Characterizing associations between household energy-related exposures to air pollution and biological indicators of respiratory health

Abstract

In 2022, household air pollution (HAP) from solid fuel combustion contributed to an estimated 3.8 million premature deaths globally, with approximately 2.3 million associated with cardiopulmonary conditions (1). These burdens fell disproportionately on low- and middle-income countries (LMICs), where biomass fuels remained predominant sources of cooking, heating, and lighting (2). Despite progress toward cleaner energy transitions, many households remained reliant on biomass due to limited access to affordable and sustainable alternatives. Consequently, exposures to fine particulate matter (PM2.5) and black carbon (BC)—primary components of HAP—continued to pose significant public health threats. Cleaner-burning cookstove interventions have offered promising solutions to reduce exposures and promote health equity. Although the biological mechanisms linking HAP to respiratory outcomes are relatively well-established—particularly for chronic obstructive pulmonary disease and asthma (3,4)—important gaps remain. Evidence demonstrating consistent respiratory health benefits following cleaner stove interventions has remained limited and variable, especially in chronically exposed LMIC populations (5-8). Few studies have incorporated sensitive biomarkers of airway inflammation or examined lung function across diverse field settings (9,10). This dissertation addressed these gaps by evaluating the association between HAP exposure and biomarkers of respiratory health using data from two field studies: a randomized improved cookstove intervention in rural Honduras and a cross-sectional baseline assessment from a clean household energy trial in Rwanda. Fractional exhaled nitric oxide (FeNO), a non-invasive biomarker of airway inflammation, and spirometry-based lung function metrics—including forced expiratory volume in one second (FEV1), forced vital capacity (FVC), mid-expiratory flow rate (FEF25–75), and FEV1/FVC ratio—were assessed to characterize respiratory impacts across these settings. Study details and dissertation aims were as follows: Honduras (FeNO and Intervention Study) The Honduras intervention trial evaluated the impact of a cleaner-burning biomass stove intervention (Justa stove) on airway inflammation, as measured by FeNO, among women in rural Honduras. The project was built on existing partnerships between Colorado State University (CSU) and Trees, Water & People (TWP) in Fort Collins, Colorado, USA, and the Honduran Association for Development (Asociación Hondureña para el Desarrollo, AHDESA, Tegucigalpa, Honduras). The study area included 10 rural communities near the town of La Esperanza in the Department of Intibucá, Honduras. Traditional stoves typically burned wood and had griddles and homemade chimneys, but did not have an engineered combustion chamber. In Aim 1, for the intent-to-treat (ITT) analysis, we assessed the association between assigned stove type and FeNO using linear mixed models. We evaluated the impact of the Justa stove intervention on FeNO among a subset of 90 women selected from 230 participants, nested within the larger randomized intervention trial. Inclusion criteria for enrollment included the following characteristics: female, 24-59 years of age, primary cook of the household, and used a wood-burning traditional stove (i.e., no evidence of an engineered combustion chamber). Exclusion criteria included self-reported pregnancy at the time of recruitment, currently smoking, or regular exposure to secondhand smoke. The study employed a stepped-wedge design between 2015 and 2018, including 6 visits to each household, with households randomized to receive the improved wood-burning Justa stove, equipped with a chimney and engineered combustion chamber, either after Visit 2 or Visit 4. This subset provided 455 observations collected over a three-year study period. FeNO was measured repeatedly using the NIOX Vero device. In addition to this ITT analysis, actual stove use was evaluated via "per protocol" analyses utilizing self-reporting, where participants indicated whether they used only the Justa stove, combined the Justa with another improved stove, stacked the Justa with a traditional stove, or used only a traditional stove. In Aim 2, we utilized the same subset of 90 rural Honduras women to characterize the exposure-response relationship between FeNO and exposure to HAP measured repeatedly using