years
Smith et al in WHO, Comparative quantification of
health risks, 2004
15
Attributable Fractions do not add to 100
Underweight 40
Poor Housing 40
Lack of Breastfeeding 10
1.7 million ARI Deaths in Children Under 5
Diarrhea 20
40 indoor outdoor?
Measles 10
Zn Deficiency 15
Lack of Case-management 65
Lack of Vaccines 50
Rough estimates only
16
4.9 million deaths/y
1.6 (1-2) million deaths/y
World Health Reports 2002, 2001
17
Problems with all PreviousALRI and IAP Studies

• Disease measures used are sensitive (not too many
  false negatives) but not specific enough, i.e.,
  there are too many false positives.
• Studies are all observational and there may be
  residual confounding, i.e., the effect may be to
  poverty-related issues not air pollution itself.
• Detailed exposure assessment has been lacking

18
Epidemiologic Research Designs
19
Acute Respiratory Infections in the World
2002 In Children 0-5 years Acute Lower
Respiratory Infections (ALRI) Acute Upper
Respiratory Infections (AURI)

WHO, Global Burden of Disease Database
20
Acute Respiratory Infections in the World
2002 In Children 0-5 years Acute Lower
Respiratory Infections (ALRI) Acute Upper
Respiratory Infections (AURI)

WHO, Global Burden of Disease Database
21
Acute Respiratory Infections in the World
2002 In Children 0-5 years Acute Lower
Respiratory Infections (ALRI) Acute Upper
Respiratory Infections (AURI)

WHO, Global Burden of Disease Database
22
Acute Respiratory Infections in the World
2002 In Children 0-5 years Acute Lower
Respiratory Infections (ALRI) Acute Upper
Respiratory Infections (AURI)

In India, if an ALRI assessment method is 82
sensitive and 88 specific, then only 31 of
the ALRI found by the method is actually ALRI.
The rest is probably AURI or nothing
Average of three best studies
WHO, Global Burden of Disease Database
23
Particulate Air Pollution and Child ARI A
randomized trial

• Main focus on ALRI among children lt18 months
• Guatemalan highlands in a Mayan Indian
  (Mam-speaking) population
• Collaborators in Bergen, Liverpool, Geneva, and
  Boston
• Funding by NIH (85), Norwegian Research Council,
  WHO, Kresge Foundation, Maxwell Chair, plus
  others

24
Drawbacks with randomized field-trials

• Time consuming Have to wait for people to become
  sick!
• Expensive! Have to go home to everyones houses
  regularly to identify the sick.
• For ethical and economic reasons, such a study
  can usually be done only once.
• Limits itself to extremely common or extremely
  dangerous diseases. (Rothman Greenland-98)

25
Overview of study design
Year 1 5500 Households total

• 530 eligible households open fire, woman
  pregnant or child less than 4 months
• Baseline survey and exposure assessment

26
Control and Intervention
Traditional 3-stone open fire
27
Indoor PM10 Levels

• Highland Guatemala
• PM10
• Kitchen
• 24 hours
• Open woodstove
• Improved woodstove with chimney
• LPG stove

Typical standards 50 ug/m3
ug/m3
2000
Mean
means
1000
Open
Improved
Gas
Stove Type
28
One reason a gas-stove was not used as the
intervention

• Wide range of particle exposures.
• Makes the study suitable for dose-response
  calculation.

29
Some innovative aspects

• First ever RCT on air pollution and health.
• Very thorough validation of ALRI health outcome
  at several levels.
• New applications of inexpensive monitoring
  equipment for exposure assessment.
• Part of a global approach to providing evidence
  based information for national governments and
  international donors.

30
(No Transcript)
31
Participants

• Eligibility criteria
• Households with open fires in defined
  geographical area
• Pregnant women or children under 4 months.
• Exclusion
• Seasonal migration.

32
Community assessment of ALRI

• Previous community studies used weekly visits, by
  IMCI trained field worker
• Concern that overestimates incidence (sensitive
  but not specific)
• Cannot be blinded
• No evidence for objective, external validation
• This study - combines (a) weekly visits by FW (b)
  blind MD assessment in a nearby community centre,
  and (c) hospital referral

33
Overview of main health outcomes assessment and
patient management process
34
Overall organization of the study

• Weekly health surveillance in all households.
  Verified by 2-3 full-time blinded physicians,
  including x-rays and pulse oximetry
• Personal CO monitoring on mother and child twice
  every season in each house and gravimetric
  particle sampling in a subset of households for
  PM1.0 and PM2.5
• Full medical care provided study households.
• Close collaboration with local Health Center and
  San Marcos Hospital (40 min away).
• 35 full-time Guatemalan field staff, 25 hired
  locally, 7 FWD vehicles, field HQ with satellite
  internet link, 4 BR house for out-of-town staff
  and visitors
• Six physicians intimately involved in project (3
  on-site) and six others as active consultants.
• A number of graduate students involved

35
Human subject concerns

• Recruitment
• Mam-speaking teams to explain the project.
• Recruitments to be as rapid as possible to avoid
  pronatal effect
• Drawbacks to participation
• Weekly visits may be perceived as invasion of
  privacy.
• Some of the monitoring equipment may be perceived
  as uncomfortable to have in the house.
• Having to wait for plancha may not be acceptable
  to control group
• Benefits of participation
• All participants receive improved stove.
• Medical treatment and health surveillance
  offered.

36
Main Add-on Studies

• Health
• Adult lung function and respiratory symptoms
• Mothers of study children
• Heart rate variability
• Older women in study households
• Acute CO poisoning from use of temescal
• Subset of study households

37
Add-on Studies, cont.

• Exposure in subsets of households
• Continuous area monitoring using UCB particle and
  HOBO CO monitors
• Personal monitoring with HOBO-CO mothers
• Time-activity measurements using recall and UCB
  personal locator
• Validation of CO breath biomarker
• Urinary biomarker for woodsmoke
• Ventilation measurement protocols
• Stove efficiency and fuel use studies
• Outdoor pollution

38
Gracias