Household air pollution and COPD: cause and effect or confounding by other aspects of poverty?

SETTING : Household air pollution (HAP) and chronic obstructive pulmonary disease (COPD) are both major public health problems, reported to cause around 4 million and 3 million deaths every year, respectively. The great majority of these deaths, as well as the burden of disease during life is felt by people in low- and middle-income countries (LMICs). OBJECTIVE AND DESIGN : The extent to which HAP causes COPD is controversial; we therefore undertook this review to offer a viewpoint on this from the Global Initiative for COPD (GOLD). RESULTS : We find that while COPD is well-defined in many studies on COPD and HAP, there are major limitations to the definition and measurement of HAP. It is thus difficult to disentangle HAP from other features of poverty that are themselves associated with COPD. We identify other limitations to primary research studies, including the use of cross-sectional designs that limit causal inference. CONCLUSION : There is substantial preventable morbidity and mortality associated with HAP, COPD and poverty, separately and together. Although it may not be possible to define clear causal links between HAP and COPD, there is a clear urgency to reduce the avoidable burden of disease these inflict on the world’s poor.


S U M M A R Y S E T T I N G :
Household air pollution (HAP) and chronic obstructive pulmonary disease (COPD) are both major public health problems, reported to cause around 4 million and 3 million deaths every year, respectively. The great majority of these deaths, as well as the burden of disease during life is felt by people in low-and middleincome countries (LMICs).

O B J E C T I V E A N D D E S I G N :
The extent to which HAP causes COPD is controversial; we therefore undertook this review to offer a viewpoint on this from the Global Initiative for COPD (GOLD). R E S U LT S : We find that while COPD is well-defined in many studies on COPD and HAP, there are major limitations to the definition and measurement of HAP. It is thus difficult to disentangle HAP from other features of poverty that are themselves associated with COPD. We identify other limitations to primary research studies, including the use of cross-sectional designs that limit causal inference. C O N C L U S I O N : There is substantial preventable morbidity and mortality associated with HAP, COPD and poverty, separately and together. Although it may not be possible to define clear causal links between HAP and COPD, there is a clear urgency to reduce the avoidable burden of disease these inflict on the world's poor. K E Y W O R D S : household air pollution; COPD; lung disease; LMICs The WHO estimates that roughly 4 million people die prematurely every year as a result of exposure to household air pollution (HAP). Of these deaths, around one third (1.5 million) are attributable to chronic obstructive pulmonary disease (COPD). The more recent Global Burden of Disease (GBD) study estimates suggest that the number of deaths related to HAP is lower than this, but nevertheless substantial, at 2.31 million deaths, 1 which accounts for 3.6% disabilityadjusted life-years (DALYs). 2,3 COPD itself is a leading cause of death globally -the third most common cause of death -responsible for just over 3 million deaths a year. 4 HAP is caused by the use of dirty-burning fuels to provide energy for cooking, heating and lighting. Almost all the people living with and dying from HAPassociated COPD are born into households in low-and middle-income countries (LMICs), by which time their exposures to COPD risk factors have already started. 5 This paper aims to highlight HAP and COPD as two major causes of avoidable global morbidity and mortality and to explore the nature of their relationship. We do this by posing four questions and suggesting priorities for action now and priorities for future research: 1) How is HAP measured in studies on HAP and COPD? 2) How is COPD diagnosed in studies on HAP and COPD? 3) Are the respiratory symptoms and spirometric abnormalities commonly observed in association with HAP exposure in LMICs due to COPD? 4) Is the phenotype of fixed airflow obstruction associated with HAP similar or different from tobacco smoke-induced COPD? relatively straightforward, its meaningful measurement has been challenging. 6 One of the big challenges is that HAP exposures can start in early life (from conception or before) and continue -but varythroughout life. There has thus been much reliance on the use of questionnaires to try to get a picture of lifecycle exposures, but these are rather blunt tools and are limited by recall bias. The development of a robustly validated estimate equivalent to cigarette smoking pack-years would be of great value; however, this has been proven elusive and is likely to remain so, given the greater difficulty of quantifying HAP exposure vs. the number of cigarettes smoked. No equivalent cooking quantity exists that is comparable to the relatively standard quantity of one cigarette. HAP exposure is often defined as 'exposed' vs. 