Science of the Total Environment 420 (2012) 54–64

Contents lists available at SciVerse ScienceDirect

Science of the Total Environment

j ourna l homepage: www.e lsev ie r .com/ locate /sc i totenv

Testing the effectiveness of two improved cookstove interventions in the Santiago deChuco Province of Peru

Christopher Fitzgerald a, Manuel Aguilar-Villalobos b, Adam R. Eppler a, Stephen C. Dorner a,Steven L. Rathbun c, Luke P. Naeher a,⁎a The University of Georgia, College of Public Health, Department of Environmental Health Science, Athens, GA, USAb Asociación del Aire Ambiental, Lima, Peruc The University of Georgia, College of Public Health, Department of Epidemiology and Biostatistics, Athens, GA, USA

⁎ Corresponding author at: The University of GeorDepartment of Environmental Health Science, EHS BuilUSA. Tel.: +1 706 542 2454(voice); fax: +1 706 542 7

E-mail address: LNaeher@uga.edu (L.P. Naeher).

0048-9697/$ – see front matter © 2011 Elsevier B.V. Alldoi:10.1016/j.scitotenv.2011.10.059

a b s t r a c t

a r t i c l e i n f o

Article history:Received 23 December 2010Received in revised form 5 October 2011Accepted 21 October 2011Available online 4 February 2012

Keywords:BiomassIndoor air pollutionPM2.5

COStove interventionsImproved stove Peru

90% of people residing in rural areas of less-developed countries rely on coal and biomass fuels for heatingand cooking, leading to high exposures to the products of incomplete combustion. Three Andean communi-ties within the Santiago de Chuco province of Peru received two different models of improved cookstoves.The impact of these stoves in reducing personal exposures and kitchen concentrations of fine particulatematter (PM2.5) and carbon monoxide (CO) was evaluated separately in 64 homes (32 with each stovemodel) using air monitoring equipment. In the community receiving stove 1, baseline aggregate 48-hpersonal exposure (n=27) and kitchen concentrations (n=26) of PM2.5 were 116.4 and 207.3 μg/m3,respectively, and 48-h personal (n=25) and kitchen (n=25) CO levels were 1.2 and 3.6 ppm. Afterintroducing the new stove to this community, those exposures reduced to 68.4 and 84.7 μg/m3, and 0.4and 0.8 ppm, representing reductions of 41.3%, 59.2%, 69.6% and 77.7% respectively. In the twocommunities receiving stove 2, corresponding levels were 126.3 μg/m3 (n=18), 173.4 μg/m3 (n=19),0.9 ppm (n=19), and 2.6 ppm (n=17) before the installation of the stoves, and they reduced to 58.3,51.1 μg/m3 and 0.6, 1.0 ppm. Overall, homes receiving stove 2 saw reductions of 53.8, 70.5, 25.8 and 63.6%.All values are statistically significant (pb0.05) with the exception of personal CO reductions in the stove 2group. Both stoves markedly reduce both kitchen and personal levels of wood smoke exposure, which webelieve has the potential to improve health and quality of life.

© 2011 Elsevier B.V. All rights reserved.

1. Introduction

More than half the world's population, and 90% of people residingin rural areas of less-developed countries rely on biomass fuels (e.g.wood, crop residues) and coal for heating and cooking needs (Bruceet al., 2002). This often involves using open fires or poorly con-structed stoves in unventilated rooms, leading to high concentrationsof pollutants in the home (Bruce and Perez-Padilla, 2000). Accordingto the World Health Organization (WHO), indoor air pollution fromsolid fuel use is responsible for approximately 1.6 million prematuredeaths each year in developing countries (Rehfuess et al., 2006) andis among the world's top ten causes of mortality and morbidity(Rehfuess, 2006).

Pollutants from solid fuel combustion are known to cause or exac-erbate many serious health problems, including reduced birth weight(Boy et al., 2002; Gouveia et al., 2004; Pope et al., 2010), high blood

gia, College of Public Health,ding, Athens, GA 30602-2102,472.

rights reserved.

pressure (Lin et al., 2009; McCracken et al., 2007), acute lower respi-ratory infections, chronic obstructive pulmonary disease and asthma(Smith et al., 2004). In poorer countries, where this issue is mostprevalent, women and children bear a vastly disproportionate burdenof such diseases, as they spend the most time at home (Smith, 2006).Specifically, women between the ages of 15 and 40 years old tend tobe most heavily involved with cooking (Balakrishnan et al., 2004).Also, children are more susceptible to the negative effects of indoorair pollution during their developmental stages (Bearer, 1995), with56% of all indoor air pollution-attributable premature deaths occur-ring in children under five years of age (Rehfuess et al., 2006).

Woodsmoke is known to contain many dangerous pollutants, andthe two most commonly used in indoor air pollution assessments areparticulate matter with an aerodynamic diameter less than 2.5 μm(PM2.5) and carbon monoxide (CO), both of which serve as indicatorsof overall woodsmoke exposure and are themselves harmful to health(Naeher et al., 2007). The recommended ambient air concentrationsfor PM2.5 and CO for the general population can be found in theNational Ambient Air Quality Standards (NAAQS) developed by theUnited States Environmental Protection Agency (USEPA) (USEPA,2005), but their target populations do not include the rural

Fig. 1. Example of improved stove 1.

Fig. 2. Example of improved stove 2.

55C. Fitzgerald et al. / Science of the Total Environment 420 (2012) 54–64

developing world, and thus these values are not completely compara-ble in such a context. Nonetheless, these standards represent guide-lines for improved health.

The most cost-effective solution to the problem of indoor air pol-lution is found in improved stove models, which when adequatelydesigned, installed and maintained are effective in reducing indoorpollution because of better combustion, improvements in ventilation(e.g. through construction of a chimney), lower emission levels andpotentially shorter cooking times (WHO, 2009). Costlier alternativesinclude the introduction of cleaner fuels like liquefied petroleum gasand the remodeling and rebuilding of homes to improve ventilation.To assess the stoves, an important first step is to understand thelevel of exposure experienced by the population to be studied.Indirect measures such as fuel type, household characteristics, orhours near a fire are often used to this end (Bruce et al., 1998), butthe incorporation of personal and area air sampling monitors is pref-erable as it offers a more detailed and quantitative understanding oftotal exposure.

Cookstove intervention programs have been implemented andstudied extensively across the globe. Beyond the significant reduc-tions of indoor air pollution offered by improved stoves (Bruce etal., 2004; Chengappa et al., 2007; Cynthia et al., 2008; Dutta et al.,2007; Ezzati et al., 2000; Masera et al., 2007; Smith et al., 2009),they have been shown to potentially reduce risk of acute lower respi-ratory infections in children under 18 months (Smith et al., 2006),reduce headaches, eye irritation (Diaz et al., 2007), and blood pres-sure levels (McCracken et al., 2007).

In Peru, approximately 33% of homes (90% in rural areas) use solidfuels for cooking and heating (WHO, 2007). Wood is the fuel of choicefor most households in rural Peru, but dung and agricultural residueare also common sources (Meier et al., 2010). The majority of thesehomes do not employ a chimney, which translates to about 8.5 millionPeruvians exposed daily to the products of solid fuel combustion (INEI,2008). In 2009 the Peruvian government began a nationwide stoveintervention programwith the objective of installing 500,000 certifiedimproved stoves by December 2011, which has great potential toimprove indoor air pollution countrywide (Cocinas Mejoradas,2010). Simultaneously, the Barrick Gold Corporation has begun pro-viding new, improved stoves as well. The objective of this research isthe same for both stove types: to test the effectiveness of the newstove in reducing indoor pollution in the natural field environment,under normal, in-home conditions. This research will provide theinvested organizations with essential data and information on thepotential impact of their respective stove intervention projects.Additionally, this research is timely in that it could provide helpfulinsight into the most effective cookstove intervention programs toguide the new improved cookstove initiative to be undertaken bythe United States, which has pledged more than $50 million towardsthis effort (Clinton, 2010).

