Article Text

Chronic airflow obstruction and ambient particulate air pollution
1. Andre F S Amaral1,
2. Peter G J Burney1,
3. Jaymini Patel1,
4. Cosetta Minelli1,
5. Filip Mejza2,
6. David M Mannino3,
7. Terence A R Seemungal4,
9. Li Cher Lo6,
10. Christer Janson7,
11. Sanjay Juvekar8,
12. Meriam Denguezli9,
13. Imed Harrabi9,
14. Emiel F M Wouters10,
16. Kevin Mortimer12,13,
17. Rain Jogi14,
18. Eric D Bateman15,
19. Elaine Fuertes1,
20. Mohammed Al Ghobain16,
21. Wan Tan17,
22. Daniel O Obaseki18,
23. Asma El Sony19,
24. Michael Studnicka20,
25. Althea Aquart-Stewart21,
26. Parvaiz Koul22,
27. Herve Lawin23,
29. Olayemi Awopeju18,
30. Gregory E Erhabor18,
31. Thorarinn Gislason25,26,
32. Tobias Welte27,
33. Amund Gulsvik28,
34. Rune Nielsen28,29,
35. Louisa Gnatiuc30,
36. Ali Kocabas31,
37. Guy B Marks32,33,
38. Talant Sooronbaev34,
39. Bertrand Hugo Mbatchou Ngahane35,
40. Cristina Barbara36,
41. A Sonia Buist37
42. The BOLD (Burden of Obstructive Lung Disease) Collaborative Research Group
1. 1National Heart and Lung Institute, Imperial College London, London, UK
2. 2Centre for Evidence Based Medicine, 2nd Department of Internal Medicine, Jagiellonian University Medical College, Krakow, Poland
3. 3Preventive Medicine and Environmental Health, University of Kentucky, Lexington, Kentucky, USA
4. 4Clinical Medical Sciences, The University of the West Indies at St Augustine, Saint Augustine, Tunapuna-Piarco, Trinidad and Tobago
5. 5Respiratory Medicine, JSS Medical College and Hospital, Mysore, Karnataka, India
6. 6Department of Medicine, RCSI & UCD Malaysia Campus, Georgetown, Pulau Pinang, Malaysia
7. 7Respiratory, Allergy and Sleep Research, Department of Medical Sciences, Uppsala University, Uppsala, Sweden
8. 8Vadu Rural Health Program, King Edward Memorial Hospital Pune, Pune, Maharashtra, India
9. 9Laboratoire de Physiologie et des Explorations Fonctionnelles, Universite de Sousse Faculte de Medecine de Sousse, Sousse, Tunisia
10. 10Department of Respiratory Medicine, Maastricht University, Maastricht, The Netherlands
11. 11Service de Epidemiologie et Medecine Preventive, Universite Badji Mokhtar Annaba Faculte de Medecine, Annaba, Algeria
12. 12Clinical Sciences, Liverpool School of Tropical Medicine, Liverpool, UK
13. 13Respiratory Medicine, Aintree University Hospitals NHS Foundation Trust, Liverpool, UK
14. 14Lung Clinic, Tartu University Hospital, Tartu, Estonia
15. 15Division of Respiratory Medicine, University of Cape Town, Rondebosch, Western Cape, South Africa
16. 16Department of Medicine, King Saud bin Abdulaziz University for Health Sciences & King Abdullah International Medical Research Centre, Riyadh, Saudi Arabia
17. 17iCAPTURE Centre, The University of British Columbia, Vancouver, Ontario, Canada
18. 18Medicine, Obafemi Awolowo University, Ile-Ife, Osun, Nigeria
19. 19Director, Epi-Lab, Khartoum, Sudan
20. 20Department of Pulmonary Medicine, Paracelsus Medical University Salzburg, Salzburg, Austria
21. 21Department of Internal Medicine, The University of the West Indies at Mona, Mona, Saint Andrew, Jamaica
22. 22Pulmonary Medicine, SKIMS, Srinagar, Jammu and Kashmir, India
23. 23Occupational and Environmental Health, University of Abomey-Calavi, Cotonou, Littoral, Benin
24. 24Community Health Sciences, Aga Khan University, Karachi, Pakistan
25. 25Department of Sleep, Landspitali University Hospital, Reykjavik, UK
26. 26Medicine, University of Iceland, Reykjavik, Iceland
27. 27Respiratory Medicine, Medizinische Hochschule Hannover, Hannover, Germany
28. 28Department of Thoracic Medicine, Haukeland University Hospital, Bergen, Norway
29. 29Department of Clinical Science, University of Bergen, Bergen, Norway
30. 30Nuffield Department of Population Health, Oxford University, Oxford, UK
31. 31Department of Chest Disease, Cukurova University, School of Medicine, Adana, Turkey
32. 32Respiratory and Environmental Epidemiology, Woolcock Institute of Medical Research, Glebe, New South Wales, Australia
33. 33South Western Sydney Clinical School, University of New South Wales, Sydney, New South Wales, Australia
34. 34Department of Respiratory Medicine, National Center for Cardiology and Internal Medicine, Bishkek, Kyrgyzstan
35. 35Internal Medicine, Douala General Hospital, Douala, Cameroon
36. 36Institute of Environmental Health, Lisbon Medical School, Lisbon University, Lisboa, Portugal
37. 37Pulmonary and Critical Care Medicine, Oregon Health & Science University, Portland, Oregon, USA
1. Correspondence to Dr Andre F S Amaral, National Heart and Lung Institute, Imperial College London, London SW3 6LR, UK; a.amaral{at}imperial.ac.uk

