Objective: We compared CO exposures and urinary PAH biomarkers pre- and postintervention with an improved biomass stove, the Patsari stove.

Methods: In a subsample of 63 women participating in a randomized controlled trial in central Mexico, we measured personal CO exposure for 8 hr during the day using continuous monitors and passive samplers. In addition, first-morning urine samples obtained the next day were analyzed for monohydroxylated PAH metabolites by gas chromatography/isotope dilution/high-resolution mass spectrometry. Exposure data were collected during the use of an open fire (preintervention) and after installation of the improved stove (postintervention) for 47 women, enabling paired comparisons.

Results: Median pre- and postintervention values were 4 and 1 ppm for continuous personal CO and 3 and 1 ppm for passive sampler CO, respectively. Postintervention measurements indicated an average reduction of 42% for hydroxylated metabolites of naphthalene, fluorene, phenanthrene, and pyrene on a whole-weight concentration basis (micrograms per liter of urine), and a 34% reduction on a creatinine-adjusted basis (micrograms per gram of creatinine). Pre- and postintervention geometric mean values for 1-hydroxypyrene were 3.2 and 2.0 μg/g creatinine, respectively.

Biomass fuels burned on open fires or traditional stoves emit a complex mixture containing thousands of chemicals present in the gaseous and/or particle phase of the smoke. Fine particulate matter (PM) [≤ 2.5 μm in aerodynamic diameter (PM_2.5)] is considered the best predictor of adverse health impacts associated with biomass smoke exposures (Naeher et al. 2007), but carbon monoxide (CO), an abundant gaseous health-damaging pollutant, is easier to measure. PM and CO are the pollutants that have been most frequently quantified as indicators of personal exposure to indoor air pollution (Smith et al. 2010). However, there is little information on how representative these indicator pollutants are of other health-damaging constituents of the smoke, for example, PAHs (Torres-Duque et al. 2008). Active PM or CO monitors are effective for accessing personal exposures, but they are expensive and difficult to implement on a large scale. Color stain diffusion tubes, originally designed for industrial hygiene applications, have been suggested as an easy, less expensive, and suitable alternative for large-scale use to assess personal CO exposures (Naeher et al. 2001; Smith et al. 2010).

An approach to reduce exposure misclassification is biomonitoring. Levels of biomarkers represent the absorbed dose of a chemical, integrated across all microenvironments and routes of exposure. Some biomarkers of exposure have been evaluated as metrics of biomass smoke exposures, for example, carboxyhemoglobin, exhaled CO, urinary methoxyphenol metabolites (Clark et al. 2007), and urinary PAH metabolites (Simpson and Naeher 2010; Torres-Dosal et al. 2008).

PAHs constitute a class of compounds composed of fused benzenoid rings and, like other pollutants from biomass combustion, are formed by the incomplete combustion of organic material. Low-molecular-weight two- and three-ring PAHs are mainly in the gas phase; four-ringed PAHs are in both gas and particle phase; and high-molecular-weight five- and six-ringed PAHs are mainly associated with particles (Bostrom et al. 2002). Urinary monohydroxylated PAHs (OH-PAHs) have been used as biomarkers to assess recent exposure to PAHs, with 1-hydroxypyrene (1-PYR) the most commonly used biomarker. Concentrations of these metabolites reflect exposure to PAHs within the previous few days. The half-life for urinary 1-PYR in humans has been reported to be in the range of 4.4–35 hr (Li et al. 2009). Information on half-lives of the other OH-PAHs is scarce; however, because these are the same group of metabolites formed under similar biological pathways, we can assume that other OH-PAHs would have similar half-lives. For nonoccupationally exposed populations, diet, ambient air, tobacco smoke, and coal-tar–containing medications are the main sources of PAH exposures [Centers for Disease Control and Prevention (CDC) 2005; Naeher et al. 2007; Torres-Duque et al. 2008]. The application of biomarkers in large population studies is limited due to the resources needed for biomonitoring assays and because of issues derived from following adequate protocols for obtaining and handling biological samples limit.