2. Methods

2.1. Purpose and design

This paper focuses on one piece of a larger improved stove inter-vention research effort in Peru which comprised of three parts. Theresearch incorporated the investigation of two new improved stovesby conducting exposure assessment through kitchen and personalair monitoring, exposure assessment using urinary biomarkers (Li etal., 2011), and the investigation of health effects using multiple indi-cator tests, results of which would be published in future. The datapresented here is a result of the air monitoring exposure assessment.One stove type (Fig. 1) was provided by the Juntos National Programand the other (Fig. 2) by the Barrick Gold Corporation's CommunityRelations Department. In addition to cooking, the traditional stoveswere used by the subjects for heating. Both types of improved stoves

could be used for heating. However, the subjects were not askedwhether they used the new stoves for this purpose. This study wasapproved by Institutional Review Boards from the University ofGeorgia (UGA), the Centers for Disease Control and Prevention(CDC), and approved by the Santiago de Chuco City Hall; informedconsent was obtained from all subjects.

56 C. Fitzgerald et al. / Science of the Total Environment 420 (2012) 54–64

2.2. Study sites and populations

This investigation took place in three locations within the Santiagode Chuco Province, in the region of La Libertad, in the Andes Mountainsof North-Central Peru (Fig. 3). The studywas conducted over thewinterseason between the first week of June and the first week of October,2008. The improved stoves offered by Juntos, hereafter referred to as“stove 1,”were installed and assessed throughout the rural communityof Huayatan, just outside the city of Santiago de Chuco. There is onemain dirt road through Huayatan, which sees very little motor vehicletraffic.

The improved stoves provided by Barrick, hereafter referred to as“stove 2,” were installed and assessed in two neighboring rural com-munities, Chaguin and Cachulla Baja. There is also only one dirt roadthrough this area, and most (83%) of these homes are located awayfrom the road, inaccessible by motor vehicle and lacking electricity.

The study was split such that two different improved stovemodelsmight be assessed independently. Every effort was made to maintaina cohesive study design and implementation across these three loca-tions. A meeting was held at each site, with local community leaders,to enlist volunteers. No incentives were used by the University ofGeorgia research team to attract study participants; however therewas a local program in place (called “Juntos”) which provided mone-tary assistance to any homes willing to comply with a list of specifica-tions, one of which was the installation of new stoves. Women werechosen to be part of the study population based upon three main cri-teria: 1) The woman uses an open woodfire for cooking indoors (withat least three and a half full walls and a roof over the kitchen), 2) shemust be of child-bearing age (18–45), and 3) she must be a willingparticipant, prepared to carry the study out to completion. Beyondthese criteria, the homes were chosen out of convenience, due tothe limited pool of potential participants. In all communities, air mon-itoring was carried out with an identical sampling protocol, and ev-erything possible was done to ensure that the new stoves wereinstalled and studied within a comparable timeframe across homesand across study locations.

Peru

La Libertad RegionS

Fig. 3. Map of Santiago de Chuco Provinc

2.2.1. Household characteristicsTo better understand the daily activities of the participants, each

woman was given a detailed questionnaire and time activity diaryboth before and after the installation of the improved cookstove.The time activity diary was used to collect information regarding out-door and indoor activities and time spent in transit, and was designedto be updated every 15 min. The numbers shown here represent thepopulation of Huayatan first, with Chaguin/Cachulla Baja given inparentheses.

The average age of the women in this study was 33 (33) years.Approximately 37% (35%) of women had more than five people intheir home. The women spend an average of 3.9 (3.7) hours cookingeach day, and additional hours in the kitchen doing other tasks. Noneof the women in any of the communities was a smoker and 0 (3)households reported other smokers in the home. These data are sum-marized in Table 1. Subjects from Huayatan were slightly more edu-cated but the difference between the two communities as computedby chi square analysis was not statistically significant.

In these communities, one of the main responsibilities of thewomen was cooking. The kitchen in most homes is a separate room,although it often serves as a bedroom as well, especially during thewinter. All homes in both regions use eucalyptus wood as their pri-mary source of fuel for cooking and heat. Their fuel source did notchange over the course of this study or with the introduction of thenew stoves.

2.3. The distribution of stoves

InHuayatan, stove 1was delivered to the homes in three pieces: a 3-hole stove-top called a plancha, an aluminum chimney, and an alumi-num tube to connect the chimney to the plancha. These pieces wereconstructed locally and cost approximately S/. 70 soles (~21 USD). Forhouseholds participating in the study, stove materials were purchasedby the local City Hall, but it was the household's responsibility to con-struct the stoves. The Juntos Program provided basic instructions onhow to build the improved stove and a timeline for completion.

Santiago de Chuco Province

antiago de Chuco

0 5 10 20 km

e in the La Libertad Region of Peru.

Table 1Baseline characteristics among all households participating in study, by region.

Huayatan(n=27)

Chaguin/Cachulla Baja(n=23)

Household characteristicsApproximate altitude above sea level,in meters

3400 3000

Dirt floor in home, number (%) 26 (96) 23 (100)Proximity of home to main road, number (%)

b50 m 12 (44) 4 (17)>50 m 15 (56) 19 (83)

People living in home, number (%)b5 17 (63) 15 (65)>5 10 (37) 8 (35)

Smoker present in home, number (%) 0 3 (13)

Participant characteristicsAge, mean (SD) 33 (6.8) 33 (7.8)Smoker 0 0Hours spent cooking each day, mean (SD) 3.9 (0.9) 3.7 (1)

Education levelPrimary (%) 23 (74) 23 (89)Secondary (%) 6 (20) 6 (11)Tertiary (%) 1 (3.2) 0 (0)

Fig. 4. Study participant wearing personal sampling vest.

57C. Fitzgerald et al. / Science of the Total Environment 420 (2012) 54–64

Participants were also asked to construct the new stove in the sameroom as the old stove had been, in order to ensure comparable pre-and post-intervention measurements. All but one subject compliedwith this requirement. The model for stove 2 was more complex, witha larger construction requiring more bricks and a 4-hole stove-top. Inlight of this, the job of building/installing the stoves was subcontractedout to a local brick mason. The improvements of both stove models onthe traditional-style stoves can be summedup in the addition of a chim-ney and the enclosure of the fuel source which aided in more efficientburning. The vast majority of homes employed little more than anopen fire for heating and cooking needs, pre-intervention.

2.4. Air monitoring

Air quality monitoring in the homes was carried out for personaland kitchen area exposure to CO and PM2.5, with a fixed samplingsite set up in the middle of town to capture ambient air levels as abackground against which to compare changes in indoor levels frompre- to post-installation of the new stoves. Although it has beenshown that CO can be used as a tracer for particulate matter exposure(Naeher et al., 2001; Northcross et al., 2010; Smith et al., 2009), wechose to include both pollutant measures for the sake of a moredetailed exposure profile. To reduce the impacts of wide within-daily and day-to-day variability on the uncertainty in our exposuremeasurements (Bhangar et al., 2004; Smith et al., 2007), the homeswere sampled for a 48 hour period. In pre-intervention (phase I) sam-pling, the homes were monitored for two days while the originalstove was still in use. The women were told to live and act normally.Once the new stoves were installed, the women were given approxi-mately three weeks to try out and adjust to using the new stoves, andthen post-intervention (phase II) was begun. In phase II air sampling,everything was done exactly the same as in phase I, with the onlycontrollable difference being the new stove. On the occasion that ameasurement was lost due to equipment failure or other factor, thehome was re-sampled in its entirety. Only homes with completepre- and post-installation measurements for at least one exposuremeasure are included in the final analyses.