'not exposed', but this ignores complexities of concentration, duration of exposure within a 24-h period, constituents, indoor vs. outdoor environments, age at which exposure starts (including prenatal), duration of cumulative exposure (e.g., cooking-years). Some of these complexities can be addressed by direct measurement of HAP exposure, for e.g., by measuring particulate matter (PM), carbon monoxide (CO), black carbon (BC), polycyclic aromatic hydrocarbons (PAHs), alveolar macrophage carbon loading. However, these involve new limitations including providing only a snapshot of a lifetime exposure, technical complexities and cost. PM 2.5 (particles with a diameter of 2.5 microns) and CO are the two pollutants that have been most commonly measured in HAP studies on respiratory health. Given the relatively large number of studies of exposure to ambient PM 2.5 and respiratory health in high-income countries, PM 2.5 has been the main focus of HAP research. Early HAP exposure studies showed 24-h indoor concentrations of PM 10 (particles with a diameter of 10 microns) generated from solid fuels in different settings to be in the range of 300-3,000 lg/m 3 , with peaks reaching as high as 20,000 lg/m 3 during cooking. [7][8][9] In a study from Guatemala, the 24-h average PM 2.5 generated by domestic wood burning was found to be 520 lg/m 3 , 10 which is similar to findings from other studies from Nepal (680 lg/m 3 ) 11 and China (489 lg/m 3 ). 12 These reported concentrations are an order of magnitude higher than the WHO global ambient air quality guidelines.
Most of the early studies linking exposure to HAP and COPD did not use measurements of pollutant levels but relied instead on self-report of cooking with solid fuels or not. One of the first studies to measure HAP in relation to measurement of airflow obstruction was conducted in Mexico. 8 In this study, peak 1-h indoor concentration of PM 10 often exceeded 2 mg/m 3 . Another early study that measured both exposure to HAP and lung function was from China and found mean kitchen concentrations of PM 10 in homes cooking with biomass fuels to be~1 mg/m 3 ; these measurements were made with a continuous reading instrument over 5-min periods. 13 One of the first studies to measure personal exposure as opposed to area concentrations of HAP was from the RESPIRE randomised controlled trial of a chimney cookstove intervention. In this study, exhaled CO (eCO) was measured repeatedly and an association between eCO and decline in forced expiratory volume in 1 sec (FEV 1 ) over an 18-month period was observed. 14 A recent study in Malawi measured personal exposure to both PM 2.5 and CO over a 48-h period at enrolment and annually over a 2year follow-up. 15 In this study, the median PM 2.5 (77 lg/m 3 ) and CO (1.27 ppm) were not associated with a decline in FEV 1 , although a caveat is that cooking is commonly done outdoors in Malawi, except when it rains. While personal monitoring is generally preferred over area monitoring, all of the studies mentioned here measured HAP for brief periods so only a ''snapshot'' assessment of exposure was obtained. Given that cumulative exposures to HAP that may contribute to the eventual development of COPD begin in childhood, [16][17][18] and perhaps even in utero or pre-conception, single or even multiple snapshot measurements of exposure in adulthood are likely insufficient.
Whether measured directly or indirectly, all measures of HAP are limited by the problem of confounding by other dimensions of poverty. It is therefore difficult to isolate the effects of HAP. This is well illustrated by the energy ladder, where the most dirty-burning (animal dung and crop residues) and clean (gas and electricity) fuels are at opposite ends of the ladder, both from HAP exposure and cost perspectives. Arguably, HAP is itself a dimension of poverty -air quality poverty -and therefore essentially impossible to disentangle from other dimensions of poverty outside of artificially created study conditions (e.g., laboratory studies, intervention trials). Many of these other dimensions of poverty (e.g., poor early life nutrition, limited access to healthcare, tobacco smoke and other noxious inhalational exposures) are themselves known to be associated with the development and progression of COPD. We are therefore left with uncertainty around the extent to which any measure of HAP is an indicator of the direct harms of this exposure vs. harms related to its association with other adverse health effects of poverty. There is, however, no doubt that HAP is a very real problem, with high and unhealthy exposures to PM 2.5 the daily norm for many people living in low-resource settings.