## Abstract

Smoking is the most well-established cause of chronic airflow obstruction (CAO) but particulate air pollution and poverty have also been implicated. We regressed sex-specific prevalence of CAO from 41 Burden of Obstructive Lung Disease study sites against smoking prevalence from the same study, the gross national income per capita and the local annual mean level of ambient particulate matter (PM2.5) using negative binomial regression. The prevalence of CAO was not independently associated with PM2.5 but was strongly associated with smoking and was also associated with poverty. Strengthening tobacco control and improved understanding of the link between CAO and poverty should be prioritised.

• COPD epidemiology

This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

## Introduction

The most important cause of chronic airflow obstruction (CAO) is tobacco smoking. The Global Burden of Disease programme has suggested that air pollution is second only to smoking in determining loss of disability-adjusted life-years due to chronic respiratory disease.1 Evidence for this was obtained by applying the risk of disease associated with air pollution exposure, as estimated from various studies, to the known distribution of fine particulate matter (PM2.5) across the world.

In this analysis, we investigated the ecological association (ie, using aggregated data)2 between the prevalence of CAO, as estimated from a large multisite study, and levels of ambient PM2.5.

## Methods

The prevalence of CAO and the prevalence of smoking were estimated for 41 sites of the Burden of Obstructive Lung Disease (BOLD) study (online supplemental file for details).3 The level of poverty of each site was estimated from the gross national income (GNI) per capita at the time of the survey, using data from the World Bank.4 Annual mean PM2.5 levels (all composition, and dust and sea-salt removed (DSSR)) for each site coordinates and a 10 km radius buffer (site as centre) were obtained from a public dataset.5 6

### Supplemental material

The unit of our analysis was the site, and the analysis was stratified by sex (online supplemental file for details).

## Results

The prevalence of CAO across sites ranged from 3.5% to 23.2% in men, and from 2% to 19.4% in women (table 1). As expected, the prevalence of CAO was substantially lower among never smokers (online supplemental table S1).

Table 1

Survey date, prevalence of chronic airflow obstruction (CAO) and smoking in men and women, gross national income (GNI) per capita and annual mean PM2.5 levels for the 41 sites of the Burden of Obstructive Lung Disease study

The prevalence of smoking varied from 4.6% to 84.4% in men and from 0% to 61.3% in women. The levels of all composition PM2.5 ranged from 4 µg/m3 in Reykjavik (Iceland) to 68 µg/m3 in Karachi (Pakistan). The GNI varied from $1120 in Malawi to$51 250 in Saudi Arabia (table 1).

Lower PM2.5 levels were weakly correlated with a higher prevalence of CAO, in both sexes (figure 1A). Among never smokers (figure 1B) and when using DSSR PM2.5, there was no correlation (figure 1C).