2.4.1. Personal samplingEach woman was fitted for two days with a vest (Fig. 4) which

held the personal air sampling equipment at or near their breathingzone. The women were told to wear the vest at all times, and only

to remove them when sleeping or if they might be exposed towater. At night, the women were asked to place the vest at breathingheight next to their bed. To measure real-time CO exposure, each vestheld a Pac III CO monitor (Draeger Safety Inc., Pittsburgh, PA), set torecord concentration levels at 30-second intervals. The instrument(reading range: 0–2000 ppm; resolution: 1 ppm) was calibratedwith 0 ppm and 50 ppm CO gas before use in the field. Forty-eight-hour time-integrated PM2.5 samples were collected on 37 mm Teflonfilters (Pall, East Hills, NY, Teflo 2.0 μm), loaded into particle-size-selective Triplex Cyclones (BGI Inc., Waltham, MA, Model SCC1.062) which were connected to SKC universal sampling pumps(SKC Inc, Eighty Four, PA, Aircheck® XR5000), set to pull air at 1.5 lper minute. After 48 h, the vests were retrieved, sampled filterswere stored in a freezer and runtimes were recorded for each pieceof equipment. The filters were analyzed in the Air Quality Lab in theDepartment of Environmental Health Science at UGA. Filters weredesiccated in climate-controlled conditions (20.6±1.4 °C; 31±13%relative humidity) for a 48-h period prior to the initial weighing ofthe unused filters and the weighing of the sampled filters. Each filterwas weighed twice before and after sampling using a Cahn C-35micro-balance (Thermo Scientific,Waltham, MA, Orion 10935-01) with a sen-sitivity of ±1 μg, exceeding the requirements of the EPA's QualityAssurance Guidance Document (USEPA, 1998).

To account for potential contamination from the loading andunloading of filters, we incorporated the use of field blanks on inter-mittent sampling days. After returning to UGA, post-weights werecalculated for these field blank filters based on a 48-h sample. Theaverage mass on the blanks was 0.88 μg/m3 (n=30), and thisamount was subtracted from every value used in our analyses toreduce the impact of human error and outside contamination fromthe reported concentrations of PM2.5.

2.4.2. Area samplingIn the kitchen of each home, a stationary sampling box (Fig. 5) was

placed within 1 m of the stove. The sampling boxes contained allmonitoring equipment, and attached to one end was a piece of PVC

Fig. 5. Stationary sampling box.

Fig. 6. Study participant using an open pit stove before intervention.

58 C. Fitzgerald et al. / Science of the Total Environment 420 (2012) 54–64

piping which rose to approximate breathing height (1.5 m). As withthe personal vests, each box contained a universal sampling pump(SKC Inc, Eighty Four, PA, Aircheck® 2000) complete with the afore-mentioned filter/cyclone sampling train attached to the piping, anda Pac III CO monitor (Draeger Safety Inc., Pittsburgh, PA) attached tothe top of the piping as well. All details described above for the han-dling of filters were carried out. In certain, randomly chosen homes,the box also contained a DustTrak™ aerosol particulate monitor (TSIInc., Shoreview, MN, Model 8520), which measures real-time PM2.5

particle concentrations using a laser photometer with a readingrange of 0–100 mg/m3 and a resolution of 0.001 mg/m3. The batteryrun time for the instrument according to the manufacturer is 16 hbut could last for 24 h. The instrument was zeroed before use in thefield. A tube ran from the intake of the unit to the top of the PVCpiping. To account for known limitations with this equipment(MacIntosh et al., 2002; Volckens et al., 1999), DustTrak values werelater reduced by 41% based on the differences seen between DustTrakand gravimetric PM2.5 samples collected during a similar study(Naeher et al., 2004) in Arequipa, Peru, with the same equipmentbeing used in this study. All reported DustTrak results reflect theseadjustments and should only be interpreted as relative measures.

2.4.3. Fixed siteA central location was chosen in each town to serve as a fixed

sampling site, providing ambient air levels of both CO and PM2.5. Asampling scheme similar to that seen in the study homes was set upoutside a window in this fixed site. To measure real-time CO, a LanganCO monitor (Langan Products Inc., Elmwood Park, NJ, model T15n)

was used instead of a Pac III. The Langan CO monitor has a readingrange of 0–200 ppm, and a resolution of 50 ppb. The monitor was zer-oed before use in the field. Forty-eight-hour time-integrated PM2.5

was measured with the same equipment used in the kitchens.

2.5. Statistical analysis

SAS version 9.1 (SAS Institute, Cary, NC) was used for all data anal-ysis. Measurement durations shorter than 42 h or longer than 54 hwere eliminated from the analysis dataset. Arc sin square root trans-formations were carried out on the percentage reduction in pollutantconcentrations before their averages were calculated since the distri-butions were right-skewed (Sendecor and Cochran, 1980). A two-way analysis of variance (ANOVA) was employed to model the trans-formed kitchen and personal exposure to PM2.5 and CO as a functionof study subject and time (pre- vs. post-intervention). In order toensure that the ANOVA assumption of homogeneity of variance is sat-isfied, variance-stabilizing log transformation was carried out for pol-lutant concentrations. Means of transformed data before and afterinstallation of the new stoves were adjusted for subject effects inorder to control for imbalance in the data resulting from missingobservations. Adjusted means (called least-squares means in SAS)are predicted values of pre and post-intervention means under thefitted ANOVA model. Statistical significance was defined by pb0.05.

3. Results

3.1. Study participation

Our original goal was a sample of 64 homes, before and after thecookstove intervention, broken down by 32 homes receiving eachstove group. We chose these numbers after calculating a conservativeestimate that a sample size of between 24 and 25 homes in each stovegroup was enough to obtain the statistical power to test our hypoth-eses. It was expected that the sample size would decrease over thecourse of the study for a variety of unforeseeable reasons, and thisassumption held true. As the purpose of our study was to investigatethe effects of a new stove, we only included in our final analysis thosehomes from which we collected complete pre- and post-interventiondata for at least one exposure measure.

Of the 35 subjects who enrolled in the study in Huayatan (Stove1), 30 of them completed both pre and post phases. Reasons for lossto follow-up included personal reasons (n=4) and traveling duringthe study time frame (n=1). Of the 32 subjects who enrolled in thestudy in Chaguin/Cachulla Baja (Stove 2), 27 of them completed

59C. Fitzgerald et al. / Science of the Total Environment 420 (2012) 54–64

both pre and post phases. Reasons for loss to follow-up included per-sonal reasons (n=2), traveling away from town for an extendedperiod of time (n=2), and illness (n=1).

Other complications associated with exposure assessment, such asequipment failure or sample loss, reduced the sample sizes further forthe final analysis. In sum, we had a valid pre- and post-interventionmeasure for homes receiving stove model 1 from: 26 homes for kitch-en PM2.5, 27 homes for personal PM2.5, 25 homes for kitchen CO, and25 homes for personal CO. We had a valid pre- and post-interventionmeasure for homes receiving stove model 2 from: 19 homes for

Table 2Kitchen concentrations and personal exposure levels for each home, stove 1 and stove 2.