HOW IS COPD DIAGNOSED IN STUDIES OF HAP AND COPD?
In contrast to the challenges of measuring exposure to HAP, there is a straightforward definition of COPD based on the results of post-bronchodilator spirom-etry. GOLD defines COPD as a ''common, preventable and treatable disease that is characterized by persistent respiratory symptoms and airflow limitation that is due to airway and/or alveolar abnormalities usually caused by significant exposure to noxious particles or gases''. 4 In many LMIC households, levels of HAP certainly qualify as such an exposure. Domestic combustion of biomass fuels generates smoke that is not dissimilar to tobacco smoke -a mixture of irritant gases and particles with complex hydrocarbons (e.g., polycyclic aromatic hydrocarbons) on their surfaces -without the nicotine.
Many studies of HAP and COPD have used robust definitions of COPD with high standards of qualityassured spirometry, although access to this for clinical practice/diagnosis is often limited in LMICs. Where airflow obstruction is found in association with HAP exposure in referral centres, it has often been found to be relatively milder than that typically observed in heavy smokers and typically without emphysema on computed tomography imaging. However, many people with heavy HAP exposure have chronic respiratory symptoms consistent with chronic bronchitis. 19 A study of airway histology in non-smoking women with exposure to biomass smoke found airway remodelling with fibrosis without much emphysema. 20 The problem of chronic respiratory symptoms/disease associated with HAP is therefore wider than spirometry-defined airflow obstruction. There is increasing evidence that small airway disease occurs mainly in non-smoking women. The type of inhalation that occurs during cooking activity is likely to lower the risk of damage spreading beyond the small airways, thus leading to an airway-predominant COPD phenotype. 21 Recent evidence suggests that small airway disease is the predominant phenotype in non-smoking women, and that oscillometry may be a useful diagnostic test. 22

ARE THE RESPIRATORY SYMPTOMS AND SPIROMETRIC ABNORMALITIES COMMONLY OBSERVED IN RELATION TO HAP EXPOSURE IN LMICS DUE TO COPD?
Although a considerable burden of disease has been attributed to COPD due to HAP exposure, this is somewhat controversial as there are conflicting data and opinions. [23][24][25][26][27][28] Given that there have been several recent systematic reviews and meta-analyses examining this association, we brought these together (see Box 1, Supplementary Data, for search strategy) to help answer this question (Table 1). These reviews draw on a similar pool of studies and data, but differ in methodology (e.g., criteria for study inclusion, definitions of exposure/outcome, etc.). These are nevertheless consistent in finding a positive association between HAP and COPD. They are also consistent in being affected by the limitations in the definition of HAP exposure (questionnaire-based) and outcome assessment (often without spirometry or spirometry without bronchodilator administration), as discussed above. Therefore, these reviews leave uncertainty as to the extent to which the association can be explained mechanistically by exposure to HAP vs. other dimensions of poverty. In addition, most of the studies included in the systematic reviews and meta-analyses are cross-sectional which limits causal inferences, and are subject to several other limitations.