Figure 1

Relation between prevalence of chronic airflow obstruction and annual mean levels of (a) PM2.5 (all composition, μg/m3) for the whole sample, (B) PM2.5 (all composition, μg/m3) for never smokers and (C) PM2.5 (dust and sea-salt removed, μg/m3) for the whole sample.

In both sexes, the prevalence of CAO was strongly positively associated with smoking and negatively associated with GNI. There was no association of prevalence of CAO with levels of PM2.5 (all composition) (table 2). The sensitivity analyses using all composition PM2.5 for a 10 km radius buffer and using DSSR PM2.5 showed no substantive difference from the main analysis (online supplemental tables S2–S4).

Table 2

Ecological negative binomial regression of chronic airflow obstruction against log(GNI), smoking and log(PM2.5), by sex

## Discussion

We were unable to show evidence of an ecological association between the prevalence of CAO and annual mean levels of PM2.5, although we have shown clear independent associations with the prevalence of smoking and GNI.

Our findings suggest that PM2.5 is unlikely to have a substantial effect on the prevalence of CAO. We have previously shown that indoor burning of solid fuels, another source of PM2.5, is also unlikely to be substantially associated with CAO,7 a conclusion supported by the findings of three large Chinese studies.8–10 Our findings are compatible with the large European ESCAPE project, which showed little evidence of an effect of any pollutant on the FEV1/FVC or its change over time.11

This analysis has several strengths. The aggregate data on prevalence of CAO and smoking were taken directly from the BOLD study. Spirometry was post-bronchodilator, and its quality was assured with a strong training programme and regular review of all spirograms in a quality control centre.

All ecological analyses have potential weaknesses. One is the temptation to ascribe the associations observed at the site level to similar associations at an individual level. In this instance, there is independent analysis showing the association of CAO with smoking12 and poverty13 at the individual level within the BOLD study.

Ecological analyses are also prone to confounding. There are strong ecological associations between the prevalence of smoking, GNI and PM2.5. The poorer countries have fewer smokers, less CAO and greater pollution levels. This probably explains the negative association of CAO with PM2.5 in the population as a whole, which was not seen for never smokers (figure 1B), or with DSSR PM2.5, or in the regression analysis adjusted for smoking prevalence and GNI.

Ecological analyses can be misleading if the average exposure in a site does not represent the exposure of those with the disease.14 Although there may be differences in pollution exposure within each site, these are likely to be small compared with the larger variation between sites, which ranged from 4 µg/m3 in Reykjavik (Iceland) to 68 µg/m3 in Karachi (Pakistan). It is unlikely that anyone living in Karachi will have exposure to ambient PM2.5 lower than any of those living in Reykjavik. The wide variation in income across sites is probably less well represented by GNI. Using the same estimate of GNI for rural and urban areas is likely to lead to more substantial errors than the approximations made for PM2.5. Nevertheless, we have found an association between poverty and CAO both at the ecological and individual levels in the BOLD study,13 and it is likely that the imprecision introduced here by using GNI to represent the site income has reduced the strength of association with CAO.

These results do not imply that air pollution is not harmful to lung growth in utero and during childhood, lung health or general health, and we clearly do not address in this study the potential of PM2.5 to cause other pathologies or to trigger acute exacerbations of disease. We cannot exclude the possibility that the toxicology of PM2.5 varies geographically, that a component of PM2.5 causes CAO but it is not always present, or that there is another pollutant that is highly correlated with PM2.5 in some sites that causes CAO. Several researchers have suggested that the properties15 or sources16 of particles may also be important in determining their effects.

This ecological study shows that, after adjustment for smoking and GNI, ambient PM2.5 is unlikely to explain a substantial amount of the prevalence of CAO, while the ecological association of smoking with CAO is strong and the association of poverty with CAO indicates that this is also likely to play an important role in its origins.

## Acknowledgments

The authors thank the participants and field workers of this study for their time and cooperation, and the BOLD (Burden of Obstructive Lung Disease) Coordinating Centre members for their technical and scientific support.

• ## Supplementary Data

This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

## Footnotes

• AFSA and PGJB are joint first authors.

• Collaborators The BOLD (Burden of Obstructive Lung Disease) Collaborative Research Group members