Home ID PM2.5 (μg/m3)

Kitchen (n=26) Personal (n=27)

PRE POST % RED* PRE POST % RED*

Stove 1 1 161.6 88.7 45.1 65.3 70.0 −7.12 82.8 36.3 56.2 40.9 24.8 39.43 228.0 28.5 87.5 152.1 46.6 69.46 – – – 186.2 164.3 11.87 150.3 48.9 67.5 72.7 73.8 −1.68 60.8 77.0 −26.6 55.8 48.4 13.39 83.4 71.3 14.5 36.7 26.2 28.5

10 85.9 62.5 27.3 49.0 89.5 −82.611 29.8 33.1 −11.1 44.9 109.5 −143.712 100.9 48.2 52.3 46.5 17.2 62.913 140.5 23.7 83.1 62.5 34.8 44.214 390.4 89.3 77.1 484.3 59.3 87.815 61.2 81.9 −33.9 108.2 44.5 58.817 1075.0 343.4 68.1 160.8 59.7 62.918 324.2 87.8 72.9 159.0 66.3 58.319 613.3 135.2 78.0 203.3 82.5 59.422 29.4 129.5 −340.4 52.9 75.7 −43.224 260.6 73.5 71.8 103.8 96.7 6.925 61.3 60.0 2.1 109.4 63.8 41.728 1331.4 376.8 71.7 262.8 63.7 75.829 73.2 9.4 87.1 221.0 139.4 36.930 1301.8 141.5 89.1 1565.2 116.6 92.631 1067.0 673.8 36.8 320.4 269.3 16.032 339.3 111.8 67.1 75.2 59.1 21.433 245.4 111.6 54.5 69.3 76.4 −10.234 1151.7 276.5 76.0 188.5 73.2 61.135 651.7 113.7 82.6 180.3 129.4 28.2

Geomean95% C.I.

207.3(163, 265)

84.7(66.4, 108)

59.2(42.4, 71.1)

116.4(94.6, 143)

68.4(55.6, 84.2)

41.3(21.2, 56

(n=19) (n=18)

Stove 2 40 199.5 3.5 98.2 185.3 77.2 58.441 184.4 70.5 61.8 – – –

43 – – – 127.2 113.7 10.645 51.2 70.1 −36.8 56.1 28.6 49.047 669.2 155.9 76.7 – – –

48 177.0 17.5 90.1 132.6 73.5 44.649 1419.4 825.3 41.9 191.1 16.6 91.350 421.2 136.3 67.6 214.5 18.8 91.251 362.3 10.5 97.1 33.3 31.8 4.552 52.1 21.8 58.2 54.9 28.6 47.953 543.1 170.3 68.6 175.6 189.7 −8.055 170.1 101.6 40.3 97.9 137.5 −40.459 41.2 59.8 −45.2 84.1 88.3 −5.060 – – – – – –

61 160.3 75.4 53.0 142.7 80.4 43.763 362.4 88.4 75.6 234.9 57.1 75.764 15.8 2.7 83.2 281.3 104.8 62.865 – – – – – –

66 – – – – – –

67 69.4 16.9 75.7 88.9 67.8 23.768 781.0 185.0 76.3 239.2 110.2 53.969 17.6 105.0 −495.5 109.6 53.8 51.071 413.5 39.0 90.6 160.5 26.3 83.6

Geomean95% C.I.

173.4(112, 268)

51.1(33.1, 78.8)

70.5(45.6, 84.1)

126.3(96.3, 166)

58.3(44.5, 76.5)

53.8(32.2, 68

The average reductions in bold font are significant at p=0.05.

kitchen PM2.5, 18 homes for personal PM2.5, 17 homes for kitchenCO, and 19 homes for personal CO (Fig. 6).

3.2. Ambient air levels

In each community, ambient air pollutant levels were measured aspreviously discussed. In Huayatan, the average ambient air CO levelfor the entire course of our study was 0.25 ppm, and average PM2.5

concentration was 12.1 μg/m3. In Chaguin/Cachulla Baja, the averageambient air CO level for the entire duration of the study was

CO (ppm)

Kitchen (n=25) Personal (n=25)

PRE POST % RED* PRE POST % RED*

4.5 1.9 58.1 1.2 0.7 37.30.9 0.1 92.3 1.7 0.1 92.14.3 0.2 94.9 2.3 0.6 74.66.7 2.6 60.3 2.0 1.9 6.0– – – 0.5 0.4 21.6– – – – – –

1.2 0.7 38.9 0.1 0.1 31.51.0 0.4 64.8 0.7 1.0 −49.82.4 0.1 95.4 0.5 2.2 −363.61.2 0.4 66.9 1.9 0.0 97.81.4 0.07 94.7 0.9 0.06 93.73.9 0.85 78.3 3.5 0.01 99.70.4 0.5 −42.2 1.2 0.3 74.810.1 4.4 56.4 1.6 0.5 71.65.7 2.9 48.9 2.5 0.6 78.45.8 1.6 73.0 1.3 0.1 94.43.2 1.2 63.3 0.4 0.5 −20.04.0 0.8 81.2 0.9 1.1 −26.90.6 0.2 64.6 0.4 0.4 19.612.0 3.6 70.1 0.8 0.0 96.92.7 0.0 99.3 – – –

11.1 1.6 85.6 3.4 1.4 58.824.8 23.7 4.5 2.9 2.3 19.15.9 1.6 72.7 0.4 0.3 4.95.1 1.1 78.9 2.5 0.8 67.713.0 4.5 65.7 2.3 0.7 70.89.6 1.6 83.5 2.6 1.4 47.0

.2)3.6(2.6, 4.9)

0.8(0.58, 1.09)

77.7(65.1, 85.7)

1.2(0.74, 1.84)

0.4(0.23, 0.56)

69.6(42.4, 84)

(n=17) (n=19)

3.6 7.3 −105.0 1.3 2.8 −118.0– – – 1.3 0.2 83.13.2 1.8 45.5 0.8 0.2 74.00.5 0.5 −7.5 0.2 0.9 −326.9– – – – – –

– – – – – –

– – – 2.5 0.7 72.05.6 1.7 69.7 2.6 1.2 52.5– – – – – –

1.2 0.2 87.8 0.3 0.0 95.39.7 4.6 52.9 1.9 2.0 −2.12.7 2.0 27.0 0.5 2.1 −276.60.5 0.9 −104.3 0.2 1.3 −526.80.7 0.3 60.3 0.2 0.5 −140.12.9 0.7 75.4 1.9 0.7 63.46.2 1.5 75.2 3.7 0.6 83.1– – – – – –

3.4 0.5 84.7 0.4 1.2 −202.42.3 0.2 92.0 1.0 0.2 78.30.9 0.2 81.9 0.2 0.2 20.612.0 3.7 68.9 2.4 1.4 39.41.8 0.7 58.8 0.9 1.1 −22.65.3 0.3 93.7 1.1 1.0 3.8

.5)2.5(1.71, 3.54)

0.8(0.59, 1.22)

65.6(42.4, 79.4)

0.8(0.53, 1.31)

0.6(0.39, 0.97)

25.9(−40, 61)

0

1

2

3

4

5

6

0

50

100

150

200

250

300

CO

(pp

m)

Geo

met

ric

mea

n (+

/-95

% C

.I.)

PM

2.5 (

ug/m

3 )

Geo

met

ric

mea

n (+

/-95

% C

.I.)