One approach to reduce the health burden related to biomass fuel use has been the provision of improved wood-burning stoves. The Patsari stove is an efficient multipot wood-burning cooking stove developed with a participatory approach to meet cooking needs, which has been disseminated in rural Mexico. This stove is built in place using a mold and locally available materials; the stove includes a big flat pan or comal, two secondary pots, and a chimney to lead the smoke to the outside. The size and inner dimensions of the stove were optimized for a more efficient combustion (Masera et al. 2007). The Patsari stove, compared with open fire, reduces median PM_2.5 concentrations by 71% near the stove and 58% in the kitchen (Zuk et al. 2007) and reduces mean greenhouse gases emissions by 74% (Johnson et al. 2008) and mean fuel consumption by 56% (Berrueta et al. 2007).

The aim of this study was to evaluate rural women’s exposure to CO and PAHs before and after installation of Patsari stoves. The study was performed using a small subsample because active personal exposure CO monitors and biomonitoring are costly and labor intensive, which limits their use for large-scale open population studies. An additional objective was to explore whether color-stain diffusion CO tubes could be used as a proxy measure of exposures to biomass burning pollutant emissions, such as PAHs, in future large-scale studies.

For the exposure biomarker evaluation study, which was performed during June–July 2005 and March–May 2006, we selected a subsample of 63 women from the Tanaco community (43 from the intervention and 20 from the control group). Tanaco is located at 2,140 m above sea level, and the average ambient temperature is 19°C (range, 8–32°C), with 48% relative humidity (range, 28–71%). There were no statistically significant differences between the women in the intervention and control groups. The participants received detailed information on the study requirements and provided written informed consent. The protocol was approved by the Bioethical and Biosecurity Committees of the Instituto Nacional de Salud Pública.

The exposure biomarker evaluation study was divided into three stages. In the preintervention stage, personal CO exposures were measured during the day while the women used a traditional open fire, and the first morning urine sample was obtained the next day to assess PAH biomarkers. During the intervention stage, a Patsari stove was installed in each participant’s home. During the postintervention stage, which took place 10 months after the Patsari stove was installed for the intervention group and 1 month after installation for the control group, CO and PAH measurements were repeated. Pre- and postintervention stages occurred in a similar season during the warmer months in this region. Information obtained during the follow-up visits for the main study indicated that some households did not fully adhere to the intervention (use of the Patsari stove) but continued using the open wood fire; these subjects were excluded in the postintervention stage of this study, leaving a total of 47 women (30 intervention and 17 controls) with complete pre- and postintervention data.

CO exposure assessment. We estimated personal CO exposures based on real-time measurements (continuous monitor) and cumulative samples (with diffusion tubes). We used continuous data-logging electrochemical CO monitors (Draeger Pac™III), with Draeger XS R CO sensors and a model D3T filter as supplied by the manufacturer (Draeger Safety Inc., Pittsburgh, PA, USA). The monitor was programmed to record CO in parts per million at 1-min intervals throughout the workday and was previously calibrated using CO concentrations of 50 ppm. We also used CO color-stain passive-diffusion tubes, with a detection range of 0–400 ppm-hr (CO Dositube; Gastec Inc., Kanagawa, Japan). The Draeger monitor and the passive diffusion tubes were attached to the clothing in the chest area of the women in the morning, before they started their daily activities in the kitchen. The average sampling time was 8 hr (range, 6–9.5 hr) for both stages. After the sampling time, the tubes were removed, capped, and read immediately by three different field technicians. Field technicians were trained to read the tubes using a standardized protocol, because reading has been reported to be the greatest source of error of this method (Harper 2001). The length of the stain in millimeters (range, 5–32 mm) was measured by each technician and converted to a cumulative exposure (parts per million-hour) according to a nonlinear calibration curve (y = 0.522x^1.7209), as done in a similar study in Guatemala (Smith et al. 2010). The average CO concentration from the three readings was divided by the sampling time to determine the passive sampler CO concentration (parts per million-hour). In addition, field technicians used a time–activity questionnaire to record the activities of the women outside and inside the kitchen at 15-min intervals during sampling.