Two large studies from China that assessed the impact of cookstove, cleaner fuel and/or ventilation interventions support a HAP-COPD association. 29,30 Evidence of reduced incidence of COPD was observed in a retrospective cohort study after a major campaign to provide chimney stoves in a rural area of China. 29 In a longitudinal study over 9 years, two interventions -biogas and kitchen ventilation -were compared to cooking as usual. 30 Use of biogas and improved ventilation were associated with a reduced decline in FEV 1 and risk of COPD; use of both interventions had greater effect than either alone. 30 To examine the association of exposure to HAP and COPD further, and in particular, to determine how to assess exposure with a possible mechanistic association with COPD, we searched for original research papers on HAP and COPD published in the last 5 years (see Box 2, Supplementary Data, for search strategy) that assessed both post-bronchodilator airflow obstruction and quantified exposure to PM 2.5 (Table 2); 44 studies were identified but only four assessed both post-bronchodilator airflow obstruction and quantified exposure to PM 2.5 . [31][32][33][34] One of these studies was the cross-sectional analysis of the baseline Malawi study discussed above. 34 A study from India showed a low prevalence of COPD and no association with biomass smoke exposure, 33 and in a study from Tanzania with a substantial prevalence of COPD, the authors were not able to show an association because 99.5% of the study population was exposed to biomass smoke. 32 In the study from Peru that did report an association between exposure to biomass PM 2.5 and COPD, there was likely confounding with tobacco smoking, high altitude and age. 31 Thus, recent studies in adults have not been particularly useful in shedding light on whether there is a mechanistic relationship between exposure to HAP and the development of COPD independent of other poverty-associated risk factors.
Looking earlier in life to the early origins of COPD, there are several studies of HAP exposure and lung function in children which provide evidence of impaired lung function in HAP-exposed children. [15][16][17]35 Results from the CRECER Study in Guatemala (a follow-up to the RESPIRE Trial) suggested that reduced exposure to HAP through the use of a chimney stove intervention during Another six studies measured COPD in women exposed to biomass fuels and women exposed to LPG, gasoline or oil; 29  pregnancy and infancy could improve lung function as measured by spirometry later in childhood. 16 In the Ghana Randomized Air Pollution and Health Study (GRAPHS), 48-h personal monitoring of CO exposures was done four times over the course of pregnancy in a cluster-randomised intervention trial of cleaner-burning cookstoves, and lung function parameters, including passive respiratory system compliance with the single-occlusion technique, were measured in the infant offspring of the monitored mothers at 1 month of age. The primary finding was that maternal CO exposure during gestation was associated with lower infant lung function. 17 In a substudy of the Cooking and Pneumonia Study (CAPS), a trial of cleaner-burning biomass-fuelled cookstoves in Malawi, children from intervention households had lower carboxyhaemoglobin and higher forced vital capacity (FVC) than controls. 15 While these studies have limitations similar to those discussed in the COPD literature, especially with regard to the snapshot nature of HAP exposure measurement, their results suggest that exposure to HAP in early life is a risk factor for poorer lung function, possibly COPD, later in life.

IS THE PHENOTYPE OF FIXED AIRFLOW OBSTRUCTION ASSOCIATED WITH HAP SIMILAR OR DIFFERENT FROM TOBACCO-SMOKE COPD?
There is a school of thought according to which COPD should be redefined according to the underlying aetiology. Such an approach would be valuable in directing attention to the non-smoking-related types of COPD about which little is known (particularly in relation to morbidity, mortality and treatment) compared to tobacco smoking-related COPD.
In contrast to some of the other types of COPD, COPD related to HAP exposure seems likely to have more in common with tobacco smoking-related COPD since tobacco is itself biomass. In fact, at least two studies have reported that the survival of spirometry-diagnosed biomass-related COPD in non-smokers is similar to that of tobacco-related COPD. [36][37][38] However, as there are clear differences between the two (see Table 3 for summary), it would be wrong to extrapolate fully from tobacco smoking-related to HAP-related COPD. An additional complexity is that in settings where HAP is common, so are other drivers of COPD, including sub-optimal lung development in early life, tobacco smoking, infections (particularly TB) and poorly treated asthma. Pure HAP-related COPD may therefore be uncommon and is more likely to be observed as part of what might be termed a mixed aetiology COPD phenotype.
Another issue on the theme of lung function and HAP is that lung function abnormalities other than fixed airflow obstruction appear to be more common in LMICs (where HAP exposure is also common). 25,39 Low FVC appears to be particularly common, although defining low FVC is highly problematic due to the limitations of the available lung function reference ranges. [40][41][42] Even less is known about the phenomenon of low FVC (e.g., in relation to morbidity, mortality, treatment) compared to airflow obstruction in LMICs. What does seem clear, however, is that low FEV 1 and low FVC are both associated with increased mortality across multiple populations, with some evidence from the PLATINO study from three Latin-American sites 43,44 that FEV 1 is a better predictor of overall and cardiovascular mortality than FVC. The high burden of impaired lung function in LMIC populations is therefore of great concern irrespective of causality.