CO

Personal

Personal

KitchenKitchen

n = 27 n = 26 n = 25

PM2.5

n = 25

A

0

1

2

3

4

5

6

0

50

100

150

200

250

300

CO

(pp

m)

Geo

met

ric

mea

n (+

/-95

% C

.I.)

PM

2.5(

ug/

m3 )

Geo

met

ric

mea

n (+

/-95

% C

.I.)

CO

Personal

Personal

Kitchen

Kitchen

n = 18 n = 19 n = 17n = 19

PM2.5

B

Fig. 7. Geometric mean reductions of kitchen and personal CO and PM2.5.

60 C. Fitzgerald et al. / Science of the Total Environment 420 (2012) 54–64

0.53 ppm, and average PM2.5 concentration was 11.6 μg/m3. Our mainpurpose in reporting ambient air levels in each community was tocatch any large changes or differences across days and across thetwo phases of our study that may have significantly impacted our re-sults. We observed no such trends.

0

20

40

60

80

100

120

10:48 AM 6:00 PM 1:12 AM 8:24 AM

Kit

chen

CO

(pp

m)

0

20

40

60

80

100

120

10:48 AM 6:00 PM 1:12 AM 8:24 AM

Per

sona

l CO

(pp

m)

01020304050607080

10:48 AM 6:00 PM 1:12 AM 8:24 AM

Kit

chen

PM

2.5

(mg/

m3)

A

B

C

Fig. 8. Pre- and post-intervention real-ti

3.3. Impact of new stoves on the indoor environment

Kitchen concentrations and personal exposure levels for eachindividual home are recorded in Table 2. Increases after the installa-tion of improved stove in indoor air and personal PM2.5 and CO

3:36 PM 10:48 PM 6:00 AM 1:12 PM

Pre-Intervention Post-Intervention

3:36 PM 10:48 PM 6:00 AM 1:12 PM

Pre-Intervention Post-Intervention

3:36 PM 10:48 PM 6:00 AM 1:12 PM

Pre-Intervention Post-Intervention

me measurements of CO and PM2.5.

Table 3Summary of air sampling results for reductions in peak concentrations.

Improvedstovemodel

Samplinglocation

Measurement N Average highest15-minconcentrationa

pre-intervention

Range Average highest 15-minconcentrationa

post-intervention

Range Average highest30-second peak:pre-interventionb

Range Average highest30-second peak:post-interventionb

Range

1 Kitchen PM2.5 (mg/m3) 19 9.2 0.27–26.3 1.7 0.3–5.0 29.0 2.0–71.8 10.0 1.3–32.9CO (ppm) 25 52.6 6.3–366 22.3 0.9–196 126.9 14–666 60.9 5–353

Personal CO (ppm) 25 19.4 3.7–51.4 10.4 1.7–35.8 67.8 18–174 38.4 4–1282 Kitchen CO (ppm) 17 31.7 3.8–135 28.4 3.6–160 58.5 14–155 62.7 9–294

Personal CO (ppm) 19 16.4 4.1–38.1 14.2 0.8–36.9 52.9 13–146 62.5 17–164

a Average highest 15-minute concentration: the average of all the highest 15-minute concentrations recorded in each home.b Highest single 30-second value recorded. Compare to the OSHA peak (or ceiling) limit for CO of 200 ppm.

61C. Fitzgerald et al. / Science of the Total Environment 420 (2012) 54–64

were observed in some cases, especially when the baseline levelswere relatively low (at or below the 25th percentile concentrations;indoor PM2.5 and CO: 73.2 μg/m3and 1.3 ppm; personal PM2.5 andCO: 65.3 μg/m3 and 0.5 ppm). Overall, there was reduction in indoorair and personal PM2.5 and CO after the improved stoves wereinstalled, and total average reductions for each indicator in bothregions are displayed in Fig. 7. All reductions shown are statisticallysignificant (pb0.05) with the exception of personal CO for stove 2.Fig. 8 demonstrates typical reductions in kitchen concentrations ofPM2.5 and reductions of both kitchen and personal exposure to COover a 48-h monitoring period before and after the installation ofstove 1 in a representative home. Times of intensive stove use areevident, and the highest (30-second) peaks occur around lunchtime,which is traditionally the largest meal in Peru.

3.3.1. Reductions in personal exposure and kitchenconcentrations — stove 1

After approximately three weeks using the new stoves, thegeometric mean 48-h personal exposure to PM2.5 in stove 1 homesreduced from 116.4 μg/m3 (95% C.I. 94.6, 143 μg/m3) to 68.4 μg/m3

(95% C.I. 55.6, 84.2 μg/m3), which represents an aggregate reductionof 41.3% (n=27, pb0.01). 48-h personal exposure to CO reducedfrom an average of 1.2 ppm (95% C.I. 0.74, 1.84 ppm) to 0.4 ppm(95% C.I. 0.23, 0.56 ppm), which is an overall reduction of 69.6%(n=25, pb0.001). The average highest single 30-second peak in per-sonal exposures to CO across all homes reduced almost 50%, droppingfrom 67.8 to 38.4 ppm. Such peaks occur most commonly when thestove is lit, fuel is adjusted, the cooking pot is placed on or removedfrom the fire, or food is stirred (Ezzati and Kammen, 2002). Furtherexplanation of these peak reductions is given in the discussion anddetails are displayed in Table 3.

The geometric mean 48-h kitchen concentration of PM2.5 reducedfrom 207.3 μg/m3 (95% C.I. 163, 265 μg/m3) to 84.7 μg/m3 (95% C.I.

Table 4Summary results for improved stove models 1 and 2 compared with traditional stoves.

Improved stove model Sampling location Measurement Stove type N

1 Kitchen PM2.5 (μg/m3) Traditional 26Improved

CO (ppm) Traditional 25Improved

Personal PM2.5 (μg/m3) Traditional 27Improved

CO (ppm) Traditional 25Improved

2 Kitchen PM2.5 (μg/m3) Traditional 19Improved

CO (ppm) Traditional 17Improved

Personal PM2.5 (μg/m3) Traditional 18Improved

CO (ppm) Traditional 19Improved

66.4, 108 μg/m3), representing an aggregate reduction of 59.2%(n=26, pb0.0001). Real-time kitchen exposure data was collected forat least part of the 48-h sampling period in a total of 19 homesreceiving stove 1. In this subsample of homes, the average highest 30-second peak reduced from 29 mg/m3 to 10 mg/m3 after installation ofthe new stove (Table 3). 48-h kitchen concentrations of CO reducedfrom 3.6 ppm (95% C.I. 2.6, 4.9 ppm) to 0.8 ppm (95% C.I. 0.58, 1.09),which is an overall reduction of 77.7% (n=25, pb0.0001). Theaverage highest 30-second peaks in kitchen area CO concentrationdropped over 50% with the introduction of the improved stove, reduc-ing from 126.9 to 60.9 ppm(Table 3). Only homeswith complete pairedbefore and after measurements were included in these analyses, andthis information is summarized in Table 4.

3.3.2. Reductions in personal exposure and kitchenconcentrations — stove 2

After approximately three weeks using the new stoves, the geomet-ric mean 48-h personal exposure to PM2.5 in stove 2 homes reducedfrom 126.3 μg/m3 (95% C.I. 96.3, 166 μg/m3) to 58.3 μg/m3 (95% C.I.44.5, 76.5 μg/m3), representing an aggregate reduction of 53.8%(n=18, pb0.001). 48-h personal exposure to CO (n=20) reducedfrom 0.9 ppm (95% C.I. 0.57, 1.31 ppm), to 0.6 ppm (95% C.I. 0.43,0.97 ppm), representing minimal change in exposure. In both stovegroups, this was the only exposure indicator that was not significantlyreduced.