Urine collection, PAH analysis, and exposure questionnaire. Containers were provided to the women to collect first morning urine samples, and they were instructed on sample collection procedures. Field workers visited each home to retrieve the urine samples and conduct a brief interview to obtain information about medication, tobacco smoke exposure, and cooking practices on the previous day. A food intake questionnaire was also administered that included questions about smoked and grilled meals (fish or meat) and coffee consumed during the previous day.

Urine samples (~ 100 mL) were refrigerated at 4°C during transportation to the field office (4–6 hr) and were then frozen at –70°C. These samples remained frozen at the Instituto Nacional de Salud Pública until they were sent to the CDC for analysis. Urine samples were analyzed for OH-PAH metabolites (Li et al. 2006). The measurements of urinary PAH metabolites were subjected to a series of quality control and quality assurance checks as described by Li et al. (2006). Creatinine concentrations were determined at the CDC by an enzymatic in vitro assay, using clinical chemistry analyzers (Creatinine Plus; Roche Diagnostics, Indianapolis, IN, USA). Metabolites were reported as fresh-weight concentrations (micrograms per liter of urine) and as creatinine-adjusted concentrations (micrograms per gram of creatinine) to correct for urine concentration/dilution in spot samples (Barr et al. 2005).

Cooking practices and other variables.
On average, the women spent 4.5 hr in the kitchen during the preintervention stage (open fire) and 5 hr during the postintervention stage (Patsari stove). The time spent for the different cooking tasks using an open fire (presented for the preintervention stage in Table 1) was similar to that in the postintervention stage for using the new Patsari stove (data not shown). We also collected information on other potential exposure sources to PAHs. Only one woman had a family member who smoked, and she reported being exposed to the second-hand smoke for < 1 hr/day. About one-third of the women drank coffee, but < 5% ate smoked or grilled foods (e.g., meat and fish) during the previous day. We found no significant differences before and after the intervention with respect to consumption of these foods. We developed linear regression models to estimate the effect of food intake variables on PAH biomarkers but found no significant associations (data not shown).

Personal CO exposure measurements. Personal exposure to CO was measured using continuous monitors and diffusion tubes (Table 2). The coefficient of variation of three readings of the tubes was 150% for preintervention and 100% for postintervention measurements. Continuous CO measurements were available for all participating women (n = 47), but valid CO passive samples were available for only 30 women.

For continuous data, we estimated average and maximum values for each sample. Summary statistics were then calculated for all samples. After the installation of the Patsari stoves, personal CO exposures using continuous monitors showed a 75% reduction in average median values and a 28% reduction in average maximum values. The intervention appeared to have a greater effect on the lower end of the exposure distribution (100% reductions at the 5th and 25th percentiles) than on the upper end of the distribution (50% reduction at the 95th percentile). Postintervention CO exposures based on personal passive samplers indicated a 66% reduction in the average median concentration. The correlation between continuous and passive average personal CO exposures was higher for the preintervention stage (Spearman rho = 0.57; p < 0.01) than for the postintervention stage (Spearman rho = 0.42; p = 0.02). The preintervention samples presented a broader range of CO concentrations than did samples from the postintervention stage.

PAH urinary biomarkers. The World Health Organization (1996) recommends that creatinine concentration in the range of 30–300 mg/dL be considered a representative sample, that is, not too dilute or too concentrated. In our case, the creatinine concentrations ranged from 36 to 304 mg/dL,
and we included all samples in the statistical evaluation of the data. We estimated parametric statistics for 10 analytes for which the frequency of detection was ≥ 60% (Barr et al. 2004), including metabolites of naphthalene, fluorene, phenanthrene, and pyrene [for creatinine-adjusted concentrations, see Table 3; for whole-weight concentrations, see Supplemental Material, Table 1 (http://dx.doi.org/10.1289/ehp.1002927)]. Nine urinary metabolites of high-molecular-weight PAHs (benzo[c]phenanthrene, benz[a]anthracene, chrysene, and benzo[a]pyrene) had a low detection frequency (0–23%) and were excluded from further statistical analysis.