the UPAS (Ultrasonic Personal Air Sampler) monitors and assessed as 24-hour gravimetric PM2.5 and BC as well as predicted long-term personal and kitchen exposures to PM2.5 and BC. For both Aims 1 and 2, we utilized linear mixed models to answer our research questions. The primary ITT included a spline for time and random effect for participant ID. Secondary ITT further adjusted for age, height, and socioeconomic status – potential confounders (i.e., to help overcome uneven distribution across arms in this subset of participants following randomization) and variables that might impact model precision. Exposure response models adjusted for age, height, socioeconomic status, kerosene use, and included a random effect for participant ID. For the per-protocol analysis, linear mixed models included a random effect for participant ID, self-reported primary stove use (categorized into three groups: Justa plus other improved stove, Justa plus traditional stove, traditional stove only) as the exposure variable of interest, and adjusted for age, kerosene use, height, socioeconomic status, and a spline for time. In the primary ITT analysis, the effect of the assigned Justa stove versus traditional stove on FeNO was negligible: percent change = 0.01% (95% CI: –2.4%, 4.3%). The secondary ITT model adjusted showed a percent change of -0.6(95% CI: –3.6%, 2.6%). Consistent inverse associations were observed in exposure-response models using 24-hour and predicted long-term exposures (all results are presented for a 25% increase in the exposure of interest). For personal PM2.5, a 25% increase in the 24-hour exposure was associated with a –0.95% change in FeNO (95% CI: –2.1%, 0.2%), and a 25% increase in predicted long-term exposure was associated with a –3.6% change in FeNO (95% CI: –6.3%, –0.8%). Kitchen PM2.5 showed similar inverse associations: –1.6% (95% CI: –2.5%, –0.8%) for 24-hour and –1.9% (95% CI: –3.4%, –0.4%) for long-term exposure. Personal BC exposure showed a –0.9% change in FeNO for 24-hour (95% CI: –1.5%, –0.3%) and –1.8% for long-term exposure (95% CI: –3.1%, –0.5%). Kitchen BC was similarly negatively associated: –1.0% (95% CI: –1.6%, –0.5%) for 24-hour and –1.4% (95% CI: –2.3%, –0.4%) for long-term exposure. Results were consistent to the exposure-response results, in per-protocol models. Compared to traditional stove users (reference, N = 236 observations), participants using both Justa and traditional stoves (N = 138 observations) showed a 3.5% increase in FeNO (95% CI: 1.3%, 5.8%), while Justa plus improved stove users (N = 81 observations) had a 2.3% increase (95% CI: –0.4%, 5.0%). These increases in FeNO, associated with reductions in HAP exposure observed over the course of the trial, were contrary to the initial hypothesis. Similar to findings in the smoking cessation literature, where quitting smoking leads to increased FeNO, these results may suggest that reducing chronic HAP exposure might similarly elevate FeNO, potentially reflecting complex feedback mechanisms involving nitric oxide synthase pathways. SHEAR: Lung Function and Cross-Sectional Exposure The Sustainable Household Energy Adoption in Rwanda (SHEAR) study is a randomized controlled trial conducted in rural Rwanda through in-country partnerships with Colorado State University, the University of Rwanda, and MeshPower Inc. The ongoing study substituted traditional household energy sources with solar power and liquefied petroleum gas (LPG) stoves to reduce household air pollution. Eligible households were recruited from the eastern lowlands of Rwanda, at least one adult and adolescent-aged child per household. Participants were followed for three years with repeated measurements of household air pollution exposure, energy use, and health outcomes. In Aim 3, we utilized baseline SHEAR data to conduct a baseline cross-sectional analysis to determine associations between HAP and lung function among adults and children (N = 1,460 individuals: 342 men, 542 women, and 579 children, from 650 households). Participants wore UPAS personal monitors to measure 48-hour PM2.5 and BC exposures. Spirometry was assessed using EasyOne software, validated by an independent reviewer, and metrics were standardized using race-neutral Global Lung Function Initiative (GLI) z-scores. Multivariable analyses included selected exposures of interest along with a weighted asset index in linear regression models. The lung function z-scores, which incorporated age and height based on the race-neutral American Thoracic Society/European