The geometric mean 48-h kitchen area concentration of PM2.5

reduced from 173.4 μg/m3 (95% C.I. 112, 268 μg/m3) to 51.1 μg/m3 (95%C.I. 33.1, 78.8 μg/m3), representing an aggregate reduction of 70.5%(n=19, pb0.001). Due to logistical issues during this portion of ourstudy, real-time kitchen area exposure data was only collected in threeof the homes receiving stove 2. As such, those data are not presentedhere. 48-h kitchen area concentration of CO reduced from 2.6 ppm (95%C.I. 1.89, 3.7 ppm) to 1.0 ppm (95% C.I. 0.69, 1.35 ppm), which is an

Geometric mean concentration (95% C.I.) % reduction (95% C.I.) Pr>F

207 (163,265) 59.2 (42.4,71.1) b0.000184.7 (66.4,108)3.6 (2.6,4.9) 77.7 (65.1,85.7) b0.00010.8 (0.6,1.1)116 (94.6,143) 41.3 (21.2,56.2) 0.001268.4 (55.6,84.2)1.2 (0.7,1.8) 69.6 (42.4,84.0) 0.0010.4 (0.2,0.6)173 (112,268) 70.5 (45.6,84.1) 0.000851.1 (33.1,78.8)2.5 (1.7,3.5) 65.6 (42.4,79.4) 0.00050.8 (0.6,1.2)126 (96.3,166) 53.8 (32.2,68.5) 0.000858.3 (44.5,76.5)0.8 (0.5,1.3) 25.9 (−40.0,61.0) 0.33710.6 (0.4, 1.0)

62 C. Fitzgerald et al. / Science of the Total Environment 420 (2012) 54–64

overall reduction of 63.6% (n=18, pb0.001). There is no evidence thatstove 2 reduced the 30-second peak exposures to CO. Again, onlyhomes with complete paired before and after measurements were in-cluded in these analyses, and the information is summarized in Table 4.

4. Discussion

Large-scale distribution of improved cookstoves has great poten-tial to reduce the trends of mortality and morbidity caused by indoorair pollution, especially in the developing world. Realizing this fact,many governmental and non-governmental organizations haveresponded with improved stove models. As every community is dif-ferent, and daily use inside a rural kitchen is quite different from alaboratory, field testing of new stoves is essential to understandingtheir potential impact. The information discussed here is meant toaid the development of improved stoves.

Improved cookstove programs around the world have consistentlyproved successful in the reduction of kitchen concentrations and per-sonal exposures to indoor air pollution (Table 5). Our research indi-cates similarly positive results from these two different stoves after3 weeks of in-home use under normal field conditions in three sepa-rate communities of North-Central, Andean Peru. Although the resultsof this study are limited due to its small scale and sampling protocol,the reductions in air pollution achieved suggest that the large-scaledissemination of these stoves throughout Peru is both feasible andadvisable. The stove's simple structure and cost-effectiveness alsoadd to the advantage they offer for a dissemination program. However,the durability of these stoves over several years and the retention ratesby stove recipients remain to be evaluated.

Table 5Summary of results from improved cookstove research around the world.

Reference/Year Study region % reduction

Naeher et al. (2001) Western highlands,Guatemala

22-h kitchen PM2.5 reduced 84%22-h personal PM2.5 reduced 46.6%22-h kitchen CO reduced 76%22-h personal CO reduced 64%

Bruce et al. (2004) La Victoria,Guatemala

24-h kitchen CO reduced 76%

Masera et al. (2007) Michoacán, Mexico 48-h kitchen PM2.5 reduced 67%48-h kitchen CO reduced 66%

Cynthia et al. (2008) Michoacán, Mexico 24-h personal PM2.5 reduced 35%24-h personal CO reduced 78%

Edwards et al.(2007)

Throughout China 24-h kitchen PM4 reduced 43%24-h kitchen CO reduced 62%

Dutta et al. (2007)a Pune, India 48-h kitchen PM2.5 reduced24, 49%48-h kitchen CO reduced 38, 39%

Chengappa et al.(2007)

Bundelkhand, India 48-h kitchen PM2.5 reduced 44%48-h kitchen CO reduced 70%

Smith et al. (2009) San Marcos,Guatemala

48-h kitchen CO reduced 90%48-h personal CO of mothersreduced 61%48-h personal CO of childrenreduced 52%

Clark et al. (2009) Santa Luciaand Suyapa, Honduras

8-h indoor PM2.5 lower by 73%8-h personal PM2.5 lower by 63%8-h indoor CO lower by 87%

Fitzgerald et al.(current study)

Santiago de Chuco,Peru

48-h kitchen PM2.5 reduced 59.2%(70.5%)b

48-h personal PM2.5 reduced41.3% (53.8%)48-h kitchen CO reduced 77.7%(65.6%)48-h personal CO reduced69.6% (25.9%)

a Two models were assessed in this study.b Values are for stove 1 (stove 2).

4.1. Effectiveness of intervention — stove 1

The results from this study represent one of the multiple stovemodels suggested by the Juntos National Program, as distributed inone of many thousand small communities throughout Peru. As such,the potential reductions in CO and PM2.5 exposures and kitchen con-centrations implied by our findings with stove 1 are most accuratelyapplicable to the communities where our study was conducted.

In this study population as a whole, paired comparisons of 48-hpersonal exposure to CO and PM2.5 indicated reductions of 69.6%and 41.3%, respectively. Similarly, paired comparisons of 48-h COand PM2.5 kitchen concentrations revealed overall reductions of77.7% and 59.2% respectively. These results are consistent withpercent reductions seen in other studies of stove interventionsaround the world (Table 5). Kitchen area and personal COexposures were reduced at levels on the high end of what isreported in the literature.

When discussing CO, one of the most notable improvementsoffered by the new stoves is the reduced intensity of the acute highexposures experienced during peak cooking times throughout theday. With a 2-day sample, the average reductions in CO exposuremay look relatively insignificant (in this case, 48-h geometric meankitchen area and personal levels only reduced from 3.6 and 1.2 ppmto 0.8 and 0.4 ppm, respectively). However, these seemingly smallaverage changes are indicative of many large changes in CO valuesseen throughout the sampling period (Fig. 8).

Beyond the actual improvements shown by our air monitoringequipment, most participants greatly appreciated their new stove.Though entirely anecdotal, many women found that they neededless wood to cook meals, which could potentially be a factor in theoverall reduction of indoor air pollution. One of the researchers hasalso visited Huayatan twice since the completion of the study andobserved 100% retention and subjects continued to be happy withtheir stoves. Personal exposures did not reduce as much as kitchenlevels, but this is consistent with other studies (Cynthia et al., 2008)and is not a surprising outcome, as women are exposed throughoutthe day to sources of pollution beyond their own kitchen (such asother homes still using traditional stoves). This highlights the limita-tions in using kitchen area concentrations alone as a proxy measurefor exposures in women and children (Smith et al., 2009). We choseto incorporate both kitchen area and personal measurements sothat we might not only see the drop in pollutant emissions withinthe home, but also to observe the extent to which that drop actuallyimpacts personal exposure.