OH-PAHs measured during the postintervention stage were significantly lower than those from the preintervention stage. Reductions in whole-weight concentrations ranged from 33% for 9_-hydroxyfluorene (9-FLUO) to 57% for 1_-hydroxynaphthalene (1-NAP), with a 42% total reduction for the sum of 10 metabolites [see Supplemental Material, Table 1 (http://dx.doi.org/10.1289/ehp.1002927)]. For creatinine-adjusted concentrations, reductions ranged from 20% (nonsignificant difference) for 9-FLUO to 48% for 1-NAP, with a 34% total reduction for the sum of 10 metabolites (Table 3).

CO measurements were associated with urinary PAHs metabolites at different indoor air pollution levels (Figure 1, Table 4). In general, we found a weak association between urinary PAH metabolites and CO personal exposures during the preintervention stage (more polluted conditions), whereas a stronger, and in many cases significant, association was found during postintervention measurements (less polluted conditions). Correlations between urinary PAH metabolites and personal CO exposures measured using passive samplers were weaker than those obtained with continuous personal CO exposures.

CO personal exposures showed significant reductions after the improved stove was introduced. The median for personal CO continuous measurements (4 ppm for open fire and 1 ppm for Patsari) resemble results from other studies performed in similar settings. A study carried out on another subsample of our intervention study population in Michoacán, in the community of Comachuén, showed mean kitchen CO concentrations of 2.3 ppm while using an open fire, and 0.5 ppm while using a Patsari stove, based on measurements made using a semicontinuous electrochemical CO sensor (HOBO monitor; Onset Corporation Inc., Bourne, MA, USA) (Armendáriz-Arnez et al. 2008). A similar study performed in Guatemala, also using the HOBO CO monitor, reported mean CO kitchen concentrations of 3.2 ppm when an open fire was in use and 0.8 ppm after an improved stove (Plancha) was installed (Clark et al. 2007).

Our findings for personal CO measured using a passive sampler (median, 3 ppm for open fire, and 1 ppm for Patsari) also resemble mean personal passive CO concentrations reported for women in the Guatemala study: 2.8 ppm for open fire and 1.4 ppm for the improved Plancha stove (Smith et al. 2010). All the results show significant reductions in CO exposures of a similar magnitude after the introduction of improved stoves.

The CO tubes are an easy-to-use, well-accepted, and accessible method for conducting personal exposure measurements, but they are not precision devices. Reading the length of stain is the single largest source of error associated with these samplers. The National Institute for Occupational Safety and Health evaluated different detector tubes and found that the accuracy of the measurements of many of them was 25–35% of the actual value (Harper 2001).

The complexity of a quantitative assessment of PAH exposures is related to the large number of individual compounds in a given mixture and by the presence of PAHs in both the gaseous and the particulate phases. The composition of emitted PAH mixtures depends on factors such as the type of fuel and its properties and the combustion technology (Bostrom et al. 2002). Comprehensive biomarker monitoring is necessary to characterize composition of the PAH mixture emitted from a biomass combustion process using either an open fire or an improved biomass stove. Data for such emission sources are scarce, and as shown here, the median reduction after an intervention is variable for different PAH metabolites.

Although four- to six-ring particulate-phase PAHs, notably benzo[a]pyrene, have been classified as known or probable human carcinogens, their concentrations in air tend to be low and difficult to measure (Sobus et al. 2009), and their metabolites are difficult to detect in urine samples (Toriba and Hayakawa 2007). Consistent with this, in the present study we did not measure detectable levels of high-molecular-weight PAH metabolites (benzo[c]phenanthrene, benz[a]anthracene, chrysene, and benzo[a]pyrene).

Generally, it is difficult to assess a carcinogenic threshold from urinary metabolites. Pyrene is not classified as a carcinogen, but it is present in the mixtures of PAHs formed during incomplete combustion (IARC 2010). The urinary metabolite 1-PYR has been widely used as a representative biomarker of PAH exposures. However, a variety of factors can influence the PAH mixture composition and benzo[a]pyrene:pyrene ratio; hence, a health-based exposure limit has not been determined (Jongeneelen 2004).

It was not possible to directly analyze PAH exposure routes in this study because we did not measure PAHs in air or food. However, there are no other important exposure sources of PAH exposure in ambient air for this study population, and we found no significant associations between PAH metabolites and the consumption of coffee or smoked and grilled meat and fish based on the dietary questionnaire.