WHERE DO WE GO FROM HERE?
There is no doubt that there is substantial preventable morbidity and mortality associated with both HAP and COPD as these are currently measured and incorporated into the GBD study and WHO estimates. They are also both strongly associated with poverty, which is itself a major cause of preventable morbidity and mortality. Ideally, we would be able to disentangle 'air quality poverty' from other forms of poverty, but these are conundrums within conundrums that limit our ability to do so. If we are to accept these imperfections and uncertainties, where does that leave us and where do we go from here? It certainly leaves us with a substantial burden of preventable morbidity and mortality in need of solutions, so we suggest that the way forward is to look for search for such solutions. We recommend that this search focuses on the development and evaluation of the clinical and cost-effectiveness of interventions to address 'air quality poverty' and other dimensions of poverty associated with COPD as these point us on the path to improved lung health and beneficial impacts. Of note, almost all trials of interventions to address HAP have, to date, concentrated on cooking stove/fuel interventions and these have not yielded the expected health benefits. 45,46 It is therefore important that interventions that go beyond cooking-related HAP exposures, explore how we can achieve clean (clean enough) air and examine how other dimensions of poverty can be addressed at the same time are developed and formally evaluated.
Interventions that are known to benefit people both at risk of developing and with established COPD are out of reach for the majority living in LMICs, and these need to be made more widely available and affordable. 4 It will be important that as interventions are developed and implemented, this is done with careful consideration of cost-benefits, as well as impacts on women, the environment and planetary health.
A large multi-country randomised controlled intervention trial of liquified petroleum gas (LPG) is ongoing and hopefully, when completed, will shed some light on the efficacy of this intervention to improve the health of both children and adults. 47 Even if shown to be efficacious, it will remain to be seen whether LPG can be feasibly provided to the poorest and most rural populations in LMICs. However, clean air during adulthood could be too late if the main impact of pollutants occurs during prenatal and postnatal lung growth and development. If this is the case, adequate trials should include at least intermediate outcomes at birth and during growth, increasing considerably the complexity and cost. 48 Finally, there is a long history of intervention implementation (e.g., 'improved cookstoves') for HAP at great cost despite minimal (or no) direct evidence of health benefits; it is especially important to avoid such wasteful approaches and concentrate on the generation of a robust evidence base to ensure appropriate use of limited resources in order to address HAP and other dimensions of poverty.

CONCLUSION
There is substantial preventable morbidity and mortality associated with HAP, COPD and poverty, separately and together. Although it may not be possible to establish clear causal links between HAP and COPD, there is a clear urgency to reduce the avoidable burden of disease they inflict on the world's poor. L'ampleur de la responsabilité de la HAP dans la COPD est controversé. Nous avons donc réalisé cette revue pour présenter le point de vue du Global Initiative for COPD (GOLD). R É S U LTAT S : Bien que la COPD soit bien définie dans de nombreusesétudes sur la COPD et la HAP, des limites considérables en matière de définition et de mesure de la HAP ontété observées. Il est donc difficile de dissocier la HAP d'autres caractéristiques de la pauvreté, qui sont elles-mêmes associéesà la COPD. Lesétudes de recherche primaire présentaient d'autres limites, dont l'utilisation de méthodes transversales qui limitent l'inférence causale. C O N C L U S I O N : Une part conséquente de la morbidité et de la mortalité associéesà la HAP, la COPD et la pauvreté peutêtreévitée, que ces dernières soient considérées séparément ou ensemble. Bien qu'il ne soit pas nécessairement possible de définir des liens de causalité clairs entre la HAP et la COPD, il est urgent de réduire le poids sanitaireévitable de ces maladies sur les pays pauvres.