Overall, this new stove offers vast improvements over traditionalmethods, greatly reducing exposure to both aggregate and acutelevels of indoor pollution. The post-intervention exposure levelsmay still seem high in comparison to the developed world, but theyare truly impressive in the context of a rural kitchen area in the de-veloping world. In a region where ambient levels of pollution alonemay often exceed WHO and USEPA air quality guidelines, the expo-sure reductions experienced by these women certainly indicategreat progress.

4.2. Effectiveness of intervention — stove 2

Chaguin/Cachulla Baja were two communities chosen to receivestove 2. The job of stove-building in these communities was subcon-tracted by Barrick to one man, which greatly reduced the variability instove construction. Concurrently, this system avoided placing respon-sibility for the stove installation on the women. As mentioned before,this portion of the study is based upon one specific stove model inspecific communities. Extrapolation of these results to other areas isonly logical if the same stove model is being distributed to a similarcommunity with similarly traditional cooking methods.

63C. Fitzgerald et al. / Science of the Total Environment 420 (2012) 54–64

On an aggregate basis, paired comparisons of 48-h personalexposure to PM2.5 indicated a reduction of 53.8%. In the kitchenareas, paired comparisons of 48-h CO and PM2.5 concentrationsrevealed overall reductions of 63.6% and 70.5%, respectively. Percentreductions for the stove 2 group again seem to match those seenelsewhere (Table 5). Personal exposure to CO was low in thesecommunities to begin with, and thus the reductions seen (25.8%)were also quite low, and not statistically significant. Also, thoughaverage levels of CO in the kitchen area did significantly improve,they did not show consistent reductions in peak levels. Closeranalysis of the data from each individual home, along with the useof field notes and time activity diaries, afforded us no viableexplanation for these results.

As with Huayatan, the post-intervention exposure levels inChaguin/Cachulla Baja may still seem high by developed world stan-dards, but the value of the stoves is best seen in looking at the reduc-tions. Increases after the installation of improved stove in indoor airand personal PM2.5 and CO were observed in some cases, especiallywhen the baseline levels were relatively low. The reasons for suchresults are not clear. The presence and increased exposure to othersources of PM2.5 and CO in the indoor and outdoor environments,such as second hand tobacco smoke and traffic, could have been a fac-tor. However, these were not consistent with cases where increasedPM2.5 and CO were measured. The overall reduction in exposure toair pollution reported here is buttressed by the results of the exposurebiomarker portion of the study. Ten hydroxyl-substituted polycyclicaromatic hydrocarbons (OH-PAHs) were analyzed in urine samplesof the subjects. The median OH-PAHs were significantly reducedby 15% to 58% in the Huayatan subjects and by 13% to 38% in theChaguin/Cachulla Baja subjects from the pre-intervention to thepost-intervention levels (Li et al., 2011).

4.3. Limitations

Research in the developing world carries with it some inherentlimitations. In the three communities there were a limited numberof households that met the study criteria and were willing to partici-pate, requiring us to employ a convenience sample. Also, seasonaldifferences could not be investigated in such a small-scale study,and while the climate remained similar across the pre- and post-intervention phases, habits and activities of some women may havechanged. Results of the study represent the effectiveness of the stovesin reducing indoor air pollution over the short term. Since the objec-tive of this study was to test the effectiveness after deployment inhomes, the durability and longer term effectiveness of the stovescould not be assessed.

The scale of this study was such that many potential factors couldnot be considered or assessed. In order to gain a more complete pic-ture of the problem of IAQ and the impact of such stoves, factorssuch as the following would have required investigation: day-to-dayvariability in exposures; effect of differences in kitchen size/design;effect of location of stove within household; the use of multiple stovesin one home; differences in ventilation within the home; contributionto personal exposure and kitchen concentrations from a neighboringhome's stove; long-term durability of new stove; and how best tomaintain stove for long-term use in the home. However, the smallscale and short term nature of this study, which was only designedto measure the potential impacts of the improved stoves in terms oftheir effectiveness in reducing indoor air pollution, precludes the con-sideration of these factors, and limits the extrapolation of the resultsfor the assessment of the long term effectiveness of the stoves andtheir performance beyond the communities where the study was per-formed. Convenience sampling was employed and the sample wasnot random. Informationwas not collected on potential subjects that re-fused to participate. Consequently our results may have been subject tobias. Homes across the three communities were however similar in

terms of the age of the women, number of people living in the homes,educational level, total time spent cooking and kitchen and personalexposure to both PM2.5 and CO at baseline. This indicates that our sam-ple may have been representative of the communities.

Subjects that continued to use the traditional cook stoves were notrecruited for the study. Therefore, results could have been biased bychange in activities that could have affected exposure from pre- topost-installation. However, observations made from time activity dia-ries that were completed indicate that there was an increase in timespent indoors by the subjects. This would have contributed to lessreduction than those observed rather than inflating the positiveresults that are reported. Moreover, ventilation, the number of peopleliving in the homes, ambient temperature and time spent cooking didnot change from pre- to post-installation.

5. Conclusions

Despite the differences in the people of Huayatan and Chaguin/Cachulla Baja, as well as the differences in the improved stove modelswe investigated, the final conclusions of this study are virtually thesame for everyone involved. Both stoves show potential to markedlyreduce both kitchen and personal levels of wood smoke exposure.

Acknowledgements

This study was generously funded by the Voluntary Contributionof Barrick Gold Corporation under an agreement with the GobiernoRegional de la Libertad. We would like to thank Barrick for all thatthey have done to support this project, financially and otherwise.Specifically: Carlos Cabanillas Bustamante, GuillermoManrique Franco,Alejandro Sanchez and Dr. Luis Perez Villasante.

Our deep appreciation goes as well to Jose Murgia Zanier, Presidentof the regional Government of La Libertad, for his limitless support ofour team. Also, Dr. Pedro Diaz Camacho from the regional Governmentof La Libertad. The local government of Santiago de Chuco for their helpand support during this research. Specifically: Professor Abner AvalosVillacorta, Mayor of Santiago de Chuco, and Dr. Alipio Mantilla linares,Director of the Hospital in Santiago de Chuco.

The research team: Manuel Aguilar-Villalobos, Maria AntonietaMelendez, Stephen Dorner, Adam Eppler, Jessica Fitzgerald, AdamGray, Kevin Horton, Elizabeth Irvin, Luke Naeher and Daniel Pope.

References

Balakrishnan K, Sambandam S, Ramaswamy P, Mehta S, Smith KR. Exposure assessmentfor respirable particulates associated with household fuel use in rural districts ofAndhra Pradesh, India. J Expo Anal Environ Epidemiol 2004;14(1):14–25.

Bearer C. How are children different from adults? Environ Health Perspect 1995;103:6.Bhangar S, Smith KR, et al. The contribution of stove type to the variability in CO

concentration in households in rural Maharashtra, India. International Society ofExposure ScienceAnnual Meeting Philadelphia.; 2004.

Boy E, Bruce N, Delgado H. Birth weight and exposure to kitchen wood smoke duringpregnancy in rural Guatemala. Environ Health Perspect 2002;110:109–14.

Bruce N, Neufeld L, Boy E, West C. Indoor biofuel air pollution and respiratory health:the role of confounding factors amongwomen in highland Guatemala. Int J Epidemiol1998;27:454–8.

Bruce N, Perez-Padilla. Indoor air pollution in developing countries: a major environ-mental and public health challenge. Bull World Health Organ 2000;78(9):1078–92.

Bruce N, Perez-Padilla R, Albalak R. Health effects of indoor air pollution exposure indeveloping countries. Protection of the Human Environment. Geneva: WorldHealth Organization; 2002.