Urinary 1-PYR has been established as a biomarker to evaluate PAH exposures (CDC 2005) because pyrene is normally abundant in PAH mixtures. Exposures occur via the diet, polluted air, and cigarette smoke, as well as in some occupational settings (Hecht 2002; Jongeneelen 2001; Toriba and Hayakawa 2007).

In the present study, even after the intervention, the urinary level for 1-PYR [GM, 2,025 ng/g creatinine; 95% confidence interval (CI), 1,621–2,528] was nine times higher than the concentration at the 95th percentile for females in the National Health and Nutrition Examination Survey (218 ng/g creatinine; 95% CI, 199–247) (CDC 2005). In a study carried out in rural Mexico, Torres-Dosal et al. (2008) evaluated a three-stage risk reduction program that included removing indoor soot adhered to roofs and internal walls, paving dirt floors, and introducing an improved wood stove with a chimney that vented smoke outside of the kitchen. In general, the introduction of an improved stove helps to reduce the soot deposits on cooking utensils and on the walls and ceiling of the kitchen, which also is a source of PAH exposure. The findings of Torres-Dosal et al. (2008) for urinary 1-PYR concentrations before and after the program (6.7 and 4.8 μmol/mol creatinine, equivalent to 13.0 and 9.3 μg/g creatinine, respectively) were much higher than our results presented here. In another study, Toriba and Hayakawa (2007) reported concentrations of urinary OH-PAHs for nonsmoking taxi drivers, traffic police officers, and rural villagers in Thailand. They found higher urinary concentrations of these metabolites in rural villagers, with mean concentrations of 1-PYR (1.2 μmol/mol creatinine, corresponding to 2.3 μg/g creatinine) that were similar to the levels in our study population. High urinary levels found for rural villagers were thought to be mainly associated with atmospheric PAHs emitted by open burning practices for agricultural purposes and with biomass (wood and charcoal) burning for cooking. Hecht (2002) reviewed studies reporting urinary 1-PYR levels for smokers in some European countries (mean levels < 1.0 μmol/mol creatinine, corresponding to 1.9 μg/g creatinine). Our results for rural women in the postintervention stage were on average higher than those reported in these studies for smokers.

Recently, attention has focused upon the more abundant gas-phase PAHs, notably naphthalene (two rings) and phenanthrene (three rings), as possible surrogates for PAH exposures (Sobus et al. 2009; Toriba and Hayakawa 2007). Generally, naphthalene, phenanthrene, and fluorene are the most abundant PAHs in the gaseous phase. Naphthalene has been classified as possibly carcinogenic to humans (group 2B), in contrast with phenanthrene and fluorene, which are not classifiable for carcinogenicity to humans (group 3) (IARC 2002, 2010). In the present study, we found the highest urine PAH metabolite concentrations for naphthalene metabolites, urinary 1-NAP and 2-NAP. In addition, we observed the largest percentage of reduction after the intervention for 1-NAP (57% for fresh weight, and 48% for creatinine-adjusted concentrations). After the Patsari stove intervention, the GMs for both naphthalene metabolites were between the 75th and the 90th percentile for the U.S. general female population (CDC 2005; Li et al. 2008). Toriba and Hayakawa (2007) reported that urinary naphthalene metabolites were higher among rural villagers than among taxi drivers and traffic police officers in Thailand (7.6 and 12.1 μmol/mol creatinine, corresponding to 9.7 and 15.5 μg/g creatinine for 1-NAP and 2-NAP, respectively). Our results for rural women were similar to those reported for rural villagers but higher than urinary 2-NAP levels reported for smokers in Korea (3.9 μmol/mol, corresponding to 5.0 μg/g creatinine) (Hecht 2002).

The postintervention GM for 2-hydroxyfluorene (2-FLUO) was similar to the 95th percentile reported for U.S. women (1,890 ng/g creatinine; 95% CI, 1,590–2,290), and 3-hydroxyfluorene (3-FLUO) levels were between the 90th (777 ng/g creatinine; 95% CI, 604–888) and 95th (1,030 ng/g creatinine; 95% CI, 923–1,270) percentiles for the U.S. general female population (CDC 2005).

(80 KB) PDF