Respiratory Society (ATS/ERS) reference equations, were used as the dependent outcomes in all models. The asset index was included as a covariate in all models, and biological sex was included in models for children to account for any potential confounding arising from sex-related differences in lung function development. Among adult men, higher PM2.5 was suggestively associated with modest increases in FEV₁ (0.152; 95% CI: –0.031, 0.335), FVC (0.065; 95% CI: –0.136, 0.266), FEF25–75 (0.171; 95% CI: –0.022, 0.365), and FEV1/FVC (0.155; 95% CI: –0.029, 0.339) z-scores. Higher BC exposure was associated with decreases in FEV1 (–0.080; 95% CI: –0.242, 0.083) and FVC (–0.223; 95% CI: –0.400, –0.047), and suggestive increases in FEF25–75 (0.122; 95% CI: –0.055, 0.300) and FEV1/FVC (0.171; 95% CI: 0.002, 0.339) z-scores. Among adult women, associations between PM2.5 and BC exposures and lung function were modest and inconsistent. For PM2.5, effect estimates were close to null for FEV1 (–0.029; 95% CI: –0.184, 0.127), FVC (–0.059; 95% CI: –0.217, 0.099) and FEV1/FVC (0.038; 95% CI: –0.106, 0.182, with a small positive trend for FEF25–75 (0.108; 95% CI: –0.076, 0.291). For BC, negative trends were observed across outcomes, though confidence intervals included the null. Among children, PM2.5 exposure was associated with small negative trends in FEV1 (–0.061; 95% CI: –0.221, 0.098), FEF25–75 (–0.092; 95% CI: –0.261, 0.078), and FEV1/FVC (–0.091; 95% CI: –0.224, 0.042), with minimal change in FVC (0.005; 95% CI: –0.157, 0.166). BC exposure showed similarly small negative associations. Effect modification plots revealed that among boys aged 10 and 13, associations between PM2.5 exposure and lung function (FEV1 and FVC) were relatively flat or slightly positive (i.e., consistent with the null association), indicating minimal impact, whereas among 16-year-old boys, there were pronounced inverse associations with both FEV1 and FVC z-scores, potentially suggesting a period of heightened susceptibility during adolescent lung development. In contrast, girls showed flatter associations across all age groups with no strong trends on lung function by exposure level. Interaction models further supported these findings, revealing significant age-by-sex modification for boys with PM2.5 exposure and both FEV1 (p-interaction = 0.0452) and FVC (p-interaction = 0.0051) z-scores. No significant effect modification was observed for BC exposures (p-interaction = 0.4632 for FEV₁ and 0.8881 for FVC). Overall, the results revealed heterogeneity in associations between exposure to PM2.5 and BC and lung function across all age and sex groups, with varying degrees of effect. Among men, exposure to BC was associated with reductions in several spirometric measures, including FEV1 and FVC, and slight increases in the FEV1/FVC ratio. These results, by which both FEV1 and FVC are decreasing with exposure and the FEV1/FVC ratio is increasing with exposure means that FVC is decreasing more than FEV1, which could suggest a restrictive process. While the observed associations in women and children were more variable and generally weaker, there were still suggestive trends of adverse effects on lung function, particularly in male children exposed to PM2.5. Conclusion This dissertation contributed novel evidence on the respiratory health effects of household air pollution in LMICs. In Honduras, while the Justa stove significantly reduced PM2.5 exposures, it did not improve airway inflammation as measured by FeNO. However, exposure-response analyses identified consistent inverse associations between PM2.5 and BC with FeNO, suggesting complex or possibly blunted inflammatory feedback responses. In Rwanda, while most results suggested no associations between HAP exposure and lung function in cross-sectional analyses, the associations observed with BC in men and with PM2.5 in the effect modification results among the children emphasize the need to consider age and sex in HAP-related exposure-response relationships. Our study is unique in that it evaluates the impact of household air pollution exposure on lung function across both adult men and adolescents, providing critical insights into demographic groups that have been historically understudied in this context. These findings underscore the importance of multifaceted approaches to evaluating health impacts of clean cooking interventions—including biomarker diversity, attention to vulnerable subgroups, and longitudinal follow-up—to inform policy and programming in LMICs.