Bruce N, McCracken J, Albalak R, et al. Impact of improved stoves, house constructionand child location on levels of indoor air pollution exposure in young Guatemalanchildren. J Expo Anal Environ Epidemiol 2004;14(1):26–33.

Cocinas Mejoradas. Campaigning for half a million improved stoves for a smokelessPeru. Retrieved December 22, 2010, from http://www.cocinasmejoradasperu.org.pe/index.html.

Chengappa C, Edwards R, Bajpai R, Shields KN, Smith KR. Impact of improved cook-stoves on indoor air quality in the Bundelkhand region of India. Energ SustainDev 2007;11(2):33–44.

Clark M, Peel J, Burch J, et al. Impact of improved cookstoves on indoor air pollution andadverse health effects amongHonduranwomen. Int J Environ Health Res 2009;19(1):357–68.

64 C. Fitzgerald et al. / Science of the Total Environment 420 (2012) 54–64

ClintonH. Remarks on Global Alliance for Clean Cookstoves at the ClintonGlobal Initiative.Retrieved December 20, 2010, from http://www.state.gov/secretary/rm/2010/09/147500.htm.

Cynthia AA, Edwards RD, Johnson M, et al. Reduction in personal exposures to particu-late matter and carbon monoxide as a result of the installation of a Patsariimproved cook stove in Michoacan Mexico. Indoor Air 2008;18(2):93-105.

Diaz E, Smith-Sivertsen T, Pope D, et al. Eye discomfort, headache and back pain amongMayan Guatemalan women taking part in a randomized stove intervention trial. JEpidemiol Community Health 2007;61(1):74–9.

Dutta K, Shields K, Edwards R, Smith KR. Impact of improved biomass cookstoves onindoor air quality near Pune, India. Energ Sustain Dev 2007;11(2):19–32.

Edwards RD, Liu Y, He G, et al. Household CO and PM measured as part of a review ofChina’s National Improved Stove Program. Indoor Air 2007;17:189–203.

Ezzati M, Mbinda BM, Kammen DM. Comparison of emissions and residential exposurefrom traditional and improved biofuel stoves in rural Kenya. Environ Sci Technol2000;34:578–83.

Ezzati M, Kammen DM. The health impacts of exposure to indoor air pollution fromsolid fuels in developing countries: knowledge, gaps, and data needs. EnvironHealth Perspect 2002;110(11):1057–68.

Gouveia N, Bremnar SA, Novaes HM. Association between ambient air pollution andbirth weight in Sao Paulo, Brazil. J Epidemiol Community Health 2004;58:11–7.

INEI (Peru National Institute of Statistics and Data). 2007 Growth and distribution ofthe population; 2008.

Li Z, Sjodin A, Romanoff LC, et al. Evaluation of exposure reduction to indoor air pollu-tion in stove intervention projects in Peru by urinary biomonitoring of polycyclicaromatic hydrocarbon metabolites. Environ Int 2011;37(7):1157–63.

Lin LY, Lin CY, Chuang KJ. The effects of indoor particles on blood pressure and heartrate among young adults in Taipei, Taiwan. Indoor Air 2009;19:482–8.

MacIntosh D, Williams P, Yanosky J. A comparison of two direct reading aerosol mon-itors with the federal reference method for PM2.5 in indoor air. Atmos Environ2002;36:107–13.

Masera O, Edwards R, Arnez CA, et al. Impact of Patsari improved cookstoves on indoorair quality in Michoacan, Mexico. Energ Sustain Dev 2007;11(2):45–56.

McCracken JP, Smith KR, Diaz A, Mittleman MA, Schwartz J. Chimney stove interven-tion to reduce long-term wood smoke exposure lowers blood pressure amongGuatemalan women. Environ Health Perspect 2007;115(7):996-1001.

Meier P, Tuntivate V, Barnes DF, Bogach SV, Farchy D. Peru: National Survey of RuralHousehold Energy Use: Special Report. Washington D.C.: The International Bankfor Reconstruction and Development; 2010

Naeher LP, Smith KR, Leaderer BP, Neufeld L, Mage DT. Carbon monoxide as a tracer forassessing exposures to particulate matter in wood and gas cookstove householdsof highland Guatemala. Environ Sci Technol 2001;35:575–81.

Naeher LP, Aguilar MM, Merry K, et al. Ambient, personal, occupational, residential,industrial, and commercial air pollution exposure assessment in Arequipa, Peru,United States Agency for International Development, CARE-Peru, and Camp Dresser& McKee, International Inc; 2004. p. 1-32.

Naeher LP, Brauer M, Lipsett M, et al. Woodsmoke health effects: a review. Inhal Toxicol2007;19(1):67-106.

Northcross A, Chowdhury Z, McCracken J, Canuz E, Smith KR. Estimating personalPM2.5 exposures using CO measurements in Guatemalan households cookingwith wood fuel. J Environ Monitoring 2010;12:873–8.

Pope DP, Mishra V, Thompson L, et al. Risk of low birth weight and stillbirth associatedwith indoor air pollution from solid fuel use in developing countries. EpidemiolRev 2010;32:70–81.

Rehfuess E, Mehta S, Pruss-Ustun A. Assessing household solid fuel use: multiple impli-cations for the millennium development goals. Environ Health Perspect2006;114(3):373–8.

Rehfuess E. Fuel for life: household energy and health. Geneva: World Health Organi-zation; 2006.

Smith KR, Mehta S, Maeusezahl-Feuz M. Indoor air-pollution from household solid fueluse. In: Ezzati M, Lopez AD, Rodgers A, Murray CJL, editors. Comparative quantifi-cation of health risks: global and regional burden of disease attributable to selectedmajor risk factors. Geneva: World Health Organization; 2004. p. 1435–93.

Smith KR. Deadly household pollution: a call to action. Indoor Air 2006;16:2.Smith KR, Bruce N, Weber MW, et al. Reduction in child pneumonia due to randomized

introduction of a chimney stove in woodburning households: results from RESPIRE.International Society for Environmental EpidemiologyAnnual Meeting Paris, France;2006.

Smith KR, Dutta K, Chengappa C, et al. Monitoring and evaluation of improved biomasscookstove programs for indoor air quality and stove performance: conclusionsfrom the household energy and health project. Energ Sustain Dev 2007;11(2):5-18.

Smith KR, McCracken J, Thompson L, et al. Personal child and mother carbon monoxideexposures and kitchen levels: methods and results from a randomized trial ofwoodfired chimney cookstoves in Guatemala (RESPIRE). J Expo Sci Environ Epidemiol2009;20(5):406–16.

Sendecor GW, Cochran WG. Statistical method. 6th ed. Ames, Iowa: The Iowa stateUniversity press; 1980.

USEPA (United States Environmental Protection Agency). National Primary andSecondary Ambient Air Quality Standards; 2005.

USEPA (United States Environmental Protection Agency). Field Standard OperatingProcedures for the PM2.5 Performance Evaluation Program, vol. 441. ResearchTriangle Park, N.C: USEPA. Office of Air Quality Planning and Standards; 1998.

Volckens J, Boundy M, Leith D, Hands D. Oil mist concentration: a comparison of sam-pling methods. Amer Ind Hyg Assoc J 1999;60:684–9.

WHO (World Health Organization). Indoor air pollution: national burden of diseaseestimates. Geneva: World Health Organization; 2007.

WHO (World Health Organization). Interventions to reduce indoor air pollution.Retrieved January 27, 2009, from http://www.who.int/indoorair/interventions/en/.